﻿
 UFDC Home myUFDC Home  |   Help
<%BANNER%>

# Finite Element Analysis Using Uniform B-Spline Approximation and Implicit Boundary Method

## Material Information

Title: Finite Element Analysis Using Uniform B-Spline Approximation and Implicit Boundary Method
Physical Description: 1 online resource (175 p.)
Language: english
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

## Subjects

Subjects / Keywords: boundary, bsplines, c1, c2, conforming, continuity, fem, grid, implicit, mesh, non, structured, xfem
Mechanical and Aerospace Engineering -- Dissertations, Academic -- UF
Genre: Mechanical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

## Notes

Abstract: The finite element method (FEM) is a widely used numerical technique for solution of boundary value problems arising in engineering applications. FEM is well studied and has several advantages; however the stress and strain solutions obtained by FEM are discontinuous across element boundaries and require smoothing algorithms to make the stress/strain appear smooth in the analysis domain. Automatic mesh generation is still a difficult task in FEM for complicated geometries and often requires user intervention for generating quality meshes. Finite elements based on uniform B-spline basis functions defined on a structured grid are developed which provide at least continuous solution through out the analysis domain. Numerical techniques are presented in this work based on structured grids and implicit equations of the boundaries of analysis domain, material interface boundaries or coarse/fine grid interfaces for solution of various engineering boundary value problems. Solution structures are constructed using approximate Heaviside step functions for imposition of Dirichlet boundary conditions, treatment of material discontinuity to perform micromechanical analysis and for local refinement of grid. The use of structured grid eliminates the need for constructing a conforming mesh and results in significant savings of time in pre-processing stage of the design cycle. Numerical examples are presented to demonstrate the performance of B-spline elements. The results are compared with analytical solutions as well as traditional finite element solutions to demonstrate the ability of B-spline elements to represent continuous stress and strain through out analysis domain. Convergence studies show that B-spline elements can provide accurate solutions for many engineering problems with fewer numbers of elements and nodes as compared to traditional FEM. Solution structure for treatment of material boundary is validated by performing a convergence analysis on a problem involving circular inclusion in a square matrix and by determining effective properties of fiber reinforced composite. Solution structure for local grid refinement is validated by analyzing classical stress concentration problems.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.

## Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0021901:00001

## Material Information

Title: Finite Element Analysis Using Uniform B-Spline Approximation and Implicit Boundary Method
Physical Description: 1 online resource (175 p.)
Language: english
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2008

## Subjects

Subjects / Keywords: boundary, bsplines, c1, c2, conforming, continuity, fem, grid, implicit, mesh, non, structured, xfem
Mechanical and Aerospace Engineering -- Dissertations, Academic -- UF
Genre: Mechanical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

## Notes

Abstract: The finite element method (FEM) is a widely used numerical technique for solution of boundary value problems arising in engineering applications. FEM is well studied and has several advantages; however the stress and strain solutions obtained by FEM are discontinuous across element boundaries and require smoothing algorithms to make the stress/strain appear smooth in the analysis domain. Automatic mesh generation is still a difficult task in FEM for complicated geometries and often requires user intervention for generating quality meshes. Finite elements based on uniform B-spline basis functions defined on a structured grid are developed which provide at least continuous solution through out the analysis domain. Numerical techniques are presented in this work based on structured grids and implicit equations of the boundaries of analysis domain, material interface boundaries or coarse/fine grid interfaces for solution of various engineering boundary value problems. Solution structures are constructed using approximate Heaviside step functions for imposition of Dirichlet boundary conditions, treatment of material discontinuity to perform micromechanical analysis and for local refinement of grid. The use of structured grid eliminates the need for constructing a conforming mesh and results in significant savings of time in pre-processing stage of the design cycle. Numerical examples are presented to demonstrate the performance of B-spline elements. The results are compared with analytical solutions as well as traditional finite element solutions to demonstrate the ability of B-spline elements to represent continuous stress and strain through out analysis domain. Convergence studies show that B-spline elements can provide accurate solutions for many engineering problems with fewer numbers of elements and nodes as compared to traditional FEM. Solution structure for treatment of material boundary is validated by performing a convergence analysis on a problem involving circular inclusion in a square matrix and by determining effective properties of fiber reinforced composite. Solution structure for local grid refinement is validated by analyzing classical stress concentration problems.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis: Thesis (Ph.D.)--University of Florida, 2008.

## Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2008
System ID: UFE0021901:00001

Full Text

FINITE ELEMENT ANALYSIS USING UNIFORM B-SPLINE APPROXIMATION AND
IMPLICIT BOUNDARY METHOD

By

RAVI KUMAR BURLA

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2008

2008 Ravi Kumar Burla

To my parents

ACKNOWLEDGMENTS

I would like to express my sincere gratitude to my advisor and chairman of my supervisory

committee, Dr. Ashok V. Kumar, for his guidance, encouragement, enthusiasm and constant

support throughout my doctoral research. I would like to thank him for the numerous insights he

provided during every stage of the research. Without his assistance it would not have been

possible to complete this dissertation.

I would also like to thank the members of my advisory committee, Dr. Bhavani V. Sankar,

Dr. Nam-Ho Kim and Dr. Jorg Peters. I am grateful for their willingness to serve on my

committee, for providing help whenever required, for involvement and valuable suggestions in

my oral qualifying examination, and for reviewing this dissertation.

My colleagues at Design and Rapid Prototyping Laboratory at the University of Florida

deserve special thanks for their help and many fruitful discussions. I would especially thank

Sanjeev Padmanabhan, Greg Freeman and Sung Uk-Zhang.

I would like to thank my parents and my brothers for their constant love and support

without which this would not have been possible.

page

A C K N O W L E D G M E N T S .............. ..... ........ ..............................................................................4

LIST OF TA BLES .............. ......... ....................................................... 8

LIST O F FIG U RE S ............................................................................... 9

ABSTRACT ............................................ .. ......... ........... 14

CHAPTER

1 INTRODUCTION ............... .............................. ............................ 16

1 .1 O v erv iew ...................................... .................................... ................ 16
1.2 Goals and Objectives .................................................. .................. ......... 19
1 .3 O u tlin e ...........................................................................................19

2 ANALYSIS TECHNIQUES FOR BOUNDARY VALUE PROBLEMS ...............................21

2 .1 F inite E lem ent M eth od ....................................................... .................. ....................2 1
2.2 B-Spline Approximation Based M ethods................................ ................................. 22
2 .3 Structured G rid M methods ......................................................................... ...................23
2.4 M eshless M methods ..................... .. ................................................. ..... ................. 24
2.5 Techniques for Determination of Effective Properties of Composite Microstructures ....27
2.6 Techniques for Local Refinem ent ............................................................................. 28

3 SOLUTION STRUCTURES FOR IMPLICIT BOUNDARY METHOD .............................30

3.1 Solution Structure for Enforcing Essential Boundary Conditions....................................31
3.1.1 Modified Weak Form for Linear Elasticity .........................................................32
3.1.2 Modified Weak Form for Steady State Heat Transfer................ ..................34
3 .1.3 D irichlet F unction .......... .............................................................. .. .... ..... .. 36
3.1.4 B boundary V alue Function............................................................. ................. 38
3.2 Solution Structure for M icromechanics................................ ......................... ........ 40
3.3 Solution Structure for Grid Refinem ent ........................................ ....................... 44
3.3.1 Solution Structure ................. ............... ..................... ........... 45
3.3.2 M modified W eak F orm ..................................................................... ..................47

4 B-SPLINE FINITE ELEMENT FORMULATION.................................... ............... 57

4.1 One D im ensional B-Spline Elem ents ........................................ ......................... 58
4.2 Two and Three Dimensional B-Spline Elements .................................. ...............60

5 NUMERICAL IMPLEMENTATIONS.....................................67

5.1 Numerical Implementations for Linear Elastic Problems ..........................................67
5.1.1 Computation of Stiffness Matrix for Internal Elements.......................................68
5.1.2 Computation of Stiffness Matrix for Boundary Elements................................. 68
5.1.3 Computation of Stiffness Matrix for Elements with Essential Boundaries............69
5.1.4 Com putation of Load V ector ....................................................... ............... 73
5.2 Numerical Implementations for Steady State Heat Transfer................ .............. ....74
5.3 Numerical Implementations for M icromechanics ................................. ................ 76
5.3.1 Computation of Stiffness Matrix in the Matrix Region.......................................79
5.3.2 Computation of Stiffness Matrix in the Inclusion Region............... .................79
5.3.3 Computation of Stiffness Matrix in the Interface Region ....................................80
5.3.4 Periodic B oundary Conditions ........................................ .......................... 82
5.3.5 Computation of Effective Properties ....................................................... 84
5.3.6 Com putation of Load V ector.............................................. .............................. 87
5.4 Numerical Implementations for Local Grid Refinement.................. .................87
5.4.1 Computation of Stiffness Matrix in Regions 1 and 2..........................................88
5.4.2 Computation of Stiffness M atrix in Region 3 ................................ ............... 89
5.4.3 Computation of Load Vector................ .... ...... ..................90
5.5 Extrapolation of Solution for Elements with Low Volume Fraction ............................90
5.5.1 Construction of Extrapolation M atrix ......................... ........... ............... .... 92
5.5.2 Construction of Extrapolation Matrix for a Boundary Element..........................94
5.5.3 Computation of Stiffness M atrix ............................................................. 94

6 A N A L Y SIS A N D R E SU L T S...................................................................... ...................103

6.1 Validation of Solution Structure for Essential Boundary Conditions ..........................103
6.1.1 P watch T est ................... ............ ... ..................... .... .... ........ 103
6.1.2 V alidation of Extrapolation Technique ............................................................. 106
6.1.3 Convergence Analysis of a Cantilever Beam ...................................................... 107
6.1.4 Square Plate with a Circular Hole ........................................................ .......... 110
6.1.5 Thick Cylinder with Internal Pressure ....................................... ................111
6.1.6 Steady State Heat Transfer Analysis ...................................... ...............112
6.1.7 A analysis of a 3D Clam p ...................................................................... 113
6.1.8 Analysis of a 3D Rib Structure ................................................................. 114
6.1.9 Analysis of a 3D Dial Bracket .................................................... ...............115
6.2 Validation of Solution Structure for Micromechanics..............................116
6.2.1 Square Plate with Circular Inclusion.................... ..... ................116
6.2.2 Determination of Effective Properties of a Fiber Reinforced Composite ..........17
6.3 Validation of Solution Structure for Grid Refinement ............................118
6.3.1 Application of Grid Refinement to Constant Stress Problem ............................1118
6.3.2 Stress Concentration Problem ..................................... ......................... .......... 119

7 CONCLUSIONS AND FUTURE WORK............................................... ............... 165

7 .1 C o n clu sio n s....................................................... ................ 16 5
7 .2 F utu re W ork ...........................................................................167

L IST O F R E F E R E N C E S .............................................................................. ..........................169

B IO G R A PH IC A L SK E T C H ............................................................................... ............... ..... 175

LIST OF TABLES

Table page

5-1 Set of periodic boundary conditions for 6 different strains .................. ............... 102

6-2 Comparison of effective properties................................................ 163

LIST OF FIGURES

Figure page

3-1 Solution structure for imposing essential boundary conditions............... ...................49

3-2 Elastic solid under equilibrium ............................................... ............................... 49

3-3 Steady state heat transfer problem ............................................. ............................ 50

3-4 Boolean operations on Dirichlet functions ............................................. ............... 50

3-5 A CSG tree representing Dirichlet function for multiple boundaries..............................51

3-6 Boundary value function using cubic B-spline basis.............................. ............... 52

3-7 Representation of grids for material discontinuity.............. ..... .................. 53

3-8 One dimensional inclusion function for treatment of material discontinuity ....................53

3-9 Two dimensional inclusion function for an elliptic inclusion in a square matrix..............54

3-10 Representation of solution structure for local grid refinement............... ................ ...54

3-11 Approximate step function for refinement................................... .......................... 55

3-12 Coarse and refined grid interface with multiple implicit equations..............................55

3-13 Step function for refinement for a 2D model.................... ..........................56

4-1 Approxim ation using ID cubic B-splines...................................... ........................ 64

4-2 Shape functions for ID quadratic B-spline element .................................. ............... 64

4-3 Shape functions for ID cubic B-spline element ..................................... .................64

4-4 Typical grid used for analysis and the mapping used for geometry and field variable.....65

4-5 Two dimensional quadratic B-spline elements in parametric space..............................66

4-6 Two dimensional cubic B-spline elements in parametric space .....................................66

5-1 Typical non-conforming grid used for analysis...................................... ............... 95

5-2 Triangulation of boundary element for volume integration........................................96

5-3 Representation of band on essential boundaries......................... ....... ........................... 96

5-4 Approximation of natural boundary for computation of load vector..............................97

5-5 Division of computational domain into three regions............................................ 98

5-6 Undeformed and deformed configuration of a set of 4 RVEs ........................ ..........98

5-7 Representative volume element used for computation of effective properties.................99

5-8 Division of computation domain into various regions...........................................99

5-9 Typical analysis grid with elements having low volume fraction .................................100

5-10 Extrapolation of external node for linear Lagrange element ...................................100

5-11 Extrapolation of external nodes for cubic B-spline element...................................100

5-12 Extrapolation of external nodes for quadratic B-spline element .................................... 101

5-13 Extapolation for boundary elem ent...................................................................... ...... 101

6-2 Non-conforming grid using one cubic B-spline element................................................121

6-3 P lot of D irichlet functions............................ .................................................. .......... 122

6-4 Effect of range parameter on the quality of solution .............. ....... ............... 123

6-5 Contour plot of o-x for one element model using cubic B-splines ...............................123

6-7 Plot of error in the tip deflection with varying aspect ratio............................................124

6-9 Plot of error in strain energy for curved beam (without extrapolation) ...........................125

6-10 Plot of error in strain energy for curved beam using extrapolation technique...............126

6-11 Contour plot of o- for Q4 elements (without extrapolation)........................... ........126

6-12 Contour plot of o- for Q4 elements using extrapolation technique .............................127

6-13 Contour plot of o-, using cubic B-spline elements .................................. ............... 128

6-14 Thick cantilever beam used for convergence analysis.........................................128

6-15 Typical non-conforming grid using cubic B-spline elements............... ..................129

6-16 Convergence plot of L2 error for cantilever beam .................................. ...............129

6-17 Convergence plot for error in strain energy for cantilever beam................................130

6-18 Comparison of deflection curves for 10 element model....................... ...............131

6-19 Comparison of normal stress along the line AB for 10 element model..........................132

6-20 Comparison of shear stress along the line AB for 10 element model............................ 133

6-21 Convergence plot for tip deflection in near incompressible limit.................................133

6-22 Analysis model for a square plate with a circular hole....................................................134

6-23 Contour plot of ox using FEM Q4 elements ............................................... ............. 134

6-24 Contour plot of o- using cubic B-spline elem ents......................................................... 135

6-25 Convergence plot for L2 error norm for stress concentration problem ..........................135

6-26 Convergence plot for error in strain energy for stress concentration problem ..............136

6-27 P lot of o- along the line 0 = ........................................................................... 136
2

6-28 Hollow thick cylinder subjected to uniform pressure............ ............... ............... 137

6-29 Contour plot of radial displacements using cubic B-spline element model.....................137

6-30 Contour plot of radial displacements using FEM Q4 elements.................. ............. 138

6-31 Comparison of radial displacements along the thickness of the cylinder........................138

6-32 Comparison of radial stresses along the thickness of the cylinder ...............................139

6-33 Comparison of tangential stresses along the thickness of the cylinder............................139

6-34 H eat conduction in a block with series of pipes ............................................................. 140

6-35 P lot of D irichlet function ....................................... ............ ............ ............................140

6-36 Plot of boundary value function.............................................. ............................. 141

6-37 Contour plot of temperature distribution using FE Q4 elements................................... 141

6-38 Contour plot of temperature distribution using cubic B-spline elements ........................142

6-39 Contour plot of temperature distribution using quadratic B-spline elements................142

6-40 Comparison of temperature along line AB ................................... .......... ................. 143

6-41 Three dimensional clamp used for analysis................................ ....................... 143

6-42 Convergence plot for strain energy for clamp model ............... ................... ...........144

6-43 Contour plot of bending stresses using H8 elements in FEM for 3D clamp ...................144

6-44 Contour plot of bending stresses using H8 elements for 3D clamp ................................. 145

6-45 Contour plot of bending stresses using cubic B-spline elements for 3D clamp ..............145

6-46 Three dimensional rib model used for analysis .................................... ............... 146

6-47 Convergence plot for strain energy for rib model .................. ................................146

6-48 Contour plot of bending stresses using tetrahedral elements for 3D rib ........................147

6-49 Contour plot of bending stresses using B-spline elements for 3D rib .............................147

6-50 Contour plot of bending stresses using H8 elements for 3D rib.................................. 148

6-51 Three dimensional dial bracket used for analysis .................. ................................148

6-52 Convergence plot for strain energy for dial bracket model ..........................................149

6-53 Contour plot of bending stresses using a cubic B-spline element model for dial
b rack et ........... ... ........ .......... .......... ............................................... 14 9

6-54 Contour plot of bending stresses using a tetrahedral element model for dial bracket.....150

6-55 Contour plot of bending stresses using a H8 element model for dial bracket ...............150

6-56 Square plate with a circular inclusion ..................................................... ............ 151

6-57 Contour plot of o using Q4 elements in IBFEM ......................................................151

6-58 Contour plot of o- using Q4 elements in FEM.........................................................152

6-59 Convergence plot for error in strain energy for square matrix with inclusion model......152

6-60 Comparison of ux along the line AB ................................................... ..................153

6-61 Comparison of stresses along the line AB .......................................................... 153

6-62 Com prison of strains along the line AB ........................................................ .............. 154

6-63 Plot of stresses for cases 1 and 2............................................. ............................. 155

6-64 Plot of stresses for cases 3 and 4............................................. ............................. 155

6-65 Plot of stresses for cases 5 and 6............................................. ............................. 155

6-66 Plot of stresses for cases 1 and 2 for 3D model .................................... ............... 156

6-67 Plot of stresses for cases 3 and 4 for 3D model .................................... ............... 156

6-68 Plot of stresses for cases 5 and 6 for 3D model .................................... ............... 156

6-69 Contour plot of u using Q4 elements and grid refinement............................................157

6-70 Contour plot of v using Q4 elements and grid refinement ..........................................157

6-71 Contour plot of u using cubic B-spline elements and grid refinement...........................158

6-72 Contour plot of v using cubic B-spline elements and grid refinement...........................158

6-73 Square plate w ith a circular hole......... .................................................. ............... 159

6-74 Contour plot of o- for local refinement using Q4 elements .......................................159

6-75 Contour plot of o- for local refinement using cubic B-spline elements.........................160

6-76 Convergence plot for L2 error in local grid refinement ..............................................160

6-77 Convergence plot for energy error in local grid refinement ................. ................161

6-78 Contour plot of o- for local refinement using Q4 and cubic B-spline elements ............161

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

FINITE ELEMENT ANALYSIS USING UNIFORM B-SPLINE APPROXIMATION AND
IMPLICIT BOUNDARY METHOD

By

Ravi Kumar Burla

May 2008

Chair: Ashok V. Kumar
Major: Mechanical Engineering

The finite element method (FEM) is a widely used numerical technique for solution of

boundary value problems arising in engineering applications. FEM is well studied and has

several advantages; however the stress and strain solutions obtained by FEM are discontinuous

across element boundaries and require smoothing algorithms to make the stress/strain appear

smooth in the analysis domain. Automatic mesh generation is still a difficult task in FEM for

complicated geometries and often requires user intervention for generating quality meshes.

Finite elements based on uniform B-spline basis functions defined on a structured grid are

developed which provide at least C1 continuous solution through out the analysis domain.

Numerical techniques are presented in this work based on structured grids and implicit equations

of the boundaries of analysis domain, material interface boundaries or coarse/fine grid interfaces

for solution of various engineering boundary value problems. Solution structures are constructed

using approximate Heaviside step functions for imposition of Dirichlet boundary conditions,

treatment of material discontinuity to perform micromechanical analysis and for local refinement

of grid. The use of structured grid eliminates the need for constructing a conforming mesh and

results in significant savings of time in pre-processing stage of the design cycle.

Numerical examples are presented to demonstrate the performance of B-spline elements.

The results are compared with analytical solutions as well as traditional finite element solutions

to demonstrate the ability of B-spline elements to represent continuous stress and strain through

out analysis domain. Convergence studies show that B-spline elements can provide accurate

solutions for many engineering problems with fewer numbers of elements and nodes as

compared to traditional FEM. Solution structure for treatment of material boundary is validated

by performing a convergence analysis on a problem involving circular inclusion in a square

matrix and by determining effective properties of fiber reinforced composite. Solution structure

for local grid refinement is validated by analyzing classical stress concentration problems.

CHAPTER 1
INTRODUCTION

1.1 Overview

Numerical methods are necessary for solving many problems arising in engineering

analysis. Even though analytical solutions are available for several engineering problems with

simple geometries, for complicated domains it is difficult to obtain analytical solutions and hence

numerical methods are needed to obtain accurate solutions. The Finite Element Method (FEM) is

widely used in engineering analysis not only due to its applicability to solve a variety of physical

problems but also due to its ability to handle complex geometry and boundary conditions. In

FEM, the analysis domain is subdivided into finite elements which approximate the geometry of

the analysis domain. The field variable is approximated on these finite elements using piecewise

continuous interpolations.

Even though FEM is a well established technique, it has certain limitations. Approximation

schemes used in FEM are typically Co continuous and hence there is a discontinuity of gradients

across the elements. The lack of gradient continuity in FEM necessitates the smoothing of

results. There are a number of approximation schemes like B-splines, moving least squares and

Hermite interpolations which provide at least C1 continuity in the domain of interest. B-spline

basis functions have compact support like traditional finite element shape functions and lead to

banded stiffness matrices. They form partition of unity and hence can exactly represent the state

of constant field in the domain of interest. The B-spline approximations have good reproducing

properties and when a cubic B-spline is used, it can represent all polynomials up to third order

exactly. B-spline approximation provides high order of continuity and is capable of providing

accurate solutions with continuous gradients through out the domain. The uniform B-spline basis

functions are polynomials and accurate integration for stiffness matrices and load vectors can be

performed by using well known techniques like Gauss Quadrature. These properties of B-splines

make them attractive for use in analysis. B-splines have not been used extensively for analysis

because they do not satisfy Kronecker-delta property and hence essential boundary conditions

cannot be imposed directly by specifying nodal values. B-splines are defined on a grid which

limits the geometry of the analysis domain to a parallelepiped.

In order to be able to use uniform B-spline basis for approximation of finite element

solutions, structured grids need to be used since the uniform B-spline basis are defined only on a

uniform structured grid. A structured grid is much easier to generate than a finite element mesh

and all the elements in the grid have regular geometry (rectangles/cuboids). The use of structured

grids for analysis eliminates the need to generate a conforming mesh which is a significant

advantage over traditional FEM. Automatic mesh generation algorithms for FEM are well

developed 2D problems but can be unreliable for 3D geometries often resulting in poor or

distorted elements in some regions that can lead to large errors in solution. Significant amount of

user intervention is needed to correct such problems making mesh generation the most time

consuming step in the design process. These problems associated with mesh generation can be

avoided with the use of a structured grid.

When structured grids are used to approximate solution in the analysis domain, the

boundaries of the domain need to be defined independent of the structured grid and this can lead

to several elements on the boundary which have partial volume inside the grid elements. Special

techniques are therefore needed to compute the stiffness matrices and load vectors in these

partial elements. B-spline approximations do not satisfy Kronecker-delta property and hence

boundary conditions cannot be applied by specifying nodal values. Furthermore, the nodes in a

structured grid are not guaranteed to lie on the boundary, therefore the traditional methods used

in FEM for applying essential boundary conditions cannot be used and new techniques are

needed for applying essential boundary conditions. In order to enforce essential boundary

conditions when C1 approximations are used on structured grids, Implicit Boundary Method is

used in this work. In Implicit Boundary Method, solution structures are constructed by using the

approximate Heaviside step functions based on the implicit equations of the boundaries for exact

imposition of essential boundary conditions.

Implicit Boundary Method is extended in this work by developing solution structure for

treatment of material discontinuity using approximate Heaviside step functions based on the

implicit equations of matrix/inclusion boundaries. This technique is further used for

determination of effective properties of fiber reinforced composite.

In many practical problems local refinement of mesh/grid is necessary to accurately

compute the stresses in the regions of stress concentrations; this problem is addressed in this

work by developing solution structure for local grid refinement using approximate Heaviside

step functions based on the implicit equations of the coarse/refined grid boundaries.

The performance and capabilities ofB-spline based finite elements are studied by solving

several problems in structural and thermal analysis in both 2D and 3D. Convergence analysis is

also performed using several examples to study the behavior of B-spline elements and it has been

demonstrated that B-spline element provide accurate solutions with fewer number of elements

when compared to Lagrange elements used in traditional FEM. Micromechanical analysis for

determining effective properties of fiber reinforced composites is performed by employing the

solution structure for treatment of material discontinuity using Lagrange finite elements. Solution

structure for local refinement is used to analyze stress concentration problems where local

resolution of solution near regions of stress concentrations is necessary for performing design

analyses.

1.2 Goals and Objectives

The goal of this research is to develop finite elements based on B-spline approximations

and to extend Implicit Boundary Method by developing solution structures for analysis of

engineering boundary value problems involving material discontinuity and local grid refinement.

The main objectives of this dissertation are

* To extend the Implicit Boundary Method for use of B-spline basis functions.

* To develop finite elements based on B-spline approximation which provides at least
C1 continuity in the analysis domain.

* To develop solution structures for modeling material discontinuity and imposition of
periodic boundary conditions involved in the micromechanical analysis of composite
microstructures and to determine effective elastic properties of fiber reinforced composites.

* To develop solution structures for local refinement of grid to accurately capture stress
concentrations and study the use of the B-spline approximations and Lagrange
interpolations in local refinement.

* To develop efficient numerical techniques for computation of various terms involved in the
Galerkin weak form.

* To study the convergence behavior of the solution structures and B-spline finite elements
using several numerical examples and to study the advantages and disadvantages of using
the B-spline based finite element.

1.3 Outline

The rest of the dissertation is organized as follows:

Chapter 2 discusses several analysis techniques that are reported in literature which use B-

splines to approximate solution. A survey on structured grid methods, meshless methods,

treatment of material discontinuity and adaptive grid/mesh refinement is also presented.

Chapter 3 discusses the construction of solution structure for exact imposition of essential

boundary conditions for structural and thermal analyses. Solution structures are developed for

modeling material discontinuity and for imposition of periodic boundary conditions which are

used for micromechanical analysis of composites. Finally, solution structure is developed for

local refinement of grid to improve the local resolution of the solution.

In chapter 4, finite elements based on B-spline approximation are developed which provide

C1 and C2 continuity in one, two and three dimensions.

Chapter 5 discusses efficient numerical implementations for computation of stiffness

matrices and load vectors involved in the modified weak form corresponding to the solution

structures presented in chapter 3.

These techniques are implemented in JAVA and several problems in structural mechanics,

steady state heat transfer, composite micromechanics and localized stress concentrations are

studied in detail in chapter 6. Convergence analyses are also performed for these problems to

study the behavior B-spline elements and solution structures in terms of accuracy and efficiency.

Finally, in chapter 7, several conclusions are drawn from the numerical experiments

performed in chapter 6. The advantages and shortcomings of proposed solution structures and B-

spline finite elements are discussed in detail. The scope for future work is presented towards the

end of the dissertation.

CHAPTER 2
ANALYSIS TECHNIQUES FOR BOUNDARY VALUE PROBLEMS

In this chapter, a detailed discussion on several analysis techniques that are used for

solution of engineering boundary value problems is presented. In the first part of this chapter, a

brief overview of the finite element method (FEM) is presented along with its advantages and

shortcomings. Techniques for employing B-spline approximations for analysis are discussed in

the next part of this chapter. A detailed discussion on structured grid methods and meshless

methods is presented in the later part of this chapter

Micromechanical analysis is an important engineering application where FEM is used

extensively. Microstructures are modeled using a representative volume element and periodic

boundary conditions are applied using multipoint constraint techniques. Meshless methods have

been extended for treatment of material discontinuity and for performing micromechanical

analysis. A survey on these techniques is presented in this chapter along with a survey on

employing FEM for determination of effective properties.

Another important feature of FEM is its capability for local refinement. A survey on the

refinement techniques available in structured grid and meshless framework is presented in the

last part of this chapter.

2.1 Finite Element Method

Finite element method (Hughes, 2000 ; Cook et. al. 2003) is a well established numerical

technique and is widely used in solving engineering problems in academia as well as in industry.

For finite element analysis, the domain of analysis is subdivided into elements and the resulting

finite element mesh approximates the geometry and is also used to approximate the solution by

piece-wise interpolation within each element. Using Galerkin's approach, the governing

differential equation in its strong form is converted into a weak form which allows the use of

interpolation/approximation schemes with lower continuity requirements than the corresponding

strong form. This leads to a set of linear equations which are then solved using Gauss

elimination, Cholesky decomposition or iterative schemes like Gauss-Siedel. FEM has been

applied to a wide variety of physical problems in linear/non-linear static and dynamic elasticity,

steady state heat transfer, electrostatics and magnetostatics.

The approximation space used in FEM is typically Lagrange interpolation that yields

Co continuous solutions. However, Co continuous solutions have discontinuous stresses and

strains (gradients) across the element boundaries. The lack of stress and strain continuity

necessitates the use of smoothing schemes to plot these quantities such that they appear to be

continuous. Solutions can be represented with higher degree of continuity using several other

basis functions like B-splines, moving least squares (MLS) and Hermite functions which provide

at least C1 continuity in the domain of interest. In this dissertation, B-spline approximations are

studied in detail which provide C1 and C2 continuous solutions in the analysis domain.

Another limitation of the FEM is the need to generate a conforming mesh, which may need

user intervention for correcting the problems associated with distorted meshes. Automatic mesh

generation algorithms work satisfactorily for 2D but are still unreliable for 3D. In order to avoid

the problems associated with mesh generation, several methods have been proposed which fall

into two categories: meshless methods and structured grid methods. In this chapter, a detailed

survey on various structured grid and meshless methods is presented.

2.2 B-Spline Approximation Based Methods

A number of researchers have studied the use of B-spline approximations to solve specific

problems where the geometry of the analysis domain is simple (e.g. beams or rectangular plates).

The boundary conditions were applied by using multiple knots at the boundary of the analysis

domain which emulates Kronecker-delta behavior on the boundaries. Leung and Au (Leung and

Au, 1990) introduced the boundary transformation method by constructing an appropriate

transformation matrix based on the boundary configuration to impose Dirichlet boundary

conditions. Kagan et. al.(1998, 2003) have provided a scheme to integrate B-spline finite element

method with a mechanically based computer aided design and the Dirichlet boundary conditions

were enforced by imposing multiple knots on the essential boundaries. They also developed

methods for adaptive refinement of B-spline elements. In summary, these methods have

limitations on the geometry of analysis domain.

Hughes and his coworkers (Hughes et. al. 2005 ; Cottrell et. al. 2006 ; Cottrell et. al. 2007)

have developed a more general approach for using B-splines called the isogeometric analysis

method, where the geometry is represented by Non-Uniform Rational B-Spline (NURBS) basis

functions to get an exact geometric representation. The same basis functions can be used for the

analysis or their order can be raised without changing the geometry or parameterization. This

enables the traditional h-, and p- adaptive refinement as well as k-refinement where the degree of

continuity of the solution can be raised. Several alternatives for imposing Dirichlet boundary

conditions were proposed including direct application by specifying control variables and use of

constraint equations, but this still remains a topic of research.

2.3 Structured Grid Methods

A number of techniques have been developed based on structured grids to approximate the

solution. Extended Finite Element Method or X-FEM was developed by Belytschko and co-

workers (Belytschko et. al. 2003). This method uses implicit equations for definition of the

geometry of the analysis domain and the Dirichlet boundary conditions are applied using

Lagrange multipliers.

Penalty boundary method developed by Clark and Anderson (Clark and Anderson, 2003a,

2003b) uses a regular grid and a constrained variational formulation to include both essential and

natural boundary conditions which is then solved using penalty method.

Shapiro (Shapiro, 1998 ; Shapiro and Tsukanov, 1999) and Hollig (Hollig et. al. 2001;

Hollig and Reif 2003 ; Hollig 2003) used R-functions to define the boundary of the problem

domain and the solution structure proposed by Kantorovich and Krylov (Kantorovich and Krylov

1958) was constructed to satisfy Dirichlet boundary conditions. Shapiro used transfinite

Lagrange interpolation to impose Dirichlet boundary conditions while Hollig used weight

functions based on distance functions and R-functions. Hollig developed extended B-spline basis

to represent the solution in the analysis domain.

The method presented in this dissertation is very similar to the approach used by Hollig,

however approximate Heaviside step function of essential boundaries are used in this work to

construct weight functions and the behavior of solution structure for very small values of range

parameter used to construct step functions is also studied. All the internal elements in our method

have identical stiffness matrices unlike Hollig's method.

2.4 Meshless Methods

A number of methods have been developed in the past two decades which avoid mesh

generation. These methods are collectively termed as meshless or meshfree methods. Some of

these methods provide continuous gradients in the analysis domain. Smoothed particle

hydrodynamics (SPH) is one of the first meshless methods in literature (Lucy, 1977 ; Gingold,

1982); it was first used to model astrophysical phenomenon and was modified to include

continuum mechanics problems (Monahan, 1982). This approach uses Shepard shape functions

as interpolants and uses point collocation method for discretization and nodal integration for

stiffness matrix computation. The drawbacks of this approach are poor accuracy and spatial

instability (tensile instability). The behavior of solution depends on the proper distribution of

nodes. So there is always an effort needed to place the nodes properly in order to avoid

instability.

Element Free Galerkin Method was developed by Belytschko and his coworkers

(Belytschko et. al., 1994 ; Belytschko et. al., 1996 ; Belytschko et. al.; 1996, Dolbow and

Belytschko, 1999 ; Krongauz and Belytschko, 1996 ; Krongauz and Belytschko; 1997 ; Krysl

and Belytschko 1997). This method is based on moving least square approximation and uses

global Galerkin method for discretization. Quadrature was performed on the background cells.

Boundary conditions were applied using Lagrange multipliers or by modification of variational

principle, where Lagrange multipliers were replaced by traction forces. The former method leads

to an increased number of algebraic equations and the stiffness matrix is not positive definite.

The latter method reduces the accuracy of the results and is not as convenient as imposing

essential boundary conditions directly. A new method for imposing essential boundary

conditions by using a string of Finite Elements around the boundary of the domain was later

incorporated into this method.

Meshless Local Petrov-Galerkin (MLPG) was introduced as a truly meshless method by

Atluri and his coworkers (Atluri et al. 1999 ; Atluri and Zhu 2000a, 2000b). This method uses

moving least square approximation as the interpolation functions. It uses the Local Petrov-

Galerkin weak form which is developed in the local sub-domain. The trial and test spaces are

different; the test functions vanish on the boundary of the sub-domain, as long as these

boundaries are internal to the problem domain. If the sub-domain intersects the boundary of

global domain, then the boundary condition is applied on this local boundary by penalty method.

The stiffness matrix obtained is banded but not symmetric. The boundary conditions are applied

using penalty method. Another advantage of this procedure over FEM is that smoothing

techniques are not required for computation of derivative. There is no need of special integration

schemes, and there is flexibility in choosing the size and shape of the local domain, which can

lead to convenient methods for integration in non-linear problems. The drawbacks of this method

are that the stiffness matrices generated by this approach are not symmetric positive definite and

penalty method imposes boundary conditions approximately. Another drawback is that we have

to make sure that the union of all the local domains should cover the global boundary.

Partition of unity based methods (PUFEM) form a class of meshless methods which use

partition of unity functions to blend the local approximation functions which represent the local

behavior (Melenk and Babuska, 1996). The main advantage of PUFEM is that we can use the

knowledge of the local behavior of solutions efficiently to get good numerical approximation.

This method also yields good results in the cases where FEM gives poor results or when FEM is

computationally very expensive. One drawback of this approach is that the stiffness matrix

obtained from this procedure can be singular and the basis function needs to be modified

appropriately or special techniques are needed to solve the singular system. Generalized Finite

Element Method was developed as a hybrid of Finite Element Method and Partition of Unity

(Babuska et. al. 2002 ; Duarte et. al. 2000). The desirable features like inter-element continuity,

local approximability and Kronecker-delta property from FEM are retained in GFEM.

Method of finite spheres (De and Bathe 2000 ; De and Bathe 2001) was developed as

meshless method which is based on partition of unity; this approach does not use background

cells for numerical quadrature. It uses compactly supported partition of unity functions and

Bubnov-Galerkin method as weighted residual scheme. Integration of weak form is performed

on spheres (circles in 2D), using specialized cubature rules. Essential boundary conditions are

imposed by placing nodes on the boundary so that the shape functions corresponding to these

nodes emulate Kronecker-delta behavior. Shepard functions are used as partition of unity

functions and quartic splines are used for weight functions.

Method of HP Clouds is a partition of unity based meshless method developed by Duarte

and Oden (Duarte and Oden 1996 ; Duarte et. al. 1996 ; Liszka et. al. 1996). HP-Cloud functions

are created by multiplying the partition of unity functions with local basis functions. Shepard

functions or MLS functions are used as partition of unity. Anisotropic approximation can be

incorporated by using the basis function which has higher order in one direction and lower order

in other direction. P-refinement is done by increasing the size of the local domain; hence more

number of points is included in the sub-domain and also by increasing the polynomial order in

the basis function. Boundary conditions are applied by modified variational principle.

Natural element method developed by Sukumar et. al.(1998) uses the natural neighbor

interpolants which are based on Voronoi tessellation of the nodes. These interpolants are smooth

over the whole domain excepting at the nodes where they are continuous. The essential boundary

conditions are applied on nodes for convex domains.

2.5 Techniques for Determination of Effective Properties of Composite Microstructures

The finite element method (FEM) has been used extensively in micromechanical analysis

of composite microstructure to determine effective properties (Sun and Vaidya; 1996 ; Taliercio,

2005 ; Xia et. al. ; 2003 ; Marrey and Sankar, 1997 ; Zhu et. al. 1998). The effective elastic

properties are determined by performing six different analyses on the representative volume

element (RVE) with unit non-zero strain component in each analysis and by imposing

appropriate periodic boundary conditions on the boundaries of the RVE. The mesh is generated

in such a way that the element boundaries conform to the matrix-inclusion interfaces. However,

for complicated microstructures, automatic mesh generation becomes difficult and special

techniques and algorithms need to be developed (Kim and Swan, 2003) to generate a quality

mesh with quadrilateral or hexahedral elements. For example, in the analysis of composite

microstructures involved in textile composites, the automatic mesh generation often becomes

difficult and one of the main problems is generating a quality mesh in the interfacial region (Kim

and Swan, 2003). This region is multiply connected and complications may arise for generation

of mesh with identical nodes on the opposite faces of the RVE in order to impose periodic

boundary conditions using multipoint constraints. Meshless methods have been extended for

treatment of material discontinuity. Element Free Galerkin Method has been extended to

incorporate discontinuous derivatives (Cai and Zhu, 2004 ; Cordes and Moran, 1996 ; Krongauz

and Belytschko, 1998) for application to material discontinuity problems and Meshless Local

Petrov-Galerkin Method has been extended to incorporate material discontinuity and for

performing micromechanical analysis (Dang and Sankar, 2007).

2.6 Techniques for Local Refinement

In many practical problems, there may be regions with stress concentrations and hence a

uniform mesh/grid for the entire model can become very expensive computationally. In order to

reduce the computational cost local mesh/grid refinement is necessary. In this section, a brief

discussion on the local refinement techniques is presented.

In the finite element analysis, local refinement is typically obtained by h-refinement or p-

refinement. In h-refinement, nodes are inserted in the regions of stress concentration which

increases the number of elements and reduces the element size in the region of interest while p-

refinement is obtained by increasing the order of polynomial basis used for interpolation.

Another type of refinement is termed as r-refinement, where the total number of nodes and

elements in the refined model is same as the coarse model but the nodes are rearranged so that

the number of nodes and elements in the regions of stress concentration is increased and k-

refinement is obtained when the order of continuity of the basis functions is increased. In

meshless methods, refinement is achieved by inserting nodes in the regions of stress

concentrations.

Hierarchical mesh refinement is studied by Krysl et. al. (2003) in the finite element method

framework. In this method, the nodal basis functions on the coarse mesh are split into the basis

functions of same order in the finer mesh. They have provided algorithms for refinement as well

as un-refinement for finite element meshes. Hollig (2003) also presented a hierarchical

refinement of B-spline basis functions, where the B-spline basis are split into the B-spline basis

of same order but on a finer grid with half the size of the coarse grid. The hierarchical refinement

provides a framework for getting a multi-level resolution, where the basis functions are split in a

recursive manner. In this dissertation, a technique for local grid refinement is presented based on

construction of solution structure using approximate step functions.

CHAPTER 3
SOLUTION STRUCTURES FOR IMPLICIT BOUNDARY METHOD

The essential boundary conditions are imposed in FEM by specifying the values at the

boundary nodes on which essential boundary conditions are specified. When structured grids are

used for discretizing the analysis domain, nodes are not guaranteed to be present on essential

boundaries; hence special techniques need to be devised to be able to impose essential boundary

conditions. In this chapter, a detail discussion on construction of solution structure for exact

imposition of Dirichlet boundary conditions using the implicit equations of boundaries is

presented. Approximate Heaviside step functions are used to construct a solution structure based

on the technique proposed by Kantorovich and Krylov (Kantorovich and Krylov, 1958 and is

termed as Implicit Boundary Method (Padmanabhan, 2006 ; Kumar et. al. 2007). Non

homogenous essential boundary conditions are imposed by constructing piecewise continuous

functions which satisfy the specified values at the essential boundaries and are termed as

boundary value functions. This solution structure is used for modifying the weak form

corresponding to linear elasticity problems and steady state heat transfer problems. Implicit

Boundary Method has been extended in this dissertation to be able to use B-spline basis.

A solution structure for treatment of material discontinuity based on the implicit equation

of the material boundaries and approximate Heaviside step function is presented later in this

chapter. Techniques for imposing periodic boundary conditions using multipoint constraints are

also discussed in detail.

In the final part of this chapter, a solution structure for adaptive refinement of grid based

on the implicit equations of the grid element boundaries and approximate Heaviside step

functions is discussed in detail. The weak form corresponding to linear elasticity is modified

using this solution structure.

3.1 Solution Structure for Enforcing Essential Boundary Conditions

The essential boundary conditions are specified as prescribed displacements or prescribed

temperatures on the essential boundaries. Let u be a trial function defined over Q c R2 or R3

that must satisfy the boundary condition u = a along F, which is the essential boundary of Q.

Here, u corresponds to a displacement component in linear elasticity and temperature in steady

state heat transfer. If D, (x) = 0 (x e Q and a : Q -> R ) is the implicit equation of the boundary

F,, then the trial function can be defined as

u(x) = (,(x)U(x) + a(x) (3-1)

The solution structure for the trial function defined in Eq. (3-1) is then guaranteed to

satisfy the condition u = a along the boundary defined by the implicit equation I (x) = 0 for

any U: Q -> R. The variable part of the solution structure is the function U(x) which can be

replaced by a finite dimensional approximate function Uh(x) defined by piece-wise

interpolation/approximation within elements of a structured grid. In the context of such solution

structures, the function (a (x) is often referred to as a weighting function. Since these functions

assist in imposing the Dirichlet boundary conditions, we have referred to these functions as

Dirichlet functions or just D-functions. D-function is constructed such that it takes a value of

zero and has non-zero gradient on the essential boundary and has a value of one far from the

essential boundary. Instead of directly using the implicit equation of the boundary as D-

functions, approximate step functions constructed using these implicit equations have been used

to construct solution structures in this work. The use of approximate step functions for

construction of D-functions localizes them to the boundary elements and hence the stiffness

matrix for all the internal matrices is identical as the D-function is unity inside them. In the

solution structure demonstrated in Figure 3-1, D represents the Dirichlet function which is zero

at the first node of element el and reaches unity within this element and remains unity through

out the domain. The boundary value function is defined by ua which takes on the value of the

prescribed boundary value of u, at the boundary and ug represents the grid variable which is

interpolated on the grid consisting of elements el-e5. The approximate solution is given by

u = Dug + ua and it can be observed that the approximated solution satisfies the boundary

condition exactly. The solution structure for linear elastostatic and steady state heat transfer

problems is first discussed in this section for a general D-function and a detail discussion on

construction D-functions is provided later in this chapter.

3.1.1 Modified Weak Form for Linear Elasticity

The governing differential equation for an elastic body under equilibrium as shown in

Figure 3-2 is given as:

Solve: V.b +b = 0 in Q

Subject to: u = u0 on F, (3-2)

a.n=to on rF

In the above equation, Q is the analysis domain bounded by essential boundary F, and

natural boundary Ft. {(} is Cauchy stress tensor and {b} is the body force and the Dirichlet

boundary conditions are displacements specified as uo on essential boundary and the natural

boundary conditions are specified traction to. The principle of virtual work is the weak form for

the boundary value problem and is given as:

fJ{} T{} dQ = Su}T {t}dF+ + Ju}T {b}dQ (3-3)
Q F, Q

Here {6S} is the virtual strain and {6u} is virtual displacement. A trial solution structure

for the displacement field can be constructed such that the Dirichlet boundary conditions are

satisfied

u}= [D]{ug}+ {u) =u)+ {u} (3-4)

In this solution structure, {ug} is a grid variable represented by piece-wise approximation

over the grid, and {u0} is the boundary value function which is a vector field constructed such

that it takes on the prescribed values of displacements at boundaries. The variable part of the

solution structure is {u"} = [D] {ug which satisfies the homogeneous boundary conditions.

[D] = diag(D,.., D ) is a diagonal matrix whose components DZ are Dirichlet functions that

vanish on boundaries on which the ith component of displacement is specified, nd is the

dimension of the problem. The same D-functions used to construct the trial solution are also used

to construct the test functions.

{(u} = [D]S6ug (3-5)

Using these solution structures, the strains and stresses can be expressed as the sum of a

homogeneous part (E s, u ) and a boundary value part ( e, o ).

S (Usj +us, )+ (u +ua = Es+ (3-6)
2 \ 1 2 J 1

o, = Csk kl = Uk = ( + a o-y (3-7)

Substituting ({c =( }+ {a in the weak form, we can rewrite it as

J~S)} {sdQ= { Su}T {t}dF + JSu}T {b}dQ- J{6} {a ) d (3-8)
Q F, Q Q

The grid variable is approximated within each element as uf = NJu, where NJ are the

shape functions and ug are the nodal values of the grid variable for ith component of

displacement. The virtual strain and stress within each element can then be expressed as

{6ST = [B]{6X'} and {s} = [C][B]{Xe}, where {Xe} and {Xe} represents the nodal values of

the grid variable for the element e and [B] is the strain-displacement matrix. Using these

notations, the weak form can be expressed in a discrete form very similar to the finite element

equations.
NE NE NBE
{86Xe}T [B]T[C][B]{Xe}d2= {6Xe} {F}+ {6XXT {N}T {t}dF (3-9a)
e=1 ^ e=1 e=1 pt

{Fe = J{N}T {b} dQ- J{B}T {fc}da (3-9b)
Q, Q,

The summation is over the number of elements (NE) and the number of boundary elements

(NBE) in the grid used for analysis. As in the finite element method these element matrices can

be assembled to form global equations, where the left hand side terms form a stiffness matrix and

the right hand side a force vector. Note that unlike in the finite element equations, there is a

forcing term that arises from the boundary value functions. Furthermore, the nodal variables to

be solved for are grid variables and not displacements. Since no assumptions have been made

restricting the choice of shape functions used to represent the grid variables, this approach can be

used with a variety of basis functions including Lagrange interpolation, B-spline basis functions

or meshless approximations.

3.1.2 Modified Weak Form for Steady State Heat Transfer

The governing differential equation for general steady state heat transfer problem shown in

Figure 3-3 is given as

Solve: V.(kVT)+f 0 in Q

Subject to: T = T on Fg (3-10)

kVTfi = qh onFh

The corresponding weak form is given as:

IVST-k-VTd f STdQ+ f qhTdF- h(T- T)TdF (3-11)
Q Q FFTh

In the above equations, F, is the portion of the boundary with prescribed temperature

while Fh is the portion with heat flux conditions and F, represents the portion with convection

conditions and To is the ambient temperature. h is the convection coefficient and k is the

thermal conductivity of the material which is assumed to be isotropic for simplicity and f is the

heat source. A trial solution structure for the temperature field is constructed such that the

essential boundary conditions are satisfied

T = T" + T = DTg + T (3-12)

Proceeding in the similar manner as for the linear elastic problems discussed in the

previous section and choosing the test function 6T = D Tg such that it vanishes on Fg, the weak

form modifies to the following:

V6T" k VT" d+ ST" hTg dF = V6T" k VTa" dA+ f6T"do
a rF a a (3-13)
+ qo dT'dF J h (T -T")dTs'dF
TFh r,

The nodal degrees of freedom for temperature are approximated as Tg = NTg and the

homogenous part of solution becomes Ts = DNTg, where D is the essential boundary function.

The gradients of temperature which are used in the weak form are constructed as VT7 = BA'T.

The weak form takes the following form when the above approximations are incorporated into

Eq. (3-13)

J[B] k[B] k {TdQ l + I[N]h[N] {T }dF =- JB k[B]{Ta}dQ,
Fa rC a, (3-14)
+ [N]Tfd + f I [N]Tq,od- f [N] h(Ta- T")dF
,e Th Fr

In the above expression, all the known quantities are moved to the right hand side and the

unknown quantities are placed in the left hand side. The first term on the right hand side is the

contribution to the load due to applied essential boundary conditions, the second term is the load

due to the heat generation, the third term is load due to applied heat flux and the last term is load

due to convection boundary conditions. The first term on the left hand side represents the

stiffness matrix while the second term represents the stiffness due to convection boundary

condition.

3.1.3 Dirichlet Function

The Dirichlet function D is a function which vanishes on all essential boundaries where

ith component of displacement (or temperature) is prescribed. In addition, DZ should have a non-

vanishing gradient on the essential boundaries to ensure that the gradients of displacements are

not constrained. Furthermore, it is necessary to ensure that D, is non-zero inside the analysis

domain so that the solution is not constrained anywhere within the domain of analysis. Implicit

equation of the curve or surface representing the boundary can be used for this purpose. However

if multiple curves are involved then it is necessary to combine them into a single equation. Some

authors have proposed the use of R-functions to construct the Boolean combination of implicit

functions (Shapiro, 1998 ; Rvachev et. al., 2001). Hollig (2003) recommends constructing a

weight function using distance functions as: w(x) = 1 max(0,1 dist(x, OD) / )' This function

varies between 0 and 1 over a narrow strip of width 6 near the boundary. Evaluating this

function requires numerically computing the distance function.

In this work, approximate step functions similar to the weighting function used by Hollig

(Hollig, 2003) are proposed that can be constructed using any type of implicit function of the

boundary. If the boundary condition is specified on a boundary with implicit equations 0 then a

step function at any given point x is defined as shown below

0 0<0

D(O)= 1- 1- 0' (3-15)<(
(8 (3-15)
1 #_>8

This function is a shifted approximate step function that tends to the Heaviside step

function in the limit as -> 0 but has a value of zero on the boundary instead of 0.5 as in

traditional approximations of step functions (Osher, 2002). Moreover, for use as a Dirichlet

function it is constructed to have non-zero gradient at # = 0. At # = S this function has Ck 1

continuity. In this paper, the step function approximation of D with 5- 0 is used as the

Dirichlet function with values of S 10 5 or smaller.

When a single boundary passes through a boundary element the D function can be

computed as in Eq. (3-15), but when multiple boundaries exist that have Dirichlet boundary

conditions on them then it is necessary to construct D functions that are Boolean combinations of

the individual step functions. The intersection, subtraction and union between two step functions

D(#1) and D(02), denoted as D12, D2 and Du2, respectively, can be defined as shown in

Eq. (3-16).

D22 =D(A)D(02) (3-16a)

D1=2 D(O)D(-2) (3-16b)

D,2 =D(1)+D(2)-D({)D(2) (3-16c)

Figure 3-4 shows plots of D-functions created as a Boolean combination of two D-

functions that represent circles. Note that the D-functions created by the Booleans inherit

important properties including zero value at boundaries, non-zero slopes at boundaries and they

plateau to a value of 1 for 0 > S Since the step function defined in Eq. (3-15) has a value equal

to zero at the boundary (and not 0.5 as in traditional approximations of step functions), the

Boolean operations defined in Eq. (3-16) do not round off corners defined by the intersection of

the implicit equations. If more than two boundaries exist within a boundary element, then the

Dirichlet function is constructed using the Boolean tree similar to CSG tree as shown in Figure

3-5.

3.1.4 Boundary Value Function

The boundary value function ua (displacement component for linear elasticity and

temperature for steady state heat transfer) must be defined such that they have value equal to the

imposed essential boundary conditions on the appropriate boundaries. If the essential boundary

conditions are specified at multiple boundaries and the assigned values are not the same for all

boundaries then it is necessary to construct a continuous function that takes on appropriate values

at respective boundaries while transitioning smoothly in between. While there is no unique

method for constructing boundary value functions, it is beneficial to ensure that the trial function

is a polynomial of the same order as the shape functions used for interpolating the grid variable.

This can be accomplished by constructing the boundary value function by piecewise

interpolation/approximation using the element basis functions. The nodal values of u" can be

constructed similar to the grid variable ug using the grid element basis functions.

u = -Nu (3-17)

In the above equation, N, are the shape functions of the grid element and u' are the nodal

values of u". It is important that the function u" should belong to the space spanned by the

shape functions used to represent the grid variable. Otherwise, the grid variable has difficulty

compensating for u~ in the solution structure leading to poor convergence. Indeed, the solution

structure will not satisfy completeness requirement, that is the ability to accurately represent

constant strains, if basis functions used to construct u" does not span a sub-space of the basis

functions that are used to represent the grid variable ug. If the basis functions N, in Eq. (3-17)

are identical to the basis functions used for approximating the grid variable then this requirement

is automatically satisfied. If solutions that have C1 or C2 continuity is desired then quadratic or

cubic B-spline basis should be used in Eq. (3-17). The nodes of the boundary elements should be

assigned nodal values of ua such that the resulting approximation will have the desired values at

the boundary passing through the element. This is very easy to achieve when the assigned value

is constant or even linearly varying along the boundary. When the specified boundary condition

is a constant value of displacement along boundary, all the support nodes of the element have to

be assigned value equal to the specified boundary condition. In the case of the 2D cubic B-spline

element, the approximation within the element is controlled by 16 support nodes. Therefore, the

value at all these nodes should be assigned to the same value to ensure that the resulting

approximation has a constant value within the element. Similarly, if the specified boundary value

has to vary linearly (as in hydrostatic pressure which varies linearly in the z-direction), the values

specified at all the support nodes should be on a plane corresponding to the linear distribution.

This is illustrated in Figure 3-6, where a simple geometry is used to demonstrate the construction

of the boundary value function. The left edge has a constant value of ua = 10 while the right

edge has a bilinear variation of u, = 10(1- x) + 5(1- y). The boundary value function is

constructed by identifying the elements for the left boundary and specifying the nodal value of

10 for all the support nodes, and then by identifying the boundary elements for right boundary

and specifying the nodal values computed using 10(1- x) + 5(1 y) on the corresponding support

nodes. The nodes which are not support nodes for the boundary elements are set to zero.

The plot for the boundary value function is shown in Figure 3-6 and it can be observed that

the boundary value function is continuous over the entire domain and it smoothly transitions

from the specified value of 10 from the left boundary to zero in the middle of the domain and

then smoothly transitioning from zero to the specified bilinear distribution along the right edge.

Even though a simple geometry is used to illustrate the concept of constructing boundary value

function, the procedure is quite general. However, care has to be taken to ensure that the grid has

sufficient density to be able to construct the appropriate boundary value function. For more

general variation of boundary values, the nodal values can be computed by least-square

minimization. The rest of the nodes in the grid that are not part of any boundary element can

have any arbitrary value and therefore can be set equal to zero. The boundary value function u"

contributes to the load computation on the right hand side of Eq. (3-10) for all those elements in

which the gradient of u" has non-zero magnitude.

3.2 Solution Structure for Micromechanics

In the finite element method, a conforming mesh is created such that the element

edges/faces form the material boundary. When structured grids as shown in Figure 3-7 are used

for analysis, the nodes may not lie on the material interface and hence there is challenge in

imposing interface conditions for material discontinuity. In Implicit Boundary Method, this is

achieved by construction of a solution structure which ensures that interface conditions are

satisfied for displacements.

The implicit equations of the inclusion boundaries are used to construct a step function for

inclusion such that it is unity inside the inclusion and smoothly blends to zero at the inclusion

boundary. Such a function can be used to patch the solution from the inclusion grid and matrix

grid to form a solution structure. Figure 3-8 shows the inclusion function in one-dimension and

Figure 3-9 shows the inclusion function for an elliptic inclusion in a square matrix.

It can be seen that the inclusion function is unity inside the inclusion and transitions to zero

at the boundary of the inclusion, the rate with which this function reaches zero is specified by the

parameter 6. The expression for inclusion function is similar to the Dirichlet function described

earlier in this chapter and shown below

0 0_<0

H"nc(0)= 1 1O' 0<0 HC O _(3-18)
1 _>8

This function inherits all the properties of Dirichlet function and is a shifted approximate

step function that tends to the Heaviside step function in the limit as 5- 0 but has a value of

zero on the boundary unlike in traditional approximations of step functions where it has value of

0.5 Moreover, for allowing normal strain discontinuity, H" nC() is constructed to have non-zero

gradient at 0 = 0. At 0 = S this function has Ck 1 continuity. In this work, the step function

approximation of HI"C with k= 2 and S 0 is used as Inclusion function with values of S w 10 2

or smaller. The Inclusion function is used to construct the solution structure for material

discontinuity as shown in Eq. (3-19).

u = (1 H'nc)ugl+ H'ncug2 (3-19)

This solution structures combines the solution from both grids to represent the solution in the

analysis domain. When the Inclusion function is unity, the solution is given by u = ug2, which is

the solution from the inclusion part and when the Inclusion function is zero, the solution is given

by u = ugl, which is the solution from the matrix. In the region where the inclusion function

varies from unity to zero, the solution is a blend of the solutions from matrix grid and inclusion

grid. This way of constructing solution structure ensures the displacement continuity of the

solution through out the analysis domain. This solution structure allows the normal strain

discontinuity; this can be seen from the gradients of the solution structure as shown below

OBu (uI' OH( l +H Ou, g2 +OH" g2
'= (1-H ) gu + Hnc _+- u (3-20)
Z ua aH1 xu8 a i

In this expression, the first term: (1-H H"n) and third term H"nc- are continuous at

H"inc Hnc""
the interface boundary while the second term: u1' and fourth term u2 are
8x Dx

discontinuous at the interface boundary and vanish in the matrix region and are non-zero in the

inclusion region. Therefore these terms provide independent slopes at both sides of the interface

resulting in discontinuous normal strains.

Modified Weak Form. The weak form of the elastostatic boundary value problem is the

principle of virtual work which can be written in the following form as shown in Eq. (3-21).

Here {6S} is the virtual strain, {0} is Cauchy stress tensor, {t} is the traction vector, {6u} is

virtual displacement and {b} is the body force. For elastostatic problems, the essential boundary

conditions are displacements {u} = { u } specified on F, and the natural boundary conditions

are traction {t} = {to} specified on Ft.

J {E T{a}d=Q S{u}T {t}dF + 6u}T {b}dQ (3-21)
n Ft,

When the solution structure shown in Eq. (3-19) is used and the displacements are

interpolated as ug1 = [A1]u l }, with N 1 = (1- H') N in grid corresponding to matrix and

ug2 = [g2 ]u 2} with Ng2 = H"nCN in inclusion grid, the expression for strains takes the

following form

Ou1
Ox,

Ei =2 [Bg B 2]{U}gl (3-22)
712 2
+-

In this expression, [B1 B2] is the modified strain displacement matrix and is a

combination of strain displacement matrices from the two grids. The expression for stress is

given as o-, = C,,c, where C, is the elasticity stiffness matrix, and in computation of stress, the

inclusion (or matrix) elasticity stiffness matrix is used when the strains are evaluated in the

inclusion (or matrix) region. The modified weak form takes the following discrete form when the

above expressions for stress and strain are incorporated into the weak form (Eq. 3-23).

AVE TE T
z(X'e} J[B, B2] [C][B, B2]{Xe}d= Z{X'e}T (I N 2} {b}d2+
el a e-l (3-23)

{Nxe} { F1 2} {t}dF
e=1 Fp

In this expression {Xe = {ugl ug2 T and the summation is over the number of elements

(NE) and the number of boundary elements (NBE) in the grid used for analysis. As in the finite

element method these element matrices can be assembled to form global equations, where the

left hand side terms form a stiffness matrix and the right hand side a force vector. The boundary

conditions arising in micromechanical analysis are periodic boundary conditions which can be

imposed in this method similar to FEM as the RVE is always a rectangle/cuboid and the nodes

are always guaranteed to be on the boundaries of the unit cell. The effective properties of the

composite microstructure are computed by performing six different analyses on the RVE

subjected to different unit strains in each case. The average values of stresses and strains in the

RVE are computed for each case and are then used to determine the effective properties. A

detailed discussion on the periodic boundary conditions and computation of effective properties

is presented in chapter 5.

3.3 Solution Structure for Grid Refinement

In many practical problems local refinement is needed to capture the localized behavior of

the solution (e.g.: stress concentrations). A uniform grid for the entire model is not suitable for

this purpose. In order to be able to refine a grid at required locations, a solution structure based

on approximate step function is presented in this chapter. The approximate step function blends

the solutions from the coarse grid and refined grid such that the continuity of the solution across

the refined grid and coarse grid interface is maintained. When higher order continuous

approximations are used for approximating the field variable, a suitable step function needs to be

used to maintain the continuity of the solution across the refine grid and course grid interface. In

Figure 3-10, the coarse grid is represented by gl and the locally refined grid is represented by g2,

it is to be noted that these grids gl and g2 are conceptual grids and analysis is performed on the

grid g.

3.3.1 Solution Structure

The approximate step function that is used to blend the solutions from the coarse and the

refined grids is constructed by using the implicit equations of the coarse/refined grid interface.

This step function for refinement is defined as:

0 0<0

i(20T o0 < <
2\)) 2
Hre' (9) 2
S 20 ( 1k (3-24)
1- 12
1 #_>8

In the above equation, # is the implicit equation of the coarse/refined grid interface, 6 is

the range parameter which defines the region in which the step function for refinement varies

from zero to one, and k is an integer which defines the order of the step function used. The step

1 S
function for refinement is such that it takes on a value of zero at # = 0, at = and one at
2 2

S= S. It has a zero slope and Ck1 continuity at both # = 0 and # = S. Hence, when B-spline

basis is used to approximate the solution, an appropriate k needs to be chosen so as to maintain

the continuity of the solution across the coarse/refined grid interface. The plot of Href and

1- Href is shown in Figure 3-11, and it can be seen that the above mentioned properties are

satisfied by the step function of refinement.

The above definition for H're is valid only when the coarse/refined grid interface is defined

by a single implicit equation. However, in many practical cases the interface is defined by

multiple implicit equations as shown in Figure 3-12. In order to define H'r for multiple implicit

equations, a Boolean tree as shown in Figure 3-12b is used to represent the step function for

refinement and the Boolean operations used to combine two step functions H're (1) and

Href (2), denoted as H ,ef H and H'f2, respectively are be defined as

Hrf = Hef ()Href (2) (3-25a)

Hre = Hef (Q1)Href(-~2) (3-25b)

Hre =Href (A)+Href (2)-Href ())Href (2 (3-25c)

The step function for refinement for the model shown in Figure 3-12 is shown in Figure 3-13.

This step function is constructed by using the Boolean operations as defined in Eq. (3-32). The

solution structure for the displacements is represented in the following form:

u= (1-Href)u +Hrefu2 (3-26)

In the above expression, Href is the step function for the coarse and refined grid interface

and it vanishes outside the refined grid and transitions from zero to unity inside the refined grid.

The transition from zero to unity takes place only across the boundary of the coarse/refined grid

interface. In Figure 3-10, the refined step function H"f varies from zero to unity across the edges

represented by AB and BC as these edges are shared between the refined elements of grid g2 and

unrefined elements of grid g .The refined step function is unity and hence does not vary across

the edges AF, FE, ED and DC as these edges are not shared between the refined and unrefined

grids. The expressions for uI and u2 have solution structure as proposed for imposition of

essential boundary conditions and can be written asuI = Dugl + Ua2 andu2 = DUg2 + a2.

3.3.2 Modified Weak Form

The solution structure for grid refinement using the step function for refinement, H'r can

be written as shown in the previous section and the virtual displacements are written as:

u = (1- Href)(D3u')+Href (D3u2) (3-27)

The grid solutions in both grids are approximated by ug' = {N1{Xel with

N1 = (1-Hr)N and ug2 = {N2Xe2} with N2 = H N, where Nrepresents the shape

functions for grid variables. The boundary values are approximated u"' = {N"' } X"' with

N" =(i-Href)N and ua2 = {a2) Xa2 } with Na2 =HrfN where N represents the basis

functions used to approximate the boundary value function. Using the above approximations, the

expressions for strain takes the following forms:

a8u

Eu X el X al

12 2, a

In the above expression, B, and B2 are modified strain displacement matrices which

contain the terms involving the gradients of H'ref, Dirichlet functions and the gradients of the

shape functions used to approximate the grid solution and Bal, Ba2 are strain displacement

matrices which contain the gradients of the shape functions used to construct boundary value

functions. The expression for stress is given as o- = C ,, where C is the elasticity stiffness

matrix, and in computation of stress. The modified weak form takes the following discrete form

when the above expressions for stress and strain are incorporated into the weak form (Eq. 3-29).

NE NE
J{8X'j}J[B, BT]2[C][B, B2]{X'jdQ {Xe}T {Fe}+
el NE e-1 (3-29a)
NBE
I X' Tf {N N2I {t} dl
e=l pt

{Fe = N N2)T {b} d2- {Bar Ba2 T {a d (3-29b)
Q, Q,

In this expression {Xe = {ug ug2 and the summation is over the number of elements

(NE) and the number of boundary elements (NBE) in the grid used for analysis. As in the finite

element method these element matrices can be assembled to form global equations, where the

left hand side terms form a stiffness matrix and the right hand side a force vector. The Dirichlet

function and the boundary value function are constructed as explained in sections 3.1.3 and

3.1.4.

In this chapter, modified weak forms for the solution structures corresponding to enforcing

essential boundary conditions, treatment of material discontinuity and local refinement of grid

are presented in detail. These modified weak forms are used in finite element implementation for

solving several engineering boundary value problems. In the next chapter, a detail discussion on

the construction ofB-spline based finite elements is presented.

u = Dug + "

a /

/ \.
/ D

- / \

te i e Uo eUI e5

Figure 3-1. Solution structure for imposing essential boundary conditions

n ot n=T

u=

Figure 3-2. Elastic solid under equilibrium

kVT. fi = qh

h,TO

T=T

Figure 3-3. Steady state heat transfer problem

Curves represented by
implicit equations

05-
0.0

Union: D1u2 = Di+D2-DiD2

11

Intersection: Din2 = D1D2 Subtract: D-2 = DiD-2

Figure 3-4. Boolean operations on Dirichlet functions

Solid Region D() = D(,l2)n3 = (D(q) + -D(() D()D(q2))(D(03))

Figure 3-5. A CSG tree representing Dirichlet function for multiple boundaries

..- ............ -- .. .. .....
o ----------i---t-- -s--

S: u=10(1-xl)+5(1-x2)
-- +---*----c- --c- *----< -r4 -- f------tr---
.----+--4-----... + --+ -
e----- ----- -- -- + ---
... -. 4-4.....4 ..... 4....... + *. *- .-"... ... o
-4 4 i ----+t--*t-----t-6- O
~~0* -4- 4 -

o 1

(a) Non-zero boundary value specified at the edges

I

1i-

5 --

- O '
C) A 0

1 5 1.5

(b) Boundary value function constructed using cubic B-spline basis

Figure 3-6. Boundary value function using cubic B-spline basis

u: o

Matrix Inclusion Grid 1

Figure 3-7. Representation of grids for material discontinuity

Zero Slope

1 HInc
I-H

HInc Non

t t t t 1
Matrix
Boundary( = o) Inclusion Boundary (0 = 0) v

Figure 3-8. One dimensional inclusion function for treatment of material discontinuity

Grid 2

latrix

1

Fcii
I

Figure 3-9. Two dimensional inclusion function for an elliptic inclusion in a square matrix

A

F E

Figure 3-10. Representation of solution structure for local grid refinement

I I

I I

Sa 8

SG Bf
Coarse Grid Boundary( = 0) Refined Grid

Figure 3-11. Approximate step function for refinement

(a) -
Figure 3-12. Coarse and refined grid interface with multiple implicit equations

04
03
05-- 0,

___ 0
06

o 4 06

--Figure 3-13. Step function for refinement for a 2D model
OB Os, 4 -02 0 02 Q04 O 88

Figure 3-13. Step function for refinement for a 2D model

CHAPTER 4
B-SPLINE FINITE ELEMENT FORMULATION

B-splines provide a methodology to approximate a given set of points with smooth

polynomial functions. The basis functions are represented by piecewise polynomials and the

approximation provides a high degree of continuity depending on the basis used. For example,

C1 continuity is obtained when quadratic basis is used and C2 continuity is obtained when cubic

basis is used and so on. It is to be noted that the Lagrange interpolations which are used in

traditional FEM are CO continuous irrespective of the basis used. B-spline basis functions are

classified into uniform, non-uniform and NURBS. Uniform B-spline basis functions have

equally spaced nodes in the parameter space and the basis functions are periodic. Adjacent nodes

in non-uniform B-splines are not equidistant in the parameter space. NURBS are generalization

of non-uniform B-splines, where the basis functions are rational functions and are capable of

representing several analytical curves exactly. In this work, uniform B-splines are used for

approximating the solution space. The discussion on B-splines in the remaining of this chapter is

limited to uniform B-splines.

The B-spline basis functions have compact support and lead to banded stiffness matrices.

B-spline approximation provides high order of continuity and is capable of providing accurate

solutions with continuous gradients through out the domain. This is a major advantage over

standard finite element shape functions. The B-spline basis functions form a partition of unity

and are able to accurately represent the state of constant field in the domain of interest. This

property is very important for convergence of the approximate solutions provided by B-splines.

The basis functions used are polynomials and accurate integration can be performed by using

Gauss quadrature. The B-spline approximation has good reproducing properties and when a

cubic B-splines are used for approximating the solution, all the polynomial solutions up to third

order can be represented exactly.

B-spline shape (basis) functions are traditionally constructed using recursive definition

(Farin, 2002). The parameter space is partitioned into elements using a knot vector (equivalent to

a collection of nodes). Very general methods are available to insert knots and elevate the order

(or degree) of the polynomial basis functions. This framework therefore can be used to

implement h-p- and k-refinement and has been studied in the context ofisogeometric analysis by

Hughes et al (Hughes, 2005).

Instead of using recursive definitions as in geometric modeling applications, polynomial

definitions for B-spline basis functions are used here. Furthermore, it is convenient to define an

independent parameter space for each element in the grid instead of a continuous parameter

space (or knot vector) for all elements. Therefore it is assumed here that for each element in the

grid, parameter values vary within [-1,1] as is done for traditional isoparametric finite elements.

This is illustrated for the one-dimensional case in Figure 4-1 where a cubic B-spline

approximation is shown. In Figure 4-1, element is defined between vertex nodes 2 and 3, but the

approximation defined over (referred to as its span) is controlled by four nodes (1-4) which can

be refer to as its support nodes.

4.1 One Dimensional B-Spline Elements

The polynomial expressions for the basis functions can be derived using the continuity

requirements between adjacent elements. A kh order B-spline has k + 1 support nodes with some

of the support nodes lying outside the element. The span corresponding to the element is defined

by the basis functions of all the support nodes. The basis functions are polynomials of the form

k
N, = ar' whose coefficients a are determined using continuity requirements. Since the
j=0

parameter range for the element is r e [-1,1], the parameter value for first support node is

k+1 k+1
r = and for the (k + 1)t support node is r = for the parameterization to be uniform.
2 2

The approximated function shown in Fig 4-1, is a linear combination of the basis functions

within each element so that the span of the element e is represented as

k k k
f (r) = -NJu = Y aIr' (4-1)
1=0 1=0 j=O-

In the above equation, u are the support nodal values and it is assumed for convenience that the

first support node for element e has index 0. B-spline approximation of order k has Ck-1

continuity at the junction between adjacent elements. This continuity requirement can be stated

as

f() f, m + ) 0,1m= ...k-1 (4-2)

where, e and e+1 are adjacent elements whose derivatives are evaluated at the end point

(r=--) and start point (r=1)respectively. m = 0 represents the continuity equation

f/(1)= f, (-1) Using Eq. (4-2), the continuity equations can be expressed as

k k k k g *]
S(j m) i (j-)! ( jm)!
(4-3)
k j! )J-m 0
tJ- ak+l 1)J k+ = 0
=0 (j m)!

Since this equation is valid for arbitrary values of u, i = O.k +1, the coefficients

corresponding to each u,i = O..k +1 should vanish. This results in a set of k + 2 linearly

independent equations. Since k+2 equations are obtained for each value of m = 0,1,..k -1, a

total of k(k + 2) linearly independent equations are obtained. An additional equation stating that

k
the basis functions form a partition of unity N, = 1 is used to obtain (k +1)2 equations needed
1=0

to solve for all coefficients of the k +1 basis functions. Using this approach, the basis functions

for quadratic and cubic B-splines can be derived as shown in Eq. (4-4) and the basis functions

are plotted in Figure 4-2.

constructed using the procedure explained above. Quadratic B-spline element in one-dimension

has three nodes and hence three shape functions. The plots of the shape functions are shown in

Figure 4-2 and the corresponding equations are shown below:

N, =(1-2r+r2);N2= (6-2r2);N3= l(1+2r+r2) (4-4)
8 8 8

Cubic B-spline element. The plots of basis functions for cubic B-spline are shown in

Figure 4-3 where it can be seen that, unlike traditional finite element shape functions, the B-

spline basis functions are not unity at the corresponding node and do not vanish at other nodes.

Therefore, these basis functions do not satisfy the Kronecker-delta property and the

approximation constructed using these basis functions does not interpolate nodal values.

1 1
N, = (1- 3r+3r2 -r);N2= (23-15r -3r2 +3r3);
48 48 (4-5)
1 1
N= --(23+15r- 3r2- 3r3); N4= (1+3r +3r2 +r3)
48 48

4.2 Two and Three Dimensional B-Spline Elements

The basis functions for the higher dimensional B-spline elements are constructed by taking

product of the basis functions for one-dimensional B-splines. A typical grid used for analysis is

illustrated in Figure 4-4. This grid corresponds to cubic B-spline elements where the span

corresponding to each element is controlled by 16 support nodes. The grid elements used here

are regular quadrilaterals (rectangle/square) or regular hexahedra (cube/cuboid), hence the

mapping for geometry of elements from parameter space to the physical space is linear. Note that

even if an iso-parametric formulation is used the mapping will degenerate to a linear map

because the element geometry in physical space is regular and can be obtained by scaling the

element geometry in parametric space. In the following sub-sections, detailed discussion on

higher dimensional quadratic and cubic B-spline elements is presented.

shown in Figure 4-5 can be constructed using the product of the one-dimensional quadratic basis

functions (expressed in terms of parametric coordinates (4,7) ).

N3 1)+ r) = N,(4)NJ (r); i = 1..3; j =1..3 (4-6)

The span is defined by 9 support nodes, as shown in Figure 4-5(a), ,but none of these are at

the vertices of the element defined in the parameter space as the square [-1,1] x [-1,1]. It is

advantageous to have nodes at the vertices of the element. In order to accomplish this, the

quadratic B-spline finite element was defined as shown in Figure 4-5(b) as an assembly of four

quadrants (Q1-Q4). This representation facilitates implementation because each element will

have nodes at its four vertices. The mapping of the geometry of element from the parameter

space to the physical space is linear, as explained earlier, and can be represented as

1- xz+ x, (4-7)
2 2

Equation (4-7) is written using index notation, where x\ and x" are the lower and upper

bounds respectively for the nodal coordinates of the element. For each quadrant, only 9-nodes

(and hence 9 basis functions) are active. For example, the active nodes of the first quadrant Q1

are 6,7,8,10,11,12,14,15 and 16 based on the node numbering shown in Figure 4-5(b). The

basis functions obtained by product of one-dimensional basis functions will have a local

coordinate system (4, q) with origin at node 11 for Q1. In order to use a parametric coordinate

(r, s) with origin at the center of the element, the (4, q) coordinate system needs to be translated

Q1: 4=r-1;u=s-1 Q2: =r+l;q=s-1
Q3: r=r-1; r=s+l Q4: =r+l;r/=s+1 (

The basis functions at any given point (r, s) are computed by first determining which

quadrant the point lies in and then by determining the corresponding support nodes and basis

functions. As an example, the basis functions for a point in Q1 are

ND (r, s) = N(r )NU (s 1); N2D (r, s) = N (r 1)N (s 1)
N2D (r, s)= N, (r 1)N (s 1); N (r, s) = N (r 1)N (s 1)
N2D(r,s)= N,(r 1)N (s -1);N (r,s)= N3(r 1)N(s 1) (4-9)
N2 (r,s)= Nl (r 1)N3(s -1); N12(r,s)= N (r 1)N, (s -1)
ND(r,s)=N3(r-1)N3(s- 1);ND (r,s)= ND(r,s)= 0
ND (r,s)= ND (r,s)= N2 (r,s)= N, (r,s)= N (r,s) = 0

N,2D(r,s) refers to bi-quadratic basis function corresponding to ith node of the element

and N, (r) represents the one-dimensional quadratic basis function as defined in Eq. (4-4).

Three dimensional quadratic B-spline element can be constructed in a similar manner and

has eight octants which are assembled to construct the 64-node element. The basis functions are

constructed using the product of ID basis functions and in any given octant, only 27 basis

functions are active. The mapping for the geometry is done as shown in Eq. (4-7).

Cubic B-Spline elements. The two dimensional cubic B-spline elements are constructed

by product of one-dimensional cubic B-splines. In this case, there are 16 support nodes as shown

in Figure 4-6 and there is a node at each vertex of the element ([-1,1] x [-1,1] in parametric

space).

N2 1)+,(r,s) =N(r)N (s); i = 1..4;j = 1..4 (4-10)

NAT1k 1)+4(J )+ ,(r,s,t)=N,(r)NJ(s)Nk(t); i = 1..4;j = 1..4;k = 1..4 (4-11)

The geometry mapping between parametric and physical space can be defined using Eq.

(4-7). The lower bounding node for this element is node 6 and the upper bounding point

corresponds to the node 11. Again, some of the support nodes lie outside the domain of the

element and are shared with neighboring elements. The 2-D bi-cubic B-spline element has 16

basis functions (Eq. 4-10) while the 3D cubic B-spline has 64 basis functions (Eq. 4-11). In these

equations, N (r) represents the cubic basis functions as defined in Eq. (4-5)

Polynomial segment(span)

C2Continuity
l Approximal
Nodal Values

1 E2 E1 3 E2 4 E3 5

Nodes
Elements

Figure 4-1. Approximation using ID cubic B-splines

I N2
I I
I I
0.5
I NI
NI N3
I I
I I
0 ---- ---'- ----- 0

1 2 mi 3
Element

Figure 4-2. Shape functions for ID quadratic B-spline element

10.5

1I 3
I I
I I
I I
0.5i

1 2 N1 3
o 0
r-1 r-1
Element

Figure 4-3. Shape functions for ID cubic B-spline element

Function

6
6

4

0 0 6 -

1 Somre noIe lis ouridie tie element geomeLry

Element
S | i Support nodes enclosed by dashed box

Boundary of analysis domain represented
by implicit equations

-- I

B-Spline Approximation for field variable
1 (1 1) ..

Reference Ele

Linear Map for Geometry of Element, S nt

Support nodes ousi e element geometry

Figure 4-4. Typical grid used for analysis and the mapping used for geometry and field variable

* 0

S
---------------------*----------e

13 14 15 16
7 8 9 r --- ---

I I
-------<-------4 I I

S19 10 Q2 Q 11 12
I I
4t---,,---> I _-._-----------
4 5 6 Q3 Q4
I I Il
---- -----.--- "---" ---- -----
SI I I
------- ----- 5 I 6 I 7 8
1 2 3 I I I_____
I-- I
I I I
I I II
(a)

1 2 (b) 3 4

Figure 4-5. Two dimensional quadratic B-spline elements in parametric space

--------------------------------
13 14 15 16

S
i--- -------I I
9 10 11
r

>------.-- -- -.
5 6 87 8

h.---------C---------~- ----------

1 2 3 4

Figure 4-6. Two dimensional cubic B-spline elements in parametric space

CHAPTER 5
NUMERICAL IMPLEMENTATIONS

The weak form was modified for various solution structures and was discussed in a

previous chapter. In this chapter, a detailed discussion on the efficient and accurate computation

of the terms involved in the modified weak form is presented. In the first part of this chapter, a

discussion on the computation of stiffness matrix and load vector corresponding to the solution

structure for imposing essential boundary conditions is presented. In the next part of this chapter,

a discussion on the construction of stiffness matrices involved in the solution structure for

micromechanical analysis is presented. Finally, a discussion on the computation of various terms

involved in the solution structure for adaptive refinement is presented.

5.1 Numerical Implementations for Linear Elastic Problems

The modified weak form in discrete form for linear elasticity was presented in a previous

chapter and was shown to be:
NE NE NBE
i{6X}T [T[C][B]{Xe}d= { } {F(X}+T + SX'} T {N}T t}dF (5-la)
e=1 ^ e=1 e=1 pt

Fe}= {N}T {b} d- f{B}T {a }dQ (5-1b)
Q, Q,

The stiffness matrix is given by the expression f [B]T [C][B]dQwhich needs to be

computed for all the elements present in the structured grid. A typical structured grid used for

analysis is shown in Figure 5-1. The grid elements can be classified into three types and the

computation of stiffness matrix depends on the type of element that is being integrated upon. In

the following, a discussion on the computation of stiffness matrix in each type of element is

presented.

5.1.1 Computation of Stiffness Matrix for Internal Elements

The internal elements lie completely inside the analysis domain and the strain

displacement relation [B] for two dimensional problems can be written as follows:

0
aN,

[B] 0 N (5-2)

3x2 32x,

In the above equation, N, are the shape functions used to approximate the displacement

solution and m is the number of nodes for the element. The stiffness matrix is computed using

[K]2m2m =l ZTZ WW [B(,) [][B (5-3)
j-1 1-1

In the above equation, nx,ny are the number of Gauss points in x and y directions

respectively, w, is the Gaussian weight associated to the Gauss point, at which the integrand is

evaluated. It is to be noted that the stiffness matrix for all the internal elements is identical and

stiffness matrix is computed only once and is used for all the internal elements. This leads to a

significant savings in computation cost associated with computation of stiffness matrices when

compared to FEM, where stiffness matrix is computed for all the elements.

5.1.2 Computation of Stiffness Matrix for Boundary Elements

The boundary elements lie partly in the analysis domain; this leads to partial elements and

hence needs special techniques for computation of stiffness matrix. The strain-displacement

matrix for these elements is identical to the strain-displacement matrix for internal elements as

shown in Eq. (5-2). To compute the stiffness matrix, the partial solid region is divided into

triangles as shown in Figure 5-2.

The stiffness matrix is computed by adding the contribution from all the triangles

generated by triangulation as shown below:

[K]mm = [K], (5-4)

In this equation, n, is the number of total number of triangles and [K], is the stiffness

matrix computed using Gauss Quadrature for ith triangle. Several algorithms are available for

automatic triangulation of a given set of points on a plane. In this work, Delaunay triangulation

algorithm is used for triangulation. It is to be noted that curved edges are approximated by

straight lines which can leads to error in the computation of stiffness matrix. However, this error

is limited to boundary elements and the errors incurred due to this approximation are similar to

the errors involved in FEM where the boundaries are approximated with conforming meshes.

Furthermore, when the grid is made finer, the errors will approach zero. In three dimensions, the

stiffness matrix is computed by tetrahedralizing the solid region and the errors involved in this

approach are same as the ones involved in triangulation in two dimensions.

5.1.3 Computation of Stiffness Matrix for Elements with Essential Boundaries

For the boundary elements containing essential boundaries, the Dirichlet functions are

active and hence the corresponding strain-displacement matrix has gradients of Dirichlet

function. Since Dirichlet functions are defined on a narrow band with 6 -> 0, their gradients can

have large values in the small band. In order to accurately compute the contribution due to

Dirichlet function and its gradient, special techniques are required. In this section, one such

technique is presented and discussed in detail. The strain-displacement matrix can be written as:

D, ON + N, O
Oa, a,

[B]3,2m 0 D D N + N (5-5)
ax,
2 2
aN 8x 8x 8xD
D, + N D, ND OD 2
ax2 ax ax ax

The above equation shows that the strain-displacement matrix has gradients of both shape

functions and Dirichlet functions. It is convenient to decompose the [B] matrix into two parts

[B] = [B] + [B2] such that [B,] contains terms involving the gradient of the shape functions and

[B2] contains terms with the gradients of the D function as shown below:

N1 0
D,
D1 0 0 0 l
D 9X2 DvI B12 (5-6)
[B1],, = 0 D2 =0 D,2 0 0 =[D B] (5-6)
DX2 0 0 D, D2 9 0

N, 0 0 .D 0
a 2 al 0 9NAx
ax,

1 xo,

[B, 32 0 N, 2 0 N = 2 0 [D2 B2l (5-7)
x, 1x 0 N, 3x2 2x2m
[ND ]Nm = o D OD

& Dx, x^, 8x,

Using the above procedure, the strain-displacement matrix can be rewritten as

[B]= [DI] [B i + D2 ][ Similar matrices can be defined for three dimensional elements

also. Using the decomposition of strain displacement matrix, the stiffness matrix is decomposed

into the following components:

[K]= f[B]T [C][B1]dV+ [B,]T [C][B]dV+ [B2 ]T [C[B1]dV+[B2 ]T[C][B]dV
Sv v (5-8)
=[K1]+ ([K2]+[K2 ]T +[K3,]

The first component [K, is computed as described earlier by decomposing the boundary

elements into triangles or tetrahedra to perform the area or volume integrals respectively. The

second and third components contain gradients of Dirichlet functions which are large near the

boundary when 6 is made very small and they vanish away from the boundary, hence the terms

in the integrals will be non-zero in a thin band along the boundary. This allows us to convert the

volume integral into a surface integral. The contribution from the gradients of finite element

shape functions can be assumed to be constant across the band but varies along the boundary and

the contribution from the gradients of Dirichlet function are constant along the boundary but vary

normal to the boundary. The band is represented in terms of two parameters t,n as shown in the

Figure5-3 and the terms in stiffness matrix take the following form.

The stiffness matrix as decomposed in Eq. (5-8) has four components; the second

component [K2] can be further decomposed as shown in the following equation.

[K= J[B T[c][B2]dV D [C] D2 d1B d
VE (5-9)

S[BJTf J[C2 [B2]dF

In this expression, the terms involving only Dirichlet functions and their gradients are

separated from the terms containing the basis functions and their gradients. This decomposition

allows us to compute the stiffness matrix efficiently and accurately. The volume integral has

been decomposed into a line integral along the normal to the boundary and a surface integral

along the boundary. The normal for the boundary represented by implicit equation 0 = 0 is given

by VO at the boundary. The component involving only Dirichlet function and its gradients is

termed as [c2 ] in Eq. (5-9) and this needs to be integrated only across the band (along the

normal to boundary), which can be done analytically as shown in Eq. (5-10). In computation of

the integral across the band, it is assumed that the gradients of the implicit function with respect

to the parametric coordinates of the element are constant; this is a valid assumption as the

integration is performed in a very thin band 8 w 105 and hence the gradients remain

approximately constant.

8D D 1 a1
D1Cl D D C12 O2 -1 1 sC12 1 00
x1 1 2 112J, dIs, 2J22 OS2

D2C2 D2
D2C12 aD1 DC OD2 C 1 as 1 aO
f[c2d= f 12x 2 2 dx2 1 2J 1 s, 2J22 OS2 (5-10)
OD 8D VsO IVs 1 O 1 O
SDC3O D3a C C33
3 2 3 1 332 22 OS2 2J I1 Os1
D D2 2 C 2 C o
D2C33 aD1D O a2 C3 1 as 1 as
3x2 3 1 2J22 OS2 2Jl, as1

In the above equation, the terms J,, denote the components of the Jacobian matrix of the

transformation described in Eq. (4.7) which transforms the element from parametric space

(s1,s2, s3) to the physical space (x,, x2, x3). It is to be noted that the Jacobian matrix is diagonal,

as the grid elements are regular quadrilaterals or hexahedra. Proceeding in the similar manner as

explained above, the third component of the decomposed stiffness matrix in Eq. (5-8) can be

written as shown in the following equation

[K,]= [B2 T [[B2]dV 2 Jf JD2] [C]] d B2]dF
VE (5-11)

J[B2fJ]T[c3}zs?[%]dl
F \sD

The components involving only the gradients of Dirichlet function is termed as [c3] and its

integral across the band is given as

1 w 4 s1 n2 4 1 & Sc n i 4
C11 +C33 1 s (C12 + C33)
CT3 e=% -rh 3( n 3s JoJ 3sh(5- (5-12)
V(1 1^ 9(-f)}f _,q f C 4C
41 4 +C134
J(12 J,,33 s s 22 2 T s 33 J s
J11J22 1 CS8 32 12 9s, 31 / 2 \ 3

The computation of the integral along the boundary F is performed numerically by Gauss

quadrature. This approach leads to accurate and efficient computation of stiffness matrix for

boundary elements with essential boundaries.

The right hand side of the modified weak form as shown in Eq(5-1) has two terms, the first

term is a volume integration for the body loads and applied non-homogenous essential boundary

conditions which is done similar to the computation of stiffness matrix as explained in the

previous sections. The second term is a surface integration over the natural boundaries.

The boundary is approximated as a set of line segments in 2D as shown in Figure 5-4 and

surface triangles in 3D and integration is performed using Gauss Quadrature as shown below:

{ft2m = f{N}T {tdF = w, {N((I)}T {t(Sl)}
L, 1=1 (5-13)

{f}2ml =Z{f}
1=]

As the grid is made finer, the approximation of the boundary approaches the exact

representation and the errors associated with approximation of boundaries with lines (planes)

approach zero. In traditional FEM, the natural boundary is approximated by the surface formed

by the nodes on the boundary and associated shape functions. This results in a load vector which

has non-zero values corresponding to the nodes which are present on the surface. However, in

the implicit boundary method, nodes are not guaranteed to be present on the surface, therefore

the load vector for a boundary element with natural boundary will have non-zero values for all

the nodes corresponding to this element. The computed load vector is a work-equivalent load

When the solid region inside the element is very small, its contribution towards the

stiffness matrix will be negligible and may lead to singular stiffness matrices. To overcome this

problem, the elements with low volume fraction, i.e. r <5%, = vol(Solid) are identified and the
vol(Element)

nodal displacements in these elements are constrained to be an extrapolation of the

displacements from the neighboring internal elements. The details of this extrapolation approach

are presented in section 5.5.

5.2 Numerical Implementations for Steady State Heat Transfer

The modified weak form for the solution structure for steady state heat transfer was

discussed in a previous chapter and was shown to be:

f I[B]T k [B] {T }d, + f J[N] h[N] {J}dr =
F r (5-14)
-Z R] k[B]{T'd + [Nrde [N]Tfdo, + [N] q dF i[N] h(Ta T.")dF
,e Fe Th rc

In this equation, the left hand side corresponds to stiffness matrix and the right hand side

corresponds to load vector. The left hand side has two terms the first term is stiffness matrix and

the second term is a contribution to the stiffness matrix due to convection boundary condition.

The right hand side has four terms, the first term is load due to applied temperature on essential

boundary condition, the second term is load due to heat generation in the body, the third term is

due to applied heat flux on natural boundary and the last term is load due to convection on

convection boundary.

Computation of Stiffness Matrix. The stiffness matrix comprises of a volume integral and a

surface integral. The volume integral is performed as explained in section 5.1.1 and 5.1.2 after

subdividing the boundary elements into triangles or tetrahedra. The surface integral for load

computation is evaluated as explained in section 5.1.4. For the boundary elements with essential

boundaries, the [B] matrix can be decomposed as shown below:

D N +N 0
[Bl] = D Ol (5-15)
0 D '2 +

In the above equation, [B] matrix contains the terms involving gradients of shape functions

and also the gradients ofD function. It is advantageous to decompose this matrix further to

[B] = [B1] + [B2] such that contains terms involving the gradient of the shape functions and [B2]

contains terms with the gradients of the D function as shown below:

D 0 0
ax, Qx,

0 D-N 0 N
Nx, Ox,

N1D 0

8D, 8x, ~N, .
[B 2]2 0 N, = N 0 =D2 B2 (5-17)
a2 00 N1 10 = 2DJLo N1
Sx2 0 N .

)2 1X
N, N-

Using the above procedure, the gradient matrix can be rewritten as [B] = D [B

+ D2 [B, Similar matrices can be defined for 3D elements also. Using this decomposition,

the stiffness matrix is decomposed into the following components:

[K,]= f[B1]T [k][B, ]dV + f[B,]T [k][B2]dV + f[B2 ]T [k][B, ]dV + f[B2 ]T [k][B2]dV
Sv (5-18)
=[K + ([K2]+[K ]T) +[K]

The first component [K, is computed as described earlier by decomposing the boundary

elements into triangles or tetrahedra to perform the area or volume integrals respectively. The

second and third components contain gradients of Dirichlet functions which are large near the

boundary when 6 is made very small and they vanish away from the boundary, hence the terms

in the integrals will be non-zero in a thin band along the boundary. This allows us to convert the

volume integral into a surface integral as explained in detail in section 5.1.3. The second term in

Eq. (5-18) can be written as:

[K2]= [B1]T [k][B2]dV = j [Bl] f[D1j[k][D2 dBB2 dF
E r (5-19)
= [B C2 2[B,] dF

This term is evaluated as a surface integral as explained in section 5.1.3.

5.3 Numerical Implementations for Micromechanics

The modified weak form for the solution structure corresponding to micromechanics was

discussed in a previous chapter and was shown to be:
NE T IT
{8Xe} JI B2] [C][B, B2] {XedQ {Xe} T{N N2} {b}d+
e1 C el (5-20)

{jXe}T IN1 N2 }{t}dF
e=1 pt

{Xe} = {g ug2 T (5-21)

In this section, a detailed discussion on the numerical implementations for the computation

of the terms involved in this weak form is presented. The left hand side of the weak form

corresponds to the stiffness matrix while the right hand side corresponds to the load vector. The

solution structure for micro-mechanics is such that the solution from the matrix grid inside the

inclusion is zero due to the step function of inclusion and similarly the solution from inclusion

grid outside of the inclusion vanishes. Therefore the nodal values of the matrix grid inside the

inclusion are set to zero, provided these nodes are outside the influence ofH'"C and the nodal

values of the inclusion grid outside the inclusion are set to zero. This leads to the reduction in the

total number of degrees of freedom for the analysis model.

The finite element solution for the grid variables are written as ug1 = [Ng' ]{ug)} and

u2 = [Ng2 ]u2 }. For two dimensional case, the strain vector using the strain displacement

relations is written as shown below:

uu

S a2 [B1 B2 Ug (5-22)
712I
+-
a 2 1,

The components of the strain displacement matrix B1,B2 are expressed as:

(1 H ) agl w 0

N 0 (1 H O Ha)

)x, 1 x, Hx, xN
a2 a21 OX1, O

HinC aNg2 + 2
H +-N 0

2, ax
n Ng2 Ngl 8H nc
B2 =0 H+"c -N, .. (5-23b)

H +-NN2 H nc _+ N+g2
a8x2, x, x ,

In the above equations, the decomposed components of strain-displacement matrix have

terms involving gradients of finite element shape functions Nk and the gradients of the
ax

inclusion function -- It will be advantageous to further split these components into a term
ix

containing gradients of shape function only and another term containing gradients of inclusion

function only. This is accomplished as shown in the following equations.

xl, &9H,
(81 H= a, 0 (1 -A)gl 0

B, =B +B,12= 0 (1 H'-) .+ 0 --- gl" (5-24)

a' a' 2H"" a2H"
(1-H"") (1-H") 'g'
ax, ax a2x, C9x

H 0 n a'g2 0
H 8x, C'

B2 = B21 +B22 = H"- .+ 0 N2 (5-25)
Cx, ax,

8x2 C9x, x2 aHx '

It can be noticed that the modified strain displacement matrices B, B12, B21 and

B22 vanish in certain regions of the computational domain and are non zero in other parts of

computational domain depending on where the inclusion function and its gradients vanish.

Therefore the computational domain is divided into three regions as shown in Figure 5-5 and

they are explained in detail in the following.

5.3.1 Computation of Stiffness Matrix in the Matrix Region

In the matrix region of the microstructure, the degrees of freedom for grid corresponding to

matrix are active, therefore the strain-displacement matrix as shown in Eq. (5-23) becomes

B = [B1 B2] = [B,1 0] with Hn" = 0 in Eqs. (5-24) and (5-25). The stiffness matrix will take

the following form in matrix region:
-tB T 1CmatB 0
K= [B1 B2 ]Tmat[B, B2]dV= B 0 V (5-26)
Q, Q, I

This stiffness matrix is computed by performing the volume integration over the elements in

matrix region. The grid element in the matrix region will be an internal element and the volume

integration is performed as discussed in the previous section 5.1.1.

5.3.2 Computation of Stiffness Matrix in the Inclusion Region

This region corresponds to the inclusion region. In this region the degrees of freedom for

grid corresponding to inclusion are active, therefore the strain-displacement matrix as shown in

Eq. (5-23) becomes B = [B B2] = [0 B21] with H" = 1 in Eqs. (5-24) and (5-25). The

stiffness matrix takes the following form:

K= f[B B2 c]Tinc [B l B2]dV= f T V (5-27)
QQ, 21, 21

The stiffness matrix is computed by performing the volume integration over the elements in

inclusion region. The grid element in this region will be an internal element and the volume

integration is performed as discussed in the previous sections 5.1.1.

5.3.3 Computation of Stiffness Matrix in the Interface Region

In this region, the step function for the inclusion is active. This means the step function is

varying in this region and strain-displacement matrix as shown in Eq. (5-23) becomes

B = [B1 B2]. The stiffness matrix is expressed as:

K=f[BB BT CB B1 CB2 ]= [ KV K12 (5-28)
2 2 B T CB, B CB2 K21 K22
Q, Q, ]Tc[B 1 2 C2 .K21 K22

In this expression C = C'"' or Cmat depending on which region is being integrated upon.

Each of the terms in this matrix can be decomposed as a sum of volume integral and a surface

integral. The procedure for this decomposition is described in detail for the first term in the

stiffness matrix K11.

K,, = J BCB1dV= (B + B12)T C(B11 +B12 dV
(5-29)
= J(BCB, + B CB12 + BCB, + B2CB12 )V
Q,

In this expression, the first term is integrated as a volume integral after the partial elements

are subdivided into triangles or tetrahedral as explained in section 5.1.2. The remaining three

terms are converted into surface integrals, as these terms are non-zero in a thin band in which the

inclusion function is non-zero. The strain-displacement matrices shown in Eq. (5-25) are

rewritten in the following form:

H a 8Nl 0 ~8Nl O
Nnc ON81 ON191
H 0 0

B21= 0 H nc = Hinc 0 = HncB21 (5-30)
8x2 a2

H ia8N'1 H nc 1 N8N81 N 8N1l
82 a2 a2 a2

aHjc aH' 0
Hinc Ng2 0 Hnc 0

AHJnC AnCF Ng2 0
B22= 0 Nn. = 0 0 = H B22 (5-31)
2 2N2 0 Ng
OHawnc 2 ti

Using the above procedure, the strain-displacement matrix can be rewritten as

B2 = HncB21 + H2B2 Similarly, the first component of strain-displacement matrix is written as

B, =(1- HincB,) H B These matrices can be defined for 3D elements also in a similar

manner.

The last three components in stiffness matrix as shown in Eq. (5-29) contain gradients of

inclusion functions which are large near the inclusion boundary when 6 is made very small and

they vanish away from the boundary. Therefore the terms in these integrals will be non-zero only

in a thin band along the boundary. This allows the volume integral to be converted to a surface

integral. The contribution from the gradients of the shape functions can be assumed to be

constant across the band but varies along the boundary and the contribution from the gradients of

inclusion function is constant along the boundary but varies normal to the boundary. The band is

represented in terms of two parameters t,n as shown in the Figure 5-3. The stiffness matrix as

decomposed in Eq. (5-29) has four components; the second component JBfCB12dV can be
E

rewritten as shown below.

J BTCB12dV= f BT ((i-Hn)CI2 )12dV

SB (5-32)
~- 1 -r --r
= B1T 1-H n,)C [-2 do B12dF f Bl1CB212dF
F 110VF

In this expression, the terms involving only inclusion functions and its gradients are

separated from the terms containing the shape functions and their gradients. This decomposition

leads to efficient and accurate computation of the stiffness matrix. The volume integral has been

decomposed into a line (surface) integral along the normal to the boundary and a surface integral

along the boundary. The normal for the boundary represented by implicit equation 0 = 0 is given

by V at the boundary.

The component involving only inclusion function and its gradients is termed as C2 in Eq.

(5-32) and is evaluated in the similar manner as explained in section 5.1.3.

5.3.4 Periodic Boundary Conditions

Composite materials have periodic arrangement of cells and their behavior can be

predicted by analyzing the behavior of the unit cell which is often termed as representative

volume element (RVE). The variations in the micro-scale (RVE) are rapid and periodic as

compared to the variations at the macro-scale. In order to analyze the behavior at the micro-

scale, the RVE is analyzed with appropriate loading and boundary conditions. The RVEs tile the

computational domain by translation and the neighboring cells fit into each other in both

deformed and un-deformed states as shown in Figure 5-6. Hence the boundary conditions for the

RVE are periodic in order to preserve the continuity of displacements, strains and stress across

each RVE. The periodic boundary conditions are expressed as linear constraints and they are

implemented as multipoint constraints in a typical finite element code. The periodic boundary

conditions are expressed in the following form:

uF -uF = const or in general uF -AuF + u= 0 (5-33a)

uF -uF = const or in general uF -AuFi + p = 0 (5-33b)

The multipoint constraints in Implicit Boundary Method can be implemented exactly as is

done in traditional finite element method (Cook et. al. 2003). It is to be noted that the RVE is

always a rectangle/cuboid, hence when Lagrange interpolations are used, the nodes are

guaranteed to be present on the boundaries on which multipoint constraints are imposed.

Furthermore, since structured grids are used, identical nodes are always guaranteed to be present

on the opposite faces and facilitate the imposition of periodic boundary conditions. However,

when FEM is used for analysis, care needs to be taken to generate mesh such that there are

identical nodes present on opposite faces.

The nodes on the faces (or edges in 2D) of the RVE can be partitioned into two parts:

independent or master nodes (s) and dependent or slave nodes (s). The nodal displacement vector

can be similarly partitioned into u = [um u], where um are displacement of the master nodes

and u, are those of slave nodes. The displacements for the slave nodes are specified in terms of

the master nodes in the following form us = Giu + g1, where gi is the non-zero periodic

boundary conditions (e.g., Case 1: u =u k + L). Hence the displacements are expressed as :

U = [Um Us= IG [um]+ = G[um]+g (5-34)

where, I denotes the identity matrix and 0 is the zero vector. The matrices G and g are

constructed globally for all the master degrees of freedom present in the analysis model. Using

this expression for displacement and a similar expression of the corresponding virtual

displacement 6u = G [6um ] in the weak form, the global stiffness matrix is transformed into

K' = GTKG and the load vector is increased by -GTKg to impose the multipoint constraints. It

is to be noted that in a general case of multipoint constraints, G and g are constructed such that

they may be a fully populated matrix and vector respectively. However, when structured grids

are used for analysis, identical nodes are guaranteed to be present on the opposite faces and

hence with proper choice of node numbers, G matrix can be constructed to be a diagonal matrix.

This simplifies the computations involved in the transformation of the stiffness matrix and hence

leads to efficient implementation. The periodic boundary conditions involved in a square array of

fibers are shown in Figure 3-10 and the corresponding multipoint constraint equations are shown

in Table 1.

5.3.5 Computation of Effective Properties

Effective properties of the composite materials with periodic distribution of

microstructures are computed by applying far field strains to the RVE as shown in Figure 5-7.

The displacements, strains and stresses are assumed to be periodic in the RVE. Hills lemma

states that the average of the strain energy in RVE is same as the strain energy due to average

stress and strains. It can be stated in the form (a : ) = () : (E) .This property is used to compute

the effective properties from an RVE. Using the stress-strain relation (a) = CH(), the effective

properties are represented as (a):(: ) = (CHE: ). The average stresses and strains are computed

by solving the boundary value problem six times with different applied strains. These average

stresses and strains are used in the relation (a) = CH(") to determine a row (column) of the

homogenized stiffness matrix CH. Six different strain states are used to determine all the

components of CH. The strain-stress relation for orthotropic materials is given as follows. Once

the homogenized stiffness matrix is determined, the effective properties can be computed by

using the strain-stress relation.

1 v21 31 0 0 0
E1 E2 E3
12 V32 0 0 0
1' E1 E2 E3 011
_22 V13 V23 1 0 0 0 o 22
E33 E1 E2 E3 22 [ t (5-35)
723 0 0 0 1 0 0 r23
731 G23 131
7122 0 0 0 0 1 0 1"12
G31

0 0 0 0 0
G12

In two-dimensional problems, the effective properties in the transverse directions are

determined by analyzing the model in plane strain conditions. The longitudinal modulus is

computed by analyzing the model in generalized plane strain conditions. The longitudinal shear

moduli are computed by analyzing the model in longitudinal shear loading conditions. Each of

these cases are explained in brief in the following sub-sections.

Plane strain. In plane strain formulation, the strains in the x3 direction (longitudinal

direction) are assumed to be zero and the stress-strain relation is constructed as follows:

r; 1-v V 0
2 (+v)(l 2v) 1 v 0 2 (5-36)
1"12 0 0 1-2v 712
2

This stress-strain relation is used in the modified weak form shown in Eq. (3.3) to solve the

boundary value problem with specified strain loading and periodic boundary conditions to

compute the transverse material properties.

Generalized plane strain. The homogenized material properties in the longitudinal

direction are computed by using a generalized plane strain model (Li and Lim, 2005), where the

strains in the longitudinal directions are specified. A constant strain in the x3 is assumed and the

strains and stresses are expressed as follows

8, 1 C11
ou
E2 an2 2 C21
2 2 and I
E3 03 C31
L712 u, 3'12 0
ax, ax,
Ox2 1Xl

Using the constant strain in 3-direction

C2 12 22 0 C2 + 23
T12 0 0 C66 12

and-3 [C13 C23 0] e2 +C33 3
12/2

0 Ez
C66 E3
C66 /12

i, the stresses are written as follows:

3 = [C] {} + Ck3

The principle of virtual work as shown in Eq. (3.3) is modified by choosing the virtual

strain vector as 6E = {(E, 8E2 712 } to get the following equation (Eq. 5-39). In this

expression, the last term is the load due to strain in the third direction. The rest of the formulation

is exactly similar to plane strain formulation.

{&6} T{( } dQ n fu}T {t}dF + f{6u}T {b} d- {6} CkE3 (5-39)
n F, Q

direction are computed by using a longitudinal shear loading, where there is only one degree of

freedom which is the displacement in the 3rd-direction. This assumption results in non-zero shear

strains in 3-2 or 3-1 directions only while all other strain and stress components are identically

zero. The equations for shear strains and shear stress are given by the following expression:

(5-37)

(5-38)

rU3
732 a 32=3 G32 OJ{732{ (5-40)
73 J 31 0 CG31 jy31

The principle of virtual work as shown in Eq. 3-3 is modified by choosing the virtual strain

vector as 6b = {Y32 731 } This formulation is used to specify unit shear strains and hence the

corresponding shear moduli are computed.

The right hand side of the modified weak form Eq. (5-20) comprises of load terms. The

first term is load due to body load and the volume integration is performed as explained in

section 5.1.1 and 5.1.2. The second term is the load term due to applied traction. The shape

function vector will be either (N1 0) or (0 N2} depending on which region the natural

boundary condition is applied. The computation of load vector is performed in a similar manner

as explained in section 5.1.4.

5.4 Numerical Implementations for Local Grid Refinement

The modified weak form for the solution structure corresponding to local grid refinement

was discussed in a previous chapter and was shown to be:
NE NE
{(X'} JI[B, B2] [C][B, B22] X'ed S6X'e} {FT
el a e=l (5-41a)
NBE
X {e}'T IN1 N2} {t}dF
e=l pt

{F'e} N1 N 2} b} dQ- (Ba, Ba2 T a d (5-41b)
X u u2 (5-41c)

{Xe = Inu1 ug2}T (5-41c)

In this section, a discussion on the numerical implementations for the computation of the

terms involved in this weak form is presented. The left hand side of the weak form corresponds

to the stiffness matrix while the right hand side corresponds to the load vector. The solution

structure for local grid refinement is such that the solution from the coarse grid inside the refined

grid is zero due to the step function for refinement. Therefore the nodal values of the coarse grid

inside the refined are set to zero, provided these nodes are outside the influence ofHrf This

leads to the reduction in the total number of degrees of freedom for the analysis model. The

computational domain can be divided into three regions as shown in Figure 5-8. It can be

observed that the solution is uI outside the refined grid and transitions from uI to u2 inside the

refined grid. It can also be observed that in the regions where step function for refinement is

active, the Dirichlet functions are inactive and vice versa. This is because of the fact the step

functions for refinement are active (varying from zero to one) only on the interface of the edges

which are shared between the refined and unrefined grid elements and the Dirichlet functions are

active only on essential boundaries and the essential boundaries cannot form the interface edges.

The computation of stiffness matrix is explained in detail in the next section.

5.4.1 Computation of Stiffness Matrix in Regions 1 and 2

The step function of refinement is not active (H"r = 0) in Region 1 and the solution

structure becomes u = Dugl + ual and hence the strain displacement matrix is given as follows:

dN dD
ON, OD
cx 'xax,
D + 0 ;[BX]32m 0 0
=[B]320 0 0 (5-42)
20 0

D + N, ,- D + N,
a2i2 s,1 1x

This matrix is similar to the one explained in section 5.1.3 and the stiffness matrix is given as:

BCB 0 (5-43)
K= [B1 B2 TC[B1 B2]dV= V (5-43)
Q, Q, 0 0

The computation of stiffness matrix is performed in a similar manner as explained in

sections 5.1.1-5.1.3.

In the Region 2, the step function of refinement is equal to one and the solution structure

becomes u = Dug2 + ua2 and the strain displacement matrix will be similar to the one shown in

Eq. (5-42) and hence the stiffness matrix becomes:

K= [B, B2]TC[B1 B2]dV= J B CB V (5-44)
Q, Q, -- 120 B CB

The stiffness matrix is computed as explained in sections 5.1.1-5.1.3.

5.4.2 Computation of Stiffness Matrix in Region 3

In the Region 3, the step function of refinement is varying and the Dirichlet functions are

unity. It is to be noted that there is a small region of the size 6 where both Dirichlet functions

and step function of refinement may be active, since this region is very small, the contribution to

the stiffness matrix in this region is negligible. The solution structure takes the form:

u = (l- Href) (ug" + ual) + Href (Ug2 + ua2) and hence the strain displacement matrix takes the

form B = [B B2]

(1- H') 0 NO 0-Ngl 0
ax, ax,
B, = B, +B,2 = 0 (1-H) x + 0 N --- (5-45)

a(1-H)2 (1-H X 2
6x2 1x, 2 1i

H"f ONg2 0 ~ N g2 0

B 21 B22, = 0 H:' + 0 N .g (5-46)
ax2 2

2 1 212 i 1

The strain displacement matrix takes the similar form as explained in section 5.3.3. Hence the

stiffness matrix will take the following form:

iTC B T C CB1 BT CB2V = [K K,
K= [BB B2 C[B B2dV= 2 V K K12 (5-47)
B, T BCB, B CB, K2 K,,

The computation of stiffness matrix is performed in the similar procedure as explained in detail

in section 5.3.3.

The load vector is computed in the similar manner as explained in section 5.3.6.

5.5 Extrapolation of Solution for Elements with Low Volume Fraction

In structured grid based methods, the grid elements do not conform to the geometry of the

analysis domain and hence there may be few elements in the model with very small solid region.

When the volume of the solid region inside the grid elements approach zero, the stiffness matrix

corresponding to these elements will have components with magnitudes which are very small in

comparison to the internal element stiffness matrices. This leads to ill-conditioning of the system

of linear equations and hence poor convergence behavior.

In order to avoid the problems associated with grid elements having small volume

fractions, an extrapolation scheme is presented in this dissertation. The degrees of freedom

corresponding to the elements with small volume fractions are constrained to be an extrapolation

of the degrees of freedom from the internal elements. Due to this extrapolation scheme, the

degrees of freedom for external nodes will not be arbitrarily large due to near singularity of the

stiffness matrix of partial elements. An added advantage of this procedure is that the total

number of degrees of freedom is reduced and hence less computational time for solution of linear

equations.

The procedure for this extrapolation strategy is explained with reference to Figure 5-9.The

external nodes in the grid are denoted by n and the internal nodes are represented by n'. The

displacement of an external node is associated with the displacement of a set of internal nodes

and the number of inner nodes that are used will depend on the approximation scheme used for

the grid variable. For example, for linear Lagrange elements, two internal nodes are associated

for a given external node and for cubic B-spline elements four internal nodes are associated with

a given external node.

The extrapolation (constraining of external nodes using internal nodes) can be written as

follows

Un = I(u,"...,Un, ) (5-48)

In the above expression, I is a function which maps the degrees of freedom of a set of

internal nodes to the displacement of an external node. This function can be expressed as an

extrapolation matrix and the transformation can be represented as a matrix multiplication:

U dxl = [C]ndxm {U, } (5-49)

In the above equation m is the number of internal nodes used to constrain the

displacements of the outer nodes and the dimension of the problem as nd and the size of [C] is

nd m.

5.5.1 Construction of Extrapolation Matrix

In this section, construction of extrapolation matrix Cis described in detail. The

extrapolation matrix is constructed by imposing the continuity requirements at the boundaries of

the elements.

Q4/H8 interpolation. The extrapolation matrix is constructed using the continuity

conditions at the nodes and is explained for a ID case and the extrapolation matrix for higher

dimensions is constructed from the ID extrapolation.

When a linear interpolation for grid variable is used, the extrapolation matrix is constructed

such that the displacement along the line joining the internal and outer nodes is linear, which

means that a C1 continuity requirement is imposed for the piecewise polynomial at the

intersection of internal and boundary element. In Figure 5-10, the interpolation between nodes

nj and n' is given by Nuj +N2u' and the interpolation between nodes n' and nis given by

Nu' + N2u. Imposing C' continuity at node n' leads to the following equation (Eq. 5-50).

Solving this equation we get u = 2u' -uz and 2D the extrapolation will be as shown in Eq. 5-51

dN1 1 + dN2 u1 dN U, +dN (5-50)
dr dr dr 2 dr

u,
ni

= 1 (5-51)
v o 0 2 0 1 u'

VI

Cubic B-spline approximation. The cubic B-spline approximations have C2 continuity.

Hence C3 continuity is imposed between the piecewise polynomial segments between internal

nodes and external nodes.

In Figure 5-11, the approximation for displacements in the internal element is given by

u = Nui + Nu + Nu + N4u' and the approximation in the boundary element is given by

U = NIu + N2,u + Nu' + N4 .

The shape functions of cubic B-splines and the their third derivative is given by

1 1
N (1 -3r+3r2 -r3);N2= (23-15r-3r2 +3r3)
48 48 (5-52)
1 1
N, -(23+15r- 3r2- 3r);N (1+ 3r + 3r2 +r3)
48 48

d3N1 -1 d3N2 3. d3N3 -3. d3N 1
(5-53)
dr3 8 dr3 8' dr3 8 dr3 8

Imposing C3 continuity at the intersection of internal element and boundary element results in

the following equation and the extrapolation relation is given as:

1 2 3 1 (5-54)
Su' = -u' + 4u t 6u' + 4u'

Using the above relation, the extrapolation matrix in two dimensions is given as:

\Uo ~-1 0 4 0 -6 0 4 01 )T}
{ o = 00 v uI v, u, v, ui, v, (5-55)
Lv, 0 -1 0 4 0 -6 0 41 "2 "' "' "

C1 continuity. Hence C2 continuity is imposed (Figure 5-12) between the piecewise polynomial

segments between internal nodes and external nodes. The extrapolation relation between the

external node and internal node is computed as:

uz = u 3u + 3u (5-56)

The extrapolation in two dimensions is constructed as:

{ = [ 7 3 03 0 .; v ; v ; (5-57)
v 0 1 0 -3 0 3 u 3

5.5.2 Construction of Extrapolation Matrix for a Boundary Element

In the above discussion, the extrapolation matrix for a single external node is constructed.

For a boundary element, there may be more than one such external node. In this section, the

construction of extrapolation matrix for a boundary element is described using the example

shown in Figure 5-13

In Figure 5-13, there are 3 external nodes for the element of interest and one internal node.

An extrapolation matrix needs to be constructed which relates the displacements of the outer

nodes using the inner nodes as shown in the following equation:

u10 = I(u4, ) = 2u, -u,4
u11 = I(u2,u5)= 3u, -2u2 (5-58)
u8 = I(u,2, ) = 2u, u2

Using the above relations, the extrapolation matrix for the boundary element shown in Figure 5-

13 is constructed as:

u8 -1 0 0 2 u2'
S7 0 1 0 0u (5-59)
u10 0 2 -1 0 4
ul -1 0 0 3 us

In the above discussion, construction of extrapolation matrix is presented for a specific

example and this approach can be extended for construction of extrapolation matrix for Lagrange

interpolations as well as B-spline approximations in both two and three dimensions.

5.5.3 Computation of Stiffness Matrix

Once the extrapolation matrix for a boundary element is constructed, it can be used to

compute the stiffness matrix in terms of the inner degrees of freedom (nodes). Assuming that the

external nodes are related to internal nodes with the relation u = Cu' and the left hand side

contains the degrees of freedom for the boundary element, the strain displacement matrix is

given as e = Bu = BCu'. Hence the stiffness matrix is given by K = CTBTDBCdV

= C' BTDBdV C. Hence if K' is the stiffness matrix in terms of the degrees of freedom of

boundary element, the modified stiffness matrix in terms of the degrees of freedom of internal

nodes is given by K' = CTKC By using this approach, the modified stiffness matrix for each
boundary element is computed and later used for assembly. Once the finite element equations are
solved for the internal degrees of freedom, the C matrix can be used to compute the
displacements of the external nodes.

Applied Traction ..n = t
Boundary Element with low
__-. Volume Fraction
/ : .

Boundary Element
Internal Element
-- Essential Boundary
u=u on Boundary

Boundary Element with
-- "Essential Boundary
Conditions
Figure 5-1. Typical non-conforming grid used for analysis

i----^.
:^===|:

Triangulation/ I

^/ N

Solid Region

Solid Region

Figure 5-3. Representation of band on essential boundaries

- Solid Region

Approximation of Boundary
Natural Boundary

Figure 5-4. Approximation of natural boundary for computation of load vector
Figure 5-4. Approximation of natural boundary for computation of load vector

Varying Hinc Region

,.. Inclusion Region
j Matrix Region

Figure 5-5. Division of computational domain into three regions

F2 j

F3 I F11
F31 Fl|
I Igure 5-6. Undeformed and deformed configuration of a set of 4 RVEs
I _F4 I

Figure 5-6. Undeformed and deformed configuration of a set of 4 RVEs

Front Face: ABEF
Back Face: CDHG
Left Face: ACGE
Right Face: BDHF
Top Face: EFHG
Bottom Face: ABDC

D

Figure 5-7. Representative volume element used for computation of effective properties

H' = 0

Boundary of refined/
coarse grid = 0
0 < H "/ < 1
H ref = 1
Thickness s

Figure 5-8. Division of computation domain into various regions

G

E I //
I / I

F

N

Region

Region

Region

Element with low
volume fraction
.Internal nodes
used for
extrapolation

Figure 5-9. Typical analysis grid with elements having low volume fraction

Impose C1 Continuity

Boundary Element

U'1 U'2 U01

Figure 5-10. Extrapolation of external node for linear Lagrange element

Impose C3 Continuity

U 1 u2 U ui3 U U4 U01
Internal Element Boundary Element

Figure 5-11. Extrapolation of external nodes for cubic B-spline element

Internal Element

Impose C2 Continuity

I I I
U'1 U2 U'3 U1
u' u 13 u 0
Internal Element Boundary Element
Figure 5-12. Extrapolation of external nodes for quadratic B-spline element

Element of Interest

Figure 5-13. Extapolation for boundary element

Table 5-1. Set of periodic boundary conditions for 6 different strains
Case Unit Front and Back Face Left and Right Face
Strain
1 C 1 u f ubk+Lu{f bk f =ubk r_ 1. __ / /
I = u I + L2' u 2 u 2 1 u 3 u 3 1 1 2 2" 3 3
2 cl uf = u bk. = bk. f bk r I r I L ur u
22 1 1 '2 2'3 3 U1 = U1 U2 2 '3 = U3
3 33 u f k ; f ; k u = = u U, =

4 2 u = u; = bk; u = Ubk u = u; u2 = uI u; r u = u +L
5 3 f bk= f bk bku L r 1. r I

1 2 = 2 3 U3 U1 U H1;U2 U2 ;U3 = u3
13 1 1 2 2 3 3 -1 1 2 2 3 3
y6 12 1 1bk;2f bk. f bk r l r l r I

Top and Bottom Face

Ht b t b t b
U =u1 ;,2 u 2;u3 u3
t b t b t b
U1 U1 U2 =U2;U3 = U3
u =;ub. t =. u t = +L
I 1 2 2 3 3
t b t b t b
U1 1 ,U2 =U2; 3 3
t b t b u b
U = U1 U2U = Uh2 = U3

I =1 b +L;U2 =, 3 =3

CHAPTER 6
ANALYSIS AND RESULTS

The solution structures developed in this dissertation have been implemented by modifying

a finite element code. The first part of this chapter presents numerical examples for validating the

solution structure corresponding to enforcing essential boundary conditions exactly for both

structural and steady state heat transfer problems. It has been be demonstrated that B-spline

elements provide good convergence capabilities with continuous stresses (heat flux) in the entire

analysis domain. The solutions obtained from B-spline elements are compared with the FE

finite elements for two dimensional problems and linear and quadratic tetrahedral elements for

three dimensional problems.

In the second part of this chapter, the solution structure for treatment of material

discontinuity and micromechanical analysis is validated. Numerical examples are presented to

study the convergence capability of this solution structure. Fiber reinforced composite is used for

determination of effective properties using 2D models and 3D models.

In the final part of this chapter, numerical examples are presented for validation of solution

structure for local refinement of grid. The classical stress concentration example is presented to

study the convergence behavior using Lagrange and B-spline approximations.

6.1 Validation of Solution Structure for Essential Boundary Conditions

6.1.1 Patch Test

This example consists of two parts. The first part demonstrates that a linear solution can be

exactly represented using B-spline elements using the solution structure presented in this

dissertation and the second part demonstrates that a cubic solution in displacement is represented

accurately with only one B-spline element using a thin cantilever beam as an example.

Uniaxial tension with constant stress. The first part of this example demonstrates the

capability of B-spline elements to represent a state of constant stress. A rectangular plate of

dimensions 2.0m by 1.0m is subjected to uniaxial loading of 1.0MPa. Only a quarter of the

plate was modeled because of symmetry as shown in Figure 6-1. The material of the plate is

assumed to be steel with modulus of elasticity equal to 200 GPa and Poisson's ratio of 0.3. The

analytical solution is given in Eq. (6-1)

u=(5.0x10-6)x ; v=(-1.5x10-6)y
(6-1)
ox=106 ; y=0 ;r =0

A non-conforming grid used for this analysis is shown in Figure 6-2. It can be seen that

several nodes lie outside the analysis domain as the B-spline elements are being used for

analysis. The Dirichlet functions for this problem are shown in Figure 6-3. The D-function for x-

component of displacement is zero on the left edge and rapidly reaches to one in the interior of

the domain as can be seen in Figure 6-3a. The D-function corresponding to the y-displacements

is zero on bottom edge and reaches to one inside the interior region as shown in Figure 6.3b.

Since the applied value of Dirichlet boundary condition is zero, the boundary value function is

zero in the entire domain.

In order to study the effect of range parameter 6 on the quality of solution, a parameter

study is performed on 6 and the solution of x-displacements along the line AB is plotted in Fig

6-4 for varying values of 6, it can be observed that for values of S <102, one element gives an

exact solution. This study demonstrates that a state of constant stress can be exactly represented

using Implicit Boundary Method with B-spline approximation. The contour plot for ox is shown

in Figure 6-5. This example serves as a patch test for implicit boundary method.

Deflection of thin cantilever beam. The second part of this example demonstrates the

ability of cubic B-spline element to represent a cubic solution accurately. A thin cantilever beam

as shown in Figure 6-6 is considered for this purpose, and according to Euler-Bernoulli's beam

theory, the deflection curve of for beam is a cubic polynomial and tip deflection is given

PL3
by 8, = where P is the end load, L is the length of beam, E is the modulus of elasticity
3El

and I is the moment of inertia of beam cross section. The tip deflection according to

P I
Timoshenko beam theory is given as 9 =- (4 + 5v) t2L + (Timoshenko and Goodier,
3EI 8

1969). When cubic B-spline elements are used to model a cantilever beam, we expect to get

accurate tip deflection as predicted by Timoshenko beam theory when only one element is used.

However, we may not get exact tip deflection as predicted by Euler-Bernoulli's beam theory as

shear deformations are neglected in this theory, while the plane stress elements account for shear

deformations. As the aspect ratio of the beam is increased, the solution from Timoshenko beam

theory and Euler-Bemoulli beam theory converge to the same solution and we expect that the

solution obtained by one element cubic B-splines converges to the same solution. In order to test

this, a cantilever beam with varying aspect ratio is modeled using one-element cubic B-splines.

The error in tip deflection for Euler-Bemoulli theory and one element model with respect

to the aspect ratio of the beam is plotted in Figure 6-7. It can be observed that the solution

obtained by one element model is accurate even for smaller aspect ratio of the beam and the

quality of solution does not degrade with increasing aspect ratio of the beam. It can also be

observed that the solution from one element model converges to the solution from Euler-

Bemoulli theory and Timoshenko beam theory at large aspect ratios of the beam. This example

also demonstrates that cubic B-spline elements do not undergo shear locking in thin limit, and

when the structures are made of thin beams, one element can be used to get acceptable solutions.

6.1.2 Validation of Extrapolation Technique

In this example, the extrapolation technique presented in this dissertation is validated. A

curved beam as shown in Figure 6-8 is used for analysis. The curved beam is a three-quarter

circular beam which is fixed on one end and loaded on the other end. The material of the beam is

assumed to be made of steel with a modulus of elasticity of 200GPa and a Poisson's ratio of 0.3.

The geometry for this beam is such that there are several elements on the boundary which

may have very small support from the nodes lying inside the geometry of the analysis domain

and hence the resulting stiffness matrix will be almost singular and can lead to ill-conditioning of

the system of linear equations and hence leads to poor convergence behavior. The extrapolation

technique presented in this dissertation is used to extrapolate the solution for the external nodes

in terms of the internal nodes.

Bilinear Lagrange interpolation are used to perform a convergence analysis where a series

of models with varying grid densities are analyzed and the error in strain energy is plotted as

shown in Figure 6-9. It can be observed that the convergence behavior is not smooth due to near-

singular stiffness matrices. Figure 6-10 shows the error in strain energy using the extrapolation

scheme. It can be observed that the convergence plot is very smooth and gives an expected

behavior.

When B-spline elements are used for modeling this example, the quality of solution was

not severely affected even when several elements are present on the boundary with small

support. This may be because B-spline elements have a wide span as compared to Lagrange

elements and hence the contribution to stiffness from a neighboring element leads to non-

singular stiffness matrices and hence does not lead to ill-conditioning. Figure 6-9 shows the plot

of error in strain energy when extrapolation technique was not used for cubic B-spline elements.

Figure 6-10 shows the plot of error in strain energy when extrapolation was used. It can be

observed that there is no appreciable difference in the convergence plots. It is to be noted that the

x-axis of the convergence plot when extrapolation technique is used shows the number of active

nodes.

The contour plot of oy without extrapolation is shown in Figure 6-11 and the contour plot

of normal stress is shown in Figure 6-12 using extrapolation technique. In both these plots, bi-

linear Lagrange elements were used for analysis and it can be observed that the stress solution

shows expected behavior only when extrapolation is used and hence it is necessary to use

extrapolation to avoid singular stiffness matrices and to obtain accurate solutions. Figure 6-13

shows the contour plot for o, using cubic B-spline elements without the use of extrapolation

technique and it can be observed that accurate solution is obtained for B-spline elements even

when extrapolation technique was not used. However, in all the remaining examples of this

section, extrapolation technique is used for Lagrange as well and B-spline elements.

6.1.3 Convergence Analysis of a Cantilever Beam

This example involves the convergence study of quadratic and cubic B-spline elements and

its comparison with four-node (Lagrange bi-linear element) and nine-node quadrilateral

(quadratic Lagrange element) elements used in traditional FEM. A cantilever beam of length

1.0m and 0.2m thickness is fixed at the one end and a uniform shear load of -0. 1MPa is

applied at the other end as shown in Figure 6-14. The exact solution for this problem is shown in

Eq. (6-2) (Timoshenko and Goodier, 1969).

u=-PY (6L-3x)x+(2+v) y2_1 D2 (6-2a)
6E 4

= 3vy2L-x)+ 4+5v)Dx+3L-x)x2 (6-2b)

P P (D2 >
cx= (L-x)y ; = -y ;oY=0 (6-2c)
I 21 4

A typical non-conforming grid used for analysis is shown in Figure 6-15. Even though this

problem can be modeled using a conforming grid, a non-conforming grid is used in order to

demonstrate the convergence behavior of the solution structure along with B-spline basis. The B-

spline elements have nodes outside the elements which form an extra layer of nodes as can be

seen in Figure 6-15. The consistent nodal loads are distributed on the nodes corresponding to the

boundary elements on which traction is specified.

A convergence analysis is performed for this model where a series of models with

increasing grid density are used for analysis and the corresponding error norms are computed and

plotted. The L2 error norm represents the errors in the displacements and the expression which

defines error is shown in Eq. (6-3). Energy error norm represents the error in stresses and strains

and the expression for defining energy error is shown in Eq (6-4). In these expressions, u, s and

a represent displacement, strain and stress respectively, while the superscript e represents

exact solution and h represents approximate solution.

L = K(ue ue ) .(u euh)Q (6-3)

E = j(ae _h(Teh)dQ2 (6-4)

The convergence plot for L2 error norm and energy error norms are computed for various

grid densities and shown in Figure 6-16 and Figure 6-17 respectively.

The plots are shown in a log-log scale, where the y-axis represents the error norm and the

x-axis represents the number of nodes in the model. B-spline elements exhibit good convergence

rates and for a given number of nodes, the errors are much smaller as compared to the bi-linear

In order to study the errors involved in deflection curve and stress distribution for a

specific model, a 10 element model with 5 elements along the length of the beam and 2 elements

through the thickness is used. Corresponding FE models with bilinear and quadratic Lagrange

elements are also used for comparison purposes. Analytical solution as shown in Eq (6-2) is used

a benchmark solution. The comparison of deflection curve is shown in Figure 6-18. It can be

observed that the deflection obtained by quadratic and cubic B-spline elements as well as

quadratic Lagrange elements used in FEM match closely with the analytical solution.

The variation of normal and shear stresses along the line AB are shown in Figure 6-19 and

Figure 6-20. It can be observed that B-spline elements as well as quadratic Lagrange elements

used in traditional FEM give accurate solutions and match closely with the analytical solutions.

However, B-spline elements provide continuous stress in the analysis domain and hence do not

need any smoothing algorithms.

In order to study the performance of B-spline elements in near incompressible limit, the

cantilever beam shown in Figure 6-14 is analyzed under plane strain conditions with a Poisson's

ratio as 0.4999999. Convergence analysis is performed by analyzing a series of models with

increasing grid densities and the tip deflection is plotted in Figure 6-21 for bi-linear Lagrange

elements in FEM and quadratic and cubic B-spline elements. It can be observed that bi-linear

Lagrange elements undergo volumetric locking while B-spline elements converge to the

expected tip deflection. This example demonstrates that the B-spline elements do not undergo

volumetric shear locking.

6.1.4 Square Plate with a Circular Hole

The performance of quadratic and cubic B-spline elements in regions of stress

concentration is studied in this example. An infinite plate with a central circular hole as shown in

Figure 6-22 is considered for this analysis. The plate is subjected to uniform unit traction (o = 1)

along the x axis at infinity. The exact solution for this problem is shown in Eq. (6-5)

(Sokolnikoff, 1956). Symmetry of the problem is exploited to model only a quadrant of the

analysis domain. The radius of the hole is chosen to be 1 (a = 1) and length and width of the

plate is chosen to be 4 (1 = w = 4a). The exact solution is used to compute the tractions on the

traction boundaries (right and top boundaries) and symmetry boundary conditions are applied on

the bottom and left edges while the inner edge is traction free.
Ia2 3 3a4
x =0 1-- -cocos20+cos40 + 4 cos40 (6-5a)
r 2 ) 2r

a 21 3a4
c L =r -- ,sin20+i4O sin440 (6-5b)
r 2 2 2r4

S= cos20- cos40 cos40 (6-5c)
y r 2 2r4

-- /-l v lFv 1 2 cosO+--- cos 3- -cos30
l+v 1 2 a 1 a2 1 a4
u=- a -- rcos+- -cos+--cos30---cos30 (6-5d)
E 1+v +v r 2 r 2 r3

l+v -v -v a2 a2 a4 (6-e)
v=-- cL rsinO-- -sinO+--sin3O-- sin30 (6-5e)
E 1+v l+v r 2 r 2 r3

The typical analysis models used for convergence analysis along with the contour plot of

stresses are shown in Figure 6-23 and Figure 6-24. The error analysis for this problem is

performed by computing the L, errors and energy errors for a series of models with varying grid

densities. For comparison purposes, errors are also computed for a series of models with FE Q4

(bi-linear) and FE Q9 (bi-quadratic) elements. The plot for L, error is shown in Figure 6-25 and

plot for energy error is shown in Figure 6-26. Cubic B-spline elements perform much better than

the quadratic B-spline elements and FE elements. For a given number of nodes, the errors for B-

splines are far smaller when compared to the traditional finite elements.

Figure 6-27 shows comparison with the analytical solution along the line 0 = It can be
2

observed that cubic B-spline element computes the stress concentration factor as 3.06 which is

very close to the theoretical value of 3.0.

6.1.5 Thick Cylinder with Internal Pressure

A hollow thick walled cylinder of internal radius 0.3in and external radius 0.5in is

modeled as shown in Figure 6-28 and for analysis purpose only a quarter of the cylinder is used.

The Dirichlet boundary conditions are applied on the straight edges of the quarter such that radial

displacement is allowed while tangential displacement is restricted. The inner curved edge is

subjected to a pressure of 500 psi. The material is assumed to be made of steel with a modulus

of elasticity as 30 x 106 psi and a Poisson's ratio of 0.3. The analytical solution is obtained from

the governing differential equations for displacements, strains and stresses in a thick walled

cylinder with open ends (Boresi et.al. 2003). The displacements and stresses as a function of

(r ((1v)+a22 (6-6a)
ur = 2 2 2- (6-6a)
E b 2-a 2)

a2 2'-b2
ar (2 b2 (6-6b)

a2 (b2 _+b2
a = p 2 (6-6c)
r2 (b2 a2)

This example demonstrates the capability of the Implicit Boundary Method to impose zero

essential boundary conditions in a non-conforming structured grid. The Dirichlet functions are

constructed such that they vanish on the boundaries where essential boundary conditions are

specified. The displacement solution for the radial displacement is shown in Figure 6-29 and

Figure 6-30.

In order to verify the accuracy of solution along the line AB, comparison of analytical

solution and the computed solution for radial displacements is shown in Figure 6-31. The

comparison of radial and tangential stresses is shown in Figure 6-32 and Figure 6-33. It can be

observed that the solutions obtained by both quadratic and cubic B-spline elements match closely

with analytical solution for displacement, radial and tangential stresses. Furthermore, the stresses

computed by B-spline elements are continuous across the element boundaries. This example

demonstrates that B-spline elements provide smooth and accurate solutions for stresses as well as

displacements with B-spline elements.

6.1.6 Steady State Heat Transfer Analysis

A series of pipes carrying hot fluids are arranged in a block to enable heat transfer as

shown in Figure 6-34. Each set of alternate pipes carry fluid with temperatures 80 C and 50 C.

The surfaces of the block are exposed to air with ambient temperature of 30 C .The heat transfer

takes place due to conduction between the pipes and the block and due to convection from the

ambient air and the surfaces of the block. The conduction coefficient of the block is assumed to

be 0.3 W / m C and the natural convection coefficient is taken to be 15 W / m2 C

This problem is analyzed as a steady state heat transfer along with convection. Symmetry

of the problem is exploited to model only a quarter as shown in Figure 6-34 (b). The Dirichlet

functions for this problem are constructed such that they vanish at the curved edges and are unity

inside the domain. The contour plot for Dirichlet function is shown in Figure 6-35.

This problem demonstrates the imposition of non-zero Dirichlet boundary conditions and

the boundary value function is constructed such that it takes the value of imposed boundary

values at the curved edges and the contour plot of the boundary value function is plotted in

Figure 6-36. The boundary value function takes on the imposed values at the curved edges and

smoothly vanishes to zero in the interior of the domain.

For validating the solution obtained by Implicit Boundary Method using B-spline

elements, the solution is compared with the solution obtained by FE using ABAQUS package.

The contour plot of the temperature distribution in the entire domain is shown in Figure 6-37,

Figure 6-38 and Figure 6-39. It can be observed that the solution obtained using cubic B-spline

elements has similar patterns when compared to FEM solution using bi-linear Lagrange

elements. The temperature distribution along the line AB is plotted in Figure 6-40. The

comparison of plots show that the temperature distribution matches exactly with the solution

obtained by FEM.

6.1.7 Analysis of a 3D Clamp

The clamp shown in Figure 6-41 is modeled using a three-dimensional model with 8-node

hexahedral elements and B-spline elements to demonstrate some advantages of using a structured

grid with undistorted elements. The structure is clamped on one end while the other end is

subjected to traction in the Y direction. Traction of 10000 Pa is applied on the flat face as shown

in Figure 6-41. The Young's modulus is assumed to be 200 GPa while the Poisson's ratio is

taken as 0.3.

This problem is analyzed using both Implicit Boundary Method (IBFEM) and using

traditional finite element software (ABAQUS). Convergence analysis for this problem was

performed by computing strain energy for a series of models with varying mesh densities as

plotted in Figure 6-42. The plot is shown on semi-log scale, where the x-axis represents the

number of nodes and the y-axis represents the strain energy. As the number of nodes increases

both methods converge to the same solution. However, in this example, the Implicit Boundary

Method provides a better estimate of the strain energy using fewer nodes and elements when

cubic B-spline elements are used. For small number of elements, it is harder to accurately

represent the geometry without having elements that are distorted in the finite element mesh,

whereas the structured grid has regular shaped elements regardless of the grid density.

The contour plot for the bending stresses is plotted in Figures 6-43, 6-44 and 6-45 using

FEM, Implicit Boundary Method with linear Lagrange elements and cubic B-spline elements

respectively. The results obtained using Implicit Boundary Method and relatively coarse grid is

able to give an answer that is close to the answer obtained by a higher density FEM mesh.

6.1.8 Analysis of a 3D Rib Structure

A 3D rib structure shown in Figure 6-46 is used to study the performance of B-spline

elements and tri-linear Lagrange hexahedral elements. A constant pressure loading of 1000Pa is

applied on the top surface and one end of the structure is clamped as shown in Figure 6-46. The

material of the structure is assumed to be steel with modulus of elasticity E = 200GPa and

Poisson's ratio v = 0.3. The geometry of the structure is quite complex and meshing with regular

hexahedral elements is difficult, so finite element models with linear tetrahedral meshes are used

as benchmark solutions.

A sequence of models with varying grid densities is modeled and strain energy is

computed for each model and the convergence in strain energy is plotted for FEM tetrahedral

elements, Implicit Boundary Method linear hexahedral elements and cubic B-spline elements is

plotted in Figure 6-47.

It can be observed that the strain energy obtained by the sequence of models converge to

the same strain energy value. In order to validate the computed stress distribution obtained by

linear hexahedral elements and B-spline elements, the contour plots of the normal stress

distributions is plotted in Figures 6-48, 6-49 and 6-50. It can be observed that the stress

distribution matches closely with the one obtained by FEM.

6.1.9 Analysis of a 3D Dial Bracket

A 3D dial bracket structure shown in Figure 6-51 is used to study the performance of B-

spline elements and linear hexahedral elements. A constant pressure loading of 1000Pa is

applied on the top flat surface and the bottom flat surface of the structure is constrained. The

material of the structure is assumed to be steel with modulus of elasticity E = 200GPa and

Poisson's ratio v = 0.3 .

The geometry of the structure is quite complex and meshing with regular hexahedral

elements is difficult, so finite element models with quadratic and linear tetrahedral meshes are

used as benchmark solutions. Convergence analysis is performed in terms of strain energy and

the results are plotted in Figure 6-52. It can be seen that the solutions obtained by B-spline

elements converge to the solution obtained from very dense FE model and the number of nodes

needed to reach a convergent solution for B-spline elements is significantly smaller as compared

to the traditional tetrahedral (linear, quadratic)elements. The contour plots of bending stresses are

shown in Figures 6-53, 6-54 and 6-55. It can be observed that stress distribution obtained by the

Implicit Boundary Method shows similar pattern as with the finite element solutions. This

example demonstrates the ease of constructing regular grid for complicated 3D geometries

without compromising the quality of solution.

6.2 Validation of Solution Structure for Micromechanics

6.2.1 Square Plate with Circular Inclusion

This example demonstrates the validity of the solution structure for treatment of material

discontinuity. A circular inclusion of volume fraction vf = 0.47 as shown in Figure6-56 is used

for analysis in this example. The matrix has material properties E, = 68.3GPa and v = 0.3 and

the fiber has material properties Ef = 379.3 GPa and vf = 0.1 and plane stress idealization is

used. The solution obtained by Implicit Boundary Method (IBFEM) is compared with the

solution obtained by FEM using ABAQUS software.

The contour plot of o-is shown in Figures 6-57 and 6-58 and it can be observed that the

stress pattern as computed by FEM matches closely with the stress pattern as computed by

IBFEM. The maximum and minimum values for the stresses also match very closely.

Convergence analysis is performed by analyzing a sequence of models and the error in strain

energy is plotted in Figure 6-59. The solution from a highly dense FE model with 12001

quadratic quadrilateral elements is used to as an exact solution for convergence analysis. The

strain energy from the dense FE model was computed to be 4.273020946 x 10-6 and the error in

strain energy for a given grid density is computed as the difference in computed strain energy

and this value. It can be observed that the solution converges as the error approaches to zero

when the grid density is increased.

The comparison of ux along the line is shown in Figure 6-60. It can be observed that the

displacement solution obtained by IBFEM matches exactly with the FEM solution. A slope

discontinuity can be observed at the inclusion boundary. Figure 6-61 shows the line plots o- and

o- along the line AB and Figure 6-62 shows the comparison of Ex and c It can be observed

that the solution for both stresses and strains match exactly with the FEM solution. The

discontinuity in stress and strain can be observed at the inclusion interface.

6.2.2 Determination of Effective Properties of a Fiber Reinforced Composite

2D model. In this example a two-dimensional model is used for computation of effective

properties of a Boron-Aluminum fiber reinforced composite with volume fraction of 0.47. For

determining the effective properties in the transverse directions, a plane strain assumption is

used. In order to determine the homogenized material properties in the longitudinal direction, a

generalized plane strain model (Li and Lim, 2005) is used. The homogenized shear properties in

the longitudinal direction are determined by using a longitudinal shear loading model with one

degree of freedom in the longitudinal direction is used. This assumption results in non-zero shear

strains in z-y and z-x directions only while all other strain and stress components are identically

zero.

The results obtained by IBFEM are compared with the results available in literature (Sun

and Vaidya, 1996, Sun and Chen, 1990, Chamis, 1984 and Whitney, 1966). The periodic

boundary conditions shown in Table 5-1 are used for analysis. The average stresses for each case

are tabulated in Table 6-1. Figures 6-63, 6-64 and 6.65 show the stress distribution for various

cases with the deformed shape. The effective properties are computed by using the average

stresses and strains and are compared in Table 6-2. It can be observed that the effective

properties determined by the current method match very closely with the effective elastic

constants as reported in literature.

3D model. In this example a three-dimensional model is used for computation of effective

properties of the fiber reinforced composite. The results obtained by three-dimensional analysis

are compared with the results available in literature. The periodic boundary conditions shown in

Table 5-1 are used for analysis. The average values of stresses and strains are tabulated in Table

6-3, and the effective elastic constants that are determined by the 3D model are tabulated in

Table. 6-2. It can be observed that the effective properties determined by 3D model matches very

closely with the ones determined by 2D model and with the ones available in literature. Figures

6-66, 6-67 and 6-68 show the stress distribution for various cases.

6.3 Validation of Solution Structure for Grid Refinement

The local grid refinement technique presented in chapter 3 is validated in this section. The

first example demonstrates the validity of solution structure for grid refinement by solving the

problem with a linear displacements and constant stress and strain as exact solution. In order to

demonstrate the performance of the refinement technique in the regions of stress concentrations,

the classic problem of square plate with a circular hole is solved and the convergence of the

solution structure is demonstrated in the second example.

6.3.1 Application of Grid Refinement to Constant Stress Problem

In this example, the validity of solution structure for grid refinement is demonstrated. A

rectangular plate with a uniform traction on one end as shown in Figure 6-1 is analyzed in this

example. Only a quarter of the plate is modeled because of symmetry, and the exact solution for

this problem is shown in Eq. (6-1).

The solution structure presented in chapter 3 for local grid refinement is used in a recursive

fashion to get multiple levels of refinement. Figure 6-69 shows the contour plot of X-

displacements using bi-linear Lagrange elements and Figure 6-70 shows the contour plot of Y-

displacements. It can be observed that the refinement is done for three levels and the exact

solution is obtained without degrading the quality of the solution.

It can be observed that at the first level and second levels of refinement, the central

element is subdivided into 3x3 elements and at the third level of refinement, the central element

is subdivided into 2x2 elements. It is to be noted that in hierarchical refinement, each element

can be subdivided only into 2x2 elements, while the method presented in this dissertation does

not have this limitation. The contour plots for X and Y displacements are shown in Figures 6-71

and 6-72 and it can be observed that for three levels of refinement, the solution is exactly

represented.

6.3.2 Stress Concentration Problem

This example involves the study of the performance of solution structure for local grid

refinement in the regions of stress concentration. An infinite plate with a central circular hole as

shown in Figure 6-73 is considered for this analysis. The plate is subjected to uniform unit

traction (o-, = 1) along the x-axis at infinity. Symmetry of the problem is exploited to model only

a quadrant of the analysis domain. The radius of the hole is chosen to be 1 (a = 1) and length and

width of the plate is chosen to be 5 (1 = w = 5a). The exact solution as shown in Eq. (6-5) is used

to compute the tractions on the traction boundaries (right and top boundaries) and symmetry

boundary conditions are applied on the bottom and left edges while the inner edge is traction

free.

Bi-linear Lagrange elements and cubic B-spline elements are used in this study and

convergence analysis is performed to demonstrate the convergence capabilities of the solution

structure. Figure 6-74 shows the contour plot of the X-stress with 4 levels of refinement using a

bi-linear Lagrange elements and it can be observed that the stress concentration factor is

computed as 2.916 which is close to the exact value of 3.0 but not exact. Figure 6-75 shows the

contour plot of o- using cubic B-spline elements for 2 levels of refinement. It can be observed

that very accurate solution is obtained by cubic B-spline approximations with relatively less

number of elements.

Convergence analysis is performed for this problem and Figure 6-76 shows the plot of

errors in L2 norm and Figure 6-77 shows the plot of energy error norm. It can be observed that

in both the plots that the solution obtained with the local grid refinement solution structure

converges and the error approaches to zero when the levels of refinement are increased. The B-

spline elements converge faster as compared to Lagrange elements and hence very accurate

solutions are obtained by B-spline approximations with relatively less number of elements.

The proposed solution structure is capable of using a mix of Lagrange and B-spline

elements in a given model. In order to test this, the stress concentration problem is modeled

using bi-linear Lagrange elements for the coarsest mesh and for the first level of refinement and

the second level of refinement is done using cubic B-spline elements. The contour plot for o0 is

shown in Figure 6-78. It can be observed that very accurate solution is obtained by using a mix

of Lagrange and B-spline elements and the stress concentration factor is computed to be 3.009

which is very close to the analytical solution of 3.0.

-A- ---15 ---- -B

Tx=1MPa

Nodes

Solid Boundary

-Element Boundary

Figure 6-2. Non-conforming grid using one cubic B-spline element

121

1.0 m

Z

"11

Df"t/udoa

9E 1

6 i

E I
ilE-2

21E.55 E-12

(a)

,E-I

(b)

Figure 6-3. Plot of Dirichlet functions (a) D, (b) D,

-6
x 106
Theoretical displacement
--- 8=1.0-1
5 -e- 8 =5.0-2
a- 8 =1.0-2
-*-- 8 =1.0-3
S4 I 8 =1.0-5
E

O.
2 -I
In

1

0 0.2 0.4 0.6 0.8
Distance from left edge of plate, m

Figure 6-4. Effect of range parameter on the quality of solution

Figure 6-5. Contour plot of o- for one element model using cubic B-splines

Stress
X
1E6

9.9E5

9.8E5

9.7E5

9.6E5

9.5E5

9.4E5

9.3E5

9.2E5

9.1E5
9E5

Aspect Ratio = L/t

10--
--- % error in CBS deflection
-e- % error in Euler-beam deflection

15
Aspect ratio of beam

Figure 6-7. Plot of error in the tip deflection with varying aspect ratio

124

--------- -S-------------------

IBFEM Q4
IBFEMCBS

-10
10

Log (Num Nodes)

Figure 6-9. Plot of error in strain energy for curved beam (without extrapolation)

IBFEM Q4
IBFEM CBS

25

20
>,

(U
" 15

10
w

Log (Num Nodes)

Figure 6-10. Plot of error in strain energy for curved beam using extrapolation technique

Stress
Y

I Ifjj4 I I I I A I I I I I I I I 14I 1
1 Pl=-

2.2962E6

1.8333E6

1.3704E6

9.0745E5

4.4455E5

1.8361E4

4.8127E5

9.4417E5

1.4071E6

-1.87E6

-2.3329E6

Figure 6-11. Contour plot of o-y for Q4 elements (without extrapolation)

126

TY-FkM I
-,----`7

Stress
y
2.9166E6
2.1803E6

1.4439E6

7.0747E5

2.8916E4

7.6531E5

1.5017E6

2.2381E6

-2.9745E6

-3.7109E6

-4.4473E6

Figure 6-12. Contour plot of o-y for Q4 elements using extrapolation technique

Stress
Y
2.9443E6

2.2016E6

1.4589E6

7.1621E5

-2.6487E4

,-7.6918E5

1.5119E6

-2.2546E6
S-2.9973E6

SA I -3.74E6

S-4.4827E6

Figure 6-13. Contour plot of o-, using cubic B-spline elements

Y
B

D X P

A

L
Figure 6-14. Thick cantilever beam used for convergence analysis

Grid Boundary

Solid Boundary

Figure 6-15. Typical non-conforming grid using cubic B-spline elements

10-5

10
101 10 10
Log (Num Nodes)

Figure 6-16. Convergence plot of L2 error for cantilever beam

CBS
QBS
FEQ4
FEQ9

CBS
QBS
11111 111-1---- FEQ4
FEQ9

2-
0 ---- I I IL L __ L_ -1 -IIIL L ---- I III IL ft--I -- I IL L

S1 I I II

m I- - --- -I L- ----------L ------L- ---------

101 102 103 10 10
Log (Num Nodes)

Figure 6-17. Convergence plot for error in strain energy for cantilever beam

0.4
Distance along the

0.6
beam length

Figure 6-18. Comparison of deflection curves for 10 element model

Sx 10

Cu
-2

c -3
0
Cu

Analytical
CBS
QBS
FEQ4
FEQ9

w
"w"
x

c
X

C

\-Yt

x 106

SAnalytical
1.5 CBS
QBS
FEQ4
1 -- FEQ9 ---------------------

S0.5--
a,/

0-
-FU 0 ------------ ; -------------
E
O
z -0.5 -----

-1------------ -------------------------

-0.1 -0.05 0 0.05 0.1
Distance along the beam thickness

Figure 6-19. Comparison of normal stress along the line AB for 10 element model

x 10,

1.5 -
SAnalytical
1 CBS
SQBS
FEQ4
0.5---------------- FEQ9

Figure 6-20.

- 1 -0.05 0 0.05 0.1
Distance along the beam thickness

Comparison of shear stress along the line AB for 10 element model

x10'

- CBS
QBS
SFEQ4
Exact

0

S- -

-2.5------ ----^--I

-4
-3.5
'^\

-3.5 -- -- -^ ^ -- -- : -- -- ---

10' 102 103 104
Log (Num Nodes)

Figure 6-21. Convergence plot for tip deflection in near incompressible limit

*4L2

a

4a

t

Figure 6-22. Analysis model for a square plate with a circular hole

IlMeo

Figure 6-23. Contour plot of o-using FEM Q4 elements
^ ^, ^ ^ ^^ ^ ^^ ^ ^ , .111U
I .mB3E

Figure 6-24. Contour plot of o using cubic B-spline elements

105
10

103
Log (Num Nodes)

Figure 6-25. Convergence plot for L2 error norm for stress concentration problem

2 766IEO

2,436E0
2.10BSEO

1.77MED

1.45B2ED

1.1211EO

70201E.1

4A29E-1

1 15WIVE-1

-10532E-1

e CBS
QBS
FEQ4
FEQ9

SCBS
QBS
-- FEQ4
FEQ9

10-4
10

103
Log (Num Nodes)

Figure 6-26. Convergence plot for error in strain energy for stress concentration problem

Plot of X-Stress along 0 = i/2

Analytical
CBS
QBS

1.5 2 2.5 3
Distance from center of circular hole, m

3.5 4

Figure 6-27. Plot of along the line 0

N
ricr
-----i~t-
r-
w
"N

(a) (b)

Figure 6-28. Hollow thick cylinder subjected to uniform pressure

1.0742E-5

Figure 6-29. Contour plot of radial displacements using cubic B-spline element model

Figure 6-29. Contour plot of radial displacements using cubic B-spline element model

1I 126E-5

FI 12E-5

I 02E

r 935SQ-6

Figure 6-30. Contour plot of radial displacements using FEM Q4 elements

04
Distance from center of cylinder, in

Figure 6-31. Comparison of radial displacements along the thickness of the cylinder

138

- Analytical
FE Q4
-- FE Q9
CBS
QBS

1 15
E

1
1 05

0 95

Analytical
FE Q4
FE Q9
CBS
SQBS

-100

Figure 6-32.

03 0 32 0 34 0 36 0 38 04 0 42 0 44 0 46 0 48 05
Distance from center of cylinder, in

Comparison of radial stresses along the thickness of the cylinder

Analytical
FE Q4
FE Q9
CBS
SQBS

S900 ----

600
800

I- 700

600

03 04 05
Distance from center of cylinder, in

Figure 6-33. Comparison of tangential stresses along the thickness of the cylinder

Ta=30, h=15

^=80 J=50 T=80

0000 0 80

Ta=30, h=15
(a)

1m
Ta=30, h=15

T=80 T=50 .5m
0.25 m

A B
(b)

Figure 6-34. Heat conduction in a block with series of pipes

Dfunction
T
1EO
9E-1
8E-1
7E-1
6E-1

2E-1
3F- 1
1E-1
OEO

Figure 6-35. Plot of Dirichlet function

Xa
T
8E1
7.2E1
6.4E1
5.6E1
4.8E1
4El
3.2E 1
F.4E1
1.6E1
8EO)
OEO

Figure 6-36. Plot of boundary value function

NT11
+8.000e+01
+7.594e+01
+7.188e+01
+6.782e+01
+6.376e+01
+5.970e+01
+5.563e+01
+5.157e+01
+4.751e+01
+4.345e+01
+3.939e+01
+3.533e+01
+3.127e+01

Figure 6-37. Contour plot of temperature distribution using FE Q4 elements

Temperature
T
8.0039E1
7.5163E1
7.0287E1
6.5411E1
6.0535E1
5.566E1
5.0784E1
4.5908E1
4.1032E1
S.6156E1
3.1281E1

Figure 6-38. Contour plot of temperature distribution using cubic B-spline elements

-- --::: i--: -J I1III-f1

1UN

U E1t:. .......

/"

Temperature
T
8.0078E1
7.5199E1
7.0319E1
6.544E1
6.0561El
5.568?FF1
5.'f0OE I
4.5923EEI

4.1044E1
I3.6165E1
3.1286E1

Figure 6-39. Contour plot of temperature distribution using quadratic B-spline elements

IgLIIIIIIIIIIIIIIIIl/l

Figure 6-40.

FE Q4
CBS
QBS

0 0.1 0.2 0.3
Distance from point A along AB, m

Comparison of temperature along line AB

Figure 6-41. Three dimensional clamp used for analysis

4.25
LU
C
2 4.2
Cl)

Log (Num Nodes)

Figure 6-42. Convergence plot for strain energy for clamp model

S, S11
(Avg: 75%)
+1.249e+05
+1.041e+05
+8. 36e-+04
S+6.2553e+04
+-4.1 9e+04
+2.100oe+'04
+2. 189e+02
-2.057e+04
-4.135e+04
-6.214e+C'4
-8.292e+C'4
-1.037e+C05
-1.245e+C'5

Figure 6-43. Contour plot of bending stresses using H8 elements in FEM for 3D clamp

Stress
xx
1.248E5

9.966E4
2",E
.I-_F I

4.937E4

2.6".1- 4

-7.635E4

1-1.015E5

-1.266E5

Figure 6-44. Contour plot of bending stresses using H8 elements for 3D clamp

Stress
xx
1.233E5

9.846E4

k6 --

4.888E4

.L ..1 LF

S-7.506E4

5 9.985E4

-1.246E5

Figure 6-45. Contour plot of bending stresses using cubic B-spline elements for 3D clamp

Figure 6-46. Three dimensional rib model used for analysis

0.245

0.24

0.235

2)
a)
W 0.23

0.225

Abaqus Linear Tetra
IBFEM H8
IBFEMCBS

0.21502
102 103
Log (Num Nodes)

Figure 6-47. Convergence plot for strain energy for rib model

S, .11
(Avg: 75%)

1- S p1 Tim 1 -
B ti "e i.-

Figure 6-48. Contour plot of bending stresses using tetrahedral elements for 3D ribr

1.1.7073E5
*q -, L 5,

St, 5.3909E 4

ncrer I Sp Time 1-1.7974E5
Figure 6-48. Contour plot of bending stresses using tetrahedral elements for 3D rib

Stress
xx

2.0967E5

9.285E4

A 5.3909E4

-2.3973E4

I-.'6.2914E4

-1.7974E5

Figure 6-49. Contour plot of bending stresses using B-spline elements for 3D rib

147

Stress
xx

2.0755E5

1.6988E5

9.4527E4

5.6852E4

1.9178E4

-1.8497E4

-5.6171E4

-9.3846E4

1-1.3152E5

1.6919E5

Figure 6-50. Contour plot of bending stresses using H8 elements for 3D rib

Figure 6-51. Three dimensional dial bracket used for analysis

Abaqus 10-Node Tetrahedron
Abaqus 4-Node Tetrahedron
IBFEM H8
IBFEM CBS

104
Log (Num Nodes)

Figure 6-52. Convergence plot for strain energy for dial bracket model

3.1274E4

2.4692E4

1.81 IE4

1.1528E4

4.9-16E3

S1.635 9E3

8.21 "SE,

1.4SE4

-2.1382E4

I-2.7963E4

-3.454SE4

Figure 6-53. Contour plot of bending stresses using a cubic B-spline element model for dial

bracket

13x10-4
10

11

2) 10
CB

9

8
8

Iz I

Figure 6-54. Contour plot of bending stresses using a tetrahedral element model for dial bracket

Stress
xx

2.9619E4

2.3235E4

I.(,- I1E4

1.0466E4

4.110%'41 4

2..3uI6EF

2.1454E4

1-2.7838E4

-3.4222E4

Figure 6-55. Contour plot of bending stresses using a H8 element model for dial bracket

S, Sl1
(Ave. Crit.: 75%)
+3. 101e+04
+2.435e+04
+1. 769e+04
+1.102e+04
+4 i -1-.13
-2. ;: : + "03 i
-8.970e+03
-1.563e+04

4 .
-4, +; .4: +'."

Traction = 1000 Pa

Inclusion Matrix

Figure 6-56. Square plate with a circular inclusion

StreX
X

1.3844E3

1.2734EI

1.162.3E1

1.0513E3

9.4026E2

b.2922E2

".181SE2

6.0715E2

4.9611E2

.h74051 E
2.7404E2

Figure 6-57. Contour plot of using Q4 elements in IBFEM

I I I I I L

F7m
1 1 1 1 1 1- 1 1 1 1 1 1 1 1 1
^,,^ ,, ^ ^ ,,

S, s51
(Avg: 75%)
+1.383e+03
+1.291e+03
+1 198e+03
+1 106e4-+03
+1.013e+03
+9.208e +02
+8 2E 3e +02
+7 358e+02
+6.434e+02
+5 509e+02
+4 54e+02
-+3 659e+02
+2.735e+02

Figure 6-58. Contour plot of o-, using Q4 elements in FEM

10-8

_I -I 1 1 --_ _-- _

W 5 - -- -- ---

I I I I I I -I I r -

W

10-1
102

103
Log (Num Nodes)

Figure 6-59. Convergence plot for error in strain energy for square matrix with inclusion model

Abaqus ux
IBFEM u

m
-- - ---- - - --- -- -- -/ -

E

c I/

S0.5-------------------
X

0 0.5 1 1.5
Distance along line AB

Figure 6-60. Comparison of ux along the line AB

1400

soo0-------------------------------

S100 -
6o 00 -- -- -- --- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- ---

Abaqus u.
1200 ---

Distance along line AB

Figure 6-61. Comparison of stresses along the line AB
E-- Abaqlus
0 400I----------------- I M -
I IBFEMI
1 200 ----

- -

-200
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Distance along line AB

Figure 6-61. Comparison of stresses along the line AB

2.5x 108
Abaqus E
Abaqus s
2 IBFEM
IBFEMs

m

CO

0)
I 0.5
o .5

(
C,-

0.2 0.4 0.6 0.8
Distance along line AB

1 1.2 1.4

Figure 6-62. Comparison of strains along the line AB

Str

3.8702EXII
3.2872Ell

.Z154GI1
.953El1'
2. E 1
.I6.'E I I
1.5114r11

A

i_2-,it

3ISIEII
_11 4 lll
7.21~111

ga

B

Figure 6-63. Plot of stresses for cases 1 and 2 (A) Plot of a,, normal stress Case 1 and (B) Plot of

,22 normal stress for Case 2

7A781E0LI

6.0 18t10

2.-199E0L
S1.4402EO1
.1044EE

n S%h.-

5.1312E10Q
."S 1110
I|.B5i6EllI

3.326E10
I :49E::

B

Figure 6-64. Plot of stresses for cases 3 and 4 (A) Plot of ,33 normal stress for Case 1 and (B)

Plot of ,23 shear stress for Case 4

Sm..
' hIT E
|WIEI
7,47SiE30
d 411F 1.1
I "
I.O IFIIrl

13599EIO
2.M5PE1B
I L640ZEIg
S.1044E9

87E 10
7.4781E10

6.4flss3l
2.018'f0l

2 I6lE I0
S140. 10E1O
9.1044E9

Figure 6-65. Plot of stresses for cases 5 and 6 (A) Plot of z31 shear stress for Case 5 and (B) Plot
of ,12 shear stress for Case 6

Uno ONNOMMISM 110500: :00 H :
sINNUIP I.M.: a.: ::.a::
REMO, I -I a, Moan MM
II.MOVIN "I' I U.M a. .00.
AM.- I- a NE,
manHIR, i:
Masson -7-lummommmm
Mmfflm:am'V Congo ::::
MMMMIII''
'Immonar, IIIIINNAMEW
now,
: IS 00 ma In, no
HHU, 12

:::::am MEN Now.
.... USMIL MWOUNIN:

is :00 off so no s no Is no 16:111
MUNK-Momma-HOUNN
M11"OFIff HHIff HUH ::::
Momommommor AU
Mir IIIHMMMHUPO- 1 :28
T man
Hir IMOV
11 L El
.0 MEE
2:0:111-!
Mae L ARMS!
Room .RNE
011006 UNIMEMEMEMILL AMMER
N ::: 601110:21111111"IN:
"MORE:
mosomm
Immmommom

HOWNUMMISIC ComomHOM:
asommossm! I'MUMNON
MUNFI, -10
:901151. 11 IMSEE
MMNNIAUI ANIU.N.86.
'Emma- -;Wmmm'
aa a 'a"HIPIL
::EMmM"6=e'IMRMMMM
HHHO"002 NAMMEME
"I'll `7 'I"'
13 p on: go
11 11 2. an I IF
EMENS6 IUH:
:::::::a.- -.11MMIN
HHUMN: 20HUNN:
MOUNMEMEMN; 4mmommommom

14l.Eill

L il3Ell
2L.2iTII

1I21-SE11l
.L5!1I1

L084ELL
1.7253Eli

I :
40JIH ]:l
.(3TELO
16I !5~SB

Figure 6-66. Plot of stresses for cases 1 and 2 for 3D model (A) Plot of 1,, normal stress for

Case 1 and (B) Plot of 72, normal stress for Case 2

Sim.

Hi 7IEll

l.086EII11
1.72T3EII
I'rIaI'

I i--E II
5i l l .l l

S.255jE10

|.412UE10

S.,
6 FLI I.'

4 6-11E 10
4*Ij!E10
4 n1(EIP

3.5163E10
9i 4EID

Figure 6-67. Plot of stresses for cases 3 and 4 for 3D model (A) Plot of 33 normal stress for Case

1 (B) Plot of a23 shear stress for Case 4

. 1 IEL

1 1s.i I[,
2.4234EI
.6949El6
I~^~

6.0657E10

I11 l"[1
14234E10
I9'4FIO

B

Figure 6-68. Plot of stresses for cases 5 and 6 for 3D model (A) Plot of -31 shear stress for Case

5 and (B) Plot of z12 shear stress for Case 6

156

Displacement
x-component

5E-6

4.5E-6

4 E 6

3.5E-6

2.5E-6

NF 6.

S1.5E-6

1E-6

5E-7

2.562E-12

Figure 6-69. Contour plot of u using Q4 elements and grid refinement

Displacement
y-component

4.689E 13

-7.501E-8

S.5E-7

-2.25E-7

-3E-7

-3.75E-7

-4.5E-7

5.25E-7

-6E-7

1-6.751E 7

-7.501E-7

Figure 6-70. Contour plot of v using Q4 elements and grid refinement

Displacement
x-component

4.5E-6

3.5E-6

WF,-6

2.SE-6

5E-7
IL b

1.562E 12

Figure 6-71. Contour plot of u using cubic B-spline elements and grid refinement

Displacement
y-component

4.687E-13

-7.5E-8

.5E-7

-2.25E-7

___-3E-7

3.75E-7

4.5E-7

5.25E-7

-6E-7

-6.75E-7

7.5E-7

Figure 6-72. Contour plot of v using cubic B-spline elements and grid refinement

5a

a

_D__<

AI~

Figure 6-73. Square plate with a circular hole

Stress
X
?.916E..
2?.15E0
2.3I5EO
7-014E0
1. 1. F 1
1.41 ZEO
1.111
,.Iln6F 1
5.099E-1

I7091E-1
-9.168E-2

Figure 6-74. Contour plot of a, for local refinement using Q4 elements

0t

X

.016E0

"\$.9 LI

2.086E0

I.'"hFb

l.46H9Ei

1.]10LIPI

.499 -1

5.405E-1

.2-2__________

Figure 6-75. Contour plot of o, for local refinement using cubic B-spline elements

10-1
CBS
SI 1 1 -- --- -I- Q4
L _ _, ,_ _, _,_ _. ,_ _, ,_ j ^

Log (Num Nodes)

Figure 6-76. Convergence plot for L2 error in local grid refinement

-I 1-14 -
\iLt

0u
t 10
a)
C
LU

10-L
100

Log (Num Nodes)

Figure 6-77. Convergence plot for energy error in local grid refinement

I

Stress
X
3.009E0

2.699E0

2.39E0

2.08E0

1.77E0

1.461F11

1.151FI1

8.416F-1

5.319E-1

12.223E-1

-8.729E-2

Figure 6-78. Contour plot of for local refinement using Q4 and cubic B-spline elements
Figure 6-78. Contour plot of ca for local refinement using Q4 and cubic B-spline elements

Case Unit oa,(GPa) o22 (GPa) r33 (GPa) 23 (GPa) r31 (GPa) 12 (GPa)
Strain

1 e-, =1 230.84 41.14 41.14 0 0 0

2 22 =1 40.35 161.20 46.22 0 0 0

3 E33 =1 40.35 46.22 161.20 0 0 0

4 23 = 1 0 0 0 46.07 0 0

5 731 = 10 0 0 0 54.44 0

6 12 = 1 0 0 0 0 0 54.44

Comparison of effective properties
IBFEM IBFEM Ref:Sun
2D 3D and Vaidya,
1996

Table 6-2.
Property

E,(GPa)
E (GPa)
E3(GPa)
G23(GPa)
G13(GPa)
G12(GPa)
v23
v13
v12

214.85
143.85
143.85
45.97
54.35
54.35
0.25
0.19
0.19

Ref: Sun and
Chen, 1990

215
144
144
45.9
57.2
57.2
0.29

0.19

Ref: Chamis,
1984

214
135
135

51.1
51.1

0.19

Ref:Whitney,
1966

214
156
156
43.6
62.6
62.6
0.31

0.20

53.9
53.9

214.83
144.07
144.07
46.07
54.44
54.44
0.25
0.20
0.20

0.19

Case Unit a-1 (GPa) 22 (GPa) -33 (GPa)
Strain
1 s,, =1 230.57 40.34 40.3L

2 22 = 1
3 33= 1
4 23 =1
5 13 =1
6 12 =1

40.35
40.35
0
0

160.93
46.25
0
0

T23 (GPa) r13 (GPa) 1,2 (GPa)

1

46.25
160.93
0
0

0
0
0
45.98
0

0
0
0
0
54.37
0

0
0
0
0
0
54.37

CHAPTER 7
CONCLUSIONS AND FUTURE WORK

7.1 Conclusions

In this dissertation, finite elements based on uniform B-spline basis defined over a

structured grid are constructed to obtain solutions with continuous gradients throughout the

analysis domain. Finite elements are developed using quadratic and cubic B-spline basis for

problems in both two and three dimensions. B-spline basis do not satisfy Kronecker-delta

property and hence Implicit Boundary Method is extended for using B-spline basis and for

applying essential boundary conditions. Solution structures are developed using the approximate

Heaviside step functions based on the implicit equations of the essential boundaries to impose

essential boundary conditions. The weak form for linear elasticity and steady state heat transfer

was modified using this solution structure such that essential boundary conditions are

automatically imposed. Methods were developed for efficiently and accurately evaluating the

terms involved in the modified weak form in the limit as the approximate Heaviside step

functions approach the exact Heaviside step functions.

The primary motivation for the method is the desire to obtain continuous gradients in the

entire analysis domain using B-spline basis on a structured grid. The B-spline elements studied

in this work exhibited good convergence behavior and the comparison with traditional FEM

solutions showed that very accurate solutions can be obtained with relatively fewer nodes. The

performance of B-spline elements does not degrade with increasing aspect ratio of the elements

and one element can be used to compute accurate solutions for beam like structures. Although B-

spline basis is used in this work to achieve continuous stress and strains, the method is quite

general and can be applied to other basis functions which may or may not satisfy Kronecker-

delta property such as Lagrange interpolations and Moving Least Squares approximation used in

many meshless methods.

An added advantage of the current method is the elimination of the mesh generation

process during the preprocessing stage of the design cycle. Generating a uniform structured grid

that encloses the geometry is easier than to generate a conforming mesh, furthermore all the

elements used in the analysis are regular and undistorted unlike traditional FEM and hence

problems associated with distorted elements are eliminated. As shown through several examples

in the dissertation, models with fewer elements can be used for getting accurate solutions. In

addition, the method would enable design engineers to perform analysis directly using solid

models instead of first generating a mesh and then applying boundary conditions on the mesh,

hence resulting in significant savings in pre-processing time. Another advantage of this method

is that all internal elements have the same stiffness matrix resulting in significant reduction in

computation required to determine the global stiffness matrix.

The Implicit Boundary Method is extended for analysis of micromechanical analysis by

developing a solution structure for treatment of material discontinuity in a structured grid

framework. A technique for imposing multipoint constraints based on the technique for

imposition of multipoint constraints in traditional FEM is presented. This approach is used for

modeling an inclusion in a matrix and convergence analysis is performed to validate the solution

structure. Lagrange interpolations are used for performing micromechanical analysis to

determine effective properties of fiber reinforced composites and it has been shown that the

solution structure presented in this work has good convergence properties and is capable of

determining the effective elastic constants with very good accuracy in both two and three

dimensions. This method holds the potential for determining effective properties of complicated

microstructures using structured grids for each inclusion and a structured grid for matrix.

The Implicit Boundary Method is extended for local grid refinement by constructing a

solution structure based on approximate Heaviside step functions of the coarse/refined grid

boundary. This solution structure is validated by analyzing the classical stress concentration

problem. Convergence analysis shows that the solution structure has good convergence

combination of Lagrange and B-spline approximations) is difficult to use, but with the proposed

solution structure, it has been demonstrated that multiple types of approximations can be used.

Using higher order elements (B-spline elements) in the regions of stress concentration and lower

order elements (Lagrange elements) far from the regions of stress concentrations makes the

analysis computationally efficient and provides accurate solutions as demonstrated in the

numerical examples.

7.2 Future Work

The B-spline elements developed in this dissertation are used for analysis of linear elastic

and steady state heat transfer problems only. B-spline based elements can be extended for

analysis of Kirchhoff plates, where the weak form requires C1 continuous approximation space.

The Implicit Boundary Method with B-spline approximations can be extending for other linear

analysis types like electro-static and magneto-static problems, non-linear analysis like large

elastic deformations and elasto-plastic deformations, fracture mechanics and contact mechanics.

The solution structure for material discontinuity can further be used to compute effective

properties of complicated microstructures and for performing damage analysis of composites.

The method can further be extended to incorporate B-spline approximations.

The solution structure for adaptive refinement can be extended for performing analysis of

3D problems. The h-, p- and k- adaptivity can be incorporated into this method. B-spline basis

provides a convenient methodology for raising the order and continuity of the approximation

space, this property can be used to enable k-adaptivity and can also be used for k-version FEM.

LIST OF REFERENCES

Adams, D.F., Crane, D.A., 1984. Finite element micromechanical analysis of a unidirectional

Amestoy P.R., Davis T.A., Duff I.S., 1996. An approximate minimum degree ordering
algorithm. SIAM Journal on Matrix Analysis and Applications 17(4) 886-905.

Atluri S.N., Kim H.G., Cho J.Y., 1999. Critical assessment of the truly Meshless Local Petrov-
Galerkin (MLPG), and Local Boundary Integral Equation (LBIE) methods. Computational
Mechanics 24(5) 348-372.

Atluri S.N., Zhu T.L., 2000a. Meshless local Petrov-Galerkin (MLPG) approach for solving
problems in elasto-statics. Computational Mechanics 25(2-3) 169-179.

Atluri S.N., Zhu T., 2000b. New concepts in meshless methods. International Journal for
Numerical Methods in Engineering 47(1) 537-556.

Babuska I., Banerjee U., Osborn, J.E., 2002. On principles for the selection of shape functions
for the Generalized Finite Element Method, Computer Methods in Applied Mechanics and
Engineering 191(49-50) 5595-5629

Belytschko T., Krongauz Y., Fleming M., Organ D., Snm L., Wing K., 1996. Smoothing and
accelerated computations in the element free Galerkin method. Journal of Computational
and Applied Mathematics. 74(5) 111-126.

Belytschko T., Krongauz Y., Organ D., Fleming M.,Krysl P., 1996. Meshless methods: an
overview and recent developments. Computer Methods in Applied Mechanics and
Engineering 139(1-4) 3-47.

Belytschko T., Lu Y.Y., Gu L., 1994. Element-free Galerkin methods. International Journal for
Numerical Methods in Engineering 37(2) 229-256.

Belytschko T., Parimi C., Moes N., Usui S., Sukumar N., 2003. Structured extended finite
element methods for solids defined by implicit surfaces. International Journal for
Numerical Methods in Engineering 56(4) 609-635.

Boresi A.P., Schmidt R.J., Sidebottom O.M., 1993. Advanced mechanics of materials. John
Wiley and Sons Inc. New York.

Cai, Y.C., Zhu, H.H., 2004. Direct imposition of essential boundary conditions and treatment of
material discontinuities in the EFG method. Computational Mechanics 34(4) 330-338.

Chamis, C.C., 1984. Simplified composite micromechanics equations for hygral, thermal and
mechanical properties. SAMPE Quarterly 14-23.

Clark B.W., Anderson D.C., 2003a. The penalty boundary method. Finite Elements in Analysis
and Design 39(5-6) 387-401.

Clark B.W., Anderson D.C., 2003b. The penalty boundary method for combining meshes and
solid models in finite element analysis. Engineering Computations 20(4) 344-365.

Cook, R.D., Malkus D.S., Plesha M.E., Witt, R.J., 2003. Concepts and Applications of Finite
Element Analysis. John Wiley and Sons Inc., New York.

Cordes, L.W., Moran, B., 1996. Treatment of material discontinuity in the element-free Galerkin
method. Computer Methods in Applied Mechanics and Engineering 139(1-4) 75-89.

Cottrell J.A., Reali A., Bazilevs Y., Hughes T.J.R., 2006. Isogeometric analysis of structural
vibrations. Computer Methods in Applied Mechanics and Engineering 195 (41-43) 5257-
5296.

Cottrell J.A., Hughes T.J.R., Reali A., 2007. Studies of refinement and continuity in
isogeometric structural analysis. Computer Methods in Applied Mechanics and
Engineering 196(41-44) 4160-4183.

Dang, T.D., Sankar, B.V., 2007. Meshless local Petrov-Galerkin formulation for problems in
composite micromechanics. AIAA Journal 45(4) 912-921.

De S., Bathe K.J., 2000. Method of finite spheres. Computational Mechanics 25(4) 329-345.

De S., Bathe K.J., 2001. The method of finite spheres with improved integration. Computers and
Structures 79(22-25) 2183-2196.

Dolbow J., Belytschko T., 1999. Numerical integration of the Galerkin weak form in meshfree
methods. Computational Mechanics 23(3) 219-230.

Duarte C.A., Oden J.T., 1996. h-p adaptive method using clouds. Computer Methods in Applied
Mechanics and Engineering 139(1-4) 237-262.

Duarte C.A., Babuska I., Oden J.T., 2000. Generalized finite element methods for three-
dimensional structural mechanics problems. Computers and Structures 77(2) 215-232

Farin G., 2002. Curves and Surfaces for CAGD: A Practical Guide. Morgan Kaufmann
Publishers, San Francisco.

Gibson, R. F., 1994. Principles of Composite Material Mechanics. McGraw- Hill, New York.

Gingold R.A., Monaghan J.J., 1982. Kernel estimates as a basis for general particle methods in
hydrodynamics. Journal of Computational Physics 46 429-53.

Hughes T.J.R., Cottrell J.A., Bazilevs Y., 2005. Isogeometric analysis: CAD, finite elements,
NURBS, exact geometry and mesh refinement. Computer Methods in Applied Mechanics
and Engineering 194 (39-41) 4135-4195.

Hughes T.J.R., 2000. The Finite Element Method: Linear Static and Dynamic Finite Element
Analysis. Dover Publications, New York.

Hillig K., Reif V., Wipper J., 2001. Weighted extended B-spline approximation of Dirichlet
problems. SIAM Journal on Numerical Analysis 39(2) 442-462.

Hillig K., Reif V., 2003. Nonuniform web-splines. Computer Aided Geometric Design 20(5)
277-294.

Hillig K., 2003. Finite element methods with B-Splines. SIAM, Philadelphia.

Kagan P., Fischer A., Bar-Yoseph P.Z., 1998. New B-spline finite element approach for
geometrical design and mechanical analysis. International Journal for Numerical Methods
in Engineering 41(3) 435-458.

Kagan P., Fischer A., Bar-Yoseph P.Z., 2003. Mechanically based models: Adaptive refinement
for B-spline finite element. International Journal for Numerical Methods in Engineering
57(8) 1145-1175.

Kantorovich L.V., Krylov V.I., 1958. Approximate Methods of Higher Analysis. Interscience,
New York.

Kim, H.J., Swan C.C., 2003. Algorithms for automated meshing and unit cell analysis of
periodic composites with hierarchical tri-quadratic tetrahedral elements. International
Journal for Numerical Methods in Engineering 58(11) 1683-1711.

Kohno H., Tanahashi T., 2004. Numerical analysis of moving interfaces using a level set method
coupled with adaptive mesh refinement. International Journal for Numerical Methods in
Fluids 45(9) 921-944.

Krongauz Y., Belytschko T., 1996. Enforcement of essential boundary conditions in meshless
approximations using finite elements. Computer Methods in Applied Mechanics and
Engineering 131(1-2) 133-145.

Krongauz, Y., Belytschko, T.,1998. EFG approximation with discontinuous derivatives.
International Journal for Numerical Methods in Engineering 41(7) 1215-1233.

Krongauz Y., Belytschko T., 1997. Consistent pseudo-derivatives of meshless methods.
Computer Methods in Applied Mechanics and Engineering 146(3-4) 371-386.

Krysl P., Grinspun E., Schroder P., 2003. Natural hierarchical refinement for finite element
methods. International Journal for Numerical Methods in Engineering 56(8) 1109-1124.

Krysl P., Belytschko T., 1997. Element-free Galerkin method: convergence of the continuous
and discontinuous shape functions. Computer Methods in Applied Mechanics and
Engineering 148(3-4) 257-277.

Kumar A.V., Padmanabhan S., Burla R., 2007. Implicit boundary method for finite element
analysis using non-conforming mesh or grid. International Journal for Numerical Methods
in Engineering. (In press).

Kumar A.V., Lee J., 2006. Step function representation of solid models and application to mesh
free engineering analysis. Journal of Mechanical Design. Transactions of the ASME
128(1) 46-56.

Kumar A.V., Lee J.H., Burla R.K., 2005. Implicit solid modeling for mesh free analysis. In:
Proceedings of International Design Engineering Technical Conferences & Computers in
Engineering IDETC/CIE Long Beach, California.

Leung A.Y.T., Au F.T.K., 1990. Spline finite elements for beam and plate. Computers and
Structures 37(5) 717-729.

Li, S., Lim, S.H., 2005. Variational principles for generalized plane strain problems and their
applications. Applied Science and Manufacturing 36(3) 353-365.

Liszka T.J., Duarte C.A.M., Tworzydlo W.W., 1996. hp-meshless cloud method. Computer
Methods in Applied Mechanics and Engineering 139(1-4) 263-288.

Liu G.R., Gu Y.T., 2001. A point interpolation method for two-dimensional solids. International
Journal for Numerical Methods in Engineering 50(4) 937-951.

Lucy L.B., 1977. A numerical approach to the testing of the fission hypothesis. The Astronomical
Journal 82(12) 1013-24.

Marrey, R.V., Sankar, B.V., 1997. A Micromechanical Model for Textile Composite Plates.
Journal of Composite Materials 31(12) 1187-1213.

Melenk J.M., Babuska I., 1996. The partition of unity finite element method: Basic theory and
applications. Computer Methods in Applied Mechanics and Engineering 139(1-4) 289-
314.

Moes N., Bechet E., Tourbier M., 2006. Imposing Dirichlet boundary conditions in the extended
finite element method. International Journal for Numerical Methods in Engineering 67(12)
1641-1669.

Monaghan J.J., 1987. An introduction to SPH. Computer Physics Communications 48(1) 89 96.

Mortenson M.E., 1997. Geometric Modeling. Wiley, New York.

Nayroles B., Touzat G., Villon P.,1992. Generalizing the finite element method: Diffuse
approximation and diffuse elements. Computational Mechanics 10(5) 307-318.

Osher S.J., Fedkiw R.P., 2002. Level Set Methods and Dynamic Implicit Surfaces. Springer

Padmanabhan, S., 2006. Analysis using non-conforming structured grid and implicit boundary
method. Doctoral Dissertation, University of Florida.

Pengcheng S., Peixiang H., 1995. Bending analysis of rectangular moderately thick plates using
spline finite element method. Computers and Structures 54(6) 1023-1029.

Rvachev V.L., Shieko T.I., 1995. R-functions in boundary value problems in mechanics. Applied
Mechanics Reviews 48(4) 151-188.

Rvachev V.L., Sheiko T.I., Shapiro V., Tsukanov I., 2001. Transfinite interpolation over
implicitly defined sets. Computer-Aided Geometric Design 18(3) 195-220.

Sankar, B.V., and Marrey, R.V., 1997. Analytical Method for Micromechanics of Textile
Composites. Composites Science and Technology 57(6) 703-713.

Shapiro V., 1998. Theory of R-functions and applications: A primer. CPA Technical Report
CPA88-3, Cornell Programmable Automation. Sibley School of Mechanical Engineering.

Shapiro V., Tsukanov I., 1999. Meshfree simulation of deforming domains. Computer Aided
Design 31(7) 459-471.

Sokolnikoff I.S., 1956. Mathematical Theory of Elasticity, McGraw-Hill, New York.

Sukumar N., Moran B., Belytschko T., 1998. Natural element method in solid mechanics.
International Journal for Numerical Methods in Engineering 43(5) 839-887.

Sun, C.T., Chen, J.L., 1990. A micromechanical model for plastic behavior of fibrous
composites. Composites Science and Technology 40 115-29.

Sun, C.T., Vaidya, R.S., 1996. Prediction of composite properties from a representative volume
element. Composites Science and Technology 56(2) 171-179.

Surana K.S., Rajwani A., Reddy J.N., 2006. The k-version finite element method for singular
boundary-value problems with application to linear fracture mechanics, International
Journal for Computational Methods in Engineering Science and Mechanics 7(3), 217-239.

Taliercio, A, 2005. Generalized plane strain finite element model for the analysis of elastoplastic
composites. International Journal of Solids and Structures 42(8) 2361-2379.

Timoshenko S.P., Goodier J.N., 1969. Theory of Elasticity, McGraw Hill, New York.

Tyrus, J.M., Gosz, M., DeSantiago, E., 2007. A local finite element implementation for imposing
periodic boundary conditions on composite micromechanical models. International Journal
of Solids and Structures 44(9) 2972-2989.

Vermeulen A.H., Heppler G.R., 1998. Predicting and avoiding shear locking in beam vibration
problems using the B-spline field approximation method. Computer Methods in Applied
Mechanics and Engineering 158(3-4) 311-327.

Whitney, J.M., Riley, M.B., 1966. Elastic properties of fiber reinforced composite materials.
A1AA Journal 4 1537-42.

Xia, Z., Zhang, Y., Ellyin, F., 2003.A unified periodical boundary conditions for representative
volume elements of composites and applications. International Journal of Solids and
Structures 40(8) 1907-1921.

Yu, W., Tang, T., 2007. A variational asymptotic micromechanics model for predicting
thermoelastic properties of heterogeneous materials. International Journal of Solids and
Structures 44(22-23) 7510-7525.

Zhu, H., Sankar, B.V., Marrey, R.V., 1998. Evaluation of Failure Criteria for Fiber Composites
Using Finite Element Micromechanics. Journal of Composite Materials 32(8) 766-782.

BIOGRAPHICAL SKETCH

Ravi Kumar Burla was born and brought up in Hyderabad, India. He graduated with a

Bachelor of Technology degree in mechanical engineering and a Master of Technology degree in

computer-aided design and automation from Indian Institute of Technology, Bombay, India in

June 2001. After graduation, he joined Geometric Software Solutions Limited, Pune, India as a

Software Engineer where he worked for 2 years. He then enrolled in the doctoral program in

Mechanical and Aerospace Engineering at the University of Florida in Fall 2003. He was

awarded an Alumni Fellowship. His areas of interest include application of B-splines for finite

element analysis, numerical methods and computer-aided geometric design.

PAGE 1

FINITE ELEMENT ANALYSIS USING UNIFORM B-SPLINE APPROXIMATION AND IMPLICIT BOUNDARY METHOD By RAVI KUMAR BURLA A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008 1

PAGE 2

2008 Ravi Kumar Burla 2

PAGE 3

To my parents 3

PAGE 4

ACKNOWLEDGMENTS I would like to express my sincere gratitude to my advisor and chairman of my supervisory committee, Dr. Ashok V. Kumar, for his guidance, encouragement, enthusiasm and constant support throughout my doctoral research. I would like to thank him for the numerous insights he provided during every stage of the research. Without his assistance it would not have been possible to complete this dissertation. I would also like to thank the members of my advisory committee, Dr. Bhavani V. Sankar, Dr. Nam-Ho Kim and Dr. Jorg Peters. I am grateful for their willingness to serve on my committee, for providing help whenever required, for involvement and valuable suggestions in my oral qualifying examination, and for reviewing this dissertation. My colleagues at Design and Rapid Prototyping Laboratory at the University of Florida deserve special thanks for their help and many fruitful discussions. I would especially thank Sanjeev Padmanabhan, Greg Freeman and Sung Uk-Zhang. I would like to thank my parents and my brothers for their constant love and support without which this would not have been possible. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS ...............................................................................................................4 LIST OF TABLES ...........................................................................................................................8 LIST OF FIGURES .........................................................................................................................9 ABSTRACT ...................................................................................................................................14 CHAPTER 1 INTRODUCTION..................................................................................................................16 1.1 Overview...........................................................................................................................16 1.2 Goals and Objectives........................................................................................................19 1.3 Outline..............................................................................................................................19 2 ANALYSIS TECHNIQUES FOR BOUNDARY VALUE PROBLMS................................21 2.1 Finite Element Method.....................................................................................................21 2.2 B-Spline Approximation Based Methods.........................................................................22 2.3 Structured Grid Methods..................................................................................................23 2.4 Meshless Methods............................................................................................................24 2.5 Techniques for Determination of Effective Properties of Composite Microstructures....27 2.6 Techniques for Local Refinement....................................................................................28 3 SOLUTION STRUCTURES FOR IMPLICIT BOUNDARY METHOD.............................30 3.1 Solution Structure for Enforcing Essential Boundary Conditions....................................31 3.1.1 Modified Weak Form for Linear Elasticity............................................................32 3.1.2 Modified Weak Form for Steady State Heat Transfer............................................34 3.1.3 Dirichlet Function...................................................................................................36 3.1.4 Boundary Value Function.......................................................................................38 3.2 Solution Structure for Micromechanics............................................................................40 3.3 Solution Structure for Grid Refinement...........................................................................44 3.3.1 Solution Structure...................................................................................................45 3.3.2 Modified Weak Form.............................................................................................47 4 B-SPLINE FINITE ELEMENT FORMULATION...............................................................57 4.1 One Dimensional B-Spline Elements...............................................................................58 4.2 Two and Three Dimensional B-Spline Elements.............................................................60 5

PAGE 6

5 NUMERICAL IMPLEMENTATIONS..................................................................................67 5.1 Numerical Implementations for Linear Elastic Problems................................................67 5.1.1 Computation of Stiffness Matrix for Internal Elements.........................................68 5.1.2 Computation of Stiffness Matrix for Boundary Elements......................................68 5.1.3 Computation of Stiffness Matrix for Elements with Essential Boundaries............69 5.1.4 Computation of Load Vector..................................................................................73 5.2 Numerical Implementations for Steady State Heat Transfer............................................74 5.3 Numerical Implementations for Micromechanics............................................................76 5.3.1 Computation of Stiffness Matrix in the Matrix Region..........................................79 5.3.2 Computation of Stiffness Matrix in the Inclusion Region......................................79 5.3.3 Computation of Stiffness Matrix in the Interface Region......................................80 5.3.4 Periodic Boundary Conditions...............................................................................82 5.3.5 Computation of Effective Properties......................................................................84 5.3.6 Computation of Load Vector..................................................................................87 5.4 Numerical Implementations for Local Grid Refinement..................................................87 5.4.1 Computation of Stiffness Matrix in Regions 1 and 2.............................................88 5.4.2 Computation of Stiffness Matrix in Region 3........................................................89 5.4.3 Computation of Load Vector..................................................................................90 5.5 Extrapolation of Solution for Elements with Low Volume Fraction...............................90 5.5.1 Construction of Extrapolation Matrix....................................................................92 5.5.2 Construction of Extrapolation Matrix for a Boundary Element.............................94 5.5.3 Computation of Stiffness Matrix............................................................................94 6 ANALYSIS AND RESULTS...............................................................................................103 6.1 Validation of Solution Structure for Essential Boundary Conditions............................103 6.1.1 Patch Test.............................................................................................................103 6.1.2 Validation of Extrapolation Technique................................................................106 6.1.3 Convergence Analysis of a Cantilever Beam.......................................................107 6.1.4 Square Plate with a Circular Hole........................................................................110 6.1.5 Thick Cylinder with Internal Pressure..................................................................111 6.1.6 Steady State Heat Transfer Analysis....................................................................112 6.1.7 Analysis of a 3D Clamp.......................................................................................113 6.1.8 Analysis of a 3D Rib Structure.............................................................................114 6.1.9 Analysis of a 3D Dial Bracket..............................................................................115 6.2 Validation of Solution Structure for Micromechanics....................................................116 6.2.1 Square Plate with Circular Inclusion....................................................................116 6.2.2 Determination of Effective Properties of a Fiber Reinforced Composite............117 6.3 Validation of Solution Structure for Grid Refinement...................................................118 6.3.1 Application of Grid Refinement to Constant Stress Problem..............................118 6.3.2 Stress Concentration Problem..............................................................................119 7 CONCLUSIONS AND FUTURE WORK...........................................................................165 7.1 Conclusions.....................................................................................................................165 7.2 Future Work....................................................................................................................167 6

PAGE 7

LIST OF REFERENCES.............................................................................................................169 BIOGRAPHICAL SKETCH.......................................................................................................175 7

PAGE 8

LIST OF TABLES Table page 5-1 Set of periodic boundary conditions for 6 different strains.............................................102 6-1 Average stresses for various loading cases......................................................................162 6-2 Comparison of effective properties..................................................................................163 6-3 Average stresses for various loading cases......................................................................164 8

PAGE 9

LIST OF FIGURES Figure page 3-1 Solution structure for imposing essential boundary conditions.........................................49 3-2 Elastic solid under equilibrium..........................................................................................49 3-3 Steady state heat transfer problem.....................................................................................50 3-4 Boolean operations on Dirichlet functions........................................................................50 3-5 A CSG tree representing Dirichlet function for multiple boundaries................................51 3-6 Boundary value function using cubic B-spline basis.........................................................52 3-7 Representation of grids for material discontinuity.............................................................53 3-8 One dimensional inclusion function for treatment of material discontinuity....................53 3-9 Two dimensional inclusion function for an elliptic inclusion in a square matrix..............54 3-10 Representation of solution structure for local grid refinement..........................................54 3-11 Approximate step function for refinement.........................................................................55 3-12 Coarse and refined grid interface with multiple implicit equations...................................55 3-13 Step function for refinement for a 2D model.....................................................................56 4-1 Approximation using 1D cubic B-splines..........................................................................64 4-2 Shape functions for 1D quadratic B-spline element..........................................................64 4-3 Shape functions for 1D cubic B-spline element................................................................64 4-4 Typical grid used for analysis and the mapping used for geometry and field variable.....65 4-5 Two dimensional quadratic B-spline elements in parametric space..................................66 4-6 Two dimensional cubic B-spline elements in parametric space........................................66 5-1 Typical non-conforming grid used for analysis.................................................................95 5-2 Triangulation of boundary element for volume integration...............................................96 5-3 Representation of band on essential boundaries................................................................96 5-4 Approximation of natural boundary for computation of load vector.................................97 9

PAGE 10

5-5 Division of computational domain into three regions........................................................98 5-6 Undeformed and deformed configuration of a set of 4 RVEs...........................................98 5-7 Representative volume element used for computation of effective properties..................99 5-8 Division of computation domain into various regions.......................................................99 5-9 Typical analysis grid with elements having low volume fraction...................................100 5-10 Extrapolation of external node for linear Lagrange element...........................................100 5-11 Extrapolation of external nodes for cubic B-spline element............................................100 5-12 Extrapolation of external nodes for quadratic B-spline element.....................................101 5-13 Extapolation for boundary element..................................................................................101 6-1 Plate subjected to uniaxial loading..................................................................................121 6-2 Non-conforming grid using one cubic B-spline element.................................................121 6-3 Plot of Dirichlet functions................................................................................................122 6-4 Effect of range parameter on the quality of solution.......................................................123 6-5 Contour plot of x for one element model using cubic B-splines..................................123 6-6 Thin cantilever beam subjected to end loading................................................................124 6-7 Plot of error in the tip deflection with varying aspect ratio.............................................124 6-8 Curved beam subjected to uniform loading.....................................................................125 6-9 Plot of error in strain energy for curved beam (without extrapolation)...........................125 6-10 Plot of error in strain energy for curved beam using extrapolation technique.................126 6-11 Contour plot of y for Q4 elements (without extrapolation)...........................................126 6-12 Contour plot of y for Q4 elements using extrapolation technique................................127 6-13 Contour plot of y using cubic B-spline elements..........................................................128 6-14 Thick cantilever beam used for convergence analysis.....................................................128 6-15 Typical non-conforming grid using cubic B-spline elements..........................................129 10

PAGE 11

6-16 Convergence plot of 2 L error for cantilever beam..........................................................129 6-17 Convergence plot for error in strain energy for cantilever beam.....................................130 6-18 Comparison of deflection curves for 10 element model..................................................131 6-19 Comparison of normal stress along the line AB for 10 element model...........................132 6-20 Comparison of shear stress along the line AB for 10 element model..............................133 6-21 Convergence plot for tip deflection in near incompressible limit....................................133 6-22 Analysis model for a square plate with a circular hole....................................................134 6-23 Contour plot of x using FEM Q4 elements....................................................................134 6-24 Contour plot of x using cubic B-spline elements..........................................................135 6-25 Convergence plot for 2 L error norm for stress concentration problem...........................135 6-26 Convergence plot for error in strain energy for stress concentration problem................136 6-27 Plot of x along the line 2 .......................................................................................136 6-28 Hollow thick cylinder subjected to uniform pressure......................................................137 6-29 Contour plot of radial displacements using cubic B-spline element model.....................137 6-30 Contour plot of radial displacements using FEM Q4 elements.......................................138 6-31 Comparison of radial displacements along the thickness of the cylinder........................138 6-32 Comparison of radial stresses along the thickness of the cylinder..................................139 6-33 Comparison of tangential stresses along the thickness of the cylinder............................139 6-34 Heat conduction in a block with series of pipes..............................................................140 6-35 Plot of Dirichlet function.................................................................................................140 6-36 Plot of boundary value function.......................................................................................141 6-37 Contour plot of temperature distribution using FE Q4 elements.....................................141 6-38 Contour plot of temperature distribution using cubic B-spline elements........................142 6-39 Contour plot of temperature distribution using quadratic B-spline elements..................142 11

PAGE 12

6-40 Comparison of temperature along line AB......................................................................143 6-41 Three dimensional clamp used for analysis.....................................................................143 6-42 Convergence plot for strain energy for clamp model......................................................144 6-43 Contour plot of bending stresses using H8 elements in FEM for 3D clamp...................144 6-44 Contour plot of bending stresses using H8 elements for 3D clamp.................................145 6-45 Contour plot of bending stresses using cubic B-spline elements for 3D clamp..............145 6-46 Three dimensional rib model used for analysis...............................................................146 6-47 Convergence plot for strain energy for rib model............................................................146 6-48 Contour plot of bending stresses using tetrahedral elements for 3D rib..........................147 6-49 Contour plot of bending stresses using B-spline elements for 3D rib.............................147 6-50 Contour plot of bending stresses using H8 elements for 3D rib......................................148 6-51 Three dimensional dial bracket used for analysis............................................................148 6-52 Convergence plot for strain energy for dial bracket model.............................................149 6-53 Contour plot of bending stresses using a cubic B-spline element model for dial bracket..............................................................................................................................149 6-54 Contour plot of bending stresses using a tetrahedral element model for dial bracket.....150 6-55 Contour plot of bending stresses using a H8 element model for dial bracket.................150 6-56 Square plate with a circular inclusion..............................................................................151 6-57 Contour plot of x using Q4 elements in IBFEM............................................................151 6-58 Contour plot of x using Q4 elements in FEM...............................................................152 6-59 Convergence plot for error in strain energy for square matrix with inclusion model......152 6-60 Comparison of along the line AB................................................................................153 xu 6-61 Comparison of stresses along the line AB.......................................................................153 6-62 Comparison of strains along the line AB.........................................................................154 6-63 Plot of stresses for cases 1 and 2......................................................................................155 12

PAGE 13

6-64 Plot of stresses for cases 3 and 4......................................................................................155 6-65 Plot of stresses for cases 5 and 6......................................................................................155 6-66 Plot of stresses for cases 1 and 2 for 3D model...............................................................156 6-67 Plot of stresses for cases 3 and 4 for 3D model...............................................................156 6-68 Plot of stresses for cases 5 and 6 for 3D model...............................................................156 6-69 Contour plot of using Q4 elements and grid refinement.............................................157 u 6-70 Contour plot of using Q4 elements and grid refinement.............................................157 v 6-71 Contour plot of using cubic B-spline elements and grid refinement...........................158 u 6-72 Contour plot of using cubic B-spline elements and grid refinement...........................158 v 6-73 Square plate with a circular hole......................................................................................159 6-74 Contour plot of x for local refinement using Q4 elements...........................................159 6-75 Contour plot of x for local refinement using cubic B-spline elements.........................160 6-76 Convergence plot for 2 L error in local grid refinement..................................................160 6-77 Convergence plot for energy error in local grid refinement............................................161 6-78 Contour plot of x for local refinement using Q4 and cubic B-spline elements............161 13

PAGE 14

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy FINITE ELEMENT ANALYSIS USING UNIFORM B-SPLINE APPROXIMATION AND IMPLICIT BOUNDARY METHOD By Ravi Kumar Burla May 2008 Chair: Ashok V. Kumar Major: Mechanical Engineering The finite element method (FEM) is a widely used numerical technique for solution of boundary value problems arising in engineering applications. FEM is well studied and has several advantages; however the stress and strain solutions obtained by FEM are discontinuous across element boundaries and require smoothing algorithms to make the stress/strain appear smooth in the analysis domain. Automatic mesh generation is still a difficult task in FEM for complicated geometries and often requires user intervention for generating quality meshes. Finite elements based on uniform B-spline basis functions defined on a structured grid are developed which provide at least continuous solution through out the analysis domain. Numerical techniques are presented in this work based on structured grids and implicit equations of the boundaries of analysis domain, material interface boundaries or coarse/fine grid interfaces for solution of various engineering boundary value problems. Solution structures are constructed using approximate Heaviside step functions for imposition of Dirichlet boundary conditions, treatment of material discontinuity to perform micromechanical analysis and for local refinement of grid. The use of structured grid eliminates the need for constructing a conforming mesh and results in significant savings of time in pre-processing stage of the design cycle. 1C 14

PAGE 15

Numerical examples are presented to demonstrate the performance of B-spline elements. The results are compared with analytical solutions as well as traditional finite element solutions to demonstrate the ability of B-spline elements to represent continuous stress and strain through out analysis domain. Convergence studies show that B-spline elements can provide accurate solutions for many engineering problems with fewer numbers of elements and nodes as compared to traditional FEM. Solution structure for treatment of material boundary is validated by performing a convergence analysis on a problem involving circular inclusion in a square matrix and by determining effective properties of fiber reinforced composite. Solution structure for local grid refinement is validated by analyzing classical stress concentration problems. 15

PAGE 16

CHAPTER 1 INTRODUCTION 1.1 Overview Numerical methods are necessary for solving many problems arising in engineering analysis. Even though analytical solutions are available for several engineering problems with simple geometries, for complicated domains it is difficult to obtain analytical solutions and hence numerical methods are needed to obtain accurate solutions. The Finite Element Method (FEM) is widely used in engineering analysis not only due to its applicability to solve a variety of physical problems but also due to its ability to handle complex geometry and boundary conditions. In FEM, the analysis domain is subdivided into finite elements which approximate the geometry of the analysis domain. The field variable is approximated on these finite elements using piecewise continuous interpolations. Even though FEM is a well established technique, it has certain limitations. Approximation schemes used in FEM are typicallycontinuous and hence there is a discontinuity of gradients across the elements. The lack of gradient continuity in FEM necessitates the smoothing of results. There are a number of approximation schemes like B-splines, moving least squares and Hermite interpolations which provide at least continuity in the domain of interest. B-spline basis functions have compact support like traditional finite element shape functions and lead to banded stiffness matrices. They form partition of unity and hence can exactly represent the state of constant field in the domain of interest. The B-spline approximations have good reproducing properties and when a cubic B-spline is used, it can represent all polynomials up to third order exactly. B-spline approximation provides high order of continuity and is capable of providing accurate solutions with continuous gradients through out the domain. The uniform B-spline basis functions are polynomials and accurate integration for stiffness matrices and load vectors can be 0C 1C 16

PAGE 17

performed by using well known techniques like Gauss Quadrature. These properties of B-splines make them attractive for use in analysis. B-splines have not been used extensively for analysis because they do not satisfy Kronecker-delta property and hence essential boundary conditions cannot be imposed directly by specifying nodal values. B-splines are defined on a grid which limits the geometry of the analysis domain to a parallelepiped. In order to be able to use uniform B-spline basis for approximation of finite element solutions, structured grids need to be used since the uniform B-spline basis are defined only on a uniform structured grid. A structured grid is much easier to generate than a finite element mesh and all the elements in the grid have regular geometry (rectangles/cuboids). The use of structured grids for analysis eliminates the need to generate a conforming mesh which is a significant advantage over traditional FEM. Automatic mesh generation algorithms for FEM are well developed 2D problems but can be unreliable for 3D geometries often resulting in poor or distorted elements in some regions that can lead to large errors in solution. Significant amount of user intervention is needed to correct such problems making mesh generation the most time consuming step in the design process. These problems associated with mesh generation can be avoided with the use of a structured grid. When structured grids are used to approximate solution in the analysis domain, the boundaries of the domain need to be defined independent of the structured grid and this can lead to several elements on the boundary which have partial volume inside the grid elements. Special techniques are therefore needed to compute the stiffness matrices and load vectors in these partial elements. B-spline approximations do not satisfy Kronecker-delta property and hence boundary conditions cannot be applied by specifying nodal values. Furthermore, the nodes in a structured grid are not guaranteed to lie on the boundary, therefore the traditional methods used 17

PAGE 18

in FEM for applying essential boundary conditions cannot be used and new techniques are needed for applying essential boundary conditions. In order to enforce essential boundary conditions when approximations are used on structured grids, Implicit Boundary Method is used in this work. In Implicit Boundary Method, solution structures are constructed by using the approximate Heaviside step functions based on the implicit equations of the boundaries for exact imposition of essential boundary conditions. 1C Implicit Boundary Method is extended in this work by developing solution structure for treatment of material discontinuity using approximate Heaviside step functions based on the implicit equations of matrix/inclusion boundaries. This technique is further used for determination of effective properties of fiber reinforced composite. In many practical problems local refinement of mesh/grid is necessary to accurately compute the stresses in the regions of stress concentrations; this problem is addressed in this work by developing solution structure for local grid refinement using approximate Heaviside step functions based on the implicit equations of the coarse/refined grid boundaries. The performance and capabilities of B-spline based finite elements are studied by solving several problems in structural and thermal analysis in both 2D and 3D. Convergence analysis is also performed using several examples to study the behavior of B-spline elements and it has been demonstrated that B-spline element provide accurate solutions with fewer number of elements when compared to Lagrange elements used in traditional FEM. Micromechanical analysis for determining effective properties of fiber reinforced composites is performed by employing the solution structure for treatment of material discontinuity using Lagrange finite elements. Solution structure for local refinement is used to analyze stress concentration problems where local 18

PAGE 19

resolution of solution near regions of stress concentrations is necessary for performing design analyses. 1.2 Goals and Objectives The goal of this research is to develop finite elements based on B-spline approximations and to extend Implicit Boundary Method by developing solution structures for analysis of engineering boundary value problems involving material discontinuity and local grid refinement. The main objectives of this dissertation are To extend the Implicit Boundary Method for use of B-spline basis functions. To develop finite elements based on B-spline approximation which provides at least continuity in the analysis domain. 1C To develop solution structures for modeling material discontinuity and imposition of periodic boundary conditions involved in the micromechanical analysis of composite microstructures and to determine effective elastic properties of fiber reinforced composites. To develop solution structures for local refinement of grid to accurately capture stress concentrations and study the use of the B-spline approximations and Lagrange interpolations in local refinement. To develop efficient numerical techniques for computation of various terms involved in the Galerkin weak form. To study the convergence behavior of the solution structures and B-spline finite elements using several numerical examples and to study the advantages and disadvantages of using the B-spline based finite element. 1.3 Outline The rest of the dissertation is organized as follows: Chapter 2 discusses several analysis techniques that are reported in literature which use B-splines to approximate solution. A survey on structured grid methods, meshless methods, treatment of material discontinuity and adaptive grid/mesh refinement is also presented. Chapter 3 discusses the construction of solution structure for exact imposition of essential boundary conditions for structural and thermal analyses. Solution structures are developed for 19

PAGE 20

modeling material discontinuity and for imposition of periodic boundary conditions which are used for micromechanical analysis of composites. Finally, solution structure is developed for local refinement of grid to improve the local resolution of the solution. In chapter 4, finite elements based on B-spline approximation are developed which provide and continuity in one, two and three dimensions. 1C 2C Chapter 5 discusses efficient numerical implementations for computation of stiffness matrices and load vectors involved in the modified weak form corresponding to the solution structures presented in chapter 3. These techniques are implemented in JAVA and several problems in structural mechanics, steady state heat transfer, composite micromechanics and localized stress concentrations are studied in detail in chapter 6. Convergence analyses are also performed for these problems to study the behavior B-spline elements and solution structures in terms of accuracy and efficiency. Finally, in chapter 7, several conclusions are drawn from the numerical experiments performed in chapter 6. The advantages and shortcomings of proposed solution structures and B-spline finite elements are discussed in detail. The scope for future work is presented towards the end of the dissertation. 20

PAGE 21

CHAPTER 2 ANALYSIS TECHNIQUES FOR BOUNDARY VALUE PROBLMS In this chapter, a detailed discussion on several analysis techniques that are used for solution of engineering boundary value problems is presented. In the first part of this chapter, a brief overview of the finite element method (FEM) is presented along with its advantages and shortcomings. Techniques for employing B-spline approximations for analysis are discussed in the next part of this chapter. A detailed discussion on structured grid methods and meshless methods is presented in the later part of this chapter Micromechanical analysis is an important engineering application where FEM is used extensively. Microstructures are modeled using a representative volume element and periodic boundary conditions are applied using multipoint constraint techniques. Meshless methods have been extended for treatment of material discontinuity and for performing micromechanical analysis. A survey on these techniques is presented in this chapter along with a survey on employing FEM for determination of effective properties. Another important feature of FEM is its capability for local refinement. A survey on the refinement techniques available in structured grid and meshless framework is presented in the last part of this chapter. 2.1 Finite Element Method Finite element method (Hughes, 2000 ; Cook et. al. 2003) is a well established numerical technique and is widely used in solving engineering problems in academia as well as in industry. For finite element analysis, the domain of analysis is subdivided into elements and the resulting finite element mesh approximates the geometry and is also used to approximate the solution by piece-wise interpolation within each element. Using Galerkins approach, the governing differential equation in its strong form is converted into a weak form which allows the use of 21

PAGE 22

interpolation/approximation schemes with lower continuity requirements than the corresponding strong form. This leads to a set of linear equations which are then solved using Gauss elimination, Cholesky decomposition or iterative schemes like Gauss-Siedel. FEM has been applied to a wide variety of physical problems in linear/non-linear static and dynamic elasticity, steady state heat transfer, electrostatics and magnetostatics. The approximation space used in FEM is typically Lagrange interpolation that yields continuous solutions. However, continuous solutions have discontinuous stresses and strains (gradients) across the element boundaries. The lack of stress and strain continuity necessitates the use of smoothing schemes to plot these quantities such that they appear to be continuous. Solutions can be represented with higher degree of continuity using several other basis functions like B-splines, moving least squares (MLS) and Hermite functions which provide at least continuity in the domain of interest. In this dissertation, B-spline approximations are studied in detail which provide and continuous solutions in the analysis domain. 0C 0C 1C 1C 2C Another limitation of the FEM is the need to generate a conforming mesh, which may need user intervention for correcting the problems associated with distorted meshes. Automatic mesh generation algorithms work satisfactorily for 2D but are still unreliable for 3D. In order to avoid the problems associated with mesh generation, several methods have been proposed which fall into two categories: meshless methods and structured grid methods. In this chapter, a detailed survey on various structured grid and meshless methods is presented. 2.2 B-Spline Approximation Based Methods A number of researchers have studied the use of B-spline approximations to solve specific problems where the geometry of the analysis domain is simple (e.g. beams or rectangular plates). The boundary conditions were applied by using multiple knots at the boundary of the analysis 22

PAGE 23

domain which emulates Kronecker-delta behavior on the boundaries. Leung and Au (Leung and Au, 1990) introduced the boundary transformation method by constructing an appropriate transformation matrix based on the boundary configuration to impose Dirichlet boundary conditions. Kagan et. al.(1998, 2003) have provided a scheme to integrate B-spline finite element method with a mechanically based computer aided design and the Dirichlet boundary conditions were enforced by imposing multiple knots on the essential boundaries. They also developed methods for adaptive refinement of B-spline elements. In summary, these methods have limitations on the geometry of analysis domain. Hughes and his coworkers (Hughes et. al. 2005 ; Cottrell et. al. 2006 ; Cottrell et. al. 2007) have developed a more general approach for using B-splines called the isogeometric analysis method, where the geometry is represented by Non-Uniform Rational B-Spline (NURBS) basis functions to get an exact geometric representation. The same basis functions can be used for the analysis or their order can be raised without changing the geometry or parameterization. This enables the traditional h-, and padaptive refinement as well as k-refinement where the degree of continuity of the solution can be raised. Several alternatives for imposing Dirichlet boundary conditions were proposed including direct application by specifying control variables and use of constraint equations, but this still remains a topic of research. 2.3 Structured Grid Methods A number of techniques have been developed based on structured grids to approximate the solution. Extended Finite Element Method or X-FEM was developed by Belytschko and co-workers (Belytschko et. al. 2003). This method uses implicit equations for definition of the geometry of the analysis domain and the Dirichlet boundary conditions are applied using Lagrange multipliers. 23

PAGE 24

Penalty boundary method developed by Clark and Anderson (Clark and Anderson, 2003a, 2003b) uses a regular grid and a constrained variational formulation to include both essential and natural boundary conditions which is then solved using penalty method. Shapiro (Shapiro, 1998 ; Shapiro and Tsukanov, 1999) and Hollig (Hollig et. al. 2001; Hollig and Reif 2003 ; Hollig 2003) used R-functions to define the boundary of the problem domain and the solution structure proposed by Kantorovich and Krylov (Kantorovich and Krylov 1958) was constructed to satisfy Dirichlet boundary conditions. Shapiro used transfinite Lagrange interpolation to impose Dirichlet boundary conditions while Hollig used weight functions based on distance functions and R-functions. Hollig developed extended B-spline basis to represent the solution in the analysis domain. The method presented in this dissertation is very similar to the approach used by Hollig, however approximate Heaviside step function of essential boundaries are used in this work to construct weight functions and the behavior of solution structure for very small values of range parameter used to construct step functions is also studied. All the internal elements in our method have identical stiffness matrices unlike Holligs method. 2.4 Meshless Methods A number of methods have been developed in the past two decades which avoid mesh generation. These methods are collectively termed as meshless or meshfree methods. Some of these methods provide continuous gradients in the analysis domain. Smoothed particle hydrodynamics (SPH) is one of the first meshless methods in literature (Lucy, 1977 ; Gingold, 1982); it was first used to model astrophysical phenomenon and was modified to include continuum mechanics problems (Monahan, 1982). This approach uses Shepard shape functions as interpolants and uses point collocation method for discretization and nodal integration for stiffness matrix computation. The drawbacks of this approach are poor accuracy and spatial 24

PAGE 25

instability (tensile instability). The behavior of solution depends on the proper distribution of nodes. So there is always an effort needed to place the nodes properly in order to avoid instability. Element Free Galerkin Method was developed by Belytschko and his coworkers (Belytschko et. al., 1994 ; Belytschko et. al., 1996 ; Belytschko et. al. ; 1996, Dolbow and Belytschko, 1999 ; Krongauz and Belytschko, 1996 ; Krongauz and Belytschko ; 1997 ; Krysl and Belytschko 1997). This method is based on moving least square approximation and uses global Galerkin method for discretization. Quadrature was performed on the background cells. Boundary conditions were applied using Lagrange multipliers or by modification of variational principle, where Lagrange multipliers were replaced by traction forces. The former method leads to an increased number of algebraic equations and the stiffness matrix is not positive definite. The latter method reduces the accuracy of the results and is not as convenient as imposing essential boundary conditions directly. A new method for imposing essential boundary conditions by using a string of Finite Elements around the boundary of the domain was later incorporated into this method. Meshless Local Petrov-Galerkin (MLPG) was introduced as a truly meshless method by Atluri and his coworkers (Atluri et al. 1999 ; Atluri and Zhu 2000a, 2000b). This method uses moving least square approximation as the interpolation functions. It uses the Local Petrov-Galerkin weak form which is developed in the local sub-domain. The trial and test spaces are different; the test functions vanish on the boundary of the sub-domain, as long as these boundaries are internal to the problem domain. If the sub-domain intersects the boundary of global domain, then the boundary condition is applied on this local boundary by penalty method. The stiffness matrix obtained is banded but not symmetric. The boundary conditions are applied 25

PAGE 26

using penalty method. Another advantage of this procedure over FEM is that smoothing techniques are not required for computation of derivative. There is no need of special integration schemes, and there is flexibility in choosing the size and shape of the local domain, which can lead to convenient methods for integration in non-linear problems. The drawbacks of this method are that the stiffness matrices generated by this approach are not symmetric positive definite and penalty method imposes boundary conditions approximately. Another drawback is that we have to make sure that the union of all the local domains should cover the global boundary. Partition of unity based methods (PUFEM) form a class of meshless methods which use partition of unity functions to blend the local approximation functions which represent the local behavior (Melenk and Babuska, 1996). The main advantage of PUFEM is that we can use the knowledge of the local behavior of solutions efficiently to get good numerical approximation. This method also yields good results in the cases where FEM gives poor results or when FEM is computationally very expensive. One drawback of this approach is that the stiffness matrix obtained from this procedure can be singular and the basis function needs to be modified appropriately or special techniques are needed to solve the singular system. Generalized Finite Element Method was developed as a hybrid of Finite Element Method and Partition of Unity (Babuska et. al. 2002 ; Duarte et. al. 2000). The desirable features like inter-element continuity, local approximability and Kronecker-delta property from FEM are retained in GFEM. Method of finite spheres (De and Bathe 2000 ; De and Bathe 2001) was developed as meshless method which is based on partition of unity; this approach does not use background cells for numerical quadrature. It uses compactly supported partition of unity functions and Bubnov-Galerkin method as weighted residual scheme. Integration of weak form is performed on spheres (circles in 2D), using specialized cubature rules. Essential boundary conditions are 26

PAGE 27

imposed by placing nodes on the boundary so that the shape functions corresponding to these nodes emulate Kronecker-delta behavior. Shepard functions are used as partition of unity functions and quartic splines are used for weight functions. Method of HP Clouds is a partition of unity based meshless method developed by Duarte and Oden (Duarte and Oden 1996 ; Duarte et. al. 1996 ; Liszka et. al. 1996). HP-Cloud functions are created by multiplying the partition of unity functions with local basis functions. Shepard functions or MLS functions are used as partition of unity. Anisotropic approximation can be incorporated by using the basis function which has higher order in one direction and lower order in other direction. P-refinement is done by increasing the size of the local domain; hence more number of points is included in the sub-domain and also by increasing the polynomial order in the basis function. Boundary conditions are applied by modified variational principle. Natural element method developed by Sukumar et. al.(1998) uses the natural neighbor interpolants which are based on Voronoi tessellation of the nodes. These interpolants are smooth over the whole domain excepting at the nodes where they are continuous. The essential boundary conditions are applied on nodes for convex domains. 2.5 Techniques for Determination of Effective Properties of Composite Microstructures The finite element method (FEM) has been used extensively in micromechanical analysis of composite microstructure to determine effective properties (Sun and Vaidya ; 1996 ; Taliercio, 2005 ; Xia et. al. ; 2003 ; Marrey and Sankar, 1997 ; Zhu et. al. 1998). The effective elastic properties are determined by performing six different analyses on the representative volume element (RVE) with unit non-zero strain component in each analysis and by imposing appropriate periodic boundary conditions on the boundaries of the RVE. The mesh is generated in such a way that the element boundaries conform to the matrix-inclusion interfaces. However, for complicated microstructures, automatic mesh generation becomes difficult and special 27

PAGE 28

techniques and algorithms need to be developed (Kim and Swan, 2003) to generate a quality mesh with quadrilateral or hexahedral elements. For example, in the analysis of composite microstructures involved in textile composites, the automatic mesh generation often becomes difficult and one of the main problems is generating a quality mesh in the interfacial region (Kim and Swan, 2003). This region is multiply connected and complications may arise for generation of mesh with identical nodes on the opposite faces of the RVE in order to impose periodic boundary conditions using multipoint constraints. Meshless methods have been extended for treatment of material discontinuity. Element Free Galerkin Method has been extended to incorporate discontinuous derivatives (Cai and Zhu, 2004 ; Cordes and Moran, 1996 ; Krongauz and Belytschko, 1998) for application to material discontinuity problems and Meshless Local Petrov-Galerkin Method has been extended to incorporate material discontinuity and for performing micromechanical analysis (Dang and Sankar, 2007). 2.6 Techniques for Local Refinement In many practical problems, there may be regions with stress concentrations and hence a uniform mesh/grid for the entire model can become very expensive computationally. In order to reduce the computational cost local mesh/grid refinement is necessary. In this section, a brief discussion on the local refinement techniques is presented. In the finite element analysis, local refinement is typically obtained by h-refinement or p-refinement. In h-refinement, nodes are inserted in the regions of stress concentration which increases the number of elements and reduces the element size in the region of interest while p-refinement is obtained by increasing the order of polynomial basis used for interpolation. Another type of refinement is termed as r-refinement, where the total number of nodes and elements in the refined model is same as the coarse model but the nodes are rearranged so that the number of nodes and elements in the regions of stress concentration is increased and k28

PAGE 29

refinement is obtained when the order of continuity of the basis functions is increased. In meshless methods, refinement is achieved by inserting nodes in the regions of stress concentrations. Hierarchical mesh refinement is studied by Krysl et. al. (2003) in the finite element method framework. In this method, the nodal basis functions on the coarse mesh are split into the basis functions of same order in the finer mesh. They have provided algorithms for refinement as well as un-refinement for finite element meshes. Hollig (2003) also presented a hierarchical refinement of B-spline basis functions, where the B-spline basis are split into the B-spline basis of same order but on a finer grid with half the size of the coarse grid. The hierarchical refinement provides a framework for getting a multi-level resolution, where the basis functions are split in a recursive manner. In this dissertation, a technique for local grid refinement is presented based on construction of solution structure using approximate step functions. 29

PAGE 30

CHAPTER 3 SOLUTION STRUCTURES FOR IMPLICIT BOUNDARY METHOD The essential boundary conditions are imposed in FEM by specifying the values at the boundary nodes on which essential boundary conditions are specified. When structured grids are used for discretizing the analysis domain, nodes are not guaranteed to be present on essential boundaries; hence special techniques need to be devised to be able to impose essential boundary conditions. In this chapter, a detail discussion on construction of solution structure for exact imposition of Dirichlet boundary conditions using the implicit equations of boundaries is presented. Approximate Heaviside step functions are used to construct a solution structure based on the technique proposed by Kantorovich and Krylov (Kantorovich and Krylov, 1958 and is termed as Implicit Boundary Method (Padmanabhan, 2006 ; Kumar et. al. 2007). Non homogenous essential boundary conditions are imposed by constructing piecewise continuous functions which satisfy the specified values at the essential boundaries and are termed as boundary value functions. This solution structure is used for modifying the weak form corresponding to linear elasticity problems and steady state heat transfer problems. Implicit Boundary Method has been extended in this dissertation to be able to use B-spline basis. A solution structure for treatment of material discontinuity based on the implicit equation of the material boundaries and approximate Heaviside step function is presented later in this chapter. Techniques for imposing periodic boundary conditions using multipoint constraints are also discussed in detail. In the final part of this chapter, a solution structure for adaptive refinement of grid based on the implicit equations of the grid element boundaries and approximate Heaviside step functions is discussed in detail. The weak form corresponding to linear elasticity is modified using this solution structure. 30

PAGE 31

3.1 Solution Structure for Enforcing Essential Boundary Conditions The essential boundary conditions are specified as prescribed displacements or prescribed temperatures on the essential boundaries. Let u be a trial function defined over that must satisfy the boundary condition 23or RR u a along u which is the essential boundary of Here, ucorresponds to a displacement component in linear elasticity and temperature in steady state heat transfer. If ( and ()0ax x :a R ) is the implicit equation of the boundary then the trial function can be defined as u ()()()()auUxxx ax (3-1) The solution structure for the trial function defined in Eq. (3-1) is then guaranteed to satisfy the condition u along the boundary defined by the implicit equation for any. The variable part of the solution structure is the function which can be replaced by a finite dimensional approximate function defined by piece-wise interpolation/approximation within elements of a structured grid. In the context of such solution structures, the function is often referred to as a weighting function. Since these functions assist in imposing the Dirichlet boundary conditions, we have referred to these functions as Dirichlet functions or just D-functions. D-function is constructed such that it takes a value of zero and has non-zero gradient on the essential boundary and has a value of one far from the essential boundary. Instead of directly using the implicit equation of the boundary as D-functions, approximate step functions constructed using these implicit equations have been used to construct solution structures in this work. The use of approximate step functions for construction of D-functions localizes them to the boundary elements and hence the stiffness matrix for all the internal matrices is identical as the D-function is unity inside them. In the a ()0ax :UR ()Ux ()hUx ()ax 31

PAGE 32

solution structure demonstrated in Figure 3-1, D represents the Dirichlet function which is zero at the first node of element e1 and reaches unity within this element and remains unity through out the domain. The boundary value function is defined by which takes on the value of the prescribed boundary value of at the boundary and au 0u g u represents the grid variable which is interpolated on the grid consisting of elements e1-e5. The approximate solution is given by g auDuu and it can be observed that the approximated solution satisfies the boundary condition exactly. The solution structure for linear elastostatic and steady state heat transfer problems is first discussed in this section for a general D-function and a detail discussion on construction D-functions is provided later in this chapter. 3.1.1 Modified Weak Form for Linear Elasticity The governing differential equation for an elastic body under equilibrium as shown in Figure 3-2 is given as: Solve: in .+b=0 Subject to: on 0u=u u (3-2) on 0.n=t t In the above equation, is the analysis domain bounded by essential boundary and natural boundary { is Cauchy stress tensor and { is the body force and the Dirichlet boundary conditions are displacements specified as on essential boundary and the natural boundary conditions are specified traction The principle of virtual work is the weak form for the boundary value problem and is given as: u t } }b 0u 0t tddTTTutub d (3-3) 32

PAGE 33

Here {is the virtual strain and { is virtual displacement. A trial solution structure for the displacement field can be constructed such that the Dirichlet boundary conditions are satisfied } }u g asuDu+uu+u a (3-4) In this solution structure, { } g u is a grid variable represented by piece-wise approximation over the grid, and { is the boundary value function which is a vector field constructed such that it takes on the prescribed values of displacements at boundaries. The variable part of the solution structure is {} }au []{} s guDu which satisfies the homogeneous boundary conditions. is a diagonal matrix whose components are Dirichlet functions that vanish on boundaries on which the component of displacement is specified, is the dimension of the problem. The same D-functions used to construct the trial solution are also used to construct the test functions. [](,..,)dindiagDDD iD thi dn g uDu (3-5) Using these solution structures, the strains and stresses can be expressed as the sum of a homogeneous part ( s ij s ij ) and a boundary value part ( aij aij ). ,,,,1122 s saasijijjiijjiijijuuuu a (3-6) () s asijijklklijklklklijijCC a (3-7) Substituting s=+ a in the weak form, we can rewrite it as tdddTTTTsautub d (3-8) 33

PAGE 34

The grid variable is approximated within each element as g gijuNu ij where are the shape functions and jN g iju are the nodal values of the grid variable for component of displacement. The virtual strain and stress within each element can then be expressed as and {}, where { and { represents the nodal values of the grid variable for the element e and [ is the strain-displacement matrix. Using these notations, the weak form can be expressed in a discrete form very similar to the finite element equations. thi {}[]{}eBX [][]{}se=CBX }T eX }eX ]B 111etNENENBEeeeeeeeeddTTTTXBCBXXFXNt (3-9a) iiedTTaFNbB d (3-9b) The summation is over the number of elements (NE) and the number of boundary elements (NBE) in the grid used for analysis. As in the finite element method these element matrices can be assembled to form global equations, where the left hand side terms form a stiffness matrix and the right hand side a force vector. Note that unlike in the finite element equations, there is a forcing term that arises from the boundary value functions. Furthermore, the nodal variables to be solved for are grid variables and not displacements. Since no assumptions have been made restricting the choice of shape functions used to represent the grid variables, this approach can be used with a variety of basis functions including Lagrange interpolation, B-spline basis functions or meshless approximations. 3.1.2 Modified Weak Form for Steady State Heat Transfer The governing differential equation for general steady state heat transfer problem shown in Figure 3-3 is given as 34

PAGE 35

Solve: T0inkf Subject to: TT on g (3-10) Tonhhkqn The corresponding weak form is given as: TTdTdTdTTTdhchkfqh (3-11) In the above equations, g is the portion of the boundary with prescribed temperature while is the portion with heat flux conditions and h c represents the portion with convection conditions and T is the ambient temperature. is the convection coefficient and is the thermal conductivity of the material which is assumed to be isotropic for simplicity and is the heat source. A trial solution structure for the temperature field is constructed such that the essential boundary conditions are satisfied h k f s agTTTDTT+ a (3-12) Proceeding in the similar manner as for the linear elastic problems discussed in the previous section and choosing the test function gTDT such that it vanishes on g the weak form modifies to the following: 0chcsssgsassasTkTdThTdTkTdfTdqTdhTTTd (3-13) The nodal degrees of freedom for temperature are approximated as gT g iiNT and the homogenous part of solution becomes sTD g iiNT where is the essential boundary function. The gradients of temperature which are used in the weak form are constructed as D s giijTBT j 35

PAGE 36

The weak form takes the following form when the above approximations are incorporated into Eq. (3-13) a0ddddTeceehcTTTggeeTTTeBkBTNhNTBkBTNfNqNh dTda (3-14) In the above expression, all the known quantities are moved to the right hand side and the unknown quantities are placed in the left hand side. The first term on the right hand side is the contribution to the load due to applied essential boundary conditions, the second term is the load due to the heat generation, the third term is load due to applied heat flux and the last term is load due to convection boundary conditions. The first term on the left hand side represents the stiffness matrix while the second term represents the stiffness due to convection boundary condition. 3.1.3 Dirichlet Function The Dirichlet function is a function which vanishes on all essential boundaries where component of displacement (or temperature) is prescribed. In addition, should have a non-vanishing gradient on the essential boundaries to ensure that the gradients of displacements are not constrained. Furthermore, it is necessary to ensure that is non-zero inside the analysis domain so that the solution is not constrained anywhere within the domain of analysis. Implicit equation of the curve or surface representing the boundary can be used for this purpose. However if multiple curves are involved then it is necessary to combine them into a single equation. Some authors have proposed the use of R-functions to construct the Boolean combination of implicit functions (Shapiro, 1998 ; Rvachev et. al., 2001). Hollig (2003) recommends constructing a weight function using distance functions as: iD thi iD iD ()1max(0,1(,)/)wxdistxD This function 36

PAGE 37

varies between 0 and 1 over a narrow strip of width near the boundary. Evaluating this function requires numerically computing the distance function. In this work, approximate step functions similar to the weighting function used by Hollig (Hollig, 2003) are proposed that can be constructed using any type of implicit function of the boundary. If the boundary condition is specified on a boundary with implicit equations then a step function at any given point x is defined as shown below 00()1101kD (3-15) This function is a shifted approximate step function that tends to the Heaviside step function in the limit as 0 but has a value of zero on the boundary instead of 0.5 as in traditional approximations of step functions (Osher, 2002). Moreover, for use as a Dirichlet function it is constructed to have non-zero gradient at 0 At this function has 1kC continuity. In this paper, the step function approximation of D with 0 is used as the Dirichlet function with values of 510 or smaller. When a single boundary passes through a boundary element the D function can be computed as in Eq. (3-15), but when multiple boundaries exist that have Dirichlet boundary conditions on them then it is necessary to construct D functions that are Boolean combinations of the individual step functions. The intersection, subtraction and union between two step functions 1D and 2D denoted as 12D 12D and 12 D respectively, can be defined as shown in Eq. (3-16). 37

PAGE 38

121DDD 2 (3-16a) 121DDD 2 (3-16b) 12121DDDDD 2 (3-16c) Figure 3-4 shows plots of D-functions created as a Boolean combination of two D-functions that represent circles. Note that the D-functions created by the Booleans inherit important properties including zero value at boundaries, non-zero slopes at boundaries and they plateau to a value of 1 for Since the step function defined in Eq. (3-15) has a value equal to zero at the boundary (and not 0.5 as in traditional approximations of step functions), the Boolean operations defined in Eq. (3-16) do not round off corners defined by the intersection of the implicit equations. If more than two boundaries exist within a boundary element, then the Dirichlet function is constructed using the Boolean tree similar to CSG tree as shown in Figure 3-5. 3.1.4 Boundary Value Function The boundary value function (displacement component for linear elasticity and temperature for steady state heat transfer) must be defined such that they have value equal to the imposed essential boundary conditions on the appropriate boundaries. If the essential boundary conditions are specified at multiple boundaries and the assigned values are not the same for all boundaries then it is necessary to construct a continuous function that takes on appropriate values at respective boundaries while transitioning smoothly in between. While there is no unique method for constructing boundary value functions, it is beneficial to ensure that the trial function is a polynomial of the same order as the shape functions used for interpolating the grid variable. This can be accomplished by constructing the boundary value function by piecewise au 38

PAGE 39

interpolation/approximation using the element basis functions. The nodal values of can be constructed similar to the grid variable au g u using the grid element basis functions. aiiiuN au (3-17) In the above equation, are the shape functions of the grid element and are the nodal values of It is important that the function should belong to the space spanned by the shape functions used to represent the grid variable. Otherwise, the grid variable has difficulty compensating for in the solution structure leading to poor convergence. Indeed, the solution structure will not satisfy completeness requirement, that is the ability to accurately represent constant strains, if basis functions used to construct does not span a sub-space of the basis functions that are used to represent the grid variable iN aiu au au au au g u If the basis functions in Eq. (3-17) are identical to the basis functions used for approximating the grid variable then this requirement is automatically satisfied. If solutions that have or continuity is desired then quadratic or cubic B-spline basis should be used in Eq. (3-17). The nodes of the boundary elements should be assigned nodal values of such that the resulting approximation will have the desired values at the boundary passing through the element. This is very easy to achieve when the assigned value is constant or even linearly varying along the boundary. When the specified boundary condition is a constant value of displacement along boundary, all the support nodes of the element have to be assigned value equal to the specified boundary condition. In the case of the 2D cubic B-spline element, the approximation within the element is controlled by 16 support nodes. Therefore, the value at all these nodes should be assigned to the same value to ensure that the resulting approximation has a constant value within the element. Similarly, if the specified boundary value has to vary linearly (as in hydrostatic pressure which varies linearly in the z-direction), the values iN 1C 2C au 39

PAGE 40

specified at all the support nodes should be on a plane corresponding to the linear distribution. This is illustrated in Figure 3-6, where a simple geometry is used to demonstrate the construction of the boundary value function. The left edge has a constant value of 10au while the right edge has a bilinear variation of The boundary value function is constructed by identifying the elements for the left boundary and specifying the nodal value of for all the support nodes, and then by identifying the boundary elements for right boundary and specifying the nodal values computed using 1 10(1)5(1)aux y 10 0(1)5(1) x y on the corresponding support nodes. The nodes which are not support nodes for the boundary elements are set to zero. The plot for the boundary value function is shown in Figure 3-6 and it can be observed that the boundary value function is continuous over the entire domain and it smoothly transitions from the specified value of 10 from the left boundary to zero in the middle of the domain and then smoothly transitioning from zero to the specified bilinear distribution along the right edge. Even though a simple geometry is used to illustrate the concept of constructing boundary value function, the procedure is quite general. However, care has to be taken to ensure that the grid has sufficient density to be able to construct the appropriate boundary value function. For more general variation of boundary values, the nodal values can be computed by least-square minimization. The rest of the nodes in the grid that are not part of any boundary element can have any arbitrary value and therefore can be set equal to zero. The boundary value function contributes to the load computation on the right hand side of Eq. (3-10) for all those elements in which the gradient of has non-zero magnitude. au au 3.2 Solution Structure for Micromechanics In the finite element method, a conforming mesh is created such that the element edges/faces form the material boundary. When structured grids as shown in Figure 3-7 are used 40

PAGE 41

for analysis, the nodes may not lie on the material interface and hence there is challenge in imposing interface conditions for material discontinuity. In Implicit Boundary Method, this is achieved by construction of a solution structure which ensures that interface conditions are satisfied for displacements. The implicit equations of the inclusion boundaries are used to construct a step function for inclusion such that it is unity inside the inclusion and smoothly blends to zero at the inclusion boundary. Such a function can be used to patch the solution from the inclusion grid and matrix grid to form a solution structure. Figure 3-8 shows the inclusion function in one-dimension and Figure 3-9 shows the inclusion function for an elliptic inclusion in a square matrix. It can be seen that the inclusion function is unity inside the inclusion and transitions to zero at the boundary of the inclusion, the rate with which this function reaches zero is specified by the parameter The expression for inclusion function is similar to the Dirichlet function described earlier in this chapter and shown below 00()1101kincH (3-18) This function inherits all the properties of Dirichlet function and is a shifted approximate step function that tends to the Heaviside step function in the limit as 0 but has a value of zero on the boundary unlike in traditional approximations of step functions where it has value of 0.5 Moreover, for allowing normal strain discontinuity, ()incH is constructed to have non-zero gradient at 0 At this function has 1kC continuity. In this work, the step function approximation of with and incH 2k 0 is used as Inclusion function with values of 210 41

PAGE 42

or smaller. The Inclusion function is used to construct the solution structure for material discontinuity as shown in Eq. (3-19). (1)incincHHg1g2uu u (3-19) This solution structures combines the solution from both grids to represent the solution in the analysis domain. When the Inclusion function is unity, the solution is given by which is the solution from the inclusion part and when the Inclusion function is zero, the solution is given by which is the solution from the matrix. In the region where the inclusion function varies from unity to zero, the solution is a blend of the solutions from matrix grid and inclusion grid. This way of constructing solution structure ensures the displacement continuity of the solution through out the analysis domain. This solution structure allows the normal strain discontinuity; this can be seen from the gradients of the solution structure as shown below g2u=u g1u=u 121(1)ggincincincgincgiiiijjjjuuuHHuHxxxxx 2ijHu (3-20) In this expression, the first term: 1(1) g incijuH x and third term 2 g incijuH x are continuous at the interface boundary while the second term: 1inc g ijHux and fourth term 2inc g ijHux are discontinuous at the interface boundary and vanish in the matrix region and are non-zero in the inclusion region. Therefore these terms provide independent slopes at both sides of the interface resulting in discontinuous normal strains. Modified Weak Form. The weak form of the elastostatic boundary value problem is the principle of virtual work which can be written in the following form as shown in Eq. (3-21). Here {is the virtual strain, { is Cauchy stress tensor, {} is the traction vector, { is } } t }u 42

PAGE 43

virtual displacement and { is the body force. For elastostatic problems, the essential boundary conditions are displacements {} specified on }b {}0u=u u and the natural boundary conditions are traction {} specified on {}0t=t t tddTTTutub d (3-21) When the solution structure shown in Eq. (3-19) is used and the displacements are interpolated as 11[]{ggiiNug1u= } with 1incHg1N N in grid corresponding to matrix and 22[]{ggiiNug2u= } with incHg2N N in inclusion grid, the expression for strains takes the following form 111222121221uxuxuuxx g 112 g 2uBBu (3-22) In this expression, 21BB is the modified strain displacement matrix and is a combination of strain displacement matrices from the two grids. The expression for stress is given as iijC j where is the elasticity stiffness matrix, and in computation of stress, the inclusion (or matrix) elasticity stiffness matrix is used when the strains are evaluated in the inclusion (or matrix) region. The modified weak form takes the following discrete form when the above expressions for stress and strain are incorporated into the weak form (Eq. 3-23). ijC 22111etNENEeeeeeNBEeeddd TTT2T111TT21XBBCBBXXNNbXNNt (3-23) 43

PAGE 44

In this expression Teg1g2X=uu and the summation is over the number of elements (NE) and the number of boundary elements (NBE) in the grid used for analysis. As in the finite element method these element matrices can be assembled to form global equations, where the left hand side terms form a stiffness matrix and the right hand side a force vector. The boundary conditions arising in micromechanical analysis are periodic boundary conditions which can be imposed in this method similar to FEM as the RVE is always a rectangle/cuboid and the nodes are always guaranteed to be on the boundaries of the unit cell. The effective properties of the composite microstructure are computed by performing six different analyses on the RVE subjected to different unit strains in each case. The average values of stresses and strains in the RVE are computed for each case and are then used to determine the effective properties. A detailed discussion on the periodic boundary conditions and computation of effective properties is presented in chapter 5. 3.3 Solution Structure for Grid Refinement In many practical problems local refinement is needed to capture the localized behavior of the solution (e.g.: stress concentrations). A uniform grid for the entire model is not suitable for this purpose. In order to be able to refine a grid at required locations, a solution structure based on approximate step function is presented in this chapter. The approximate step function blends the solutions from the coarse grid and refined grid such that the continuity of the solution across the refined grid and coarse grid interface is maintained. When higher order continuous approximations are used for approximating the field variable, a suitable step function needs to be used to maintain the continuity of the solution across the refine grid and course grid interface. In Figure 3-10, the coarse grid is represented by g1 and the locally refined grid is represented by g2, 44

PAGE 45

it is to be noted that these grids g1 and g2 are conceptual grids and analysis is performed on the grid g. 3.3.1 Solution Structure The approximate step function that is used to blend the solutions from the coarse and the refined grids is constructed by using the implicit equations of the coarse/refined grid interface. This step function for refinement is defined as: 0012022()1212221krefkH (3-24) In the above equation, is the implicit equation of the coarse/refined grid interface, is the range parameter which defines the region in which the step function for refinement varies from zero to one, and is an integer which defines the order of the step function used. The step function for refinement is such that it takes on a value of zero at k 0 12 at 2 and one at It has a zero slope and continuity at both 1kC 0 and Hence, when B-spline basis is used to approximate the solution, an appropriate needs to be chosen so as to maintain the continuity of the solution across the coarse/refined grid interface. The plot of and is shown in Figure 3-11, and it can be seen that the above mentioned properties are satisfied by the step function of refinement. k refH 1refH The above definition for is valid only when the coarse/refined grid interface is defined by a single implicit equation. However, in many practical cases the interface is defined by refH 45

PAGE 46

multiple implicit equations as shown in Figure 3-12. In order to define ref H for multiple implicit equations, a Boolean tree as shown in Figure 3-12b is used to represent the step function for refinement and the Boolean operations used to combine two step functions 1refH and 2refH denoted as and respectively are be defined as 12refH 12refH 12refH 121refrefrefHHH 2 (3-25a) 121refrefrefHHH 2 (3-25b) 12121refrefrefrefrefHHHH 2H (3-25c) The step function for refinement for the model shown in Figure 3-12 is shown in Figure 3-13. This step function is constructed by using the Boolean operations as defined in Eq. (3-32). The solution structure for the displacements is represented in the following form: (1)refrefHH1uu 2ua2a2 (3-26) In the above expression, is the step function for the coarse and refined grid interface and it vanishes outside the refined grid and transitions from zero to unity inside the refined grid. The transition from zero to unity takes place only across the boundary of the coarse/refined grid interface. In Figure 3-10, the refined step functionvaries from zero to unity across the edges represented by AB and BC as these edges are shared between the refined elements of grid and unrefined elements of grid .The refined step function is unity and hence does not vary across the edges AF, FE, ED and DC as these edges are not shared between the refined and unrefined grids. The expressions for and have solution structure as proposed for imposition of essential boundary conditions and can be written as and. refH refH 2g 1g 1u 2u 1g1u=Du+u 2g2u=Du+u 46

PAGE 47

3.3.2 Modified Weak Form The solution structure for grid refinement using the step function for refinement, ref H can be written as shown in the previous section and the virtual displacements are written as: 1refrefHH g 1 g 2uDuD u (3-27) The grid solutions in both grids are approximated by g 11eu=NX 1 with 1refH1NN and g 22eu=NX 2 with refH2N N where represents the shape functions for grid variables. The boundary values are approximated N a1a1a1u=NX with 1refHa1NN and a2a2a2u=NX with refHa2NN where represents the basis functions used to approximate the boundary value function. Using the above approximations, the expressions for strain takes the following forms: N 1112222121221uxuxuuxx e1a11a1a2e2a2XBBBBXX Xj (3-28) In the above expression, and are modified strain displacement matrices which contain the terms involving the gradients of Dirichlet functions and the gradients of the shape functions used to approximate the grid solution and,are strain displacement matrices which contain the gradients of the shape functions used to construct boundary value functions. The expression for stress is given as 1B 2B refH a1B a2B iijC where is the elasticity stiffness matrix, and in computation of stress. The modified weak form takes the following discrete form when the above expressions for stress and strain are incorporated into the weak form (Eq. 3-29). ijC 47

PAGE 48

22111etNENEeeeeNBEeedd ee TTT11TT12XBBCBBXXFXNNt (3-29a) iiedTT12aa1a2FNNbBB d (3-29b) In this expression Teg1g2X=uu and the summation is over the number of elements (NE) and the number of boundary elements (NBE) in the grid used for analysis. As in the finite element method these element matrices can be assembled to form global equations, where the left hand side terms form a stiffness matrix and the right hand side a force vector. The Dirichlet function and the boundary value function are constructed as explained in sections 3.1.3 and 3.1.4. In this chapter, modified weak forms for the solution structures corresponding to enforcing essential boundary conditions, treatment of material discontinuity and local refinement of grid are presented in detail. These modified weak forms are used in finite element implementation for solving several engineering boundary value problems. In the next chapter, a detail discussion on the construction of B-spline based finite elements is presented. 48

PAGE 49

Figure 3-1. Solution structure for imposing essential boundary conditions Figure 3-2. Elastic solid under equilibrium 49

PAGE 50

Figure 3-3. Steady state heat transfer problem Figure 3-4. Boolean operations on Dirichlet functions 50

PAGE 51

Figure 3-5. A CSG tree representing Dirichlet function for multiple boundaries 51

PAGE 52

Figure 3-6. Boundary value function using cubic B-spline basis 52

PAGE 53

Figure 3-7. Representation of grids for material discontinuity Figure 3-8. One dimensional inclusion function for treatment of material discontinuity 53

PAGE 54

Figure 3-9. Two dimensional inclusion function for an elliptic inclusion in a square matrix Figure 3-10. Representation of solution structure for local grid refinement 54

PAGE 55

Figure 3-11. Approximate step function for refinement Figure 3-12. Coarse and refined grid interface with multiple implicit equations 55

PAGE 56

Figure 3-13. Step function for refinement for a 2D model 56

PAGE 57

CHAPTER 4 B-SPLINE FINITE ELEMENT FORMULATION B-splines provide a methodology to approximate a given set of points with smooth polynomial functions. The basis functions are represented by piecewise polynomials and the approximation provides a high degree of continuity depending on the basis used. For example, continuity is obtained when quadratic basis is used and continuity is obtained when cubic basis is used and so on. It is to be noted that the Lagrange interpolations which are used in traditional FEM are continuous irrespective of the basis used. B-spline basis functions are classified into uniform, non-uniform and NURBS. Uniform B-spline basis functions have equally spaced nodes in the parameter space and the basis functions are periodic. Adjacent nodes in non-uniform B-splines are not equidistant in the parameter space. NURBS are generalization of non-uniform B-splines, where the basis functions are rational functions and are capable of representing several analytical curves exactly. In this work, uniform B-splines are used for approximating the solution space. The discussion on B-splines in the remaining of this chapter is limited to uniform B-splines. 1C 2C 0C The B-spline basis functions have compact support and lead to banded stiffness matrices. B-spline approximation provides high order of continuity and is capable of providing accurate solutions with continuous gradients through out the domain. This is a major advantage over standard finite element shape functions. The B-spline basis functions form a partition of unity and are able to accurately represent the state of constant field in the domain of interest. This property is very important for convergence of the approximate solutions provided by B-splines. The basis functions used are polynomials and accurate integration can be performed by using Gauss quadrature. The B-spline approximation has good reproducing properties and when a 57

PAGE 58

cubic B-splines are used for approximating the solution, all the polynomial solutions up to third order can be represented exactly. B-spline shape (basis) functions are traditionally constructed using recursive definition (Farin, 2002). The parameter space is partitioned into elements using a knot vector (equivalent to a collection of nodes). Very general methods are available to insert knots and elevate the order (or degree) of the polynomial basis functions. This framework therefore can be used to implement h-pand k-refinement and has been studied in the context of isogeometric analysis by Hughes et al (Hughes, 2005). Instead of using recursive definitions as in geometric modeling applications, polynomial definitions for B-spline basis functions are used here. Furthermore, it is convenient to define an independent parameter space for each element in the grid instead of a continuous parameter space (or knot vector) for all elements. Therefore it is assumed here that for each element in the grid, parameter values vary within [1as is done for traditional isoparametric finite elements. This is illustrated for the one-dimensional case in Figure 4-1 where a cubic B-spline approximation is shown. In Figure 4-1, element is defined between vertex nodes 2 and 3, but the approximation defined over (referred to as its span) is controlled by four nodes (1-4) which can be refer to as its support nodes. ,1] 4.1 One Dimensional B-Spline Elements The polynomial expressions for the basis functions can be derived using the continuity requirements between adjacent elements. A order B-spline has thk 1k support nodes with some of the support nodes lying outside the element. The span corresponding to the element is defined by the basis functions of all the support nodes. The basis functions are polynomials of the form whose coefficients are determined using continuity requirements. Since the 0kjiijjNa r ija 58

PAGE 59

parameter range for the element is [1,1]r the parameter value for first support node is 12kr and for the ( support node is 1)thk 12kr for the parameterization to be uniform. The approximated function shown in Fig 4-1, is a linear combination of the basis functions within each element so that the span of the element e is represented as 000()kkkjeiiijiij i f rNuar u (4-1) In the above equation, are the support nodal values and it is assumed for convenience that the first support node for element has index 0. B-spline approximation of order has iu e k 1kC continuity at the junction between adjacent elements. This continuity requirement can be stated as 1(1)(1)mmeemmffrr 0,1...1k ,m (4-2) where, e and e+1 are adjacent elements whose derivatives are evaluated at the end point (r=-1) and start point (r=1)respectively. 0m represents the continuity equation Using Eq. (4-2), the continuity equations can be expressed as 1(1)(1)eeff 000100110!!!(1)()!()!()!!(1)0()!kkkkjmjijijijjkjmkjkjjjjauaaujmjmjmjaujm ji1 (4-3) Since this equation is valid for arbitrary values of ,0..iuik the coefficients corresponding to each should vanish. This results in a set of linearly independent equations. Since equations are obtained for each value of a total of linearly independent equations are obtained. An additional equation stating that ,0..iuik 1) 2k 2k 0,1,..1mk (2kk 59

PAGE 60

the basis functions form a partition of unity 01kiiN is used to obtain 2(1)k equations needed to solve for all coefficients of the basis functions. Using this approach, the basis functions for quadratic and cubic B-splines can be derived as shown in Eq. (4-4) and the basis functions are plotted in Figure 4-2. 1k Quadratic B-spline element. The shape functions for quadratic B-spline element are constructed using the procedure explained above. Quadratic B-spline element in one-dimension has three nodes and hence three shape functions. The plots of the shape functions are shown in Figure 4-2 and the corresponding equations are shown below: 22123111(12);(62);(12)888NrrNrNr 2r (4-4) Cubic B-spline element. The plots of basis functions for cubic B-spline are shown in Figure 4-3 where it can be seen that, unlike traditional finite element shape functions, the B-spline basis functions are not unity at the corresponding node and do not vanish at other nodes. Therefore, these basis functions do not satisfy the Kronecker-delta property and the approximation constructed using these basis functions does not interpolate nodal values. 23231223233411(133);(231533);484811(231533);(133)4848NrrrNrrNrrrNrr rr (4-5) 4.2 Two and Three Dimensional B-Spline Elements The basis functions for the higher dimensional B-spline elements are constructed by taking product of the basis functions for one-dimensional B-splines. A typical grid used for analysis is illustrated in Figure 4-4. This grid corresponds to cubic B-spline elements where the span corresponding to each element is controlled by 16 support nodes. The grid elements used here are regular quadrilaterals (rectangle/square) or regular hexahedra (cube/cuboid), hence the 60

PAGE 61

mapping for geometry of elements from parameter space to the physical space is linear. Note that even if an iso-parametric formulation is used the mapping will degenerate to a linear map because the element geometry in physical space is regular and can be obtained by scaling the element geometry in parametric space. In the following sub-sections, detailed discussion on higher dimensional quadratic and cubic B-spline elements is presented. Quadratic B-Spline elements. The basis functions for the bi-quadratic B-spline element shown in Figure 4-5 can be constructed using the product of the one-dimensional quadratic basis functions (expressed in terms of parametric coordinates (,) ). 23(1)(,)()();1..3;1..3DjiijNNNij (4-6) The span is defined by 9 support nodes, as shown in Figure 4-5(a), ,but none of these are at the vertices of the element defined in the parameter space as the square [1 ,1][1,1] It is advantageous to have nodes at the vertices of the element. In order to accomplish this, the quadratic B-spline finite element was defined as shown in Figure 4-5(b) as an assembly of four quadrants (Q1-Q4). This representation facilitates implementation because each element will have nodes at its four vertices. The mapping of the geometry of element from the parameter space to the physical space is linear, as explained earlier, and can be represented as 1122liiirr uii x x x (4-7) Equation (4-7) is written using index notation, where li x and ui x are the lower and upper bounds respectively for the nodal coordinates of the element. For each quadrant, only 9-nodes (and hence 9 basis functions) are active. For example, the active nodes of the first quadrant Q1 are based on the node numbering shown in Figure 4-5(b). The basis functions obtained by product of one-dimensional basis functions will have a local 6,7,8,10,11,12,14,15 and 16 61

PAGE 62

coordinate system (,) with origin at node 11 for Q1. In order to use a parametric coordinate with origin at the center of the element, the (, (,)rs ) coordinate system needs to be translated for each quadrant as 1:1;12:1;13:1;14:1;1QrsQrsQrsQrs (4-8) The basis functions at any given point are computed by first determining which quadrant the point lies in and then by determining the corresponding support nodes and basis functions. As an example, the basis functions for a point in Q1 are (,)rs 2261172122831101222112212322214131523(,)(1)(1);(,)(1)(1)(,)(1)(1);(,)(1)(1)(,)(1)(1);(,)(1)(1)(,)(1)(1);(,)(1)(1DDDDDDDDNrsNrNsNrsNrNsNrsNrNsNrsNrNsNrsNrNsNrsNrNsNrsNrNsNrsNrNs22216331222222345913)(,)(1)(1);(,)(,)0(,)(,)(,)(,)(,)0DDDDDDDDNrsNrNsNrsNrsNrsNrsNrsNrsNrs (4-9) 2(,)DiNrs refers to bi-quadratic basis function corresponding to node of the element and represents the one-dimensional quadratic basis function as defined in Eq. (4-4). thi ()jNr Three dimensional quadratic B-spline element can be constructed in a similar manner and has eight octants which are assembled to construct the 64-node element. The basis functions are constructed using the product of 1D basis functions and in any given octant, only 27 basis functions are active. The mapping for the geometry is done as shown in Eq. (4-7). Cubic B-Spline elements. The two dimensional cubic B-spline elements are constructed by product of one-dimensional cubic B-splines. In this case, there are 16 support nodes as shown in Figure 4-6 and there is a node at each vertex of the element ([1 ,1][1,1] in parametric space). 62

PAGE 63

24(1)(,)()();1..4;1..4DjiijNrsNrNsij (4-10) 316(1)4(1)(,,)()()();1..4;1..4;1..4DkjiijkNrstNrNsNtijk (4-11) The geometry mapping between parametric and physical space can be defined using Eq. (4-7). The lower bounding node for this element is node 6 and the upper bounding point corresponds to the node 11. Again, some of the support nodes lie outside the domain of the element and are shared with neighboring elements. The 2-D bi-cubic B-spline element has 16 basis functions (Eq. 4-10) while the 3D cubic B-spline has 64 basis functions (Eq. 4-11). In these equations, represents the cubic basis functions as defined in Eq. (4-5) ()jNr 63

PAGE 64

Figure 4-1. Approximation using 1D cubic B-splines Figure 4-2. Shape functions for 1D quadratic B-spline element Figure 4-3. Shape functions for 1D cubic B-spline element 64

PAGE 65

Figure 4-4. Typical grid used for analysis and the mapping used for geometry and field variable 65

PAGE 66

Figure 4-5. Two dimensional quadratic B-spline elements in parametric space Figure 4-6. Two dimensional cubic B-spline elements in parametric space 66

PAGE 67

CHAPTER 5 NUMERICAL IMPLEMENTATIONS The weak form was modified for various solution structures and was discussed in a previous chapter. In this chapter, a detailed discussion on the efficient and accurate computation of the terms involved in the modified weak form is presented. In the first part of this chapter, a discussion on the computation of stiffness matrix and load vector corresponding to the solution structure for imposing essential boundary conditions is presented. In the next part of this chapter, a discussion on the construction of stiffness matrices involved in the solution structure for micromechanical analysis is presented. Finally, a discussion on the computation of various terms involved in the solution structure for adaptive refinement is presented. 5.1 Numerical Implementations for Linear Elastic Problems The modified weak form in discrete form for linear elasticity was presented in a previous chapter and was shown to be: 111etNENENBEeeeeeeeeddTTTTXBCBXXFXNt T (5-1a) iiedTTaFNbB d (5-1b) The stiffness matrix is given by the expression ed TBCB which needs to be computed for all the elements present in the structured grid. A typical structured grid used for analysis is shown in Figure 5-1. The grid elements can be classified into three types and the computation of stiffness matrix depends on the type of element that is being integrated upon. In the following, a discussion on the computation of stiffness matrix in each type of element is presented. 67

PAGE 68

5.1.1 Computation of Stiffness Matrix for Internal Elements The internal elements lie completely inside the analysis domain and the strain displacement relation for two dimensional problems can be written as follows: B 11132211210..0..mNxNxNNxxB .. (5-2) In the above equation, are the shape functions used to approximate the displacement solution and is the number of nodes for the element. The stiffness matrix is computed using Gauss Quadrature as shown below: iN m 2211(,)(,)yxnnijijijmmjiww TK=BCB (5-3) In the above equation, are the number of Gauss points in x and y directions respectively, is the Gaussian weight associated to the Gauss point ,xynn iw i at which the integrand is evaluated. It is to be noted that the stiffness matrix for all the internal elements is identical and stiffness matrix is computed only once and is used for all the internal elements. This leads to a significant savings in computation cost associated with computation of stiffness matrices when compared to FEM, where stiffness matrix is computed for all the elements. 5.1.2 Computation of Stiffness Matrix for Boundary Elements The boundary elements lie partly in the analysis domain; this leads to partial elements and hence needs special techniques for computation of stiffness matrix. The strain-displacement matrix for these elements is identical to the strain-displacement matrix for internal elements as 68

PAGE 69

shown in Eq. (5-2). To compute the stiffness matrix, the partial solid region is divided into triangles as shown in Figure 5-2. The stiffness matrix is computed by adding the contribution from all the triangles generated by triangulation as shown below: 221tnmmiiK=K (5-4) In this equation, is the number of total number of triangles and tn iK is the stiffness matrix computed using Gauss Quadrature for triangle. Several algorithms are available for automatic triangulation of a given set of points on a plane. In this work, Delaunay triangulation algorithm is used for triangulation. It is to be noted that curved edges are approximated by straight lines which can leads to error in the computation of stiffness matrix. However, this error is limited to boundary elements and the errors incurred due to this approximation are similar to the errors involved in FEM where the boundaries are approximated with conforming meshes. Furthermore, when the grid is made finer, the errors will approach zero. In three dimensions, the stiffness matrix is computed by tetrahedralizing the solid region and the errors involved in this approach are same as the ones involved in triangulation in two dimensions. thi 5.1.3 Computation of Stiffness Matrix for Elements with Essential Boundaries For the boundary elements containing essential boundaries, the Dirichlet functions are active and hence the corresponding strain-displacement matrix has gradients of Dirichlet function. Since Dirichlet functions are defined on a narrow band with 0 their gradients can have large values in the small band. In order to accurately compute the contribution due to Dirichlet function and its gradient, special techniques are required. In this section, one such technique is presented and discussed in detail. The strain-displacement matrix can be written as: 69

PAGE 70

111111123221221112112122110.[]0....mNDDNxxNDDNxxNDNDDNDNxxxxB (5-5) The above equation shows that the strain-displacement matrix has gradients of both shape functions and Dirichlet functions. It is convenient to decompose the B matrix into two parts such that contains terms involving the gradient of the shape functions and contains terms with the gradients of the D function as shown below: 1[][][]BBB 2] 1[]B 2[B 1111111211113222344212122111212110..0..0..000[]0..0000..00..0..mmNxNDNxDxNDDNxDDxNNDDNxxxBD B (5-6) 1111112222321322212212121121210..00..[]0..00....mmDDNxxNDDNNxxDDDDNNxxxx2BD B (5-7) Using the above procedure, the strain-displacement matrix can be rewritten as 1122BDBDB Similar matrices can be defined for three dimensional elements also. Using the decomposition of strain displacement matrix, the stiffness matrix is decomposed into the following components: 70

PAGE 71

eeeeeVVVVdVdVdVdVTTTT111221221223TKBCBBCBBCBBCBK+K+K+K (5-8) The first component 1K is computed as described earlier by decomposing the boundary elements into triangles or tetrahedra to perform the area or volume integrals respectively. The second and third components contain gradients of Dirichlet functions which are large near the boundary when is made very small and they vanish away from the boundary, hence the terms in the integrals will be non-zero in a thin band along the boundary. This allows us to convert the volume integral into a surface integral. The contribution from the gradients of finite element shape functions can be assumed to be constant across the band but varies along the boundary and the contribution from the gradients of Dirichlet function are constant along the boundary but vary normal to the boundary. The band is represented in terms of two parameters as shown in the Figure5-3 and the terms in stiffness matrix take the following form. ,tn The stiffness matrix as decomposed in Eq. (5-8) has four components; the second component 2K can be further decomposed as shown in the following equation. 2221ETTTVTdVdddd1121212KBCBBDCDBBCB (5-9) In this expression, the terms involving only Dirichlet functions and their gradients are separated from the terms containing the basis functions and their gradients. This decomposition allows us to compute the stiffness matrix efficiently and accurately. The volume integral has been decomposed into a line integral along the normal to the boundary and a surface integral along the boundary. The normal for the boundary represented by implicit equation 0 is given 71

PAGE 72

by at the boundary. The component involving only Dirichlet function and its gradients is termed as 2C in Eq. (5-9) and this needs to be integrated only across the band (along the normal to boundary), which can be done analytically as shown in Eq. (5-10). In computation of the integral across the band, it is assumed that the gradients of the implicit function with respect to the parametric coordinates of the element are constant; this is a valid assumption as the integration is performed in a very thin band 510 and hence the gradients remain approximately constant. 1211111211121211122212212222122212111221233133133222211223323321DD11DCDCCCxx22DD11DCDCCCxx2211DD1CCDCDC2xxDDDCDCxx 2 J sJ s J sJddJs s 2C 3311133332221111211CC22 J s J sJ s (5-10) In the above equation, the terms ii J denote the components of the Jacobian matrix of the transformation described in Eq. (4.7) which transforms the element from parametric space to the physical space 123(,,)sss 123(,,) x xx It is to be noted that the Jacobian matrix is diagonal, as the grid elements are regular quadrilaterals or hexahedra. Proceeding in the similar manner as explained above, the third component of the decomposed stiffness matrix in Eq. (5-8) can be written as shown in the following equation 31ETTTVTdVdddd222232222KBCBBDCDBBCB (5-11) 72

PAGE 73

The components involving only the gradients of Dirichlet function is termed as 3C and its integral across the band is given as 221133123322121122121122322123322332211221212112214141433311414133CCCCssJJssJJdCCCCJJssssJJ 43 C (5-12) The computation of the integrals along the boundary is performed numerically by Gauss quadrature. This approach leads to accurate and efficient computation of stiffness matrix for boundary elements with essential boundaries. 5.1.4 Computation of Load Vector The right hand side of the modified weak form as shown in Eq(5-1) has two terms, the first term is a volume integration for the body loads and applied non-homogenous essential boundary conditions which is done similar to the computation of stiffness matrix as explained in the previous sections. The second term is a surface integration over the natural boundaries. The boundary is approximated as a set of line segments in 2D as shown in Figure 5-4 and surface triangles in 3D and integration is performed using Gauss Quadrature as shown below: ii211211()()LilniimiLnimidwTTfNtNtf=f (5-13) As the grid is made finer, the approximation of the boundary approaches the exact representation and the errors associated with approximation of boundaries with lines (planes) approach zero. In traditional FEM, the natural boundary is approximated by the surface formed by the nodes on the boundary and associated shape functions. This results in a load vector which has non-zero values corresponding to the nodes which are present on the surface. However, in 73

PAGE 74

PAGE 75

Computation of Stiffness Matrix. The stiffness matrix comprises of a volume integral and a surface integral. The volume integral is performed as explained in section 5.1.1 and 5.1.2 after subdividing the boundary elements into triangles or tetrahedra. The surface integral for load computation is evaluated as explained in section 5.1.4. For the boundary elements with essential boundaries, the [B matrix can be decomposed as shown below: ] 1111211220.[]0.mNDDNxxNDDNxxB ..2 (5-15) In the above equation, [Bmatrix contains the terms involving gradients of shape functions and also the gradients of D function. It is advantageous to decompose this matrix further to such that contains terms involving the gradient of the shape functions and contains terms with the gradients of the D function as shown below: ] 1[][][]BBB 2[]B 111112211220..0..[]0..0..mmNNDxxDNNDxx D BB (5-16) 11111222212221221211210..00..[]0..0..0..mmDNDxxNDNNDxxDDNNxx 2BD B (5-17) Using the above procedure, the gradient matrix can be rewritten as B 1DB 22DB Similar matrices can be defined for 3D elements also. Using this decomposition, the stiffness matrix is decomposed into the following components: 75

PAGE 76

eeeeeVVVVkdVkdVkdVkdTTTT111221221223TKBBBBBBBBK+K+K+K V (5-18) The first component 1K is computed as described earlier by decomposing the boundary elements into triangles or tetrahedra to perform the area or volume integrals respectively. The second and third components contain gradients of Dirichlet functions which are large near the boundary when is made very small and they vanish away from the boundary, hence the terms in the integrals will be non-zero in a thin band along the boundary. This allows us to convert the volume integral into a surface integral as explained in detail in section 5.1.3. The second term in Eq. (5-18) can be written as: 2221ETTTVTkdVkdddd1121212KBBBDDBBCB (5-19) This term is evaluated as a surface integral as explained in section 5.1.3. 5.3 Numerical Implementations for Micromechanics The modified weak form for the solution structure corresponding to micromechanics was discussed in a previous chapter and was shown to be: 22111etNENEeeeeeNBEeeddd TTT2T111TT21XBBCBBXXNNbXNNt (5-20) Teg1g2X=uu (5-21) In this section, a detailed discussion on the numerical implementations for the computation of the terms involved in this weak form is presented. The left hand side of the weak form 76

PAGE 77

corresponds to the stiffness matrix while the right hand side corresponds to the load vector. The solution structure for micro-mechanics is such that the solution from the matrix grid inside the inclusion is zero due to the step function of inclusion and similarly the solution from inclusion grid outside of the inclusion vanishes. Therefore the nodal values of the matrix grid inside the inclusion are set to zero, provided these nodes are outside the influence of inc H and the nodal values of the inclusion grid outside the inclusion are set to zero. This leads to the reduction in the total number of degrees of freedom for the analysis model. The finite element solution for the grid variables are written as and For two dimensional case, the strain vector using the strain displacement relations is written as shown below: 11[]{ggiiNug1u= }} 22[]{ggiiNug2u= 111222121221uxuxuuxx 2 g11g2uBBu (5-22) The components of the strain displacement matrix are expressed as: 12B,B 1111112211112211(1)0..0(1)(1)(1)..gincincgiigincincgiiggincincincgincgiiiiNHHNxxNHHNxxNNHHHNHNxxxx1B .. (5-23a) 77

PAGE 78

22111222212222110.0..gincincgiigincincgiiggincincincgincgiiiiNHHNxxNHHNxxNNHHHNHNxxxx2B ... (5-23b) In the above equations, the decomposed components of strain-displacement matrix have terms involving gradients of finite element shape functions g kijN x and the gradients of the inclusion function incjH x It will be advantageous to further split these components into a term containing gradients of shape function only and another term containing gradients of inclusion function only. This is accomplished as shown in the following equations. 1111112211112121(1)0..0.0(1)..0...(1)(1)..gincincgiigincincgiiggincincggincinciiiiNHHNxxNHHNxxNNHHNNHHxxxx .. 11112B=B+B= (5-24) 22111222212221210..0.0..0....gincincgiigincincgiiggincincggincinciiiiNHHNxxNHHxxNNHHNNHHxxxx ...N 22122B=B+B= (5-25) It can be noticed that the modified strain displacement matrices ,,and vanish in certain regions of the computational domain and are non zero in other parts of computational domain depending on where the inclusion function and its gradients vanish. 11B 12B 21B 22B 78

PAGE 79

Therefore the computational domain is divided into three regions as shown in Figure 5-5 and they are explained in detail in the following. 5.3.1 Computation of Stiffness Matrix in the Matrix Region In the matrix region of the microstructure, the degrees of freedom for grid corresponding to matrix are active, therefore the strain-displacement matrix as shown in Eq. (5-23) becomes 2111B=BB=B0 with 0incH in Eqs. (5-24) and (5-25). The stiffness matrix will take the following form in matrix region: 22eedV TmatTmat111111BCB0K=BBCBBdV=00 (5-26) This stiffness matrix is computed by performing the volume integration over the elements in matrix region. The grid element in the matrix region will be an internal element and the volume integration is performed as discussed in the previous section 5.1.1. 5.3.2 Computation of Stiffness Matrix in the Inclusion Region This region corresponds to the inclusion region. In this region the degrees of freedom for grid corresponding to inclusion are active, therefore the strain-displacement matrix as shown in Eq. (5-23) becomes 212 1 B= with BB=0B1incH in Eqs. (5-24) and (5-25). The stiffness matrix takes the following form: 22eedV Tinc11Tinc212100K=BBCBBdV=0BCB (5-27) The stiffness matrix is computed by performing the volume integration over the elements in inclusion region. The grid element in this region will be an internal element and the volume integration is performed as discussed in the previous sections 5.1.1. 79

PAGE 80

5.3.3 Computation of Stiffness Matrix in the Interface Region In this region, the step function for the inclusion is active. This means the step function is varying in this region and strain-displacement matrix as shown in Eq. (5-23) becomes 21B=BB The stiffness matrix is expressed as: 22eedV TTT1112111211TT21222122KKBCBBCBK=BBCBBdV=KKBCBBCB (5-28) In this expression depending on which region is being integrated upon. Each of the terms in this matrix can be decomposed as a sum of volume integral and a surface integral. The procedure for this decomposition is described in detail for the first term in the stiffness matrix. orincmatC=CC 11K 2222eeedVdVTT111111121112TTTT111111111111K=BCBdV=B+BCB+B=BCB+BCB+BCB+BCB (5-29) In this expression, the first term is integrated as a volume integral after the partial elements are subdivided into triangles or tetrahedral as explained in section 5.1.2. The remaining three terms are converted into surface integrals, as these terms are non-zero in a thin band in which the inclusion function is non-zero. The strain-displacement matrices shown in Eq. (5-25) are rewritten in the following form: 11111122111122220..0..0..0......gginciiggincincinciiggggincinciiiiNNHxxNNHHxxNNNNHHxxxx2121BB H (5-30) 80

PAGE 81

2112222222221210..00..0..00....incincgigincincgiigiincincincincggiiHHNxxNHHNxxNHHHHNNxxxx2222BH B (5-31) Using the above procedure, the strain-displacement matrix can be rewritten as 2incH21222BBH B Similarly, the first component of strain-displacement matrix is written as 11incH11111BB HB These matrices can be defined for 3D elements also in a similar manner. The last three components in stiffness matrix as shown in Eq. (5-29) contain gradients of inclusion functions which are large near the inclusion boundary when is made very small and they vanish away from the boundary. Therefore the terms in these integrals will be non-zero only in a thin band along the boundary. This allows the volume integral to be converted to a surface integral. The contribution from the gradients of the shape functions can be assumed to be constant across the band but varies along the boundary and the contribution from the gradients of inclusion function is constant along the boundary but varies normal to the boundary. The band is represented in terms of two parameters as shown in the Figure 5-3. The stiffness matrix as decomposed in Eq. (5-29) has four components; the second component can be rewritten as shown below. ,tn EdVT1112BCB 12122111eeincincdVHdVHddTT1112112TT1121112BCBBCHBBCHBBC dB (5-32) 81

PAGE 82

In this expression, the terms involving only inclusion functions and its gradients are separated from the terms containing the shape functions and their gradients. This decomposition leads to efficient and accurate computation of the stiffness matrix. The volume integral has been decomposed into a line (surface) integral along the normal to the boundary and a surface integral along the boundary. The normal for the boundary represented by implicit equation 0 is given by at the boundary. The component involving only inclusion function and its gradients is termed as 2C in Eq. (5-32) and is evaluated in the similar manner as explained in section 5.1.3. 5.3.4 Periodic Boundary Conditions Composite materials have periodic arrangement of cells and their behavior can be predicted by analyzing the behavior of the unit cell which is often termed as representative volume element (RVE). The variations in the micro-scale (RVE) are rapid and periodic as compared to the variations at the macro-scale. In order to analyze the behavior at the micro-scale, the RVE is analyzed with appropriate loading and boundary conditions. The RVEs tile the computational domain by translation and the neighboring cells fit into each other in both deformed and un-deformed states as shown in Figure 5-6. Hence the boundary conditions for the RVE are periodic in order to preserve the continuity of displacements, strains and stress across each RVE. The periodic boundary conditions are expressed as linear constraints and they are implemented as multipoint constraints in a typical finite element code. The periodic boundary conditions are expressed in the following form: 13FFuuconst or in general 130FFuu (5-33a) 24FFuuconst or in general 240FFuu (5-33b) 82

PAGE 83

The multipoint constraints in Implicit Boundary Method can be implemented exactly as is done in traditional finite element method (Cook et. al. 2003). It is to be noted that the RVE is always a rectangle/cuboid, hence when Lagrange interpolations are used, the nodes are guaranteed to be present on the boundaries on which multipoint constraints are imposed. Furthermore, since structured grids are used, identical nodes are always guaranteed to be present on the opposite faces and facilitate the imposition of periodic boundary conditions. However, when FEM is used for analysis, care needs to be taken to generate mesh such that there are identical nodes present on opposite faces. The nodes on the faces (or edges in 2D) of the RVE can be partitioned into two parts: independent or master nodes (s) and dependent or slave nodes (s). The nodal displacement vector can be similarly partitioned into where are displacement of the master nodes and are those of slave nodes. The displacements for the slave nodes are specified in terms of the master nodes in the following form whereis the non-zero periodic boundary conditions (e.g., Case 1: msu=[uu] mu su s1mu=Gu+g 1L 1g 11fbkuu ). Hence the displacements are expressed as : msmm11I0u=[uu]=u=GugGg (5-34) where, I denotes the identity matrix and is the zero vector. The matrices Gand are constructed globally for all the master degrees of freedom present in the analysis model. Using this expression for displacement and a similar expression of the corresponding virtual displacement 0 g m u=Gu in the weak form, the global stiffness matrix is transformed into and the load vector is increased by to impose the multipoint constraints. It is to be noted that in a general case of multipoint constraints, and gare constructed such that 'TK=GKG TGKg G 83

PAGE 84

they may be a fully populated matrix and vector respectively. However, when structured grids are used for analysis, identical nodes are guaranteed to be present on the opposite faces and hence with proper choice of node numbers, matrix can be constructed to be a diagonal matrix. This simplifies the computations involved in the transformation of the stiffness matrix and hence leads to efficient implementation. The periodic boundary conditions involved in a square array of fibers are shown in Figure 3-10 and the corresponding multipoint constraint equations are shown in Table 1. G 5.3.5 Computation of Effective Properties Effective properties of the composite materials with periodic distribution of microstructures are computed by applying far field strains to the RVE as shown in Figure 5-7. The displacements, strains and stresses are assumed to be periodic in the RVE. Hills lemma states that the average of the strain energy in RVE is same as the strain energy due to average stress and strains. It can be stated in the form :=: .This property is used to compute the effective properties from an RVE. Using the stress-strain relation H=C the effective properties are represented as The average stresses and strains are computed by solving the boundary value problem six times with different applied strains. These average stresses and strains are used in the relation H:=C: H=C to determine a row (column) of the homogenized stiffness matrix Six different strain states are used to determine all the components of The strain-stress relation for orthotropic materials is given as follows. Once the homogenized stiffness matrix is determined, the effective properties can be computed by using the strain-stress relation. HC HC 84

PAGE 85

312112332121111123222213233312322232323313112123112100010001000100000100000100000EEEEEEEEEGGG-1C (5-35) In two-dimensional problems, the effective properties in the transverse directions are determined by analyzing the model in plane strain conditions. The longitudinal modulus is computed by analyzing the model in generalized plane strain conditions. The longitudinal shear moduli are computed by analyzing the model in longitudinal shear loading conditions. Each of these cases are explained in brief in the following sub-sections. Plane strain. In plane strain formulation, the strains in the 3 x direction (longitudinal direction) are assumed to be zero and the stress-strain relation is constructed as follows: 11212121010(1)(12)12002E 2 (5-36) This stress-strain relation is used in the modified weak form shown in Eq. (3.3) to solve the boundary value problem with specified strain loading and periodic boundary conditions to compute the transverse material properties. Generalized plane strain. The homogenized material properties in the longitudinal direction are computed by using a generalized plane strain model (Li and Lim, 2005), where the 85

PAGE 86

strains in the longitudinal directions are specified. A constant strain in the 3 x is assumed and the strains and stresses are expressed as follows 1111121311221222322231323333366121212122100and0000uxCCCuCCCxCCCCuuxx 123 (5-37) Using the constant strain in 3-direction, the stresses are written as follows: 1111211321222223312661213132323331200000and 0kCCCCCCCCCCCC 3 (5-38) The principle of virtual work as shown in Eq. (3.3) is modified by choosing the virtual strain vector as to get the following equation (Eq. 5-39). In this expression, the last term is the load due to strain in the third direction. The rest of the formulation is exactly similar to plane strain formulation. 1212T 3tTkddd C TTTutub (5-39) Longitudinal shear loading. The homogenized shear properties in the longitudinal direction are computed by using a longitudinal shear loading, where there is only one degree of freedom which is the displacement in the 3 rd -direction. This assumption results in non-zero shear strains in 3-2 or 3-1 directions only while all other strain and stress components are identically zero. The equations for shear strains and shear stress are given by the following expression: 86

PAGE 87

323232323231313131310;0uxGGux (5-40) The principle of virtual work as shown in Eq. 3-3 is modified by choosing the virtual strain vector as This formulation is used to specify unit shear strains and hence the corresponding shear moduli are computed. 3231T 5.3.6 Computation of Load Vector The right hand side of the modified weak form Eq. (5-20) comprises of load terms. The first term is load due to body load and the volume integration is performed as explained in section 5.1.1 and 5.1.2. The second term is the load term due to applied traction. The shape function vector will be either 01N or 20N depending on which region the natural boundary condition is applied. The computation of load vector is performed in a similar manner as explained in section 5.1.4. 5.4 Numerical Implementations for Local Grid Refinement The modified weak form for the solution structure corresponding to local grid refinement was discussed in a previous chapter and was shown to be: 111etNENEeeeeNBEeedd ee TTT1212TT12XBBCBBXXFXNNt (5-41a) iiedTT12aa1a2FNNbBB d (5-41b) Teg1g2X=uu (5-41c) 87

PAGE 88

In this section, a discussion on the numerical implementations for the computation of the terms involved in this weak form is presented. The left hand side of the weak form corresponds to the stiffness matrix while the right hand side corresponds to the load vector. The solution structure for local grid refinement is such that the solution from the coarse grid inside the refined grid is zero due to the step function for refinement. Therefore the nodal values of the coarse grid inside the refined are set to zero, provided these nodes are outside the influence of ref H This leads to the reduction in the total number of degrees of freedom for the analysis model. The computational domain can be divided into three regions as shown in Figure 5-8. It can be observed that the solution is outside the refined grid and transitions from to inside the refined grid. It can also be observed that in the regions where step function for refinement is active, the Dirichlet functions are inactive and vice versa. This is because of the fact the step functions for refinement are active (varying from zero to one) only on the interface of the edges which are shared between the refined and unrefined grid elements and the Dirichlet functions are active only on essential boundaries and the essential boundaries cannot form the interface edges. The computation of stiffness matrix is explained in detail in the next section. 1u 1u 2u 5.4.1 Computation of Stiffness Matrix in Regions 1 and 2 The step function of refinement is not active ( 0refH ) in Region 1 and the solution structure becomes and hence the strain displacement matrix is given as follows: 1g1au=Du+u 1111111213221232221112112122110..00..[]0..;[]00..00....mmNDDNxxNDDNxxNDNDDNDNxxxx BB (5-42) This matrix is similar to the one explained in section 5.1.3 and the stiffness matrix is given as: 88

PAGE 89

eedV TT111212BCB0K=BBCBBdV=00 (5-43) The computation of stiffness matrix is performed in a similar manner as explained in sections 5.1.1-5.1.3. In the Region 2, the step function of refinement is equal to one and the solution structure becomes and the strain displacement matrix will be similar to the one shown in Eq. (5-42) and hence the stiffness matrix becomes: g2a2u=Du+u 22eeTdV T121200K=BBCBBdV=0BCB (5-44) The stiffness matrix is computed as explained in sections 5.1.1-5.1.3. 5.4.2 Computation of Stiffness Matrix in Region 3 In the Region 3, the step function of refinement is varying and the Dirichlet functions are unity. It is to be noted that there is a small region of the size where both Dirichlet functions and step function of refinement may be active, since this region is very small, the contribution to the stiffness matrix in this region is negligible. The solution structure takes the form: and hence the strain displacement matrix takes the form 221refrefHHg1a1gau=u+u+u+u 12BBB 1111112211112121(1)0..0.0(1)..0...(1)(1)..grefrefgiigrefrefgiiggrefrefggrefrefiiiiNHHNxxNHHNxNNHHNNHHxxxx ..x 11112B=BB= (5-45) 89

PAGE 90

22111222212221210..0.0..0....grefrefgiigrefrefgiiggrefrefggrefrefiiiiNHHNxxNHHxxNNHHNNHHxxxx ...N 22122B=B+B= (5-46) The strain displacement matrix takes the similar form as explained in section 5.3.3. Hence the stiffness matrix will take the following form: 22eedV TTT1112111211TT21222122KKBCBBCBK=BBCBBdV=KKBCBBCB (5-47) The computation of stiffness matrix is performed in the similar procedure as explained in detail in section 5.3.3. 5.4.3 Computation of Load Vector The load vector is computed in the similar manner as explained in section 5.3.6. 5.5 Extrapolation of Solution for Elements with Low Volume Fraction In structured grid based methods, the grid elements do not conform to the geometry of the analysis domain and hence there may be few elements in the model with very small solid region. When the volume of the solid region inside the grid elements approach zero, the stiffness matrix corresponding to these elements will have components with magnitudes which are very small in comparison to the internal element stiffness matrices. This leads to ill-conditioning of the system of linear equations and hence poor convergence behavior. In order to avoid the problems associated with grid elements having small volume fractions, an extrapolation scheme is presented in this dissertation. The degrees of freedom corresponding to the elements with small volume fractions are constrained to be an extrapolation of the degrees of freedom from the internal elements. Due to this extrapolation scheme, the degrees of freedom for external nodes will not be arbitrarily large due to near singularity of the 90

PAGE 91

stiffness matrix of partial elements. An added advantage of this procedure is that the total number of degrees of freedom is reduced and hence less computational time for solution of linear equations. The procedure for this extrapolation strategy is explained with reference to Figure 5-9.The external nodes in the grid are denoted by and the internal nodes are represented by The displacement of an external node is associated with the displacement of a set of internal nodes and the number of inner nodes that are used will depend on the approximation scheme used for the grid variable. For example, for linear Lagrange elements, two internal nodes are associated for a given external node and for cubic B-spline elements four internal nodes are associated with a given external node. on in The extrapolation (constraining of external nodes using internal nodes) can be written as follows oi1mnnnu=I(u,...,u) i (5-48) In the above expression, is a function which maps the degrees of freedom of a set of internal nodes to the displacement of an external node. This function can be expressed as an extrapolation matrix and the transformation can be represented as a matrix multiplication: I 1[]{}iondndmnu=Cu (5-49) In the above equation is the number of internal nodes used to constrain the displacements of the outer nodes and the dimension of the problem as and the size of [C is m dn ] dnm 91

PAGE 92

5.5.1 Construction of Extrapolation Matrix In this section, construction of extrapolation matrix Cis described in detail. The extrapolation matrix is constructed by imposing the continuity requirements at the boundaries of the elements. Q4/H8 interpolation. The extrapolation matrix is constructed using the continuity conditions at the nodes and is explained for a 1D case and the extrapolation matrix for higher dimensions is constructed from the 1D extrapolation. When a linear interpolation for grid variable is used, the extrapolation matrix is constructed such that the displacement along the line joining the internal and outer nodes is linear, which means that a continuity requirement is imposed for the piecewise polynomial at the intersection of internal and boundary element. In Figure 5-10, the interpolation between nodes is given by and the interpolation between nodes is given by Imposing continuity at node leads to the following equation (Eq. 5-50). Solving this equation we get and 2D the extrapolation will be as shown in Eq. 5-51 1C 1andin 2iniono1iu 1122iNuNu 2andin 122iNuNu 1C 2in 22-oiuu 121212211iiiorrdNdNdNdNuuuudrdrdrdr (5-50) 112220100201iiooiinnnnnnuvuvuv (5-51) Cubic B-spline approximation. The cubic B-spline approximations have continuity. Hence continuity is imposed between the piecewise polynomial segments between internal nodes and external nodes. 2C 3C 92

PAGE 93

In Figure 5-11, the approximation for displacements in the internal element is given by and the approximation in the boundary element is given by 11223344iiiuNuNuNuNu io 12233441iiiuNuNuNuNu The shape functions of cubic B-splines and the their third derivative is given by 23231223233411(133);(231533)484811(231533);(133)4848NrrrNrrNrrrNrr rr (5-52) 33333123333133;;;888dNdNdNdNdrdrdrdr 4181o3i (5-53) Imposing continuity at the intersection of internal element and boundary element results in the following equation and the extrapolation relation is given as: 3C 1234234112343333464iiiiiiioiiiiuuuuuuuuuuuuu (5-54) Using the above relation, the extrapolation matrix in two dimensions is given as: 112233441040604001040604oiiiiiiiioTnnnnnnnnnnuuvuvuvuvv (5-55) Quadratic B-spline approximation. The quadratic b-spline approximations have continuity. Hence continuity is imposed (Figure 5-12) between the piecewise polynomial segments between internal nodes and external nodes. The extrapolation relation between the external node and internal node is computed as: 1C 2C 11233oiiuuuu (5-56) The extrapolation in two dimensions is constructed as: 112233103030010303ooTniiiiiinuuvuvuvv (5-57) 93

PAGE 94

5.5.2 Construction of Extrapolation Matrix for a Boundary Element In the above discussion, the extrapolation matrix for a single external node is constructed. For a boundary element, there may be more than one such external node. In this section, the construction of extrapolation matrix for a boundary element is described using the example shown in Figure 5-13 In Figure 5-13, there are 3 external nodes for the element of interest and one internal node. An extrapolation matrix needs to be constructed which relates the displacements of the outer nodes using the inner nodes as shown in the following equation: 104,774112,55282,552()2()32()2uIuuuuuIuuuuuIuuuu (5-58) Using the above relations, the extrapolation matrix for the boundary element shown in Figure 5-13 is constructed as: 82774105111002010002101003uuuuuuuu i (5-59) In the above discussion, construction of extrapolation matrix is presented for a specific example and this approach can be extended for construction of extrapolation matrix for Lagrange interpolations as well as B-spline approximations in both two and three dimensions. 5.5.3 Computation of Stiffness Matrix Once the extrapolation matrix for a boundary element is constructed, it can be used to compute the stiffness matrix in terms of the inner degrees of freedom (nodes). Assuming that the external nodes are related to internal nodes with the relation and the left hand side contains the degrees of freedom for the boundary element, the strain displacement matrix is ou=Cu 94

PAGE 95

given as Hence the stiffness matrix is given by Hence if is the stiffness matrix in terms of the degrees of freedom of boundary element, the modified stiffness matrix in terms of the degrees of freedom of internal nodes is given by By using this approach, the modified stiffness matrix for each boundary element is computed and later used for assembly. Once the finite element equations are solved for the internal degrees of freedom, the C matrix can be used to compute the displacements of the external nodes. o=Bu=BCu iC dVTTKCBDBC TdVTCBDB oK iToKCKC Figure 5-1. Typical non-conforming grid used for analysis 95

PAGE 96

Figure 5-2. Triangulation of boundary element for volume integration Figure 5-3. Representation of band on essential boundaries 96

PAGE 97

Figure 5-4. Approximation of natural boundary for computation of load vector 97

PAGE 98

Figure 5-5. Division of computational domain into three regions Figure 5-6. Undeformed and deformed configuration of a set of 4 RVEs 98

PAGE 99

Figure 5-7. Representative volume element used for computation of effective properties Figure 5-8. Division of computation domain into various regions 99

PAGE 100

Figure 5-9. Typical analysis grid with elements having low volume fraction Figure 5-10. Extrapolation of external node for linear Lagrange element Figure 5-11. Extrapolation of external nodes for cubic B-spline element 100

PAGE 101

Figure 5-12. Extrapolation of external nodes for quadratic B-spline element Figure 5-13. Extapolation for boundary element 101

PAGE 102

Table 5-1. Set of periodic boundary conditions for 6 different strains Case Unit Strain Front and Back Face Left and Right Face Top and Bottom Face 1 11 112233;; f bkfbkfbkuuLuuuu 112233;;rlrlrluuuuuu 11223;;tbtbtuuuuuu 3b 2 22 112233;; f bkfbkfbkuuuuuu 112233;;rlrlrluuuuLuu 11223;;tbtbtuuuuuu 3b3 33 112233;; f bkfbkfbkuuuuuu 112233;;rlrlrluuuuuu 112233;;tbtbtbuuuuuuL 4 23 112233;; f bkfbkfbkuuuuuu 112233;;rlrlrluuuuuuL 11223;;tbtbtuuuuuu 3b5 13 112233;;fbkfbkfbkuuuuuuL 112233;;rlrlrluuuuuu 11223;;tbtbtuuuuuu 3b6 12 112233;; f bkfbkfbkuuuuuu 112233;;rlrlrluuuuuu 11223;;tbtbtuuLuuuu 3b 102

PAGE 103

CHAPTER 6 ANALYSIS AND RESULTS The solution structures developed in this dissertation have been implemented by modifying a finite element code. The first part of this chapter presents numerical examples for validating the solution structure corresponding to enforcing essential boundary conditions exactly for both structural and steady state heat transfer problems. It has been be demonstrated that B-spline elements provide good convergence capabilities with continuous stresses (heat flux) in the entire analysis domain. The solutions obtained from B-spline elements are compared with the FE results obtained using bi-linear (4-noded quadrilateral) and bi-quadratic (9-noded quadrilateral) finite elements for two dimensional problems and linear and quadratic tetrahedral elements for three dimensional problems. In the second part of this chapter, the solution structure for treatment of material discontinuity and micromechanical analysis is validated. Numerical examples are presented to study the convergence capability of this solution structure. Fiber reinforced composite is used for determination of effective properties using 2D models and 3D models. In the final part of this chapter, numerical examples are presented for validation of solution structure for local refinement of grid. The classical stress concentration example is presented to study the convergence behavior using Lagrange and B-spline approximations. 6.1 Validation of Solution Structure for Essential Boundary Conditions 6.1.1 Patch Test This example consists of two parts. The first part demonstrates that a linear solution can be exactly represented using B-spline elements using the solution structure presented in this dissertation and the second part demonstrates that a cubic solution in displacement is represented accurately with only one B-spline element using a thin cantilever beam as an example. 103

PAGE 104

Uniaxial tension with constant stress. The first part of this example demonstrates the capability of B-spline elements to represent a state of constant stress. A rectangular plate of dimensions by 1.0 is subjected to uniaxial loading of 1.0 2.0m m M Pa Only a quarter of the plate was modeled because of symmetry as shown in Figure 6-1. The material of the plate is assumed to be steel with modulus of elasticity equal to and Poissons ratio of The analytical solution is given in Eq. (6-1) 200GPa 0.3 666(5.010);(1.510)10;0;0xyxyuxv y (6-1) A non-conforming grid used for this analysis is shown in Figure 6-2. It can be seen that several nodes lie outside the analysis domain as the B-spline elements are being used for analysis. The Dirichlet functions for this problem are shown in Figure 6-3. The D-function for x-component of displacement is zero on the left edge and rapidly reaches to one in the interior of the domain as can be seen in Figure 6-3a. The D-function corresponding to the y-displacements is zero on bottom edge and reaches to one inside the interior region as shown in Figure 6.3b. Since the applied value of Dirichlet boundary condition is zero, the boundary value function is zero in the entire domain. In order to study the effect of range parameter on the quality of solution, a parameter study is performed on and the solution of x-displacements along the line AB is plotted in Fig 6-4 for varying values of it can be observed that for values of 210 one element gives an exact solution. This study demonstrates that a state of constant stress can be exactly represented using Implicit Boundary Method with B-spline approximation. The contour plot for x is shown in Figure 6-5. This example serves as a patch test for implicit boundary method. 104

PAGE 105

Deflection of thin cantilever beam. The second part of this example demonstrates the ability of cubic B-spline element to represent a cubic solution accurately. A thin cantilever beam as shown in Figure 6-6 is considered for this purpose, and according to Euler-Bernoullis beam theory, the deflection curve of for beam is a cubic polynomial and tip deflection is given by 33tipPL E I where is the end load, P L is the length of beam, E is the modulus of elasticity and I is the moment of inertia of beam cross section. The tip deflection according to Timoshenko beam theory is given as 2314538tipPtLLEI (Timoshenko and Goodier, 1969). When cubic B-spline elements are used to model a cantilever beam, we expect to get accurate tip deflection as predicted by Timoshenko beam theory when only one element is used. However, we may not get exact tip deflection as predicted by Euler-Bernoullis beam theory as shear deformations are neglected in this theory, while the plane stress elements account for shear deformations. As the aspect ratio of the beam is increased, the solution from Timoshenko beam theory and Euler-Bernoulli beam theory converge to the same solution and we expect that the solution obtained by one element cubic B-splines converges to the same solution. In order to test this, a cantilever beam with varying aspect ratio is modeled using one-element cubic B-splines. The error in tip deflection for Euler-Bernoulli theory and one element model with respect to the aspect ratio of the beam is plotted in Figure 6-7. It can be observed that the solution obtained by one element model is accurate even for smaller aspect ratio of the beam and the quality of solution does not degrade with increasing aspect ratio of the beam. It can also be observed that the solution from one element model converges to the solution from Euler-Bernoulli theory and Timoshenko beam theory at large aspect ratios of the beam. This example 105

PAGE 106

also demonstrates that cubic B-spline elements do not undergo shear locking in thin limit, and when the structures are made of thin beams, one element can be used to get acceptable solutions. 6.1.2 Validation of Extrapolation Technique In this example, the extrapolation technique presented in this dissertation is validated. A curved beam as shown in Figure 6-8 is used for analysis. The curved beam is a three-quarter circular beam which is fixed on one end and loaded on the other end. The material of the beam is assumed to be made of steel with a modulus of elasticity of and a Poissons ratio of 0.3. 200GPa The geometry for this beam is such that there are several elements on the boundary which may have very small support from the nodes lying inside the geometry of the analysis domain and hence the resulting stiffness matrix will be almost singular and can lead to ill-conditioning of the system of linear equations and hence leads to poor convergence behavior. The extrapolation technique presented in this dissertation is used to extrapolate the solution for the external nodes in terms of the internal nodes. Bilinear Lagrange interpolation are used to perform a convergence analysis where a series of models with varying grid densities are analyzed and the error in strain energy is plotted as shown in Figure 6-9. It can be observed that the convergence behavior is not smooth due to near-singular stiffness matrices. Figure 6-10 shows the error in strain energy using the extrapolation scheme. It can be observed that the convergence plot is very smooth and gives an expected behavior. When B-spline elements are used for modeling this example, the quality of solution was not severely affected even when several elements are present on the boundary with small support. This may be because B-spline elements have a wide span as compared to Lagrange elements and hence the contribution to stiffness from a neighboring element leads to non-singular stiffness matrices and hence does not lead to ill-conditioning. Figure 6-9 shows the plot 106

PAGE 107

of error in strain energy when extrapolation technique was not used for cubic B-spline elements. Figure 6-10 shows the plot of error in strain energy when extrapolation was used. It can be observed that there is no appreciable difference in the convergence plots. It is to be noted that the x-axis of the convergence plot when extrapolation technique is used shows the number of active nodes. The contour plot of y without extrapolation is shown in Figure 6-11 and the contour plot of normal stress is shown in Figure 6-12 using extrapolation technique. In both these plots, bi-linear Lagrange elements were used for analysis and it can be observed that the stress solution shows expected behavior only when extrapolation is used and hence it is necessary to use extrapolation to avoid singular stiffness matrices and to obtain accurate solutions. Figure 6-13 shows the contour plot for y using cubic B-spline elements without the use of extrapolation technique and it can be observed that accurate solution is obtained for B-spline elements even when extrapolation technique was not used. However, in all the remaining examples of this section, extrapolation technique is used for Lagrange as well and B-spline elements. 6.1.3 Convergence Analysis of a Cantilever Beam This example involves the convergence study of quadratic and cubic B-spline elements and its comparison with four-node (Lagrange bi-linear element) and nine-node quadrilateral (quadratic Lagrange element) elements used in traditional FEM. A cantilever beam of length and 0. thickness is fixed at the one end and a uniform shear load of 1.0m 2m 0.1 M Pa is applied at the other end as shown in Figure 6-14. The exact solution for this problem is shown in Eq. (6-2) (Timoshenko and Goodier, 1969). 22163264PyuLxxyEI D (6-2a) 107

PAGE 108

221345364PvyLxDxLxxEI 2 (6-2b) 22;24xxyPPDLxyyII ;0y (6-2c) A typical non-conforming grid used for analysis is shown in Figure 6-15. Even though this problem can be modeled using a conforming grid, a non-conforming grid is used in order to demonstrate the convergence behavior of the solution structure along with B-spline basis. The B-spline elements have nodes outside the elements which form an extra layer of nodes as can be seen in Figure 6-15. The consistent nodal loads are distributed on the nodes corresponding to the boundary elements on which traction is specified. A convergence analysis is performed for this model where a series of models with increasing grid density are used for analysis and the corresponding error norms are computed and plotted. The 2 L error norm represents the errors in the displacements and the expression which defines error is shown in Eq. (6-3). Energy error norm represents the error in stresses and strains and the expression for defining energy error is shown in Eq (6-4). In these expressions, u and represent displacement, strain and stress respectively, while the superscript represents exact solution and represents approximate solution. e h 122LTehehu-u.u-u d (6-3) 1212ETeheh-d (6-4) The convergence plot for error norm and energy error norms are computed for various grid densities and shown in Figure 6-16 and Figure 6-17 respectively. 2L 108

PAGE 109

The plots are shown in a log-log scale, where the y-axis represents the error norm and the x-axis represents the number of nodes in the model. B-spline elements exhibit good convergence rates and for a given number of nodes, the errors are much smaller as compared to the bi-linear and quadratic Lagrange elements. In order to study the errors involved in deflection curve and stress distribution for a specific model, a 10 element model with 5 elements along the length of the beam and 2 elements through the thickness is used. Corresponding FE models with bilinear and quadratic Lagrange elements are also used for comparison purposes. Analytical solution as shown in Eq (6-2) is used a benchmark solution. The comparison of deflection curve is shown in Figure 6-18. It can be observed that the deflection obtained by quadratic and cubic B-spline elements as well as quadratic Lagrange elements used in FEM match closely with the analytical solution. The variation of normal and shear stresses along the line AB are shown in Figure 6-19 and Figure 6-20. It can be observed that B-spline elements as well as quadratic Lagrange elements used in traditional FEM give accurate solutions and match closely with the analytical solutions. However, B-spline elements provide continuous stress in the analysis domain and hence do not need any smoothing algorithms. In order to study the performance of B-spline elements in near incompressible limit, the cantilever beam shown in Figure 6-14 is analyzed under plane strain conditions with a Poissons ratio as 0.4999999. Convergence analysis is performed by analyzing a series of models with increasing grid densities and the tip deflection is plotted in Figure 6-21 for bi-linear Lagrange elements in FEM and quadratic and cubic B-spline elements. It can be observed that bi-linear Lagrange elements undergo volumetric locking while B-spline elements converge to the 109

PAGE 110

expected tip deflection. This example demonstrates that the B-spline elements do not undergo volumetric shear locking. 6.1.4 Square Plate with a Circular Hole The performance of quadratic and cubic B-spline elements in regions of stress concentration is studied in this example. An infinite plate with a central circular hole as shown in Figure 6-22 is considered for this analysis. The plate is subjected to uniform unit traction ( 1 ) along the x axis at infinity. The exact solution for this problem is shown in Eq. (6-5) (Sokolnikoff, 1956). Symmetry of the problem is exploited to model only a quadrant of the analysis domain. The radius of the hole is chosen to be 1 ( 1a ) and length and width of the plate is chosen to be 4 (). The exact solution is used to compute the tractions on the traction boundaries (right and top boundaries) and symmetry boundary conditions are applied on the bottom and left edges while the inner edge is traction free. 4lwa 242331cos2cos4cos422xarr 4a (6-5a) 24213sin2sin4sin422xyarr 4a (6-5b) 24213cos2cos4cos422yarr 4a (6-5c) 224311211coscoscos3cos31122aaaurErr r (6-5d) 22431111sinsinsin3sin31122aaavrErr r (6-5e) The typical analysis models used for convergence analysis along with the contour plot of stresses are shown in Figure 6-23 and Figure 6-24. The error analysis for this problem is 110

PAGE 111

performed by computing the 2 L errors and energy errors for a series of models with varying grid densities. For comparison purposes, errors are also computed for a series of models with FE Q4 (bi-linear) and FE Q9 (bi-quadratic) elements. The plot for 2 L error is shown in Figure 6-25 and plot for energy error is shown in Figure 6-26. Cubic B-spline elements perform much better than the quadratic B-spline elements and FE elements. For a given number of nodes, the errors for B-splines are far smaller when compared to the traditional finite elements. Figure 6-27 shows comparison with the analytical solution along the line 2 It can be observed that cubic B-spline element computes the stress concentration factor as 3.06 which is very close to the theoretical value of 3.0. 6.1.5 Thick Cylinder with Internal Pressure A hollow thick walled cylinder of internal radius 0.3and external radius 0.5 is modeled as shown in Figure 6-28 and for analysis purpose only a quarter of the cylinder is used. The Dirichlet boundary conditions are applied on the straight edges of the quarter such that radial displacement is allowed while tangential displacement is restricted. The inner curved edge is subjected to a pressure of The material is assumed to be made of steel with a modulus of elasticity as and a Poissons ratio of 0.3. The analytical solution is obtained from the governing differential equations for displacements, strains and stresses in a thick walled cylinder with open ends (Boresi et.al. 2003). The displacements and stresses as a function of radial position ris: in in 500psi 63010psi 22222211r p abruparEba (6-6a) 111

PAGE 112

222222rrarbprba (6-6b) 222222arbprba (6-6c) This example demonstrates the capability of the Implicit Boundary Method to impose zero essential boundary conditions in a non-conforming structured grid. The Dirichlet functions are constructed such that they vanish on the boundaries where essential boundary conditions are specified. The displacement solution for the radial displacement is shown in Figure 6-29 and Figure 6-30. In order to verify the accuracy of solution along the line AB, comparison of analytical solution and the computed solution for radial displacements is shown in Figure 6-31. The comparison of radial and tangential stresses is shown in Figure 6-32 and Figure 6-33. It can be observed that the solutions obtained by both quadratic and cubic B-spline elements match closely with analytical solution for displacement, radial and tangential stresses. Furthermore, the stresses computed by B-spline elements are continuous across the element boundaries. This example demonstrates that B-spline elements provide smooth and accurate solutions for stresses as well as displacements with B-spline elements. 6.1.6 Steady State Heat Transfer Analysis A series of pipes carrying hot fluids are arranged in a block to enable heat transfer as shown in Figure 6-34. Each set of alternate pipes carry fluid with temperatures 80 and 50. The surfaces of the block are exposed to air with ambient temperature of 30.The heat transfer takes place due to conduction between the pipes and the block and due to convection from the oC oC oC 112

PAGE 113

ambient air and the surfaces of the block. The conduction coefficient of the block is assumed to be 0.3and the natural convection coefficient is taken to be /oWmC 215/oWmC This problem is analyzed as a steady state heat transfer along with convection. Symmetry of the problem is exploited to model only a quarter as shown in Figure 6-34 (b). The Dirichlet functions for this problem are constructed such that they vanish at the curved edges and are unity inside the domain. The contour plot for Dirichlet function is shown in Figure 6-35. This problem demonstrates the imposition of non-zero Dirichlet boundary conditions and the boundary value function is constructed such that it takes the value of imposed boundary values at the curved edges and the contour plot of the boundary value function is plotted in Figure 6-36. The boundary value function takes on the imposed values at the curved edges and smoothly vanishes to zero in the interior of the domain. For validating the solution obtained by Implicit Boundary Method using B-spline elements, the solution is compared with the solution obtained by FE using ABAQUS package. The contour plot of the temperature distribution in the entire domain is shown in Figure 6-37, Figure 6-38 and Figure 6-39. It can be observed that the solution obtained using cubic B-spline elements has similar patterns when compared to FEM solution using bi-linear Lagrange elements. The temperature distribution along the line AB is plotted in Figure 6-40. The comparison of plots show that the temperature distribution matches exactly with the solution obtained by FEM. 6.1.7 Analysis of a 3D Clamp The clamp shown in Figure 6-41 is modeled using a three-dimensional model with 8-node hexahedral elements and B-spline elements to demonstrate some advantages of using a structured grid with undistorted elements. The structure is clamped on one end while the other end is 113

PAGE 114

subjected to traction in the Y direction. Traction of 10000 Pa is applied on the flat face as shown in Figure 6-41. The Youngs modulus is assumed to be 200 GPa while the Poissons ratio is taken as 0.3. This problem is analyzed using both Implicit Boundary Method (IBFEM) and using traditional finite element software (ABAQUS). Convergence analysis for this problem was performed by computing strain energy for a series of models with varying mesh densities as plotted in Figure 6-42. The plot is shown on semi-log scale, where the x-axis represents the number of nodes and the y-axis represents the strain energy. As the number of nodes increases both methods converge to the same solution. However, in this example, the Implicit Boundary Method provides a better estimate of the strain energy using fewer nodes and elements when cubic B-spline elements are used. For small number of elements, it is harder to accurately represent the geometry without having elements that are distorted in the finite element mesh, whereas the structured grid has regular shaped elements regardless of the grid density. The contour plot for the bending stresses is plotted in Figures 6-43, 6-44 and 6-45 using FEM, Implicit Boundary Method with linear Lagrange elements and cubic B-spline elements respectively. The results obtained using Implicit Boundary Method and relatively coarse grid is able to give an answer that is close to the answer obtained by a higher density FEM mesh. 6.1.8 Analysis of a 3D Rib Structure A 3D rib structure shown in Figure 6-46 is used to study the performance of B-spline elements and tri-linear Lagrange hexahedral elements. A constant pressure loading of 1000 is applied on the top surface and one end of the structure is clamped as shown in Figure 6-46. The material of the structure is assumed to be steel with modulus of elasticity and Poissons ratio Pa 200EG Pa 0.3 The geometry of the structure is quite complex and meshing with regular 114

PAGE 115

hexahedral elements is difficult, so finite element models with linear tetrahedral meshes are used as benchmark solutions. A sequence of models with varying grid densities is modeled and strain energy is computed for each model and the convergence in strain energy is plotted for FEM tetrahedral elements, Implicit Boundary Method linear hexahedral elements and cubic B-spline elements is plotted in Figure 6-47. It can be observed that the strain energy obtained by the sequence of models converge to the same strain energy value. In order to validate the computed stress distribution obtained by linear hexahedral elements and B-spline elements, the contour plots of the normal stress distributions is plotted in Figures 6-48, 6-49 and 6-50. It can be observed that the stress distribution matches closely with the one obtained by FEM. 6.1.9 Analysis of a 3D Dial Bracket A 3D dial bracket structure shown in Figure 6-51 is used to study the performance of B-spline elements and linear hexahedral elements. A constant pressure loading of 1000 is applied on the top flat surface and the bottom flat surface of the structure is constrained. The material of the structure is assumed to be steel with modulus of elasticity and Poissons ratio Pa 200EG Pa 0.3 The geometry of the structure is quite complex and meshing with regular hexahedral elements is difficult, so finite element models with quadratic and linear tetrahedral meshes are used as benchmark solutions. Convergence analysis is performed in terms of strain energy and the results are plotted in Figure 6-52. It can be seen that the solutions obtained by B-spline elements converge to the solution obtained from very dense FE model and the number of nodes needed to reach a convergent solution for B-spline elements is significantly smaller as compared 115

PAGE 116

to the traditional tetrahedral (linear, quadratic)elements. The contour plots of bending stresses are shown in Figures 6-53, 6-54 and 6-55. It can be observed that stress distribution obtained by the Implicit Boundary Method shows similar pattern as with the finite element solutions. This example demonstrates the ease of constructing regular grid for complicated 3D geometries without compromising the quality of solution. 6.2 Validation of Solution Structure for Micromechanics 6.2.1 Square Plate with Circular Inclusion This example demonstrates the validity of the solution structure for treatment of material discontinuity. A circular inclusion of volume fraction 0.47fv as shown in Figure6-56 is used for analysis in this example. The matrix has material properties 68.3m E GPa and 0.3m and the fiber has material properties 379.3fEG P a and 0.1f and plane stress idealization is used. The solution obtained by Implicit Boundary Method (IBFEM) is compared with the solution obtained by FEM using ABAQUS software. The contour plot of x is shown in Figures 6-57 and 6-58 and it can be observed that the stress pattern as computed by FEM matches closely with the stress pattern as computed by IBFEM. The maximum and minimum values for the stresses also match very closely. Convergence analysis is performed by analyzing a sequence of models and the error in strain energy is plotted in Figure 6-59. The solution from a highly dense FE model with 12001 quadratic quadrilateral elements is used to as an exact solution for convergence analysis. The strain energy from the dense FE model was computed to be and the error in strain energy for a given grid density is computed as the difference in computed strain energy and this value. It can be observed that the solution converges as the error approaches to zero when the grid density is increased. -64.27302094610 116

PAGE 117

The comparison of along the line is shown in Figure 6-60. It can be observed that the displacement solution obtained by IBFEM matches exactly with the FEM solution. A slope discontinuity can be observed at the inclusion boundary. Figure 6-61 shows the line plots xu x and y along the line AB and Figure 6-62 shows the comparison of x and y It can be observed that the solution for both stresses and strains match exactly with the FEM solution. The discontinuity in stress and strain can be observed at the inclusion interface. 6.2.2 Determination of Effective Properties of a Fiber Reinforced Composite 2D model. In this example a two-dimensional model is used for computation of effective properties of a Boron-Aluminum fiber reinforced composite with volume fraction of 0.47. For determining the effective properties in the transverse directions, a plane strain assumption is used. In order to determine the homogenized material properties in the longitudinal direction, a generalized plane strain model (Li and Lim, 2005) is used. The homogenized shear properties in the longitudinal direction are determined by using a longitudinal shear loading model with one degree of freedom in the longitudinal direction is used. This assumption results in non-zero shear strains in z-y and z-x directions only while all other strain and stress components are identically zero. The results obtained by IBFEM are compared with the results available in literature (Sun and Vaidya, 1996, Sun and Chen, 1990, Chamis, 1984 and Whitney, 1966). The periodic boundary conditions shown in Table 5-1 are used for analysis. The average stresses for each case are tabulated in Table 6-1. Figures 6-63, 6-64 and 6.65 show the stress distribution for various cases with the deformed shape. The effective properties are computed by using the average stresses and strains and are compared in Table 6-2. It can be observed that the effective 117

PAGE 118

properties determined by the current method match very closely with the effective elastic constants as reported in literature. 3D model. In this example a three-dimensional model is used for computation of effective properties of the fiber reinforced composite. The results obtained by three-dimensional analysis are compared with the results available in literature. The periodic boundary conditions shown in Table 5-1 are used for analysis. The average values of stresses and strains are tabulated in Table 6-3, and the effective elastic constants that are determined by the 3D model are tabulated in Table. 6-2. It can be observed that the effective properties determined by 3D model matches very closely with the ones determined by 2D model and with the ones available in literature. Figures 6-66, 6-67 and 6-68 show the stress distribution for various cases. 6.3 Validation of Solution Structure for Grid Refinement The local grid refinement technique presented in chapter 3 is validated in this section. The first example demonstrates the validity of solution structure for grid refinement by solving the problem with a linear displacements and constant stress and strain as exact solution. In order to demonstrate the performance of the refinement technique in the regions of stress concentrations, the classic problem of square plate with a circular hole is solved and the convergence of the solution structure is demonstrated in the second example. 6.3.1 Application of Grid Refinement to Constant Stress Problem In this example, the validity of solution structure for grid refinement is demonstrated. A rectangular plate with a uniform traction on one end as shown in Figure 6-1 is analyzed in this example. Only a quarter of the plate is modeled because of symmetry, and the exact solution for this problem is shown in Eq. (6-1). The solution structure presented in chapter 3 for local grid refinement is used in a recursive fashion to get multiple levels of refinement. Figure 6-69 shows the contour plot of X118

PAGE 119

displacements using bi-linear Lagrange elements and Figure 6-70 shows the contour plot of Y-displacements. It can be observed that the refinement is done for three levels and the exact solution is obtained without degrading the quality of the solution. It can be observed that at the first level and second levels of refinement, the central element is subdivided into 3x3 elements and at the third level of refinement, the central element is subdivided into 2x2 elements. It is to be noted that in hierarchical refinement, each element can be subdivided only into 2x2 elements, while the method presented in this dissertation does not have this limitation. The contour plots for X and Y displacements are shown in Figures 6-71 and 6-72 and it can be observed that for three levels of refinement, the solution is exactly represented. 6.3.2 Stress Concentration Problem This example involves the study of the performance of solution structure for local grid refinement in the regions of stress concentration. An infinite plate with a central circular hole as shown in Figure 6-73 is considered for this analysis. The plate is subjected to uniform unit traction ( 1x ) along the x-axis at infinity. Symmetry of the problem is exploited to model only a quadrant of the analysis domain. The radius of the hole is chosen to be 1 () and length and width of the plate is chosen to be 5 ( 1a 5lwa ). The exact solution as shown in Eq. (6-5) is used to compute the tractions on the traction boundaries (right and top boundaries) and symmetry boundary conditions are applied on the bottom and left edges while the inner edge is traction free. Bi-linear Lagrange elements and cubic B-spline elements are used in this study and convergence analysis is performed to demonstrate the convergence capabilities of the solution structure. Figure 6-74 shows the contour plot of the X-stress with 4 levels of refinement using a 119

PAGE 120

bi-linear Lagrange elements and it can be observed that the stress concentration factor is computed as 2.916 which is close to the exact value of 3.0 but not exact. Figure 6-75 shows the contour plot of x using cubic B-spline elements for 2 levels of refinement. It can be observed that very accurate solution is obtained by cubic B-spline approximations with relatively less number of elements. Convergence analysis is performed for this problem and Figure 6-76 shows the plot of errors in 2 L norm and Figure 6-77 shows the plot of energy error norm. It can be observed that in both the plots that the solution obtained with the local grid refinement solution structure converges and the error approaches to zero when the levels of refinement are increased. The B-spline elements converge faster as compared to Lagrange elements and hence very accurate solutions are obtained by B-spline approximations with relatively less number of elements. The proposed solution structure is capable of using a mix of Lagrange and B-spline elements in a given model. In order to test this, the stress concentration problem is modeled using bi-linear Lagrange elements for the coarsest mesh and for the first level of refinement and the second level of refinement is done using cubic B-spline elements. The contour plot for x is shown in Figure 6-78. It can be observed that very accurate solution is obtained by using a mix of Lagrange and B-spline elements and the stress concentration factor is computed to be 3.009 which is very close to the analytical solution of 3.0. 120

PAGE 121

Figure 6-1. Plate subjected to uniaxial loading Figure 6-2. Non-conforming grid using one cubic B-spline element 121

PAGE 122

Figure 6-3. Plot of Dirichlet functions (a) (b) uD vD 122

PAGE 123

0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5 6x 10-6 Distance from left edge of plate, m X displacement, m Theoretical displacement = 1.0-1 = 5.0-2 = 1.0-2 = 1.0-3 = 1.0-5 Figure 6-4. Effect of range parameter on the quality of solution Figure 6-5. Contour plot of x for one element model using cubic B-splines 123

PAGE 124

Figure 6-6. Thin cantilever beam subjected to end loading 0 5 10 15 20 25 30 -10 -8 -6 -4 -2 0 2 4 6 8 10 Aspect ratio of beam% error % error in CBS deflection% error in Euler-beam deflection Figure 6-7. Plot of error in the tip deflection with varying aspect ratio 124

PAGE 125

Figure 6-8. Curved beam subjected to uniform loading 101 102 103 104 -10 0 10 20 30 40 50 60 70 80 90 Log (Num Nodes)% Error in Strain Energy IBFEM Q4IBFEM CBS Figure 6-9. Plot of error in strain energy for curved beam (without extrapolation) 125

PAGE 126

101 102 103 104 -5 0 5 10 15 20 25 30 Log (Num Nodes)% Error in Strain Energy IBFEM Q4IBFEM CBS Figure 6-10. Plot of error in strain energy for curved beam using extrapolation technique Figure 6-11. Contour plot of y for Q4 elements (without extrapolation) 126

PAGE 127

Figure 6-12. Contour plot of y for Q4 elements using extrapolation technique 127

PAGE 128

Figure 6-13. Contour plot of y using cubic B-spline elements Figure 6-14. Thick cantilever beam used for convergence analysis 128

PAGE 129

Figure 6-15. Typical non-conforming grid using cubic B-spline elements 101 102 103 104 105 10-8 10-7 10-6 10-5 Log (Num Nodes)L2 Error CBSQBSFEQ4 FEQ9 Figure 6-16. Convergence plot of 2 L error for cantilever beam 129

PAGE 130

101 102 103 104 105 10-2 10-1 100 Log (Num Nodes) Energy Error CBSQBSFEQ4 FEQ9 Figure 6-17. Convergence plot for error in strain energy for cantilever beam 130

PAGE 131

0 0.2 0.4 0.6 0.8 1 -6 -5 -4 -3 -2 -1 0x 10-5 Distance along the beam lengthDeflection along neutral axis AnalyticalCBSQBSFEQ4 FEQ9 Figure 6-18. Comparison of deflection curves for 10 element model 131

PAGE 132

-0.1 -0.05 0 0.05 0.1 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2x 106 Distance along the beam thicknessNormal Stress AnalyticalCBSQBSFEQ4 FEQ9 Figure 6-19. Comparison of normal stress along the line AB for 10 element model 132

PAGE 133

-0.1 -0.05 0 0.05 0.1 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2x 105 Distance along the beam thicknessShear Stress AnalyticalCBSQBSFEQ4 FEQ9 Figure 6-20. Comparison of shear stress along the line AB for 10 element model 101 102 103 104 105 -4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0x 10-5 Log (Num Nodes)Tip Deflection CBSQBSFEQ4Exact Figure 6-21. Convergence plot for tip deflection in near incompressible limit 133

PAGE 134

Figure 6-22. Analysis model for a square plate with a circular hole Figure 6-23. Contour plot of x using FEM Q4 elements 134

PAGE 135

Figure 6-24. Contour plot of x using cubic B-spline elements 101 102 103 104 105 10-5 10-4 10-3 10-2 Log (Num Nodes)L2 Error CBSQBSFEQ4 FEQ9 Figure 6-25. Convergence plot for 2 L error norm for stress concentration problem 135

PAGE 136

101 102 103 104 105 10-4 10-3 10-2 10-1 Log (Num Nodes) Energy Error CBSQBSFEQ4 FEQ9 Figure 6-26. Convergence plot for error in strain energy for stress concentration problem 1 1.5 2 2.5 3 3.5 4 0.5 1 1.5 2 2.5 3 3.5 Plot of X-Stress along = /2Distance from center of circular hole, mX-Stress AnalyticalCBSQBS Figure 6-27. Plot of x along the line 2 136

PAGE 137

Figure 6-28. Hollow thick cylinder subjected to uniform pressure Figure 6-29. Contour plot of radial displacements using cubic B-spline element model 137

PAGE 138

Figure 6-30. Contour plot of radial displacements using FEM Q4 elements 0.3 0.4 0.5 0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3x 10-5 Distance from center of cylinder, inRadial displacement ur AnalyticalFE Q4FE Q9CBSQBS Figure 6-31. Comparison of radial displacements along the thickness of the cylinder 138

PAGE 139

0.3 0.32 0.34 0.36 0.38 0.4 0.42 0.44 0.46 0.48 0.5 -500 -400 -300 -200 -100 0 100 Distance from center of cylinder, inRadial stress r AnalyticalFE Q4FE Q9CBSQBS Figure 6-32. Comparison of radial stresses along the thickness of the cylinder 0.3 0.4 0.5 500 600 700 800 900 1000 1100 Distance from center of cylinder, inTangential stress t AnalyticalFE Q4FE Q9CBSQBS Figure 6-33. Comparison of tangential stresses along the thickness of the cylinder 139

PAGE 140

Figure 6-34. Heat conduction in a block with series of pipes Figure 6-35. Plot of Dirichlet function 140

PAGE 141

Figure 6-36. Plot of boundary value function Figure 6-37. Contour plot of temperature distribution using FE Q4 elements 141

PAGE 142

Figure 6-38. Contour plot of temperature distribution using cubic B-spline elements Figure 6-39. Contour plot of temperature distribution using quadratic B-spline elements 142

PAGE 143

0 0.1 0.2 0.3 0.4 0.5 45 50 55 60 65 70 75 80 85 Distance from point A along AB, mTemperature FE Q4CBSQBS Figure 6-40. Comparison of temperature along line AB Figure 6-41. Three dimensional clamp used for analysis 143

PAGE 144

102 103 104 105 4 4.05 4.1 4.15 4.2 4.25 4.3 4.35 4.4 4.45x 10-3 Log (Num Nodes)Strain Energy Abaqus H8IBFEM H8IBFEM CBS Figure 6-42. Convergence plot for strain energy for clamp model Figure 6-43. Contour plot of bending stresses using H8 elements in FEM for 3D clamp 144

PAGE 145

Figure 6-44. Contour plot of bending stresses using H8 elements for 3D clamp Figure 6-45. Contour plot of bending stresses using cubic B-spline elements for 3D clamp 145

PAGE 146

Figure 6-46. Three dimensional rib model used for analysis 102 103 104 0.215 0.22 0.225 0.23 0.235 0.24 0.245 Log (Num Nodes)Strain Energy Abaqus Linear TetraIBFEM H8IBFEM CBS Figure 6-47. Convergence plot for strain energy for rib model 146

PAGE 147

Figure 6-48. Contour plot of bending stresses using tetrahedral elements for 3D rib Figure 6-49. Contour plot of bending stresses using B-spline elements for 3D rib 147

PAGE 148

Figure 6-50. Contour plot of bending stresses using H8 elements for 3D rib Figure 6-51. Three dimensional dial bracket used for analysis 148

PAGE 149

102 103 104 105 106 6 7 8 9 10 11 12 13x 10-4 Log (Num Nodes)Strain Energy Abaqus 10-Node TetrahedronAbaqus 4-Node TetrahedronIBFEM H8IBFEM CBS Figure 6-52. Convergence plot for strain energy for dial bracket model Figure 6-53. Contour plot of bending stresses using a cubic B-spline element model for dial bracket 149

PAGE 150

Figure 6-54. Contour plot of bending stresses using a tetrahedral element model for dial bracket Figure 6-55. Contour plot of bending stresses using a H8 element model for dial bracket 150

PAGE 151

Figure 6-56. Square plate with a circular inclusion Figure 6-57. Contour plot of x using Q4 elements in IBFEM 151

PAGE 152

Figure 6-58. Contour plot of x using Q4 elements in FEM 102 103 104 10-10 10-9 10-8 Log (Num Nodes)Error in Strain Energy Figure 6-59. Convergence plot for error in strain energy for square matrix with inclusion model 152

PAGE 153

0 0.5 1 1.5 0 0.5 1x 10-8 Distance along line ABX-Displacement along line AB Abaqus uxIBFEM ux Figure 6-60. Comparison of along the line AB xu 0 0.2 0.4 0.6 0.8 1 1.2 1.4 -200 0 200 400 600 800 1000 1200 1400 Distance along line ABStress components along line AB Abaqus xxAbaqus yyIBFEM xxIBFEM yy Figure 6-61. Comparison of stresses along the line AB 153

PAGE 154

0 0.2 0.4 0.6 0.8 1 1.2 1.4 -1 -0.5 0 0.5 1 1.5 2 2.5x 10-8 Distance along line ABStrain components along line AB Abaqus xxAbaqus yyIBFEM xxIBFEM yy Figure 6-62. Comparison of strains along the line AB 154

PAGE 155

A B Figure 6-63. Plot of stresses for cases 1 and 2 (A) Plot of 11 normal stress Case 1 and (B) Plot of normal stress for Case 2 22 A B Figure 6-64. Plot of stresses for cases 3 and 4 (A) Plot of 33 normal stress for Case 1 and (B) Plot of shear stress for Case 4 23 A B Figure 6-65. Plot of stresses for cases 5 and 6 (A) Plot of 31 shear stress for Case 5 and (B) Plot of shear stress for Case 6 12 155

PAGE 156

A B Figure 6-66. Plot of stresses for cases 1 and 2 for 3D model (A) Plot of 11 normal stress for Case 1 and (B) Plot of normal stress for Case 2 22 A B Figure 6-67. Plot of stresses for cases 3 and 4 for 3D model (A) Plot of 33 normal stress for Case 1 (B) Plot of shear stress for Case 4 23 A B Figure 6-68. Plot of stresses for cases 5 and 6 for 3D model (A) Plot of 31 shear stress for Case 5 and (B) Plot of shear stress for Case 6 12 156

PAGE 157

Figure 6-69. Contour plot of u using Q4 elements and grid refinement Figure 6-70. Contour plot of using Q4 elements and grid refinement v 157

PAGE 158

Figure 6-71. Contour plot of u using cubic B-spline elements and grid refinement Figure 6-72. Contour plot of using cubic B-spline elements and grid refinement v 158

PAGE 159

Figure 6-73. Square plate with a circular hole Figure 6-74. Contour plot of x for local refinement using Q4 elements 159

PAGE 160

Figure 6-75. Contour plot of x for local refinement using cubic B-spline elements 100 101 102 103 10-4 10-3 10-2 10-1 Log (Num Nodes)L2 Error CBSQ4 Figure 6-76. Convergence plot for 2 L error in local grid refinement 160

PAGE 161

100 101 102 103 10-3 10-2 10-1 Log (Num Nodes)Energy Error CBSQ4 Figure 6-77. Convergence plot for energy error in local grid refinement Figure 6-78. Contour plot of x for local refinement using Q4 and cubic B-spline elements 161

PAGE 162

Table 6-1. Average stresses for various loading cases Case Unit Strain 11()GPa 22()GPa 33()GPa 23()GPa 31()GPa 12()GPa 1 111 230.84 41.14 41.14 0 0 0 2 221 40.35 161.20 46.22 0 0 0 3 331 40.35 46.22 161.20 0 0 0 4 231 0 0 0 46.07 0 0 5 311 0 0 0 0 54.44 0 6 121 0 0 0 0 0 54.44 162

PAGE 163

Table 6-2. Comparison of effective properties Property IBFEM 2D IBFEM 3D Ref:Sun and Vaidya, 1996 Ref: Sun and Chen, 1990 Ref: Chamis, 1984 Ref:Whitney, 1966 1() E GPa 214.83 214.85 215 214 214 215 2() E GPa 144.07 143.85 144 135 156 123 3() E GPa 144.07 143.85 144 135 156 123 23()GGPa 46.07 45.97 45.9 43.6 13()GGPa 54.44 54.35 57.2 51.1 62.6 53.9 12()GGPa 54.44 54.35 57.2 51.1 62.6 53.9 23 0.25 0.25 0.29 0.31 13 0.20 0.19 12 0.20 0.19 0.19 0.19 0.20 0.19 163

PAGE 164

Table 6-3. Average stresses for various loading cases Case Unit Strain 11()GPa 22()GPa 33()GPa 23()GPa 13()GPa 12()GPa 1 111 230.57 40.34 40.34 0 0 0 2 221 40.35 160.93 46.25 0 0 0 3 331 40.35 46.25 160.93 0 0 0 4 231 0 0 0 45.98 0 0 5 131 0 0 0 0 54.37 0 6 121 0 0 0 0 0 54.37 164

PAGE 165

CHAPTER 7 CONCLUSIONS AND FUTURE WORK 7.1 Conclusions In this dissertation, finite elements based on uniform B-spline basis defined over a structured grid are constructed to obtain solutions with continuous gradients throughout the analysis domain. Finite elements are developed using quadratic and cubic B-spline basis for problems in both two and three dimensions. B-spline basis do not satisfy Kronecker-delta property and hence Implicit Boundary Method is extended for using B-spline basis and for applying essential boundary conditions. Solution structures are developed using the approximate Heaviside step functions based on the implicit equations of the essential boundaries to impose essential boundary conditions. The weak form for linear elasticity and steady state heat transfer was modified using this solution structure such that essential boundary conditions are automatically imposed. Methods were developed for efficiently and accurately evaluating the terms involved in the modified weak form in the limit as the approximate Heaviside step functions approach the exact Heaviside step functions. The primary motivation for the method is the desire to obtain continuous gradients in the entire analysis domain using B-spline basis on a structured grid. The B-spline elements studied in this work exhibited good convergence behavior and the comparison with traditional FEM solutions showed that very accurate solutions can be obtained with relatively fewer nodes. The performance of B-spline elements does not degrade with increasing aspect ratio of the elements and one element can be used to compute accurate solutions for beam like structures. Although B-spline basis is used in this work to achieve continuous stress and strains, the method is quite general and can be applied to other basis functions which may or may not satisfy Kronecker165

PAGE 166

delta property such as Lagrange interpolations and Moving Least Squares approximation used in many meshless methods. An added advantage of the current method is the elimination of the mesh generation process during the preprocessing stage of the design cycle. Generating a uniform structured grid that encloses the geometry is easier than to generate a conforming mesh, furthermore all the elements used in the analysis are regular and undistorted unlike traditional FEM and hence problems associated with distorted elements are eliminated. As shown through several examples in the dissertation, models with fewer elements can be used for getting accurate solutions. In addition, the method would enable design engineers to perform analysis directly using solid models instead of first generating a mesh and then applying boundary conditions on the mesh, hence resulting in significant savings in pre-processing time. Another advantage of this method is that all internal elements have the same stiffness matrix resulting in significant reduction in computation required to determine the global stiffness matrix. The Implicit Boundary Method is extended for analysis of micromechanical analysis by developing a solution structure for treatment of material discontinuity in a structured grid framework. A technique for imposing multipoint constraints based on the technique for imposition of multipoint constraints in traditional FEM is presented. This approach is used for modeling an inclusion in a matrix and convergence analysis is performed to validate the solution structure. Lagrange interpolations are used for performing micromechanical analysis to determine effective properties of fiber reinforced composites and it has been shown that the solution structure presented in this work has good convergence properties and is capable of determining the effective elastic constants with very good accuracy in both two and three 166

PAGE 167

dimensions. This method holds the potential for determining effective properties of complicated microstructures using structured grids for each inclusion and a structured grid for matrix. The Implicit Boundary Method is extended for local grid refinement by constructing a solution structure based on approximate Heaviside step functions of the coarse/refined grid boundary. This solution structure is validated by analyzing the classical stress concentration problem. Convergence analysis shows that the solution structure has good convergence capabilities. In traditional adaptive techniques, multiple types of approximations (Eg. combination of Lagrange and B-spline approximations) is difficult to use, but with the proposed solution structure, it has been demonstrated that multiple types of approximations can be used. Using higher order elements (B-spline elements) in the regions of stress concentration and lower order elements (Lagrange elements) far from the regions of stress concentrations makes the analysis computationally efficient and provides accurate solutions as demonstrated in the numerical examples. 7.2 Future Work The B-spline elements developed in this dissertation are used for analysis of linear elastic and steady state heat transfer problems only. B-spline based elements can be extended for analysis of Kirchhoff plates, where the weak form requires continuous approximation space. The Implicit Boundary Method with B-spline approximations can be extending for other linear analysis types like electro-static and magneto-static problems, non-linear analysis like large elastic deformations and elasto-plastic deformations, fracture mechanics and contact mechanics. 1C The solution structure for material discontinuity can further be used to compute effective properties of complicated microstructures and for performing damage analysis of composites. The method can further be extended to incorporate B-spline approximations. 167

PAGE 168

The solution structure for adaptive refinement can be extended for performing analysis of 3D problems. The h-, pand kadaptivity can be incorporated into this method. B-spline basis provides a convenient methodology for raising the order and continuity of the approximation space, this property can be used to enable k-adaptivity and can also be used for k-version FEM. 168

PAGE 169

LIST OF REFERENCES Adams, D.F., Crane, D.A., 1984. Finite element micromechanical analysis of a unidirectional composite including longitudinal shear loading. Computers and Structures 6 1153-1165. Amestoy P.R., Davis T.A., Duff I.S., 1996. An approximate minimum degree ordering algorithm. SIAM Journal on Matrix Analysis and Applications 17(4) 886-905. Atluri S.N., Kim H.G., Cho J.Y., 1999. Critical assessment of the truly Meshless Local Petrov-Galerkin (MLPG), and Local Boundary Integral Equation (LBIE) methods. Computational Mechanics 24(5) 348-372. Atluri S.N., Zhu T.L., 2000a. Meshless local Petrov-Galerkin (MLPG) approach for solving problems in elasto-statics. Computational Mechanics 25(2-3) 169-179. Atluri S.N., Zhu T., 2000b. New concepts in meshless methods. International Journal for Numerical Methods in Engineering 47(1) 537-556. Babuska I., Banerjee U., Osborn, J.E., 2002. On principles for the selection of shape functions for the Generalized Finite Element Method, Computer Methods in Applied Mechanics and Engineering 191(49-50) 5595-5629 Belytschko T., Krongauz Y., Fleming M., Organ D., Snm L., Wing K., 1996. Smoothing and accelerated computations in the element free Galerkin method. Journal of Computational and Applied Mathematics. 74(5) 111-126. Belytschko T., Krongauz Y., Organ D., Fleming M.,Krysl P., 1996. Meshless methods: an overview and recent developments. Computer Methods in Applied Mechanics and Engineering 139(1-4) 3-47. Belytschko T., Lu Y.Y., Gu L., 1994. Element-free Galerkin methods. International Journal for Numerical Methods in Engineering 37(2) 229-256. Belytschko T., Parimi C., Moes N., Usui S., Sukumar N., 2003. Structured extended finite element methods for solids defined by implicit surfaces. International Journal for Numerical Methods in Engineering 56(4) 609-635. Boresi A.P., Schmidt R.J., Sidebottom O.M., 1993. Advanced mechanics of materials. John Wiley and Sons Inc. New York. Cai, Y.C., Zhu, H.H., 2004. Direct imposition of essential boundary conditions and treatment of material discontinuities in the EFG method. Computational Mechanics 34(4) 330-338. Chamis, C.C., 1984. Simplified composite micromechanics equations for hygral, thermal and mechanical properties. SAMPE Quarterly 14-23. Clark B.W., Anderson D.C., 2003a. The penalty boundary method. Finite Elements in Analysis and Design 39(5-6) 387-401. 169

PAGE 170

Clark B.W., Anderson D.C., 2003b. The penalty boundary method for combining meshes and solid models in finite element analysis. Engineering Computations 20(4) 344-365. Cook, R.D., Malkus D.S., Plesha M.E., Witt, R.J., 2003. Concepts and Applications of Finite Element Analysis. John Wiley and Sons Inc., New York. Cordes, L.W., Moran, B., 1996. Treatment of material discontinuity in the element-free Galerkin method. Computer Methods in Applied Mechanics and Engineering 139(1-4) 75-89. Cottrell J.A., Reali A., Bazilevs Y., Hughes T.J.R., 2006. Isogeometric analysis of structural vibrations. Computer Methods in Applied Mechanics and Engineering 195 (41-43) 5257-5296. Cottrell J.A., Hughes T.J.R., Reali A., 2007. Studies of refinement and continuity in isogeometric structural analysis. Computer Methods in Applied Mechanics and Engineering 196(41-44) 4160-4183. Dang, T.D., Sankar, B.V., 2007. Meshless local Petrov-Galerkin formulation for problems in composite micromechanics. AIAA Journal 45(4) 912-921. De S., Bathe K.J., 2000. Method of finite spheres. Computational Mechanics 25(4) 329-345. De S., Bathe K.J., 2001. The method of finite spheres with improved integration. Computers and Structures 79(22-25) 2183-2196. Dolbow J., Belytschko T., 1999. Numerical integration of the Galerkin weak form in meshfree methods. Computational Mechanics 23(3) 219-230. Duarte C.A., Oden J.T., 1996. h-p adaptive method using clouds. Computer Methods in Applied Mechanics and Engineering 139(1-4) 237-262. Duarte C.A., Babuska I., Oden J.T., 2000. Generalized finite element methods for three-dimensional structural mechanics problems. Computers and Structures 77(2) 215-232 Farin G., 2002. Curves and Surfaces for CAGD: A Practical Guide. Morgan Kaufmann Publishers, San Francisco. Gibson, R. F., 1994. Principles of Composite Material Mechanics. McGrawHill, New York. Gingold R.A., Monaghan J.J., 1982. Kernel estimates as a basis for general particle methods in hydrodynamics. Journal of Computational Physics 46 429-53. Hughes T.J.R., Cottrell J.A., Bazilevs Y., 2005. Isogeometric analysis: CAD, finite elements, NURBS, exact geometry and mesh refinement. Computer Methods in Applied Mechanics and Engineering 194 (39-41) 4135-4195. Hughes T.J.R., 2000. The Finite Element Method: Linear Static and Dynamic Finite Element Analysis. Dover Publications, New York. 170

PAGE 171

Hllig K., Reif V., Wipper J., 2001. Weighted extended B-spline approximation of Dirichlet problems. SIAM Journal on Numerical Analysis 39(2) 442-462. Hllig K., Reif V., 2003. Nonuniform web-splines. Computer Aided Geometric Design 20(5) 277-294. Hllig K., 2003. Finite element methods with B-Splines. SIAM, Philadelphia. Kagan P., Fischer A., Bar-Yoseph P.Z., 1998. New B-spline finite element approach for geometrical design and mechanical analysis. International Journal for Numerical Methods in Engineering 41(3) 435-458. Kagan P., Fischer A., Bar-Yoseph P.Z., 2003. Mechanically based models: Adaptive refinement for B-spline finite element. International Journal for Numerical Methods in Engineering 57(8) 1145-1175. Kantorovich L.V., Krylov V.I., 1958. Approximate Methods of Higher Analysis. Interscience, New York. Kim, H.J., Swan C.C., 2003. Algorithms for automated meshing and unit cell analysis of periodic composites with hierarchical tri-quadratic tetrahedral elements. International Journal for Numerical Methods in Engineering 58(11) 1683-1711. Kohno H., Tanahashi T., 2004. Numerical analysis of moving interfaces using a level set method coupled with adaptive mesh refinement. International Journal for Numerical Methods in Fluids 45(9) 921-944. Krongauz Y., Belytschko T., 1996. Enforcement of essential boundary conditions in meshless approximations using finite elements. Computer Methods in Applied Mechanics and Engineering 131(1-2) 133-145. Krongauz, Y., Belytschko, T.,1998. EFG approximation with discontinuous derivatives. International Journal for Numerical Methods in Engineering 41(7) 1215-1233. Krongauz Y., Belytschko T., 1997. Consistent pseudo-derivatives of meshless methods. Computer Methods in Applied Mechanics and Engineering 146(3-4) 371-386. Krysl P., Grinspun E., Schroder P., 2003. Natural hierarchical refinement for finite element methods. International Journal for Numerical Methods in Engineering 56(8) 1109-1124. Krysl P., Belytschko T., 1997. Element-free Galerkin method: convergence of the continuous and discontinuous shape functions. Computer Methods in Applied Mechanics and Engineering 148(3-4) 257-277. Kumar A.V., Padmanabhan S., Burla R., 2007. Implicit boundary method for finite element analysis using non-conforming mesh or grid. International Journal for Numerical Methods in Engineering. (In press). 171

PAGE 172

Kumar A.V., Lee J., 2006. Step function representation of solid models and application to mesh free engineering analysis. Journal of Mechanical Design. Transactions of the ASME 128(1) 46-56. Kumar A.V., Lee J.H., Burla R.K., 2005. Implicit solid modeling for mesh free analysis. In: Proceedings of International Design Engineering Technical Conferences & Computers in Engineering IDETC/CIE Long Beach, California. Leung A.Y.T., Au F.T.K., 1990. Spline finite elements for beam and plate. Computers and Structures 37(5) 717-729. Li, S., Lim, S.H., 2005. Variational principles for generalized plane strain problems and their applications. Applied Science and Manufacturing 36(3) 353-365. Liszka T.J., Duarte C.A.M., Tworzydlo W.W., 1996. hp-meshless cloud method. Computer Methods in Applied Mechanics and Engineering 139(1-4) 263-288. Liu G.R., Gu Y.T., 2001. A point interpolation method for two-dimensional solids. International Journal for Numerical Methods in Engineering 50(4) 937-951. Lucy L.B.,1977. A numerical approach to the testing of the fission hypothesis. The Astronomical Journal 82(12) 1013-24. Marrey, R.V., Sankar, B.V., 1997. A Micromechanical Model for Textile Composite Plates. Journal of Composite Materials 31(12) 1187-1213. Melenk J.M., Babuska I., 1996. The partition of unity finite element method: Basic theory and applications. Computer Methods in Applied Mechanics and Engineering 139(1-4) 289-314. Moes N., Bchet E., Tourbier M., 2006. Imposing Dirichlet boundary conditions in the extended finite element method. International Journal for Numerical Methods in Engineering 67(12) 1641-1669. Monaghan J.J., 1987. An introduction to SPH. Computer Physics Communications 48(1) 89 96. Mortenson M.E., 1997. Geometric Modeling. Wiley, New York. Nayroles B., Touzat G., Villon P.,1992. Generalizing the finite element method: Diffuse approximation and diffuse elements. Computational Mechanics 10(5) 307-318. Osher S.J., Fedkiw R.P., 2002. Level Set Methods and Dynamic Implicit Surfaces. Springer Verlag, Philadelphia. Padmanabhan, S., 2006. Analysis using non-conforming structured grid and implicit boundary method. Doctoral Dissertation, University of Florida. 172

PAGE 173

Pengcheng S., Peixiang H., 1995. Bending analysis of rectangular moderately thick plates using spline finite element method. Computers and Structures 54(6) 1023-1029. Rvachev V.L., Shieko T.I., 1995. R-functions in boundary value problems in mechanics. Applied Mechanics Reviews 48(4) 151-188. Rvachev V.L., Sheiko T.I., Shapiro V., Tsukanov I., 2001. Transfinite interpolation over implicitly defined sets. Computer-Aided Geometric Design 18(3) 195-220. Sankar, B.V., and Marrey, R.V., 1997. Analytical Method for Micromechanics of Textile Composites. Composites Science and Technology 57(6) 703-713. Shapiro V., 1998. Theory of R-functions and applications: A primer. CPA Technical Report CPA88-3, Cornell Programmable Automation. Sibley School of Mechanical Engineering. Shapiro V., Tsukanov I., 1999. Meshfree simulation of deforming domains. Computer Aided Design 31(7) 459-471. Sokolnikoff I.S., 1956. Mathematical Theory of Elasticity, McGraw-Hill, New York. Sukumar N., Moran B., Belytschko T., 1998. Natural element method in solid mechanics. International Journal for Numerical Methods in Engineering 43(5) 839-887. Sun, C.T., Chen, J.L., 1990. A micromechanical model for plastic behavior of fibrous composites. Composites Science and Technology 40 115-29. Sun, C.T., Vaidya, R.S., 1996. Prediction of composite properties from a representative volume element. Composites Science and Technology 56(2) 171-179. Surana K.S., Rajwani A., Reddy J.N., 2006. The k-version finite element method for singular boundary-value problems with application to linear fracture mechanics, International Journal for Computational Methods in Engineering Science and Mechanics 7(3), 217-239. Taliercio, A, 2005. Generalized plane strain finite element model for the analysis of elastoplastic composites. International Journal of Solids and Structures 42(8) 2361-2379. Timoshenko S.P., Goodier J.N., 1969. Theory of Elasticity, McGraw Hill, New York. Tyrus, J.M., Gosz, M., DeSantiago, E., 2007. A local finite element implementation for imposing periodic boundary conditions on composite micromechanical models. International Journal of Solids and Structures 44(9) 2972-2989. Vermeulen A.H., Heppler G.R., 1998. Predicting and avoiding shear locking in beam vibration problems using the B-spline field approximation method. Computer Methods in Applied Mechanics and Engineering 158(3-4) 311-327. Whitney, J.M., Riley, M.B., 1966. Elastic properties of fiber reinforced composite materials. AlAA Journal 4 1537-42. 173

PAGE 174

Xia, Z., Zhang, Y., Ellyin, F., 2003.A unified periodical boundary conditions for representative volume elements of composites and applications. International Journal of Solids and Structures 40(8) 1907-1921. Yu, W., Tang, T., 2007. A variational asymptotic micromechanics model for predicting thermoelastic properties of heterogeneous materials. International Journal of Solids and Structures 44(22-23) 7510-7525. Zhu, H., Sankar, B.V., Marrey, R.V., 1998. Evaluation of Failure Criteria for Fiber Composites Using Finite Element Micromechanics. Journal of Composite Materials 32(8) 766-782. 174

PAGE 175

BIOGRAPHICAL SKETCH Ravi Kumar Burla was born and brought up in Hyderabad, India. He graduated with a Bachelor of Technology degree in mechanical engineering and a Master of Technology degree in computer-aided design and automation from Indian Institute of Technology, Bombay, India in June 2001. After graduation, he joined Geometric Software Solutions Limited, Pune, India as a Software Engineer where he worked for 2 years. He then enrolled in the doctoral program in Mechanical and Aerospace Engineering at the University of Florida in Fall 2003. He was awarded an Alumni Fellowship. His areas of interest include application of B-splines for finite element analysis, numerical methods and computer-aided geometric design. 175

xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E20101204_AAAAAN INGEST_TIME 2010-12-04T08:55:58Z PACKAGE UFE0021901_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 2013 DFID F20101204_AAALKJ ORIGIN DEPOSITOR PATH burla_r_Page_109.txt GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
b7f1435d878c9614859dd7de701f91cd
SHA-1
57aa55bea4c1922407b3a45bcf0c1cce0d4888dc
2039 F20101204_AAALJV burla_r_Page_093.txt
3d72355d7c85c62355006b5575ee5025b8d80730
40909 F20101204_AAAKHH burla_r_Page_140.jpg
d895aa0b445a52c6057e6550f6341588
b45a678a949c603c75f161bc669f3a1947e2dd72
2327 F20101204_AAAKGT burla_r_Page_001thm.jpg
02273a41f775588b43780c5f1b4a983b
14789bfd23ef98bbefe92219e59f2ef67f554e8d
2216 F20101204_AAALKK burla_r_Page_110.txt
332a8d6779242d59795fc0a6bdcfdee7
1860 F20101204_AAALJW burla_r_Page_094.txt
20b8e96bcff812c8f421aaf3d5c94472905a3ec2
37730 F20101204_AAAKHI burla_r_Page_086.pro
ef03bcc37b8b47660104c0c3552a9f4a
3ec3e0f157c739dc5c9273758205d9dabdbe3b2b
21187 F20101204_AAAKGU burla_r_Page_032.QC.jpg
14bed6d2d4aa926edb0e77333777eb22cdb6b2a4
1167 F20101204_AAALLA burla_r_Page_127.txt
a8c35511e08963ec2556afd5ace0ff2c
436c2494fa7d3d45f775477285c117f762bde50b
1979 F20101204_AAALKL burla_r_Page_111.txt
8a8116a0d8c43dc46dd9576e815fe204
1a488cca5ee4e3b512bf8ba971fd1c375c37cc47
1354 F20101204_AAALJX burla_r_Page_095.txt
4fd1a8b143d016e75cbbb0ed6d0325bf
0e98b71e256c80e419999a1fd9f33ce32ca9505e
9897 F20101204_AAAKHJ burla_r_Page_157.pro
4f67e056c2474dc167b92a5f1451508c
c3f5d7dcb44a6113bbe2a8755b78a606673e7aa2
18820 F20101204_AAAKGV burla_r_Page_037.QC.jpg
7e72f18f789f2143d423c6200f03b58f
84641bf6c43361774ebbcd25ee20d7bbb026184d
1211 F20101204_AAALLB burla_r_Page_128.txt
41a81e76a9f4255dcc3f611e8ed5f8a0
71e1a5d6391876df7728696149a83819f002b43f
2018 F20101204_AAALKM burla_r_Page_113.txt
84699ea1a7cc708074f2557965e5ec31
5922829739ab0cdbe32389780f406dac08b8499c
251 F20101204_AAALJY burla_r_Page_096.txt
766878ab2f4d7c6c7e96af854cd0e4fa
b20f334bfcb11b3befd146b383d624d9735cfb83
161955 F20101204_AAAKHK burla_r_Page_007.jp2
f2232044e1ced3f5123a67ac1690edec
25271604 F20101204_AAAKGW burla_r_Page_051.tif
169169f461d8ba37bd6a960c8b691b00
7577ca24c6259e0d89b036b0bcc806e85f8460cf
474 F20101204_AAALLC burla_r_Page_129.txt
a3f44e806678c3728c5b1ca47ff43223
fd3184daeaed6eed838db09664f61efd45547486
2070 F20101204_AAALKN burla_r_Page_114.txt
06576947369f3f71d053508eb069f9b0
38be8b35b0245e060fc8a1f24e7c98733c84ef0a
361 F20101204_AAALJZ burla_r_Page_097.txt
ac7f978679f9041ba4498f8e79279984
23c46dda46efbc9470170bee486fe03da79048a6
1053954 F20101204_AAAKIA burla_r_Page_046.tif
0a923fd55fd5f463876fbc10a60c935b
69ccfea957a7dfeca0d412a27b0bf3e34ca36283
54591 F20101204_AAAKHL burla_r_Page_075.jpg
641f19b349eafcb2e35555204e8ed3ae
670f182c6370e3374efac2c9ab62725df5f57cfa
40111 F20101204_AAAKGX burla_r_Page_125.jpg
3532a7af9b0c9b632cc69fd8f8745583
e389fb5b4c1ac69bfe2f031ae1b49da7ce0284c2
646 F20101204_AAALLD burla_r_Page_130.txt
879412fcd46d29956f644c973f470ce8
e4b9971300fc5d4ee899edcfcc6e91d9b05d48e6
1928 F20101204_AAALKO burla_r_Page_115.txt
2a1d0eaf1996d81b4c68c5e099781407
3dbf07c5ec633d7533d60e8986abaf11b91018bb
48861 F20101204_AAAKIB burla_r_Page_133.jp2
a0e35b874c85d6832576996977c92aae
9653 F20101204_AAAKHM burla_r_Page_049.QC.jpg
F20101204_AAAKGY burla_r_Page_119.tif
00cc174bb192f5b85ea49af38b65bfc8d18c8767
688 F20101204_AAALLE burla_r_Page_132.txt
954f12cf535734a220b1b84fd10f6bc3
1c19723272799a8c25d8a3da07d1f7b274882dd3
1995 F20101204_AAALKP burla_r_Page_116.txt
2f1233261ff175ae53176e3e364ff167
a82274a78632440eb624e67b8dc4cf3f3db37192
F20101204_AAAKIC burla_r_Page_136.tif
d890bdc80e8a0b8dcdbf5c7cc12088b8
e8d9f7183ac411a370e75f3496462891ecfac295
F20101204_AAAKHN burla_r_Page_028.tif
9cc7d32376494c4c43a01cb084a67339
8463650bfaf9445d49aec81a08ce741ff971bc01
88208 F20101204_AAAKGZ burla_r_Page_047.jp2
344a3ee078d42c2b37765e7acc86d472
e4c47db5b12ab15ed153453e8dc7db02469c2a47
1000 F20101204_AAALLF burla_r_Page_133.txt
4e2c5c69147c1a013462162cb9b0941f
f42d8b956291a1100164c54dc3e55e71f6e9be4e
2019 F20101204_AAALKQ burla_r_Page_117.txt
66cb32f0032a47db5dac9626e5b8f800509c3997
70298 F20101204_AAAKID burla_r_Page_156.jpg
f0cf1517d5f137cfcb6f2d1bd369a3f8
903be19b8ed2f20a6abda7312487e562367ba86f
72160 F20101204_AAAKHO burla_r_Page_116.jpg
9338083a74fb5a4f61acef62b861d0fa
05f0d80f4686f2c15c3df707e75ac07a0e5f02a1
380 F20101204_AAALLG burla_r_Page_134.txt
2aec57253fe0e715e580a18f66f5a954
e041794944243d49a4c0bf371eeb6e39bc5a0f42
2149 F20101204_AAALKR burla_r_Page_118.txt
fb89dca6f2dc28596b20b57ef48e2fd2
e133e2b0b3d419f56a6b10cd63d6b4bd2bdaeb82
20244 F20101204_AAAKIE burla_r_Page_120.QC.jpg
0d70872676d1f77041e687989d87831a
0bcc3d80e71e027b989a6738e788e46b5779300b
25170 F20101204_AAAKHP burla_r_Page_131.jpg
47a83b31da16b01b72e7d253d0ba7864
684b373a414b425e0864d78b854c3e4c9882a6dc
485 F20101204_AAALLH burla_r_Page_135.txt
f2f9a9fbfdd3cbb664abf53c29680140
a541bef09dd0b0073f73a6cdb0a2f24885d39b7c
2017 F20101204_AAALKS burla_r_Page_119.txt
fc6935529066a1e3ebae1a02c1331ed0
80861 F20101204_AAAKIF burla_r_Page_091.jp2
99592e879427d48287d71cf84ea0bcdc
662a08bc34603d9b5ce1c47a80d1f273c67b2e96
55304 F20101204_AAAKHQ burla_r_Page_091.jpg
6d7989431b4a32c07d28272791af79ae
23d2edd7ba2baf19dc9cb935450439d15fdfafb0
523 F20101204_AAALLI burla_r_Page_136.txt
d9d8f725e98ac974cfe640b02f99651c
1643 F20101204_AAALKT burla_r_Page_120.txt
81363991ee9c46effb20e97a1ecc1d5f
9fc34b5ee8112a25c4af923a6cdcfc16107ffe85
5268 F20101204_AAAKIG burla_r_Page_085thm.jpg
c15a1a06db92f564288e985890309a59
58538 F20101204_AAAKHR burla_r_Page_095.jpg
063a2880b21b8b7257992108f46e3797
7d8a47d72912c08677a5b5d377e4b77ea9e39e06
355 F20101204_AAALKU burla_r_Page_121.txt
5fc98f99582d572db04d5e4354ba6308
d5a66714438af1140764ec4113bab751c249b5e2
673 F20101204_AAAKHS burla_r_Page_100.txt
b33c21d565a114d7428083e5cc7821c0
8713d0e23a82ca1221646f14bcee898b026da4d0
792 F20101204_AAALLJ burla_r_Page_138.txt
7a4711dee7f1006f2533a87939b25ecd
676 F20101204_AAALKV burla_r_Page_122.txt
e4399fa98da59bfc3ccf9618b72edffa
20827 F20101204_AAAKIH burla_r_Page_110.QC.jpg
9f3f44a8572b0be17b558b17e49367bd
521d84acd972011fb41229a54dfa011a6db502a3
58289 F20101204_AAAKHT burla_r_Page_172.pro
a610443842f493dc6970c4fec91b30f3
4101a62aef694f79e9cd4f3bd5a17060e0ab11d4
743 F20101204_AAALLK burla_r_Page_139.txt
bc8e4b68b7e3b850acf5b2684ed651fd
a80e8932bc0bd9bd235f453aa812e51a0aa3daf1
714 F20101204_AAALKW burla_r_Page_123.txt
5ecdb175ac7f0f5238940a62ab39d08b
0c7fdaef45dcc022e8b6bc01644005241f05ac55
47144 F20101204_AAAKII burla_r_Page_084.pro
4a7e2003fd8fd61fc9f1653702dc4c14
5916a54750e9b00dc0cc2439f36d8a9e0b8b31cb
1051974 F20101204_AAAKHU burla_r_Page_145.jp2
c79aff83cae952c59ce5eddbe4ee707e
494 F20101204_AAALMA burla_r_Page_157.txt
f77ca375e31cedde0ebe0ede0d6d7156
43a214ddd5b5ab69a119678deccda0ba9c67ccbc
481 F20101204_AAALLL burla_r_Page_140.txt
73f602825a32e1ae5b3e9e9ef0de6e7d
0d88f7d92f1824193c9ff9c88382c205e9146530
527 F20101204_AAALKX burla_r_Page_124.txt
780d73be58e211d772c833311c9e4a58
877810e8e0fed95f7965d1064110ff1e7673a841
30502 F20101204_AAAKIJ burla_r_Page_055.jpg
3c8d319d5f17dc7ed97fb70f3da140aac720507d
134769 F20101204_AAAKHV burla_r_Page_170.jp2
42c5d7ab0b49ba5712f73824ce0f25fb
ae3fa7c6c93730cee4462927da2bcec73639d0d2
2346 F20101204_AAALMB burla_r_Page_158.txt
21aa2c38e4651908fac49e237bac07e9
1387 F20101204_AAALLM burla_r_Page_141.txt
52ac0d466c55359a8da1709e489bcc7d9ef04f27
271 F20101204_AAALKY burla_r_Page_125.txt
d8b8a9ee8401072158e6981d0cf4ec9b
83d3bfde1fb7af9647270ec63021c387a6ea7db9
25329 F20101204_AAAKIK burla_r_Page_039.QC.jpg
c61a57305d818e1847b5686c52d6c908
2049 F20101204_AAAKHW burla_r_Page_036.txt
5fc1ae9a1d85534f4bf2082cd7f449e4
5583b9bc866a33633695e0b01e54763d7e9067c3
345 F20101204_AAALMC burla_r_Page_159.txt
c964926df9480a5cdb89cb7f6aa0eaa1
313 F20101204_AAALLN burla_r_Page_143.txt
e168f1fcc6ffa357e6ab5497753022e2
d189b481a9c25c34312c9a53fb8c29ce153b3d27
737 F20101204_AAALKZ burla_r_Page_126.txt
ff00cd5f98da58a3e6a69fb0f90341f6
37a4ee995fac69d6c998d3238c6fdf2f539c1d64
83373 F20101204_AAAKIL burla_r_Page_010.jpg
75e8169ff30188a1d9d2fae718b98cb9
9a837195bfc66d5362c7456280489837cbb98c18
2030 F20101204_AAAKHX burla_r_Page_023.txt
3e9a45b4a4889cbd49606c738a53f36e
5c9c8fe396c395a90c0302d7b742db729fd00777
2165 F20101204_AAAKJA burla_r_Page_073.txt
fe68891eb985722af04323bdb42e90bb
0fab85d529732b57206c176c1a27a0d198f40511
1188 F20101204_AAALMD burla_r_Page_160.txt
a3414ae8fc3944c943eabe2ed9c5bff9
5c713b2abd74b9cbe1760c170b3a98946f12445c
495 F20101204_AAALLO burla_r_Page_144.txt
f49fdefe5f19c2a1bdbf9d7b7c4727d9
50559 F20101204_AAAKIM burla_r_Page_136.jp2
e210d31ba737504dc75b496a9b5d6183
0def4519b00429e88093a0a753205147b43543a4
76894 F20101204_AAAKHY burla_r_Page_075.jp2
cd6c71f64b31a7227689da30aa00fa0e
f533a8c90edabd40dfbf139914af9bfeb49b420d
330234 F20101204_AAAKJB burla_r_Page_124.jp2
0c79eab0135ea557fee47187333e0b66
c13d6a219f3d53821d3366ee105552b9bafc9ec6
530 F20101204_AAALME burla_r_Page_161.txt
e187ded670c1da3c99e8063bdfe7f040
dd1604c2d5a40bd7e4023dcf322323a8e8e647b0
2152 F20101204_AAALLP burla_r_Page_145.txt
d201bbcd98eaedcd7db814c12d81fc4e
0913f85987b73389b80aa1181cc90e671cb0977e
6817 F20101204_AAAKIN burla_r_Page_030thm.jpg
e73ed592c3def6b983274122459dbf73
ae01afb665d5ebc9d7bdcd8854bf1737c521fa21
F20101204_AAAKHZ burla_r_Page_054.tif
dcb3b6db831a95d73f764d4099753a8c
9579764f76495b0edd5e7e52da4e07bd00581d49
49228 F20101204_AAAKJC burla_r_Page_083.pro
dd910d792c9bb47cd9e54683430beb08
3d8bb4f47af747ed7bdd57ea6cefd2b3c5dae65f
617 F20101204_AAALMF burla_r_Page_162.txt
0ca91481ca1f5ea6c270ec26c179fcb4
395b2a535a628b6c77ce4b3a5d913b0b30d3875b
463 F20101204_AAALLQ burla_r_Page_146.txt
df2c43c0abe5894050875cb2f5a26ce4
914666 F20101204_AAAKIO burla_r_Page_128.jp2
992fef97f9c9404b37087c46a80a2253
5603f4ced576cd97065cbb2f2aa931423acabc97
40158 F20101204_AAAKJD burla_r_Page_045.pro
8e47729107b6d328aa8475f25e506813
1f6162e759c7c089028b35c06bc5538e0ca430b6
648 F20101204_AAALMG burla_r_Page_163.txt
0749fb0e066df2ab98380943f47829bb
1724128fa65d238d03e59ff25fb72f4341152b1f
1549 F20101204_AAALLR burla_r_Page_147.txt
3c075ef27e11388959250c0df13601d3
529979bc5fa98403e157b5b484df2a1a731fda08
84292 F20101204_AAAKIP burla_r_Page_092.jp2
3a9907a47732f0bc53ff69ba247b8659
4d2db03db95eab0084e339702e83fdd5e540b842
F20101204_AAAKJE burla_r_Page_114.tif
b207b2f5bdd72e11a7da7203e696ff39
abbaf85191cd13b41705f3f9f5a180d4912bbc42
524 F20101204_AAALMH burla_r_Page_164.txt
a3865af28d069289f96cf932185882f6
a2cd45df40df71b7d06c82b10635a27765ee9eb4
F20101204_AAALLS burla_r_Page_148.txt
f31450812506a6a224b5578eee7a2c4e
d5d6436794a3f56acd25d0f35275162864aef849
10259 F20101204_AAAKIQ burla_r_Page_124.QC.jpg
e32ed16540bbc33dcc795389e57fa66b
c6cb95d84c07b5bf60da06b5264a83492f7ab33d
37797 F20101204_AAAKJF burla_r_Page_035.pro
bc576d39526427be9d3b05517d5b7896
e29cd67e0edd143037d7b9b64b05b5f7ff2fa239
2057 F20101204_AAALMI burla_r_Page_165.txt
60c72fef1ce9365419e93f9b5337c993
ec445a681a26bb41e0cea06e1d102b9574e6c5d5
585 F20101204_AAALLT burla_r_Page_149.txt
c065f1dd797694687fe5c0e422f47d26c1f2a83e
880 F20101204_AAAKIR burla_r_Page_015.txt
376bb95ea35fb6651a6dd6757c5fba16
2c1662c8a1c7e89147e963619fd300322781ce9a
F20101204_AAAKJG burla_r_Page_014.tif
9637fd227578fa4f6ed8c06634aabf59
2c745348689f1f38599d6f3f67470f12f69552d6
2023 F20101204_AAALMJ burla_r_Page_166.txt
caecf73123cec349f8b6095576a99672c918ca40
644 F20101204_AAALLU burla_r_Page_150.txt
cfd698bdf35e12b511dab807f49df96c
c7f55335d6b0348268392a9344dd932d1381b64f
5448 F20101204_AAAKIS burla_r_Page_125.pro
9b979647e60d88584368986f0e11bb8d
4870a479ca6a72b0c1ea0e77865981a84534135e
14407 F20101204_AAAKJH burla_r_Page_149.QC.jpg
732f953d0ff467c14a126e1cb7342364
5b940d157b39c660908cfc46f3ae35cb396c860e
462 F20101204_AAALLV burla_r_Page_151.txt
e403cc6068a4822791fdd6c1365ef1d383825504
21675 F20101204_AAAKIT burla_r_Page_014.QC.jpg
0358a31f03656e12a2331a2a0db61edd
4a7434bc818a4f087bed9c46cd2337001091c8dc
2095 F20101204_AAALMK burla_r_Page_167.txt
0818281deb22aeb5acfc354d37428bcb
ea434353ec5f434cd6f12ea93545d2e8f30fb02f
706 F20101204_AAALLW burla_r_Page_152.txt
b1ab9ae5248d248510f527713f633c96
09789984a3960dd4500eca6dba2022370e3f6e45
F20101204_AAAKIU burla_r_Page_001.tif
5ce0a4cc8aae94346dfdebdc727b4b8c
610021 F20101204_AAAKJI burla_r_Page_137.jp2
daab4495831b7f3f7eccc256ea786a6a
18c5021eeda12fcc2d6c8fb20283de9770e9ce8a
23968 F20101204_AAALNA burla_r_Page_044.QC.jpg
404 F20101204_AAALML burla_r_Page_168.txt
7edaa2fcce5c993bee2d2ca401a5be74
faf46ec616d8bd68dec9009da8a23500da4e20c6
332 F20101204_AAALLX burla_r_Page_154.txt
97d84c69f72335e05a645aa37d3b1cef
52a0f745ac8e21a56eaab95c0c2fde33a4fdbe89
F20101204_AAAKIV burla_r_Page_167.tif
3e3fb16b742a028b8f5de7db74c39ce0
9f70bed29bd5c7931d4603e9825294c23b540bbd
9695 F20101204_AAAKJJ burla_r_Page_158.pro
34d1913e883e584e6501d8bfb65be751
2466e0fcbf8dfa0eb13e0fdbf114d8ab1c84d382
8226 F20101204_AAALNB burla_r_Page_102.QC.jpg
cc9ea1a11813014428caecee05d6e22214389c20
2557 F20101204_AAALMM burla_r_Page_169.txt
8e1940d432fd431a8cf2277400e83180
ced49256ab64ecaae3eb2b1f6d4d580cc4ab804e
2330 F20101204_AAALLY burla_r_Page_155.txt
ffe40cc04740453c32ee35cbd5294e16
95f22355e4d5f0630784277f11f65d44850e0599
F20101204_AAAKIW burla_r_Page_175.tif
b0444ee82f9d1572722aa0a7a5222627
0543a6c2b08ba755fe87d649a6bb9bb8d0dfa51e
17220 F20101204_AAAKJK burla_r_Page_020.QC.jpg
c934f4828a1b0d5344b6df16c08a86cf
521259e28ee51fe6159fd059fb05934991a2baff
22730 F20101204_AAALNC burla_r_Page_060.QC.jpg
b0ba411905bd11145e0cc0cb64da60d3
618fee9a76ea3263a9606ddef4b38eee07f9e646
2534 F20101204_AAALMN burla_r_Page_170.txt
c81f6e2e343a47eefccb6bfb6abea6e7
f6a7a41f66bdb366b9e80b644e108f8243e4145d
1076 F20101204_AAALLZ burla_r_Page_156.txt
eeeac2f06e6939869b8cddbe1e5824e7
659e0d303bb910ac2579098257a96a2b2680a9f8
5658 F20101204_AAAKIX burla_r_Page_045thm.jpg
2c9885144e317f841afd80cc3da3c742
fd105c0462476cba94781be768be6e099be2ea0d
4150 F20101204_AAAKKA burla_r_Page_029thm.jpg
cef5a8f5f57a65986828ea089ffacb7eeac956bd
F20101204_AAAKJL burla_r_Page_095.tif
30ae42fef91fd86a45dd8f3a58fa4a8ece518c31
19352 F20101204_AAALND burla_r_Page_075.QC.jpg
44ef4261b7a748b002d0ac109db58a57
8d7c8d58b9d9f74648d1a2df47d03608eb93ef59
2541 F20101204_AAALMO burla_r_Page_171.txt
6e3f664c74c0e3f14b35505d7279de3f
21389 F20101204_AAAKIY burla_r_Page_111.QC.jpg
74ffa0f2f1bf8ec2e1876515ba8673b1
fa810ce8037ac056de00517d408c4550054e8a8e
F20101204_AAAKKB burla_r_Page_048.tif
dc39ed3aedb26e111eb07f8dc10b53d3
f0f14c9a3d53c418d03e50aab8efa0de47bf61c4
13363 F20101204_AAAKJM burla_r_Page_142.pro
6dc03f4e1e5da19f58c488ddb014a810
ba513f000847b52eb52660f82cd802401505de54
2632 F20101204_AAALNE burla_r_Page_056thm.jpg
5238441688a55feed57368316a3393d0
c9168fa0401f20cff0c5c400136a3f70706d958f
2671 F20101204_AAALMP burla_r_Page_173.txt
3cba7f7cae87087fb0accfcc93170510
c7f4740b9c7fdbfc916bf2b22a69d934072a7650
24972 F20101204_AAAKIZ burla_r_Page_105.QC.jpg
32758db9219f54f61418eeb131f00981
a9b573aa08c437745837216d5dd2456a7861f456
1051967 F20101204_AAAKKC burla_r_Page_148.jp2
34174cf2156af601f67bf8ba15cb2f4c
eb7db25144b1b70763a211a62265d76869801742
1941 F20101204_AAAKJN burla_r_Page_067.txt
0deb9bd5f0c9bbae6a33ef7819af99f9
306da8331dba64bd161f8bd503f60c09fb2b6222
20366 F20101204_AAALNF burla_r_Page_156.QC.jpg
3a5000c17763df9f8d41802d0317a5389b726871
645 F20101204_AAALMQ burla_r_Page_174.txt
cfc6a73769a035a90f634d3f16bf64ee
587e1ba16d875b36ddf0aa44df1138f705e9de1c
6223 F20101204_AAAKKD burla_r_Page_046thm.jpg
368086ef9dde6a68848311a7b1ebb3da
7a701d0c986bbe10560ffb8b1834bcfdce9d6aa7
57977 F20101204_AAAKJO burla_r_Page_092.jpg
8a1f6469c43294ee61a927c1b514c667
6b2571ec52c5a720c16eab65413d7772a60b875a
23693 F20101204_AAALNG burla_r_Page_165.QC.jpg
e10a85a3685ee31ed747d1bd03efb0d9
116c19607a0de23e4f2d0e4a3d923d72bf92c2f5
802 F20101204_AAALMR burla_r_Page_175.txt
9bfeb755b6dba93d232b5a400675d7d3
d19bdebb714033cf09e9af000b3864531111c06e
14709 F20101204_AAAKKE burla_r_Page_133.pro
a28b4f7553ff489a1c5ac385b0e7a5cb
5751fcd87503f447efd49a026918a23d40ae4441
118132 F20101204_AAAKJP burla_r_Page_017.jp2
3c4ba6112b685d9ed2c891b38b882e07
63b71f6f2d46a683b3cf088cd7cf2fe4eb8d8acd
6446 F20101204_AAALNH burla_r_Page_071thm.jpg
2ee5f59c70f24680f9458471663473c5
1cf5b28b126e97acaffde79fc167fc49b63b8385
5254789 F20101204_AAALMS burla_r.pdf
b38e9f49e656abfefb896ef301c0cd7c
affdf17caf6ce9faa7527540f2b5f2c357c6da6f
1623 F20101204_AAAKKF burla_r_Page_091.txt
922db3bd9689c574ca15a4136ae4f1c9
43989eb84ef15816f27a71448edd6174c956e9fc
5334 F20101204_AAAKJQ burla_r_Page_075thm.jpg
a5ea36b02efd68a0bb6b56eda25180fc2f83b500
6839 F20101204_AAALNI burla_r_Page_122.QC.jpg
7662e4df9dca4e23b7365fa7a241962a
3c46e25de4270a3082793c5574f62457fe73608e
24226 F20101204_AAALMT burla_r_Page_167.QC.jpg
432067e63fc198403851c3eca7c96d2b
117403 F20101204_AAAKKG burla_r_Page_024.jp2
45a81c6739fa14f720d0b9eeced486c1518053e1
6911 F20101204_AAAKJR burla_r_Page_006thm.jpg
7bc983a814c005d4ebd1ac38f54d5c78
6529 F20101204_AAALNJ burla_r_Page_011thm.jpg
518feb1dd2015379de270da3dd2c684d
f8b4553f41916085baa4428d85ef39f9074fd449
6987 F20101204_AAALMU burla_r_Page_051.QC.jpg
82c1ef8939a1841874fc8877279a4c04
4a5eec16bbbc621449ba1c5faf6e017de8b60efc
346 F20101204_AAAKKH burla_r_Page_131.txt
b59de58fda7286cc4fc870e89e828a21
3c746971842c8c43b779a63ce9fa061c6682680a
475 F20101204_AAAKJS burla_r_Page_137.txt
57d3ec6d81ec6c9e3371db23577cef96
a17538bb0773a88298a09340f2d785455c8c895e
19219 F20101204_AAALNK burla_r_Page_095.QC.jpg
3e125799dc5b3d8cec3c6401b3d592d4
361ef09d8a524806d37595eb563042e2cf1b1d14
5706 F20101204_AAALMV burla_r_Page_095thm.jpg
bb4f7548a4751bd5cebd13dd2c11ef98
9c50f06e0a27a2772aafda6844dd3b8261ba8fcb
2105 F20101204_AAAKKI burla_r_Page_042.txt
8c83bd0d5209752d52aee90781083626
5c1311f82022f9715f1dca13c3f87ce626c94fa1
F20101204_AAAKJT burla_r_Page_124.tif
77a4623f98eb9f91e86602cd8e77cbb5
0004f9c064a8df7444a9298115241834e7f84660
9918 F20101204_AAALMW burla_r_Page_096.QC.jpg
786777aeb4da26d115697aa01ce45e8499409f0e
19574 F20101204_AAAKJU burla_r_Page_043.QC.jpg
91580bbaf16dd89d7247ae73d90d489b
d820bfec15a5ce29cd23ec646962e6fccdfc41a6
4322 F20101204_AAALOA burla_r_Page_152thm.jpg
3a31c2c32190ef606d1b42d2801b3c9c
33ff53f4ebfe1d34576010aee5927a1b89beae24
9807 F20101204_AAALNL burla_r_Page_153.QC.jpg
539c385e6df210cc431692c62c1ca2b4
3256 F20101204_AAALMX burla_r_Page_159thm.jpg
60b866001006f2611ab2947d7f389670
04a79dbdd5af77fd7cafc3114b9ba05ca6a0c11a
15728 F20101204_AAAKKJ burla_r_Page_147.QC.jpg
c6731ebb89311fee90c9ba8239b184ef
238517 F20101204_AAAKJV burla_r_Page_056.jp2
309830d31cc9dcbde05a5b7050362115ec3f484e
7273 F20101204_AAALOB burla_r_Page_170thm.jpg
a3ccaf8f88e6ffbd30eee3bc1d700163
bea046f9a1747ecec9812917e28f05b95f43c6b4
22988 F20101204_AAALNM burla_r_Page_019.QC.jpg
815b5200d81ca6e3d29dffa08b72fedc
384d7a96e8b8739be3c7e41e36da616aed9d99a3
4519 F20101204_AAALMY burla_r_Page_004thm.jpg
77007c92321852db6ec22d29b79f0477
398c12cbcfef2d0706a42d00b036d8d582a4b63c
47827 F20101204_AAAKKK burla_r_Page_038.pro
e30476dfcb31bbaec045de645ec1c943
7584fb8331af266d2d80d265917cf840bb749bc5
95588 F20101204_AAAKJW burla_r_Page_042.jp2
4c24deb36d459cc0a318f1b0087ca7b0
cb8b6ac42e18529dbce68e918b3b83be57691444
21822 F20101204_AAALOC burla_r_Page_036.QC.jpg
3fb97303a57f9598ef637c286b097c8c
9749b1fdeb907dc17244fd519118e9abcaa36560
6503 F20101204_AAALNN burla_r_Page_166thm.jpg
479660967f435fde7d55c723ff647552
2bb49ac3328ba40b4773d44e6c7342dbf182eb8f
16395 F20101204_AAALMZ burla_r_Page_077.QC.jpg
943f17fc6d7956e5146045c2d0a28dab
f4d6a9cfe7b6e7a53325ea1e05e963429cdcec82
47930 F20101204_AAAKLA burla_r_Page_004.jpg
9395e0f314dc70ba6ebe95bdc37dbf06
756db3f57b6b399bdb95f7d6b3cdee70c41e8608
F20101204_AAAKKL burla_r_Page_019.tif
24cd3fddffb4e2cf5f4d5e1881d4a3be
68c6f1540aeef78019704b00de3504084555edf3
5645 F20101204_AAAKJX burla_r_Page_037thm.jpg
c01110c03a27f30b4d19cf4c51afcacf
f85f93b516187784b3cf536fa4c36f37900f31ea
5102 F20101204_AAALOD burla_r_Page_145thm.jpg
8db57faf1e8cc409c010db99454f660d
62d499d3b543d64897cfe69071a3fd5351a05f06
26465 F20101204_AAALNO burla_r_Page_026.QC.jpg
e94e35372c976e7c3f2c27a050cb5d5e
db38a7f31303cbaa34b4d728e29f7b9b150216c2
79685 F20101204_AAAKLB burla_r_Page_005.jpg
c52432230a3c1182ea12c1276008b6ae
701269c11720c444abfe55409e3fda3bb0978186
79391 F20101204_AAAKKM burla_r_Page_033.jp2
0a38590e8b0596be73e1aa4f21779124
6c0789166d4990897b5f956ccf4c5eabfe298bfc
2517 F20101204_AAAKJY burla_r_Page_131thm.jpg
9d633a591e1bb1d9f2ee971f2238b337
6784 F20101204_AAALOE burla_r_Page_022thm.jpg
04ea0bb318a1967248b8918c71092ec1
0abe0ff8dff807a949e3aa564a27034b53887d9a
20837 F20101204_AAALNP burla_r_Page_046.QC.jpg
92fc5fba4f3c1916a3df44fcc7634419
a8eeb3e68e32b9aa4eb6386116b2c1e157421b04
21420 F20101204_AAAKLC burla_r_Page_008.jpg
ac2bc4f34112b36b0290be77921d12aa
aac6a7ae3143aedb67e31d7d8412a5f84c55117e
623 F20101204_AAAKKN burla_r_Page_003.pro
4db026d4c68709d4c9162bf547e81641
b5cbac57328fd135999a7fb2915ce4b8c1afbf0e
73914 F20101204_AAAKJZ burla_r_Page_105.jpg
2ab405ee2be26bda9b1c66897bd1cf53
f5658ffbf08901d8fb959b080ac9e70eb10f16da
16335 F20101204_AAALOF burla_r_Page_145.QC.jpg
fd0ee741108a8695fbc2c0ce1dcf6fd7
82c5a565e6fe718a9f3aa72e0b02a3987985cf2a
25030 F20101204_AAALNQ burla_r_Page_010.QC.jpg
d596a33b2e1e8651015dc358e75939dd
318c664ae3523d4e632b03bebc1b3babb8a49343
82509 F20101204_AAAKLD burla_r_Page_009.jpg
a42b51753bf7acfd4e625f368f758516
65495 F20101204_AAAKKO burla_r_Page_150.jpg
bd93a7eae30c60a384086e418323c1e6
da609bb71e212c6b4eb0a7ae49063c4eafaedfdf
6981 F20101204_AAALOG burla_r_Page_040thm.jpg
74510b28d98fd4d03f98a5535c97b4634705c6c5
17824 F20101204_AAALNR burla_r_Page_086.QC.jpg
3df5f90f1d199596c6fcdd2ddfcb7d2d
dc216d744e131e717fb48b55b089d4777311dc27
85747 F20101204_AAAKLE burla_r_Page_011.jpg
744e0265d6f9e3c44d5164637464d21a
6758 F20101204_AAAKKP burla_r_Page_061thm.jpg
47844e9ed96ee9ee7692192b5c19d593
833f9fcbf4084d674d5bf40f0ae2c2ca2cfdd5c3
24686 F20101204_AAALOH burla_r_Page_030.QC.jpg
5c9cb5a68b33a4cc6e55f008f97b8c16
dda8a2b54c458b819796ac24325550cd1c337e9c
3942 F20101204_AAALNS burla_r_Page_064thm.jpg
d98ca025bf02042bbb11828031371b2c
2f7dd5800d32c7e251cd83cf3db10940549e894f
52860 F20101204_AAAKLF burla_r_Page_013.jpg
aa9d33f9541b3bfaabea2beb2534984c
1b2afc58ffcff11d815595613d71a3a2fec58ce0
63197 F20101204_AAAKKQ burla_r_Page_170.pro
cd2098c4968f6a655fa25d89b848130e
d3b6a75e0055f3311286fa8f003715a4cf62ba8c
6933 F20101204_AAALOI burla_r_Page_028thm.jpg
99f4ac80fa4e82fe4299beed37fb47c6
94faf9695ff11976a2eac80111c7847e13bddcc7
11253 F20101204_AAALNT burla_r_Page_098.QC.jpg
311eeb7638350b251a792d1aa50f8722
9ea4f9f294b14b6db303ba0e9bc19245c9bfc8e9
69578 F20101204_AAAKLG burla_r_Page_014.jpg
e13f92a2aa50d5095b5a7eb6aa0926c2
40dee7edbf75886487da2dae3d8c4fc1e8aa46bd
F20101204_AAAKKR burla_r_Page_115.tif
077d02a34d8d4548ffa6c98fea07b7f2
15578 F20101204_AAALOJ burla_r_Page_152.QC.jpg
ec59b7401a6875e6c8720525735aed42
2368 F20101204_AAALNU burla_r_Page_168thm.jpg
326b64b50dbfb70b789cdcc6607b1e4c
901e16e57326b96793838d286938eaca8c5673fd
F20101204_AAAKKS burla_r_Page_122.tif
beebf0072aa396f48fb266690e6537c5
34dd4c3a65c8d89fd24ea348081c8fbb25d3410f
36480 F20101204_AAAKLH burla_r_Page_015.jpg
0f42c691a0dcecdb2706544ff7202e5c
d5f5b3c316a02dec3706b09fcf3fc6cca7588464
3097 F20101204_AAALOK burla_r_Page_002.QC.jpg
44a27d2433321ae422f559f645e5e646
b63f11ce98acbe1fb1834d7d077a68d8fba937d2
6363 F20101204_AAALNV burla_r_Page_104thm.jpg
7f4111dce2e2c693d8e6e59ef69c01df
4b65783527e08aa39c6a052fc8d7761bc3aef354
6896 F20101204_AAAKKT burla_r_Page_148.pro
23970359fed88a7ee6ca31fe4a6513d4
6ed893525f90485950faa9caa4858044b34acc16
73618 F20101204_AAAKLI burla_r_Page_016.jpg
dd4260309b43c98e7fec403902c1ddc1
41d7b146b874555ef976d4f68b598eb575c52772
5058 F20101204_AAALOL burla_r_Page_147thm.jpg
70eac54d8273be667c51db555e5c58da
ae9a426f5b32a3b71b9c636cbdb3d529dd73c66d
5413 F20101204_AAALNW burla_r_Page_141thm.jpg
5838929a370380707186fea535420707
17c45bb539e8c31560cab814e6d3dc6428f00862
32525 F20101204_AAAKKU burla_r_Page_102.jp2
950b2223842d9cb9d208769ed2898a26
79a3947cfc8a8984f4a410bf28f6ba3ae10fab28
78353 F20101204_AAAKLJ burla_r_Page_017.jpg
9868f2875f0a56dede41209a47b6090d
50b8302f1725aa3040ca95587050af41b61820e1
5598 F20101204_AAALPA burla_r_Page_080thm.jpg
20b0513fe170df67b3f8691aece287f6
3ce5ae30d9acabb54dc3efa11cbf2652dcf02e6a
6177 F20101204_AAALNX burla_r_Page_038thm.jpg
ea962f0a15c36a69462fde9fcd49bf3c76cddf32
200347 F20101204_AAAKKV UFE0021901_00001.mets FULL
d05632fc03585f5f0759684269a83f81
fc4af0fae7a6193d20c9dac50d4eba3399d6e752
5669 F20101204_AAALPB burla_r_Page_081thm.jpg
06e603564d245c494b47fe6aca82bc4e
40c4b56899069caa5bef3615df406dbf8ddc9eb5
12048 F20101204_AAALOM burla_r_Page_123.QC.jpg
136dedb01fd8e88c1e9fb750ed658f075c65df1d
6492 F20101204_AAALNY burla_r_Page_167thm.jpg
77319f41a91021bcfd1e14192a907150
74056 F20101204_AAAKLK burla_r_Page_018.jpg
9ca5ce7be1dda42a80dbaa5fc7ee66a6
57bf50354b71bd161002bc892a9ac3b260ec0451
4309 F20101204_AAALPC burla_r_Page_013thm.jpg
c720ce5ef011b7576a9cd944f565b8ab
1a6c05db057b6e5de175cc848ba31b5bd8fbce84
20448 F20101204_AAALON burla_r_Page_034.QC.jpg
b7060c136057d608b38c1bfee1e4d08e
3bb874b63fb234a7ce9175f4780cb80a6412630e
23277 F20101204_AAALNZ burla_r_Page_104.QC.jpg
cb5cea16ebc2c453c8b57d7d47508496
53035 F20101204_AAAKMA burla_r_Page_035.jpg
c05728b7dcff69eecfed1604962bedfd
1355c9c086c3c4823b556c0d439bf8b21d57f91c
73770 F20101204_AAAKLL burla_r_Page_019.jpg
01cb7b7635beb4fe9f318bcd52ae6260
7c677e6ea2b7675e998248935d9413cd3183af9e
12492 F20101204_AAALPD burla_r_Page_053.QC.jpg
a047ee00048a46489376ec35232c5d92
3ea02b2ce88af92f7f6a2a5824213abd8359021f
F20101204_AAALOO burla_r_Page_060thm.jpg
70c6e1242cae01c27e725013e6d7aa23
3a19053ee6cfb4e461f02a96290ab31eb3542e6e
23488 F20101204_AAAKKY burla_r_Page_001.jpg
0bb77c94c4621fb0f5b890dc2e65bf75
69648 F20101204_AAAKMB burla_r_Page_036.jpg
8ccc86fe57e55bed60df314066534662
51781 F20101204_AAAKLM burla_r_Page_020.jpg
41be979ff69c827b1057a90beb778d13
a1f5bf4d6913d21d2709e37d6b76eedaeba107d5
6025 F20101204_AAALPE burla_r_Page_047thm.jpg
d5395d43a82ab73a13be6ca4bcdb9c3c
908a8dcffc647425348c5ba0bf978bb6b7bee065
6398 F20101204_AAALOP burla_r_Page_069thm.jpg
f435cf8aa0ea6404404ec71de7a037f6
652c58d2bc7d919854752dde01048fbcf9edbeaf
9956 F20101204_AAAKKZ burla_r_Page_002.jpg
9142f6e0473fa508fb87d67e2957b7c8
23b7926ecea9eaf9fdc9362788880e7cebe72f21
66224 F20101204_AAAKMC burla_r_Page_038.jpg
0e50111f4cf83c3f413b5105e8b5c4b9525df2c3
72817 F20101204_AAAKLN burla_r_Page_021.jpg
d48fce350d0a8b36cfc74bdf3afe1d59
0dcd182805c9ae71d5026c89dd7c7319b448d277
5813 F20101204_AAALPF burla_r_Page_120thm.jpg
b4b0c4a7db4159e14ec9fbbda179141a
5410 F20101204_AAALOQ burla_r_Page_142thm.jpg
9d0c208c85cdfcb899faa5ecd4e100a3
3d63231039755d39f45243fe4c4a55cc705bac34
74210 F20101204_AAAKMD burla_r_Page_039.jpg
0b222a0e6c96f82a0916f958719385b0
655b3400b7c9964e45854bfe7800b30606129c43
76468 F20101204_AAAKLO burla_r_Page_022.jpg
c1456ca88fb779f594ae31f22d0ef0e3
8282be9f1c4c348f3a552a5c8376f558b5195310
26681 F20101204_AAALPG burla_r_Page_169.QC.jpg
47631087326b07c0b4b9188ce6e8a004
99bd92f62efb45b23a2dee93ae5c8affaca71cb8
9914 F20101204_AAALOR burla_r_Page_134.QC.jpg
51fc34d0b54259f48994a2d35c91daa1
64317 F20101204_AAAKME burla_r_Page_041.jpg
4a218226091e8ba91242eea92366d406
8febb27593883c92a435db169300755ec3a4cc13
72764 F20101204_AAAKLP burla_r_Page_023.jpg
39797c72d30f6d6954ce168a70b3c3ca2d996f2c
5537 F20101204_AAALPH burla_r_Page_091thm.jpg
f489212c68a5abfdf1186e932b7b1e39
29634 F20101204_AAALOS burla_r_Page_006.QC.jpg
d559bf9f19a78263da2480d2aea33565
86fa29c1870a4604d4f851a5e6a31fb960f508cd
64366 F20101204_AAAKMF burla_r_Page_042.jpg
1546359ab9eeb8f070951253eebe3064
77291 F20101204_AAAKLQ burla_r_Page_024.jpg
60b914cd018f9317ff0ec91ca8d90af9
9524e7dd3329711e71c1d235dc956b6e29e2465d
24178 F20101204_AAALPI burla_r_Page_021.QC.jpg
1cecde06d35f68092aa46dfa6f230613607a87af
4755 F20101204_AAALOT burla_r_Page_048thm.jpg
2cd7dcd4c858e516f6de2c13b3dff561
75243 F20101204_AAAKMG burla_r_Page_044.jpg
fd2befced239451572d6064aa98c2525
c8d35e27e8cbd394395447c3337266065d7305c0
75199 F20101204_AAAKLR burla_r_Page_025.jpg
f9643612df3ddbe345b4a47555c152af
16d51f082a91980048e88973b749308ab78ddc03
12083 F20101204_AAALPJ burla_r_Page_065.QC.jpg
216174b58d4d50eb3cb3fec7bb761804
e0d62d07264fac723537c10301695d16f5dff1c8
F20101204_AAALOU burla_r_Page_062.QC.jpg
68b14c1836f4b1589c33e1267144824a
ff4cd83ed1b00e055b3aecd7931cbc589c3e7610
57339 F20101204_AAAKMH burla_r_Page_045.jpg
53f966ea97f0b80d3cf9ceee3a9a5117
8f7ed7dd14ca0712d4a54dcbc7bf307b304056fc
80838 F20101204_AAAKLS burla_r_Page_026.jpg
c6c82ed30a9e8afb2acf9df81b9de0e8
4f92aee10f1c39e43ca454b384f87172a3abe966
6502 F20101204_AAALPK burla_r_Page_155thm.jpg
15e4be3f47c84e075924f9a6f1616192
eb3beab96fdf2895f8dbe2dc6f07841ba0fcaf57
5831 F20101204_AAALOV burla_r_Page_097.QC.jpg
cd0d88caf09b5303a7ab7c00589ec0d2
f5e3d02cf4b68a9575b39e60a415a135d7f91405
63093 F20101204_AAAKMI burla_r_Page_046.jpg
2988fbe73099beb33d122619cbf80263
e942045a4c0b798c68d39052d41f369a8a494ca0
80140 F20101204_AAAKLT burla_r_Page_027.jpg
44d94150f06ee2d869fcf1a663d4fe17
76bc30547e2372765dfcd7ab28c8e11a808fbb1b
1582 F20101204_AAALPL burla_r_Page_007thm.jpg
955f0e5eb3286d78dd88d23c14aeb7fe
d887f3a59ba94bddf738401cc608854129731e8a
5763 F20101204_AAALOW burla_r_Page_072thm.jpg
bb6c3a2ab225884cef1996346c44588f
67c8e920faef8e9228b9b1d11a484e115d46f583
60438 F20101204_AAAKMJ burla_r_Page_047.jpg
7024d3bfdceb03b994e7244d2e975b82
79301 F20101204_AAAKLU burla_r_Page_028.jpg
fbd31097a76c040a364cf63c3511b837
f7bec488523d1063a2cc5c2650daef0e17fdb1c1
6748 F20101204_AAALQA burla_r_Page_119thm.jpg
e1b6459bbf5485f94d5e99b4fede7a31
458c4e53601d72922bf53884d328b1bb54cd7106
13995 F20101204_AAALPM burla_r_Page_158.QC.jpg
1c6f09aa330b2a4309422947400eebb1
7411979ce06008901d47dc455c64e1848bfb5437
4034 F20101204_AAALOX burla_r_Page_129thm.jpg
c776ce464c475dc4bd8f641614480579
5b4ecb94910d33bfdd1e26bd5e2ce22cea3625ee
47467 F20101204_AAAKMK burla_r_Page_048.jpg
d21b821e56b77219fd7bb62bc0d2640c
8f91bcd877372782915e05f4293c5d6561318484
42854 F20101204_AAAKLV burla_r_Page_029.jpg
a43c6b9941ec41bffc2c1aa45e518729
26173 F20101204_AAALQB burla_r_Page_028.QC.jpg
b7236b394ce90654fbedb0d01cc5514f
10897 F20101204_AAALOY burla_r_Page_138.QC.jpg
f5e7527f18443730e0512c33795b7c6b
75388 F20101204_AAAKLW burla_r_Page_030.jpg
f741204139a572b971aaae91b95bb552
1e1a00d46600938633f04f310e0743bcf8c54920
6690 F20101204_AAALQC burla_r_Page_116thm.jpg
397d73ede3fb9b44512e62e97a6f26ce
2556e6ea38aa274621a771bcbec945a1d57869a0
3749 F20101204_AAALPN burla_r_Page_161thm.jpg
046b04b72a8fe8c5f7f7171644d823c8
6836 F20101204_AAALOZ burla_r_Page_044thm.jpg
4c70ba63b903e0b29f52038a184c6b424f89537e
69928 F20101204_AAAKNA burla_r_Page_069.jpg
8030a8c32dc43708f2d685b7ca17fcef
dfe3e90d898359a85cfd2f1028ce2a750b802e2a
27730 F20101204_AAAKML burla_r_Page_049.jpg
9627ffd149dbebd57d998ef1e9ed5b77
e83e6f33b709121ac896445e3ee3204bc63de9d2
72318 F20101204_AAAKLX burla_r_Page_031.jpg
3ff6c1efe226b7e17c5be3d417902c3d
490e9e274e2f44924ff3e8efdb72f5fdd047cd43
17335 F20101204_AAALQD burla_r_Page_085.QC.jpg
b50bdfd7d0b7b7b93719287525bb0625
72089eb89e696177b2bd8e6683a4455011027383
10304 F20101204_AAALPO burla_r_Page_054.QC.jpg
fca458145bc8e487330ff707705e4e72021f2306
48005 F20101204_AAAKNB burla_r_Page_070.jpg
7227303c6848c06cd7e501c93927b8a9
36281 F20101204_AAAKMM burla_r_Page_050.jpg
2b41679060a802fef0204b3241dc00a4
fef18ed1b0574bb202f032ffe0c11e12bdfaab1f
56507 F20101204_AAAKLY burla_r_Page_033.jpg
a19747f26b46843fbf635aea35cb3e2d
936dccf029783f01cce3f832b7842a73d286d19c
F20101204_AAALQE burla_r_Page_034thm.jpg
888bc70b2d82f7a2ed1213a8a16399f9
71e0f3bf12f0ab4a49360150ee40186580e77af0
11131 F20101204_AAALPP burla_r_Page_137.QC.jpg
df15e22526260b891a0aacc86cc869f4
b2952cf553a170d774e92e2033a38b1298d9c833
59793 F20101204_AAAKNC burla_r_Page_072.jpg
3d86064f30dedf0595dcbc7790a18bf0
22500 F20101204_AAAKMN burla_r_Page_051.jpg
227e4d440f943f76b63014db6db75e8c
af51c6402d7ac65c259fefef8fd0dded4233e8fc
65445 F20101204_AAAKLZ burla_r_Page_034.jpg
9c2c7d8694c62f9e116067c537940016
3278 F20101204_AAALQF burla_r_Page_096thm.jpg
4844eec36cbb165b48a67d9070ff1ea23cee1a23
4582 F20101204_AAALPQ burla_r_Page_151thm.jpg
0af67dfb2d7c1dcbb7681508d2287d25
53ea3961a14a1a0bdc93011650961cc9fe74428c
65453 F20101204_AAAKND burla_r_Page_073.jpg
e62b091a9cc3d895a589d2674bcdd316
1d395f1b2aee267341e51aee14cf7967cbd0dd31
37574 F20101204_AAAKMO burla_r_Page_052.jpg
14422 F20101204_AAALQG burla_r_Page_029.QC.jpg
2583f9f9d17a800bb877001575f5b373
d7503ea9886dab24ae155773ea468baaa908744d
7073 F20101204_AAALPR burla_r_Page_024thm.jpg
0a32247cd66723f7ee8a90253a721bd0
4239cdfbaf348b40f55baba1eb455930baa68988
69390 F20101204_AAAKNE burla_r_Page_074.jpg
b322fd76876e2ecfeda2d96c4df599bc85907532
30871 F20101204_AAAKMP burla_r_Page_054.jpg
87e25c27ebd370f23d409995e10a8d1d
b59f5952c06680e8959d1ba22cb3450c2fc6d154
4333 F20101204_AAALQH burla_r_Page_149thm.jpg
be67d44518156451c820b4e614eabcab
18a9e3c60544b32d7a0454029befe66ebf7cdcbe
2525 F20101204_AAALPS burla_r_Page_051thm.jpg
8311e0936c9974f2b35e50e6dc1a84f1
58788 F20101204_AAAKNF burla_r_Page_076.jpg
21ae241596138c84472265b6025c4333
4ffc9cc87e489569c34428690b4c348f38bd265c
75040 F20101204_AAAKMQ burla_r_Page_057.jpg
aa5ef3e30f38891878aee36689f25831
4117 F20101204_AAALQI burla_r_Page_053thm.jpg
7d6422386e18f734869e3e6f1dcc1125
bb6a9fae29984ee474d57125116f6dfcd9f4d3d0
27567 F20101204_AAALPT burla_r_Page_170.QC.jpg
9401557f382bf3836a7d00009a153060143194f3
49352 F20101204_AAAKNG burla_r_Page_077.jpg
68594 F20101204_AAAKMR burla_r_Page_060.jpg
fd369a22045bd62c0f2129b70106d41e
3256276ae1cdf6d29acdafb20513b81f842dc88f
3795 F20101204_AAALQJ burla_r_Page_137thm.jpg
88e749ece47d23264de58bf1679df260
447313719967286d9925db4f3fbe200bd91d1473
6803 F20101204_AAALPU burla_r_Page_057thm.jpg
20a053c33457e6fb5711f7f77726378e
c032a774a0ef82253e81b0f103c0b77e348fe4c1
48627 F20101204_AAAKNH burla_r_Page_078.jpg
9f220f4290658034f40738ef666aecd0
005672e334a268769961ea4a8ac18fb0e3cd43cd
71179 F20101204_AAAKMS burla_r_Page_061.jpg
3000f29a492f59f6672df38ed5bd7d03
cbf9e69f4ebee9ac977a677d73b744ae17f54bcc
5866 F20101204_AAALQK burla_r_Page_032thm.jpg
ab5920e200a874ae6bf4033759c25b53
5f0b6a2107b059779360bf69647fda79921f7d33
23353 F20101204_AAALPV burla_r_Page_115.QC.jpg
5cc865cd8d529c7e0b98036723a04e6a
5566b5d2f7ac760cc1fe32ac2df88890ec04a5a0
61215 F20101204_AAAKNI burla_r_Page_079.jpg
1f39d346b6873d3144556700216705c9
e4f8d51a49ccaee6f3a1a380b6e7105bf94ca7c2
65674 F20101204_AAAKMT burla_r_Page_062.jpg
a826316c3e339a82d00fabd51e96fa44
19e07c2875039471e6bf3bf34781b7edc2ab52d6
3571 F20101204_AAALQL burla_r_Page_127thm.jpg
4826baee861b937675d77b4a445b90f0
7c742b4c46182a97a83a231f9c0f7326c176dc0b
6739 F20101204_AAALPW burla_r_Page_025thm.jpg
6d6d94e8a830788bf47b5040f583eb5a
57018 F20101204_AAAKNJ burla_r_Page_080.jpg
30fa38a98d19ddcd4ab6faae98d66fc4
940ff50249a6fb248c1ac524f089fd79d0294af2
30428 F20101204_AAAKMU burla_r_Page_063.jpg
426d1633e71bdafe4b7c204a9192a550
82ee0d03d618edc591043ef7b5fa91cab95bdf34
6795 F20101204_AAALRA burla_r_Page_172thm.jpg
dc41fec0951f8eb1c46e48fa38e8a3b2
5d63f04e8465f5ff3dbd8617cc459242c86f96db
24690 F20101204_AAALQM burla_r_Page_025.QC.jpg
b1bb49695d20c2ed063bb8185e40574f
e5013c89b14b5176c6cd3e45c164af30ea026964
26696 F20101204_AAALPX burla_r_Page_027.QC.jpg
d68a0a97ea93d16b06bfacc4263292f1
2f1d242e65b8fa853062b8d6676dc8c06025cba0
55552 F20101204_AAAKNK burla_r_Page_081.jpg
c765f08933c5df35fc1dedc6ed1611a3
0db3019f215111ed23f5bae60b8db8366c2977e4
35795 F20101204_AAAKMV burla_r_Page_064.jpg
b8b7eabcedf57b959f94cc407dc2704e
f0d720e88dfb8a9c16df81d5954d03c865d05a85
6155 F20101204_AAALRB burla_r_Page_082thm.jpg
4b7637e95e9c2dd8e6f89698c125ca44
1aac4a2bd77f4368887ba3e96f30f7c842fcc16b
22974 F20101204_AAALQN burla_r_Page_071.QC.jpg
588bd64e3269ba920b4e7019eccc0f87
e7b54cb206a6bee92dfe903fc4850feffff6fa40
15003 F20101204_AAALPY burla_r_Page_148.QC.jpg
bc7fa9fa3a0bb8f30f0ae4ceef5a0a65d728b47f
67844 F20101204_AAAKNL burla_r_Page_082.jpg
f1052534244997aaa9f3127d182d9761
bae7a6a40c16c3f12ef21dfa2043f2684676d3e8
36105 F20101204_AAAKMW burla_r_Page_065.jpg
2d49380d6b51b1dc6722bc5a6144e742
c6acd777550d7036a1c8a8f934b872facbacca35
16087 F20101204_AAALRC burla_r_Page_151.QC.jpg
c28cd158b5437a759cea54fe72944fac
264d73214c56433ff13595b92f33c7817957cfd9
24576 F20101204_AAALPZ burla_r_Page_016.QC.jpg
511b26d3eb2c3fe8301712beb74088ca
c2900fb03c68ef412aa2ec8bf77a0d9f0ddf43b9
38475 F20101204_AAAKMX burla_r_Page_066.jpg
339b5fe612bf94e4fc8d254e207fa97a
23882 F20101204_AAAKOA burla_r_Page_101.jpg
715c60dfcc4be39ba3b28ed30618fc53
c0491ed082f1acacde53f06a6e2d03ebdb13d32a
5508 F20101204_AAALRD burla_r_Page_086thm.jpg
4b4c801cb299f5d2145a6354f52e0484
3041482a3c4a05e427f0420c4d487c6ed8546cdb
22368 F20101204_AAALQO burla_r_Page_084.QC.jpg
79c5ac3899feb6d5ee8ca4e19fb0f062
9c53ce7ee2d3d0898e815e418030a61b213591b3
71483 F20101204_AAAKNM burla_r_Page_083.jpg
7c6734044be516467605aa119fd4dca2
aa366f0779822bc2a7dbb2d11f195601aebbafcb
64129 F20101204_AAAKMY burla_r_Page_067.jpg
d28891b96bb42ed28b83a41cbb48a662
48d732cfbb85a25297dd4c87dabc06b224481be2
72302 F20101204_AAAKOB burla_r_Page_103.jpg
93ce9629e98d24f5b96b5498b2d4ef10
1a491cc927ea5d520d9ecbcafe53dd3274ac8046
9705 F20101204_AAALRE burla_r_Page_063.QC.jpg
921d64731654521b05b6a8555ed7c39c
e15da92824098a4a459154650527a08fb49f6eb7
3577 F20101204_AAALQP burla_r_Page_015thm.jpg
5649569ffcd4cb90081e7f0ec4fa1d6d
8b979178d22a97023953cc81f49d1dcceb18347e
66913 F20101204_AAAKNN burla_r_Page_084.jpg
fa47066072be09025e4a1f6da336a723
25c211eb82bcec4e188c6210f9eedf6f4d9e5be4
57789 F20101204_AAAKMZ burla_r_Page_068.jpg
2748d65a8091f2f087275dc1c8e2347317b851dc
69137 F20101204_AAAKOC burla_r_Page_104.jpg
67fcdfb2d0610a5e1958c1fe899011a2
793ba502c922964e880f0c16fdfc6c6c7630c1cb
17422 F20101204_AAALRF burla_r_Page_087.QC.jpg
13046 F20101204_AAALQQ burla_r_Page_157.QC.jpg
8bda3cb9da7494f741b2c7ea4296627ca98ddeda
52948 F20101204_AAAKNO burla_r_Page_085.jpg
ebb44bedcf93f8f7ed0dab559c98bccb
9d916b44172775011cd3aaff64341c7950521ec9
77621 F20101204_AAAKOD burla_r_Page_106.jpg
226c7bdb4fda860676544815e4144be8edc6dcf5
2176 F20101204_AAALRG burla_r_Page_008thm.jpg
2574a7a12c7a525db468ac1b7d961449
048cccab3a50cface872507c53ca3b85f7c7a161
24424 F20101204_AAALQR burla_r_Page_118.QC.jpg
916c23500a03a7b8aa316176bef22bec
b36f6278ccd654e44928b4912507cd6a1e19133d
54610 F20101204_AAAKNP burla_r_Page_086.jpg
0221b347385fc9c2f17e035c6eb06e63
1b011ee3b1ac50a8e48b874b9016ec0997614519
70464 F20101204_AAAKOE burla_r_Page_107.jpg
659e9443ac7d12b0a72986722364c45c
1335 F20101204_AAALRH burla_r_Page_002thm.jpg
be48b49b9a68366149aeb2370662b133
a7fd13862e02a38308cb41c9bef08d2af10728cb
25712 F20101204_AAALQS burla_r_Page_106.QC.jpg
4394239cb80a7ae6cda742480eb9a2cb
00c147782480703ff2cf35bcbd51eea5b62aa226
49781 F20101204_AAAKNQ burla_r_Page_087.jpg
bfacc14cd6e703a7ca26ab3f4b8073b4afe2731d
60468 F20101204_AAAKOF burla_r_Page_108.jpg
8a59d226479c1d1dd6c3b9c127abd739
8c38643f04e4d0a3a10ee520a95e9cd51abfb3a1
4262 F20101204_AAALRI burla_r_Page_066thm.jpg
9c6acee3cf490aa83bf101a23b04325a
423c6df14f2fcd269294ecd0bc62e877fbc0436f
13169 F20101204_AAALQT burla_r_Page_140.QC.jpg
3d9d37c98242842ae306c9a0358611643a937d05
70500 F20101204_AAAKNR burla_r_Page_088.jpg
e844a6c9c8ec8857b7955e8597a2a913
62e7b437d8f380d5e5378b2cd522d3cf778321fe
73373 F20101204_AAAKOG burla_r_Page_109.jpg
587f7109a092ca583f07efb749d9027a
10056 F20101204_AAALRJ burla_r_Page_055.QC.jpg
b772017524a16b1b80611e6ee0388d2a
26ae4dc7b1e8ce2e6f8d4d3473f6ebc430297793
22294 F20101204_AAALQU burla_r_Page_090.QC.jpg
8f2d66944108a924ced46a6de67982ae
51052 F20101204_AAAKNS burla_r_Page_089.jpg
74b8057fed61a2bec0ddc63e09f32150
1ef845756e55f95582155df399c27045f924953e
60278 F20101204_AAAKOH burla_r_Page_110.jpg
f1ab22e89e9e2be274ec224231654da1
75c409ed3ef95451e2cb263084d4b2293b74ce4a
6681 F20101204_AAALRK burla_r_Page_088thm.jpg
9a647345aedb55f1527d960530d11d99
6404 F20101204_AAALQV burla_r_Page_008.QC.jpg
b511430c7c34d5e300212f785510fbe6
83e926f60abd1db29a95d4fa792ac69c37832ab2
66418 F20101204_AAAKNT burla_r_Page_090.jpg
d66d22ef55d8dd042927322af510a389
ea76bea2a4e50a61017f3b00bd6ab3fc5b3cbbb5
63488 F20101204_AAAKOI burla_r_Page_111.jpg
07e81b670b91f2a802bae14f4f737660
d648495a3911bd37489f0f44d6891d60032cdefd
15542 F20101204_AAALRL burla_r_Page_144.QC.jpg
d29906a21503d11eb3b56448e56b70b1
2a149f3aebbee4c4310ae09f41189f4912b32075
6102 F20101204_AAALQW burla_r_Page_073thm.jpg
77cfe9e059a8b1276ca317f90597f1ef
806219d6a49376f1daa6a7e138f1ff5d0cc07610
60858 F20101204_AAAKNU burla_r_Page_094.jpg
442be6f6132aa5e7dc583de155fcbb3a
42d25f609dc0816729b84f05fb68026bfe58633b
73563 F20101204_AAAKOJ burla_r_Page_113.jpg
1e4d931221274bc20b406a1d8dda697a
468fb0e172fac9c8172e12ac11a90fce2cf5389a
25684 F20101204_AAALSA burla_r_Page_017.QC.jpg
5e693db6724cae21b9b6f2d3344bdbb0
4592 F20101204_AAALRM burla_r_Page_158thm.jpg
63d5413e084df3a31dc580fcdcf91cec
92b8a2e9815238c391896d7a10cd6ec138f2ae63
25721 F20101204_AAALQX burla_r_Page_172.QC.jpg
50500e730fbf00d4cf866bbd0a4110c4
1212f20f9d97b95e294f4e70953be65f8af5157c
27444 F20101204_AAAKNV burla_r_Page_096.jpg
2da405cd90f8ccf525c3b5920d73e1d0
0810fd60ac2947cb7e91f4efd0dc5ff5532f08a9
69673 F20101204_AAAKOK burla_r_Page_115.jpg
d999ef19ab261ca53eb4f7f3d727a158
3c5e3d5df0f03b366d6a380a91ffefcbb4aba6bc
17521 F20101204_AAALSB burla_r_Page_141.QC.jpg
82bd635c7cb69d50052ec4c009183828
a582d29d6622b536e6d920db4a216cc00d26ea2c
19150 F20101204_AAALRN burla_r_Page_108.QC.jpg
85e9a3e77cab011b19a9723bce683325
24521 F20101204_AAALQY burla_r_Page_009.QC.jpg
30f4e7c18864ae47fbc62c4c8507e3b4
fd14081884b50ce7afeb65e0c297f3103a5fa451
17510 F20101204_AAAKNW burla_r_Page_097.jpg
b58fdca3736d8c7ae3b82a4edc15ce25
2382ed8b9a3d0e47b4d2758012eef97dbe278785
73456 F20101204_AAAKOL burla_r_Page_117.jpg
8246e28842e6323b4743045c6f26b080
eafa33870f4d46422ae47c94384b447a190daed7
9201 F20101204_AAALSC burla_r_Page_159.QC.jpg
d67655be147edd86b98b46ce31691320
3a53863350414c7f85619524911edd20c71c3047
25967 F20101204_AAALRO burla_r_Page_012.QC.jpg
08a7838a2ba0a9e714f5760aa3988f93
6971 F20101204_AAALQZ burla_r_Page_171thm.jpg
6ef3fb7dc5bce4b765378c2dce10d114207008fe
32985 F20101204_AAAKNX burla_r_Page_098.jpg
de38de4a0cedf345b12cf4e057ea46dc
f445ee5c13455154e44747f106b51b75a7ffce74
34196 F20101204_AAAKPA burla_r_Page_137.jpg
9b1e93155fd3f1b56a89513bb9f98b79
4c9824b45db60a4a236db7e23f9979fb73fe1ec1
60173 F20101204_AAAKOM burla_r_Page_120.jpg
cdc46f3937f3d57bf7472f24711645ff
17241 F20101204_AAALSD burla_r_Page_142.QC.jpg
bdc5dd2e8ac5bef7ac01e70651e7f29f
34358 F20101204_AAAKNY burla_r_Page_099.jpg
38fa98555b34726d7ff16969b1965e47
32866 F20101204_AAAKPB burla_r_Page_138.jpg
64ca001273484e8703efb284733aacc7
cbd7665176058a837a5ce05a682f3d9f60dbd8eb
5770 F20101204_AAALSE burla_r_Page_059thm.jpg
ecc705ab701c770696037d655c2ea4c1
7ec609d839c1d7817b84bf4cca4aee72ae4a9cd6
6950 F20101204_AAALRP burla_r_Page_017thm.jpg
9f35fa8db1ce6abedc35e4ea5876bcd6
dccc59a79f45b46845008eb2fa152043d692f6f5
42462 F20101204_AAAKNZ burla_r_Page_100.jpg
d04bd893307f1c5125125e7b948a80c2
36769 F20101204_AAAKPC burla_r_Page_139.jpg
cbff6079f2d8279bf4187f247df91248
e7c13d770a840092186529cd81a0ec6383c75bb9
25436 F20101204_AAAKON burla_r_Page_121.jpg
96c0f7cff966f481c75f1536cc7ce86ccc523e23
11548 F20101204_AAALSF burla_r_Page_052.QC.jpg
8920ac4433576a7ff1dbacc1b611d6c7
085544b41ab00765944d0ee7552405001348af1c
23856 F20101204_AAALRQ burla_r_Page_088.QC.jpg
80c124b587b588012402a757b2d652a2
21815b3d3fe7469e073232cc2b1bf7c6d42710fa
61570 F20101204_AAAKPD burla_r_Page_141.jpg
e66cb9396ffee9ffe46e96de050bcebe
04c141107f7c9fcb6711ae0120813180ce9e27f9
19163 F20101204_AAAKOO burla_r_Page_122.jpg
273cf2ed6351edc3b764d9439499a214
18668b289345fae619c7fde1b9791cc6f16577bc
6097 F20101204_AAALSG burla_r_Page_042thm.jpg
f6d1fb3d29ece69ff3ab0c5ef388ba9d
63e52fc5742c1ed56c40e6b44bd5d85cbfdf8149
6003 F20101204_AAALRR burla_r_Page_112thm.jpg
1d878ae4457e9e8bd64833f39463c00f
79b324981269aee495caeb51cc6cd00ab0a42cf7
58723 F20101204_AAAKPE burla_r_Page_142.jpg
c1d687f42cecf9dec2abc16a043c8453
a4e2a01e010ee0e11cb6422633a86426a4eeef54
35730 F20101204_AAAKOP burla_r_Page_123.jpg
3ff71d8728274f8c2734b2709882ae82
3d9a522ae04af2472a103e41d61736ece3fa6ebb
3089 F20101204_AAALSH burla_r_Page_146thm.jpg
f7a1cd22a87256dfe428a2af104e8909
0a4bf33e6368a0f09378e73c71052cce376a4bc4
6249 F20101204_AAALRS burla_r_Page_090thm.jpg
2d2d4a557d50026d0028a1cabb1078f6
28245 F20101204_AAAKPF burla_r_Page_143.jpg
b6134c38f857b851ec4a45d31c21a5c3
33781 F20101204_AAAKOQ burla_r_Page_124.jpg
0b7a356af3fd532312616942bc008167
aa6510bf5a64bffde8556aab9254b0b3bc303652
6827 F20101204_AAALSI burla_r_Page_021thm.jpg
9ce28a3f2dfecfe6119dccdd42ef7b51
29a064a23d8bf7196000e9c93f82a6e1e95f7139
16258 F20101204_AAALRT burla_r_Page_135.QC.jpg
b648d9c76290150ee2ee9032b70ee9eb
566d0261292e4fe7abcc47706c851e2a42436af0
57209 F20101204_AAAKPG burla_r_Page_144.jpg
3bf16396aca1c1925275251ecbc99cba
2d1a0cf88521df2415bef5ee3d54c1bb4975d421
53968 F20101204_AAAKOR burla_r_Page_126.jpg
22ec2c91ca464d9bf2c7e2712718f5b7
e2b8940580709ce45131cc07fe41e94b4e53201e
15894 F20101204_AAALSJ burla_r_Page_013.QC.jpg
564bc09feccb1d3bd474a69276473979
fc0d224e53d3f81731131a9f6e310445f4336bfd
6792 F20101204_AAALRU burla_r_Page_012thm.jpg
03dfa82e09cd4244ff90f27816e115f3
1fc663171272c4d802350d7684dddf13e02a8a42
55702 F20101204_AAAKPH burla_r_Page_145.jpg
d6a94bbc374c41034b8daba84fab8a72
86b4d6e544f089d7e7c329723e9ff6e84d16000c
35117 F20101204_AAAKOS burla_r_Page_127.jpg
5554c82547c258d75db237e32e43b3cc
4f2faf97cecf9546bf142af71c1458fbf5fafa40
24233 F20101204_AAALSK burla_r_Page_113.QC.jpg
f47ac7830364a6d718294ec7f5287b81
1ac25af5b84c20db3bacf659bd957d3dc2c4465a
24240 F20101204_AAALRV burla_r_Page_057.QC.jpg
961c88de17a27b93b05367eeecb862fa
40f08fa4e23098713f52cd05ffa8bcbafb2c64d4
31815 F20101204_AAAKPI burla_r_Page_146.jpg
a29a1b691c03ce577d5035e74b6de720
765dfe54562ab361c4d2643a69bc07a8dbafc3eb
45420 F20101204_AAAKOT burla_r_Page_128.jpg
ff381c12bd2af883c9f686bf455b3299
2529 F20101204_AAALSL burla_r_Page_122thm.jpg
57d7dc67989638e6cce735b77e0237b9
eedb78125531bd6f7fc6ee65c78c46961757595a
19305 F20101204_AAALRW burla_r_Page_080.QC.jpg
03cc7e641cbf750a9323d942f083f092
b6296471f65b85fc53c6de49e18ee3d85a63e5ba
56896 F20101204_AAAKPJ burla_r_Page_147.jpg
0a3fafbe583777559930956d1003bc1a
bf493cf27835e77c062edf8f8e5b0b5210b3b635
42388 F20101204_AAAKOU burla_r_Page_129.jpg
108ecc1f86a9f2e2eb55b94f346e9fee
6ba5961bb2da2325934a7c8aa942d9eaf0b3f922
9392 F20101204_AAALTA burla_r_Page_163.QC.jpg
cf6e138480c0a78f44197ff11428177d
d5a87f448f3cf822c127753ce565d5a33056fab7
4043 F20101204_AAALSM burla_r_Page_007.QC.jpg
ca4abbe85812e6db161db2d975b4c194
a58d45c40e85472e7718b8e197af04bc3dee9585
19774 F20101204_AAALRX burla_r_Page_072.QC.jpg
39aef41935aaa1c25384a0d4e9b73081
1b7631eaaaff782f049ec40bda237700ff2a0556
50944 F20101204_AAAKPK burla_r_Page_148.jpg
937ef10cdb06de266cd01531d9a52caa
7fe93bb262fa6116f3558e0af9a7c36da04e336c
35846 F20101204_AAAKOV burla_r_Page_130.jpg
3b17043ed0a28d803dd8036904549fb6
15455af33eaa6e1882387425f70d0a6c3264c163
21656 F20101204_AAALTB burla_r_Page_038.QC.jpg
9f6e52f7b3372fe667ff47a69d21ddd6
8caf7ea8947861e83dfa2868ec31cae42bcec10b
2501 F20101204_AAALSN burla_r_Page_102thm.jpg
2557aec4f5ddb3f6702e539b55f55ede
bf7ff0077abb226a30fe3d1811886ef42e7f0c1e
3481 F20101204_AAALRY burla_r_Page_098thm.jpg
3bcafa193faa65743f1bf2a59f029eff
9daaf831d361907f22d501f9bf9e434f64437dcc
48640 F20101204_AAAKPL burla_r_Page_149.jpg
b1cc4c10bace8a8b826168707cba78c1
56fe1abff2f6c8f7d181f146a2dddff2299ebd70
22838 F20101204_AAAKOW burla_r_Page_132.jpg
5c2736d9cb07f084a17cb2dbbb04a2f5
548b5417427ecaf977bc3fe6b187d4c38cdba65d
2493 F20101204_AAALTC burla_r_Page_162thm.jpg
f53e02e48452e45b47ea06007df15c85
2aa358e641dc0eacf194c82120b67eded4e90cab
2676 F20101204_AAALSO burla_r_Page_174thm.jpg
69b1fc686762a280ae827c363c7b59b3
3bb7acf4f8f47c58761e5c7e834f0a6265ecbae4
21575 F20101204_AAALRZ burla_r_Page_042.QC.jpg
169045ecf215126e48e68f109902fc4a
306abeb96eab8d602afaa21d0d0b3638325b6f53
75222 F20101204_AAAKQA burla_r_Page_167.jpg
97d45a9eff64affbddc4ce5129a85f69
551b90e68c5fa310f7795be3834de4d09564982a
55120 F20101204_AAAKPM burla_r_Page_151.jpg
b1966012f6f357a621cf0016db55d7e8
97a4cfcba6131f2dc7f9cfe4e59e7d5fb32c4d3c
31320 F20101204_AAAKOX burla_r_Page_134.jpg
215f465e9ba0275cd28e952476ca8de6
91d09ff8e282991607893bcc2da351912d7dd8db
7256 F20101204_AAALTD burla_r_Page_164.QC.jpg
668d29eecba8674a9856e63d889382d9
10002 F20101204_AAALSP burla_r_Page_127.QC.jpg
3a5b7f32ae3219f7d79877eebb0dc003
23c7b365db218e4c7742df487be63feb6cf5796f
21264 F20101204_AAAKQB burla_r_Page_168.jpg
0c638e2e85db9a5ca2b22c3fb1a5f6fb
3b8f1e43e0536f56313e359baf949fff6a85849d
57423 F20101204_AAAKPN burla_r_Page_152.jpg
b6251ec1dfa95e47ccd90610774e2fb2
f893d7ecd8f6f3a0720277714ecf591ee78ff4ff
59444 F20101204_AAAKOY burla_r_Page_135.jpg
a3fb2c9dc12d6de4b40c085f9ab50310
b93f57bef1ed15ece0276fdc35fd2104a86ddd85
24781 F20101204_AAALTE burla_r_Page_031.QC.jpg
d386ce54c58783b58fe883c01aed36b8
5cdcf0e48ced950c4f92aa0384cbb83db6293e53
84107 F20101204_AAAKQC burla_r_Page_169.jpg
c254fcb7cc5768b5a0bc28ca5c2468dd
bf6405eb6f3701792369ff83922f4f66b86afdac
43653 F20101204_AAAKOZ burla_r_Page_136.jpg
b5864f775d63520c98c3ff847be69fd3
8684d34c94e2fa71b3086a7a38453bbf10a20beb
24310 F20101204_AAALTF burla_r_Page_018.QC.jpg
da23774a806ba96b7160094748708399
cc712ec3fb008cbc7e8405ff8f29945fb700745e
5953 F20101204_AAALSQ burla_r_Page_067thm.jpg
f34caabe3a9b37970cf47dbaa3d0bb0d91178d5b
89535 F20101204_AAAKQD burla_r_Page_170.jpg
86e49f775ea5e223ef25e9e0ebfa5d37
31111 F20101204_AAAKPO burla_r_Page_153.jpg
d85d76bcdd8d9693d8b5374122f09351
099662b8e587392ce09876e874bd52621ebe55c8
3892 F20101204_AAALTG burla_r_Page_138thm.jpg
2de2711af451567ccdefa99c9eaf8895
e90552d3c92127f5fe6c1f5c24293c98bd7ee21f
7044 F20101204_AAALSR burla_r_Page_039thm.jpg
4b111c953c4f9acdf00e6c5aac141deb
5b28ecbb39564eac087573c6f90d86cf475c1cbe
22128 F20101204_AAAKPP burla_r_Page_154.jpg
87282 F20101204_AAAKQE burla_r_Page_171.jpg
53423e598507f08716b176b88ac99834
8f998b4bf9d16190960f843d955a15f66c7931e7
25679 F20101204_AAALTH burla_r_Page_024.QC.jpg
9a94369700b4b6aaca6c86e25ac3c20bbdc1084a
16047 F20101204_AAALSS burla_r_Page_048.QC.jpg
86cf366cd588afc0b92d8bc78266b3ef
d4db73ac0acfe1a7871f79795ab3bd8b3269a851
76086 F20101204_AAAKPQ burla_r_Page_155.jpg
93835200e562af49be8abc8f3ee5e8a8
0180ef6d2ce72670292ffa028291a233f4424c42
85539 F20101204_AAAKQF burla_r_Page_172.jpg
bc2bbf626dbc2eda0886c571a46a73cc
dd6542b8be0d319a1c230525dc9b3f33220461a8
14009 F20101204_AAALTI burla_r_Page_100.QC.jpg
546be3170c5cfbdeaccebf7ea80bf1a53fbf3a29
12305 F20101204_AAALST burla_r_Page_161.QC.jpg
195a4d2b106fb73f6f876aedbfa1e844
78e577e5ef3e43896a8c54fb004cc6742a4e28f2
37475 F20101204_AAAKPR burla_r_Page_157.jpg
ede6c123ed09cce2af56e62cbb105864
49e44e0fe187d59e63875da5f8f395ba38ced40b
92073 F20101204_AAAKQG burla_r_Page_173.jpg
1eee72a44c887c19672d909b3e42d611
d85b8286aa6238d324d59224d5de7ab360097d37
3667 F20101204_AAALTJ burla_r_Page_055thm.jpg
586d00498301659412e957d884dfe281
90b2d7071d9e18498bbeec3af9235a8b5c1234e8
16120 F20101204_AAALSU burla_r_Page_070.QC.jpg
0b230413c1c03aff38ac17fbdeb88ff3
4920b6bf3d0d17e47057162511b7976996659f2d
37736 F20101204_AAAKPS burla_r_Page_158.jpg
9853e9b8e89d8c0b95bf267df4eeb573
dec7d1d65649bba34503394b8e25c9c665e89351
30201 F20101204_AAAKQH burla_r_Page_174.jpg
a5006f59280f3ff8f94a6d6f8583efe3
f452f6c64614e9e5e8ec3e64080efbbf0da351c9
23597 F20101204_AAALTK burla_r_Page_107.QC.jpg
aa9688ab5c00210fb02d714ca08609ea
db458bf11a9cf8e54e6b3cc27cf01bc86737ee70
11644 F20101204_AAALSV burla_r_Page_015.QC.jpg
42c17d4f8aefe558051940a87f366efa
cf31c699d768675fd04da26d9177d6ecec34c95a
27784 F20101204_AAAKPT burla_r_Page_159.jpg
94bf17cce4ae21dafce8ec0ee86034c8
3ac68df0f15305ed7a5b4a02a8b2c1f4fb2d2908
34856 F20101204_AAAKQI burla_r_Page_175.jpg
28bf9c7dd46cc28940f6c9b1a346e89d
de5ae19be9efd4d1b133d53fb50cb7e2903f7d2f
6635 F20101204_AAALTL burla_r_Page_018thm.jpg
e20e1fc1c3de999d60ecf2f279e21d10
677c03ba3985bcb09ca86d74ef092ce798dedc65
15532 F20101204_AAALSW burla_r_Page_004.QC.jpg
b05aed3ac5e58871676bdf0b7c42da06
da83af610d2c332a900c2f65e2dc2bc55af027e7
39693 F20101204_AAAKPU burla_r_Page_160.jpg
292e4489041ac35ac880e866f13deff1
e2d36b68fc7a9b2cd534dde7848a8bdfa4faa681
25555 F20101204_AAAKQJ burla_r_Page_001.jp2
0a10ecfed2cae4d3c6bd755253bfee20
542e082121df813dc829749c9a07c778c01b9b77
6843 F20101204_AAALUA burla_r_Page_016thm.jpg
823542352c64dc584a24b46d934b42f9
d284696e7a84ef057facb6a06c330e07334c19de
6150 F20101204_AAALTM burla_r_Page_084thm.jpg
30698f3e0341cd5faca686617cedf8ce
27aebe9e24f15bb5715edf103ed458d622498a10
25113 F20101204_AAALSX burla_r_Page_022.QC.jpg
77a00e2c6ec11871377fe04c0a38bbff
3426eff8fd700e7cb50756a73408779f870eaee2
38498 F20101204_AAAKPV burla_r_Page_161.jpg
30318d2d81ac878578925084d2077bdd
b52dab02a1c4728aab07b08c0a1b3f09169b4b43
4428 F20101204_AAAKQK burla_r_Page_003.jp2
1cc6cdd9bfd482db92f5e16b054a17ef
70c105cc67ba5681a0a70a84bb09f2de66d12dc5
6430 F20101204_AAALUB burla_r_Page_019thm.jpg
8c166d8bed377df22b878acfb7d7892c
ec7bf3d7463d5d63e0ef3140e76616376be2b1d8
17672 F20101204_AAALTN burla_r_Page_093.QC.jpg
5989d3e6edfd48e8d11e723e31935664
c68f14a7a275cdc6462a0547c01fe474170fa6fa
22027 F20101204_AAALSY burla_r_Page_041.QC.jpg
275fd5c83d91a7b3998ae6be45a1c209
249d4ed9114afc0d480841c8c72531413569ae15
21292 F20101204_AAAKPW burla_r_Page_162.jpg
c26f791aa5b558d4522eac07f8356b08
e9c780e868b943d6e453b6ea6daf5c3abfc000b0
68015 F20101204_AAAKQL burla_r_Page_004.jp2
82c93011e4365aeec6e5d7e230429c89
548e6c407f08f0f8fde8f1799eb58534350fd1cc
24555 F20101204_AAALUC burla_r_Page_023.QC.jpg
9c0b8ab40653d3143c3d9d7fd90a6245
238d24fd9d1bde16b1426200d1edee078eafd352
22623 F20101204_AAALTO burla_r_Page_082.QC.jpg
b55a4debb1491d049ffd947266efaf8b
6431 F20101204_AAALSZ burla_r_Page_036thm.jpg
278e3aed157409f072287f26ffd4367d
a0ce2a14862e20cdbd095ff3de2ab9f2d55e379a
28007 F20101204_AAAKPX burla_r_Page_163.jpg
2d9d0f32535a7db48c85d21c639c55eb
6bc711689553d0d257a8f54d9f9ab93ea3ed2d82
108663 F20101204_AAAKRA burla_r_Page_021.jp2
70382f5b80fe10e48eb09b24b2b6e904841920ba
1051986 F20101204_AAAKQM burla_r_Page_005.jp2
83ff8937ef344acc1035f7249fcb6f91058e4be4
6707 F20101204_AAALUD burla_r_Page_023thm.jpg
5b3f5c36a691d167618e5083866fb042
03dff7c023d83e321a4338706bd50bf6a0ab6468
20572 F20101204_AAALTP burla_r_Page_112.QC.jpg
915a1067fe8bd39c5d945587866a792a
90b535000cd4cd464da1a5d831954f990658db84
22018 F20101204_AAAKPY burla_r_Page_164.jpg
24aff2c7fe154eab2b54f0903d82c550
116615 F20101204_AAAKRB burla_r_Page_022.jp2
d203cb984aa51e436113ac4a7873ef7a
a6ca13384410224783b2b42c0c4a36011cf7c77d
1051979 F20101204_AAAKQN burla_r_Page_006.jp2
e59aae6da68caa024f79601a4ea3b7f9
7ac5659e95866cc15e79a1108fc097f7cd043ebc
7033 F20101204_AAALUE burla_r_Page_026thm.jpg
c11bb3d0d2b250a0af4d17a1ab2f18c2
f60542db6995dfc4c4ca2432c812797f287eb24d
3509 F20101204_AAALTQ burla_r_Page_054thm.jpg
01014e042ea7844bcbc7c2d4a55b0386
7240efa367365d8a8a59565070114ac2d82ef381
73525 F20101204_AAAKPZ burla_r_Page_166.jpg
5266e6cfb747998e41bd4c46bd114cac
0797d3f92c11cd7d247bacdfc6475aaafca3f024
111617 F20101204_AAAKRC burla_r_Page_023.jp2
f3273e4f2ed947636d50d7c5d18f8270525c3716
399795 F20101204_AAAKQO burla_r_Page_008.jp2
f74672a41734b6a04de876dac65abfd6
fe96046526fb9d43e75be9f1589432f9e9fa87e1
7216 F20101204_AAALUF burla_r_Page_027thm.jpg
758b0ab99889351a17f0c6fa3e2ba82d
6195def9d4b9dc5dc925374485b6879ae36edd90
112398 F20101204_AAAKRD burla_r_Page_025.jp2
d2f37d112ae1534a8daa11ea2d980f63
402f264573d1525f82179440facbd04d23d4afc6
6535 F20101204_AAALUG burla_r_Page_031thm.jpg
e0ceec86abe41736f8e3d88876647ef0
25eac736a8aa6a15843c00e63d7b3950b876faa2
259644 F20101204_AAALTR UFE0021901_00001.xml
1e8fc0216246093f17ab4ba83737db8b
a47443534f66a1d701c0c00c9dac8f51336d01ed
121392 F20101204_AAAKRE burla_r_Page_026.jp2
455e3ec92d3409f3f2f4367ee9e531ce
b3cfa10397bb45df442ba0d98547247fe8f12d16
F20101204_AAAKQP burla_r_Page_009.jp2
ddcdd68fe7969b36f53a35cd1cc2ded2
98f77e8989d20dbed37d9cc4fb739b2a4ec0e5c6
20584 F20101204_AAALUH burla_r_Page_033.QC.jpg
7f755b70f4317c0edd058d296acc98f7
9878321b697cede4a85041143aa73e18357cc655
7448 F20101204_AAALTS burla_r_Page_001.QC.jpg
161a635de3ba7b4d1b4d5d7d52d94fb7
20f40928a5ca534ee3808ace9cbf31d7cdd11aae
121495 F20101204_AAAKRF burla_r_Page_027.jp2
7ffcd99b580068ed52a86f8118f4acfc
1051980 F20101204_AAAKQQ burla_r_Page_010.jp2
6433e0036e8156698d3b5982f1d9434b
47431527a1e37df53eae0034e2ce7bfdf458bceb
16007 F20101204_AAALUI burla_r_Page_035.QC.jpg
c78d049fb7456c83640e589cb4005ede
2962 F20101204_AAALTT burla_r_Page_003.QC.jpg
e510ce90ccef69428d9586aef334e098
62d91ce967ea1e7d878de51da262316478851137
119431 F20101204_AAAKRG burla_r_Page_028.jp2
a496c7b7f85e0ce318c6bc27d45db32d
6e7a4a2d3f43ec377aed7801c1bd9042bbe1f0c6
F20101204_AAAKQR burla_r_Page_011.jp2
373e5700428d633856acfe9c9ac7ed75
06b4df4606671627b39e86c1c2d6c50eb2f318a7
5543 F20101204_AAALUJ burla_r_Page_035thm.jpg
6376f53e0b7b8df3dfce6176286f7845
488da7944d796ab4502fd98d86a129f4b5830ec2
1307 F20101204_AAALTU burla_r_Page_003thm.jpg
5a8729e3347e751a6cd371e337929e0f431b122e
60067 F20101204_AAAKRH burla_r_Page_029.jp2
c883b7b38d4f7ec3ca864aefcd678d12
7dd76c66ba91c25aa17f0b369f3bd6312094aeaf
1051985 F20101204_AAAKQS burla_r_Page_012.jp2
caa79ffc74924da0da50d44f648fda81
d1854bd6b6e150038fe225f7a11f494d2f8ce0b1
25796 F20101204_AAALUK burla_r_Page_040.QC.jpg
12faf73ec723b773a61dc76d91889e69
acdd6c9661ea5cf2c09d37ac5f6a47771af51439
22378 F20101204_AAALTV burla_r_Page_005.QC.jpg
5f43e21f850d4d88729102a798a3857a
de906809378b987573d681bcb6c5b1ecd868fc8e
112334 F20101204_AAAKRI burla_r_Page_030.jp2
fe641cca758d391575607bf23c2c7267
e945b4614e6f8223c1fa72d0f87b6e0bf839876f
1051977 F20101204_AAAKQT burla_r_Page_013.jp2
07c866b4d91e1e83afd3fd5e601e70c1
aac5a2fe5523a7590e9b498cc0370be1b8945095
6340 F20101204_AAALUL burla_r_Page_041thm.jpg
4ffe12d36909e33c664634f490a7725f76cfc48a
5627 F20101204_AAALTW burla_r_Page_005thm.jpg
1325938eca92bc2d55ed686505050838
a54317dfabec04fd0d5d16dcf1d47d8492f6d89f
106261 F20101204_AAAKRJ burla_r_Page_031.jp2
4b22a0f77e0fe6748b1a309f12c70cdd
311b756af309ed13379000d6cf3e587f0442dd26
101290 F20101204_AAAKQU burla_r_Page_014.jp2
2fe708cef20043b4ba8c93b03bbef6ec
464d9fe636a48752db25737d2fc3c40574a1d934
22993 F20101204_AAALVA burla_r_Page_069.QC.jpg
0ff9fc69ca62ac12242f1f09f75603e1
5388 F20101204_AAALUM burla_r_Page_043thm.jpg
8bf90c7dfd636130f213b4bb534568d45e9917c5
6932 F20101204_AAALTX burla_r_Page_010thm.jpg
39431aacfaf1454977010a64c4573052
86871 F20101204_AAAKRK burla_r_Page_032.jp2
b67f97276e9f2d50702014d139145a73
50516 F20101204_AAAKQV burla_r_Page_015.jp2
a2a8d92378ecc818401e1f375174309e
99a9273ce85c8d0c7e3fdbea826138651e9e1ee3
4841 F20101204_AAALVB burla_r_Page_070thm.jpg
00a994dee3f11b0cfbf9153ce124f497
6aa82a80ede7b43f198b3d2b89c94821a6a1b1f4
18908 F20101204_AAALUN burla_r_Page_045.QC.jpg
e10c3f76de1d47bf4d4c913181284796
ace361b31d3381f3baee032df9320c390a924a15
26238 F20101204_AAALTY burla_r_Page_011.QC.jpg
54ca201c2568f52208a48e0873b39cd3
b6a884771de607282cc9f11baea13190b550316e
96978 F20101204_AAAKRL burla_r_Page_034.jp2
fd06afbba2f77057037216f70257fac1
169c7e6736ef6fea78b336577ab26b8e27acdef7
111721 F20101204_AAAKQW burla_r_Page_016.jp2
6a5aa1a43c4eaa9b021ba01cf9c1a2bc
65a280a38c982bf3f9e98965bef5ac9a440baac0
20951 F20101204_AAALVC burla_r_Page_073.QC.jpg
7d0acf5733e433fa48f97be7b28b853f
7afdbc6ecbb7a86ff94f5a3eb6a91c7811db7e08
20540 F20101204_AAALUO burla_r_Page_047.QC.jpg
d36f7f2b4c1cf4ee161e6f84668809d2
6211 F20101204_AAALTZ burla_r_Page_014thm.jpg
ab92e9d61e75ca129e132b11be2ba323
23773bfba8ced5a29cd06f7240fd3405f8fc7555
339775 F20101204_AAAKSA burla_r_Page_053.jp2
e90a219bd9725f6d0e4b09503b39ecda
76697 F20101204_AAAKRM burla_r_Page_035.jp2
557599d41535357d6c5e17aa939baa92
f07243575d9b9e41c5ffe10497137a06388f837b
112076 F20101204_AAAKQX burla_r_Page_018.jp2
e1406e031e6e715cf2110ee5fce88b1b
22876 F20101204_AAALVD burla_r_Page_074.QC.jpg
80380e1bd7e136dd12bf6e0ec1ca334e
fa7b462ab108ea6238959e279ca4a02320587bf7
3411 F20101204_AAALUP burla_r_Page_049thm.jpg
602136aa5f67f359d6d93c69b1c6d0a3
5d3a97dac01dd0c5dd0db4f98fb0a80027aeb854
427014 F20101204_AAAKSB burla_r_Page_054.jp2
1bfc456223bfe7fd5bf586a63ec050b8359aa26f
102742 F20101204_AAAKRN burla_r_Page_036.jp2
2d672de6dd9c011e63151836e1c3987a
c6b530185f91302dbbc463c35de7be118c650310
111961 F20101204_AAAKQY burla_r_Page_019.jp2
381eda17b614c5335f093b48a93201e9
97ee9d95889a463a696105e7742350d236ea2855
18929 F20101204_AAALVE burla_r_Page_076.QC.jpg
7ea96119b679228d1a365730ef888870
37b2c4b760a7104085587cf48c578b4c0bf0c854
13362 F20101204_AAALUQ burla_r_Page_050.QC.jpg
7d6808d8de0ca20709f59af814173ba4
f82f58f42d925ae58255d39beed6ba53288bcfdf
110908 F20101204_AAAKSC burla_r_Page_057.jp2
0ddc9c097843661663084ccdf6b62a20
ae903f43e00a4a5a83c0604153f4d4682ba42d30
87717 F20101204_AAAKRO burla_r_Page_037.jp2
8669aa388bf218b317ce65f36fbf3e40
dc9e81c771448dd6d14cb13518b59256edb0c6f0
75848 F20101204_AAAKQZ burla_r_Page_020.jp2
9c78b38a17e9c11ed5511e6644803d84
af1a9a0e8fd17b762f30f078bab931ca0f61e1d6
5668 F20101204_AAALVF burla_r_Page_076thm.jpg
87ea41f6c05650f4ff3fe5d77b65256c
1f936583ea9952bd8b11e429d21f67061c4b8ba7
3774 F20101204_AAALUR burla_r_Page_052thm.jpg
ee8ddb7b6be91b60cf7acd3e0d2370c5
3b9679d8e27d7434ae4460b08ffdb3fd71fe8e22
109733 F20101204_AAAKSD burla_r_Page_058.jp2
f7873cfb21420ae5a516d7ee4a156d85
9bd858723a59b5e00782e84e06a745a3f1c57bc6
99033 F20101204_AAAKRP burla_r_Page_038.jp2
051a051e644a6708e23d267dcedfa0f7
2a0b5442602ec806f9af62e5cb934079c1e0f6d8
4691 F20101204_AAALVG burla_r_Page_077thm.jpg
624f886641f1dd6186b54554238ac20b
c5bf44d0778fcc4e36b3f09db73ac674a6f3d99c
84545 F20101204_AAAKSE burla_r_Page_059.jp2
e647aa358c08e92a72611bb48ff7243d
cb488bec5bf5afa76852d5502becc72c3710a883
5041 F20101204_AAALVH burla_r_Page_078thm.jpg
67e6699029061c06ef5537348ed61360
23b25b9819f46e31861385e91e7332c30c4dd8bd
24264 F20101204_AAALUS burla_r_Page_058.QC.jpg
43353eabf9676c973d3ae1a3ea464092
c51baf1f8eb719193fa73c57b4be52485434573d
101222 F20101204_AAAKSF burla_r_Page_060.jp2
13ffb1c2104ea87e8ae887be32f4c91f
65ef1cc2f13902aa5c94876fe5bbd5b22c5244dd
114455 F20101204_AAAKRQ burla_r_Page_040.jp2
d108962758d44b6a2c611b9b73f8a352
553efb5bb2d9ff986334e1fb235d6bbdea3cc0a9
5991 F20101204_AAALVI burla_r_Page_079thm.jpg
49b33efa30984f3370902cc502105e9f
1cdc904dea6f5f144e74feddca68ccdea8c0d506
6658 F20101204_AAALUT burla_r_Page_058thm.jpg
55e22e3d0f5a03e081dce60f8414e930
4ef54449a24cf537139daa204291a6c50b4ea7d8
107172 F20101204_AAAKSG burla_r_Page_061.jp2
8f653758d3445c369ed5b45af1273908
76d288ca47338f39a8cabbd30e2c293fe6104dc4
97019 F20101204_AAAKRR burla_r_Page_041.jp2
3d3b90e5186c1e476fe64b39815701f9
c0b6eb0e64326539fe659c1f0c61b493a03d677b
19063 F20101204_AAALVJ burla_r_Page_081.QC.jpg
79dce864b906a1b8f239b38f7af5d673986c16d4
19429 F20101204_AAALUU burla_r_Page_059.QC.jpg
69cc86a6d5b31ba0ca840526c7ebbbaf
642be3de4210bc021b6be2fdae29fca2d1c88704
95538 F20101204_AAAKSH burla_r_Page_062.jp2
d423f865c54302ba48c3ae31bc4af361
3aa28d2dd2f2f3719817fb069bdba2238969d562
78739 F20101204_AAAKRS burla_r_Page_043.jp2
ec1241da035af17a36fd50603248738b
25788 F20101204_AAALVK burla_r_Page_083.QC.jpg
1a3c628623aaaf006cf766a0ee704ce1
6760b6558d58db13d2b843b216fa67fc47ef358a
6023 F20101204_AAALUV burla_r_Page_062thm.jpg
903e81b7491f2dc16a8a901b0481cffa
378d573888331bfa20ba03c89065fa6cc813bc8f
371587 F20101204_AAAKSI burla_r_Page_064.jp2
03af8e46cc0f54542233e5bf29520419
77a83dd0d6a47a45bb567620da278858c6ceba45
111297 F20101204_AAAKRT burla_r_Page_044.jp2
6ea97f1673fb38c6f0ccb5792b827eae
55c1c41d2fe0dc98b364f055a6cb3563693bd3f9
6558 F20101204_AAALVL burla_r_Page_083thm.jpg
0170779ec19484a7efd016b6fbf25dc1
f3db4ba117aec9ecc4aa5fa5b45a36419f5e8b86
12166 F20101204_AAALUW burla_r_Page_064.QC.jpg
89fedbeb8c92edd512be5cdeafe28988
095fbf42f5a75fe603f92719d318ddb387b6f75a
513440 F20101204_AAAKSJ burla_r_Page_065.jp2
ba1d3eac95ae550c681eba58521b0297
38e7c102224907a4e0cd628b83779a4cea2f8de3
83453 F20101204_AAAKRU burla_r_Page_045.jp2
ddbb90432303622565effc3d0c75e827
05378ccc0a12419bd17b21473a9086e1b1bcef5c
6624 F20101204_AAALWA burla_r_Page_103thm.jpg
7e8b7710025ba81c162a20ebdb2096ef
6d22bece1505a2c8e30fd178063f643f46b4af29
5415 F20101204_AAALVM burla_r_Page_087thm.jpg
b200c59f187df4c0d7a04fe0ae017fc9
95a0f2ec13de8b20a76cbe9ed8a9e76628cd6f9a
13602 F20101204_AAALUX burla_r_Page_066.QC.jpg
9bc3a4c04aa31c7bfb736c4e84a54a41
58f2253a83be476e7426a28d96d593618a84bbde
391224 F20101204_AAAKSK burla_r_Page_066.jp2
29c5bd5440013fa4a4f08f9707dc867f
19ce66eac5cf708c2377b718b94cbcbef8ff10c4
94053 F20101204_AAAKRV burla_r_Page_046.jp2
4c92f2de94390b86bd1c7aaee2e22bed
357c6d2bc93a406def0dbdb8a662af552886aaa9
6730 F20101204_AAALWB burla_r_Page_105thm.jpg
1c535b07b25ef155ba702fecc19fbc2e
b7666f5fabec9d994616bfb0f81d4f37870a7812
17591 F20101204_AAALVN burla_r_Page_089.QC.jpg
f4216eb34a8f015e378fbd5bcdc1a389
20148 F20101204_AAALUY burla_r_Page_067.QC.jpg
a427471890d99ed93a028976879fa6fc
5496c4d8bdeb0ed09e18bb1f5647964a9b52c7b1
92399 F20101204_AAAKSL burla_r_Page_067.jp2
4cf72f49fc54dfe381aed18f2fbcffe0
aee8885d13759ff6f044d95af4edfcb7a563a2c6
266409 F20101204_AAAKRW burla_r_Page_049.jp2
3c2c0876b14fa47325d320800f818355
6477 F20101204_AAALWC burla_r_Page_107thm.jpg
a85d98841daa2a3e44dd60e4d1fc50f8
4238c2bdceda5ab90b36529165339d34160f03cd
5409 F20101204_AAALVO burla_r_Page_089thm.jpg
c97b8e18b379b0a21693300a2efbbe48
c9cc9f6d147a9ee3f5b9d5237a754b82178252a3
20216 F20101204_AAALUZ burla_r_Page_068.QC.jpg
a0b24ebf466efbc58d768a85be438cda
457753f231cea7c9423d1a8570d44f9078329681
83848 F20101204_AAAKSM burla_r_Page_068.jp2
643fb846f4520792b7c633c38c5cb8f0
468c877a3935f17e285f82d3231f3ecda69a29c0
596445 F20101204_AAAKRX burla_r_Page_050.jp2
6cd9f64d1a8b2874d63fee017b9db887
11365ae31905e5eb5d1c17164ea690e7d0b3998b
81674 F20101204_AAAKTA burla_r_Page_086.jp2
681b5fe721f4a216dd62312fd0720eb18dcbf9a2
5725 F20101204_AAALWD burla_r_Page_108thm.jpg
df43c6abc9140b51d78ce665ebbb4b0d
af1d668955a0ede55598bbb118a70e905e5d5da1
18454 F20101204_AAALVP burla_r_Page_091.QC.jpg
7daac754e082c69de6fd5bff90d24e7c
4eb0c61be2ae8b7b451b16b99fce8d184a720dec
104441 F20101204_AAAKSN burla_r_Page_069.jp2
004a25953a42e9ca512e7c28e66773e6
b414b3240368094e82e530730fb74222940467fb
229559 F20101204_AAAKRY burla_r_Page_051.jp2
68e33c410532384c69762156fb7b89b1
909a2d4fa709fb35b7b0fe70a42fbf448337689d
74072 F20101204_AAAKTB burla_r_Page_087.jp2
1a4ac480e05c0b1d0eaf962b214b030a
55cff9b644b0532a8ff0678cacb458c2161527ac
24156 F20101204_AAALWE burla_r_Page_109.QC.jpg
5642 F20101204_AAALVQ burla_r_Page_092thm.jpg
b037a99d4524bf4d2aa6a435799d5a18
9648677d6029456b7d09fc81d1f2d48296d04e99
64537 F20101204_AAAKSO burla_r_Page_070.jp2
691d0cfae30755ca553d7a75d03a57eb
2dd3b9d6966bf5bd59c544dd4f52a3ef59377eb8
737915 F20101204_AAAKRZ burla_r_Page_052.jp2
e078db3e9008cf36dd3cf9abaffd00e8
0747eea1250e4e883480dd5f9f6d8f1b1624b637
106115 F20101204_AAAKTC burla_r_Page_088.jp2
f84fe644727f9fccc70871f5f39ae679
42d98d26ca11228b84cdb5b739b1696ee56b25b1
6830 F20101204_AAALWF burla_r_Page_109thm.jpg
9ec54aee4174f2d7295c92bca6519269
3eb87047a895eaf7f9821a1322fa673e9331cdb7
F20101204_AAALVR burla_r_Page_093thm.jpg
9a0a54ec63f2b8c7224e95bca8ed4eca714640a3
100192 F20101204_AAAKSP burla_r_Page_071.jp2
7c4f2396c9567ef88b6b5f91dcf3ac3f
28d0a932f920eb1c9ecf8ea0a506c78779497a3f
71032 F20101204_AAAKTD burla_r_Page_089.jp2
9c4a6c3f9abeb4ba4f29333d9dcdb73c
d5253219f0e6c7e2c3378ac4e351ae9f84bc1d4c
5816 F20101204_AAALWG burla_r_Page_110thm.jpg
b893258f462bf4de84b8021bf6fef0e8
a774c21c1904201194c53a9e1a39bf900950f200
19873 F20101204_AAALVS burla_r_Page_094.QC.jpg
084ebb22fe26f9def3d4bee2c9303170
bd09983bf5b22ee40ee249f7cc5100e6d60ec5d4
87161 F20101204_AAAKSQ burla_r_Page_072.jp2
45eab719015d26796d2c62554a551a84
96383 F20101204_AAAKTE burla_r_Page_090.jp2
c2c49f5e9a0cc007b5ae23e84da43e54
6130 F20101204_AAALWH burla_r_Page_111thm.jpg
dd55efb46831a46a83b799dd2f4eb85e
90801 F20101204_AAAKTF burla_r_Page_094.jp2
dd4531685e20aa47574e034218a4c4b8
712e1d908db94f12771ea4f05a38e1f29337f6fb
6607 F20101204_AAALWI burla_r_Page_113thm.jpg
43d6f85bfcd682b9db944220db51e1b6
16fe8cf84c161a5836f01f2003ed031542b4bba5
5947 F20101204_AAALVT burla_r_Page_094thm.jpg
60a4ff09621477575500312c11faaef2441b29a6
96504 F20101204_AAAKSR burla_r_Page_073.jp2
df3b4567da74abc1a8c79b2dfd7c8359
b66f71993531dd4b46b79b5f9e0f8fc50c67f080
682382 F20101204_AAAKTG burla_r_Page_095.jp2
d8fe962efd354fac754c418bb822f34e
6fec95f5069ed4958202f3656c24177e600cbb4f
25504 F20101204_AAALWJ burla_r_Page_114.QC.jpg
0180a8962a6cdb79459509076b6a4e7765177b09
2203 F20101204_AAALVU burla_r_Page_097thm.jpg
acf921a893ee0393df35fdf586f34988
b3c96ed116eee60d7110d45d6e829d48f258233c
101267 F20101204_AAAKSS burla_r_Page_074.jp2
b1135bebe25985e0de9aace897f9d15b
49c0cc0b2eb008d494aedfd5a735226b51b17085
236018 F20101204_AAAKTH burla_r_Page_096.jp2
13002d7347f90b78c4ec015a2c63f2bc
c8ffb9971e31d444205063295e7422453fc69094
6780 F20101204_AAALWK burla_r_Page_114thm.jpg
2a5c1157751af18efa20cf091890acbd
bb330bdc29ef95421dcb4ce31bf1bab9a021cac0
11831 F20101204_AAALVV burla_r_Page_099.QC.jpg
9d741871496d4bd906f19082bfaa9708
1f85c82ce61bd223715e6ca91a20aec4c94cc819
82628 F20101204_AAAKST burla_r_Page_076.jp2
c065bc808cb55e3c37574e6e26346856
5c90cb5d30f3c042e7940f22642a11ceea4e2a21
133246 F20101204_AAAKTI burla_r_Page_097.jp2
e551e5d8fb376480a71c626420b23eab
795a1255dc0f4c13e039b03b70cd44e3e6302bea
6354 F20101204_AAALWL burla_r_Page_115thm.jpg
04b8b270c6cf4c8ec77e63414d9a93a1
5fd98c89a40d5726c2e290f22f5ff6eaa1eaa1b4
3840 F20101204_AAALVW burla_r_Page_099thm.jpg
323ef4ccb08c8453dda25f8a55b486ca
fe61243bb0d496e6f3326f72846110c2b501c74b
69802 F20101204_AAAKSU burla_r_Page_077.jp2
8b8b92c3a34ec0fb08c73c7b2f18009c
1c357a6b7f2d51e1fe458cc2b612785e9eb65a02
310241 F20101204_AAAKTJ burla_r_Page_098.jp2
3ec975a17363800491b6f419c59f7087
6bb258c549cd2fb8784febb8e71c19ba1b9882f1
7571 F20101204_AAALXA burla_r_Page_131.QC.jpg
9759c61c9805ce5fc77edd8cc64eb456
2f4df142fa23a5a9677790086f0112d840e814cb
24477 F20101204_AAALWM burla_r_Page_116.QC.jpg
294aa7258f8a2965199bb46a0ff7628f
F20101204_AAALVX burla_r_Page_100thm.jpg
0c51d73e20afce03009d67f4cb1440d2
c619a6b41d68db2963376175dbdf4b474b4175c3
72270 F20101204_AAAKSV burla_r_Page_078.jp2
0cec6dfef67db337832964b4ac8cfb25
bff605aacafaa6fd18e40ece58c82207b9abd224
372311 F20101204_AAAKTK burla_r_Page_099.jp2
bb23faf200513afcb045ce6cd4720122
11688 F20101204_AAALXB burla_r_Page_133.QC.jpg
c410acaa0f194091de4eca66a76bf4a9
6861aebd589ab5b40b21a8193095af431f0a3640
6515 F20101204_AAALWN burla_r_Page_117thm.jpg
d532e55eca506f7a75090d74304959b7899b023f
7899 F20101204_AAALVY burla_r_Page_101.QC.jpg
d50ef768e53b38e97d1a76579cb94852
20e6e87fb8b55dedd4b49bd0807fbce62aa62080
87751 F20101204_AAAKSW burla_r_Page_079.jp2
7efa6193e007a2767e6e962148db7afe
c55107c21bafabca6a7a02d54517546e44aa0671
441308 F20101204_AAAKTL burla_r_Page_100.jp2
55db0100e25dc7cfe57e0a83d1104ec04a12591c
3607 F20101204_AAALXC burla_r_Page_133thm.jpg
c506afcf0c93f10b1f006124b0744986
409bd236c0d9ef54df0a32b0e6e6322798703c44
6984 F20101204_AAALWO burla_r_Page_118thm.jpg
2645711501f37b2e4f96cbde6a398718
efe99c810973657a52fd6d0de1f8f81cf36e5e93
23610 F20101204_AAALVZ burla_r_Page_103.QC.jpg
77653226ee2d3d8db489b500b1c95e85
78896 F20101204_AAAKSX burla_r_Page_080.jp2
5984ecc6d5c64af375a70755c718a66e
d95745fe89a84a099f75dd2a46a8232dbdd8b6bc
108792 F20101204_AAAKUA burla_r_Page_117.jp2
5105bdf57e6495644368544f26737550
151998b21992b0da27270ac6b8e1016a41b09839
209841 F20101204_AAAKTM burla_r_Page_101.jp2
367338b110cb097ba3b28d3959e1dd67
a7f353e7a8daae05e806aa6d5ea594a3c2a5ae7e
3342 F20101204_AAALXD burla_r_Page_134thm.jpg
608037e750dd3980977e22ea28891dd6
25579c98e3d560525cae4fb7af0a49e462e3bcb5
23674 F20101204_AAALWP burla_r_Page_119.QC.jpg
932245513de61666dcee6661fe58a3f0
07868797203816c4b589d0806a6561036119f08e
82484 F20101204_AAAKSY burla_r_Page_081.jp2
efe19414942fc747d9ba661a6d0cb110
04672b5f7664706f60305bb83ffccf742e4086d4
112516 F20101204_AAAKUB burla_r_Page_118.jp2
4e52b989735414e9d8ef88773c6a5ab7
5611b45e3d91b37f16038da6ba0ee86aa133033c
108993 F20101204_AAAKTN burla_r_Page_103.jp2
ae149b204ec88c093612ca9e1493921f
12109 F20101204_AAALXE burla_r_Page_136.QC.jpg
5159c24d523ebbff50f38f1d29877d23
8265 F20101204_AAALWQ burla_r_Page_121.QC.jpg
b61230632d82fd5bf73a2693d753358d
75426 F20101204_AAAKSZ burla_r_Page_085.jp2
40b496e88dd396952e7977f8934fcaf5
106912 F20101204_AAAKUC burla_r_Page_119.jp2
66f4fbb4fb8fb814634524b341962572
7bdaca5527d8020d69fd3232185222443790bd99
101687 F20101204_AAAKTO burla_r_Page_104.jp2
b6d118f084245c5fdd506322756da2e0
46491713dbaf4875ccec055f477852b357d025fd
3633 F20101204_AAALXF burla_r_Page_139thm.jpg
a9632f1c71f299b0393285086e21c60f
23f90f91d580f6aa807c97b24ef4a5d3a2d083f2
4136 F20101204_AAALWR burla_r_Page_123thm.jpg
9aed775e350ab7d47e37568deec6afef7b1c66e7
89614 F20101204_AAAKUD burla_r_Page_120.jp2
882dd85dc0e677c9a1b0f7bc72bcbd07
c672f53624d3b79ebae47fd5e3b52345f64be612
110627 F20101204_AAAKTP burla_r_Page_105.jp2
597c0ba78feab0bb40877c9f2114df7e
4964c01113a49a8efff42cf986ac8602b6a39055
8866 F20101204_AAALXG burla_r_Page_143.QC.jpg
d51604fb0af0f2482329fc5782024560a3363d35
12583 F20101204_AAALWS burla_r_Page_125.QC.jpg
3448dd1aca43659b4e10fb0d4b446879
a1f3f7190e7e5f56ac7f6abfaac5404b8a15635b
221228 F20101204_AAAKUE burla_r_Page_121.jp2
799e18ed5b23883a31b96eb64fa36faf
2dca83c78d0156b79d046c09fa2650d0a1fd3aee
117229 F20101204_AAAKTQ burla_r_Page_106.jp2
fb14b9fd12950b1407297994cd09b21b
a146ecb1d95df831921e5fb32e1a94d0720f6096
2984 F20101204_AAALXH burla_r_Page_143thm.jpg
2eeb32c5283155867989d2bdb96b755f
474d948f3674706ec457c0e03486aa6206ea5f3a
3786 F20101204_AAALWT burla_r_Page_125thm.jpg
ed05b1471d87f953f2218b7caca265e7
31c2025de252a26342102a12466a295714b2f221
420650 F20101204_AAAKUF burla_r_Page_123.jp2
2c631d3ed4b72e658c2d5bec26dd5f1f
2c2912b17877940912d6678e01e8ea27cc0ab93d
104340 F20101204_AAAKTR burla_r_Page_107.jp2
6b261a7c73d1bdcb0c9f1631765e166f
9702 F20101204_AAALXI burla_r_Page_146.QC.jpg
ac238c37fe346354bf515a1d8ccca9e6
e3797aaa08bf59315a38d5b582bfbec709e23c11
499026 F20101204_AAAKUG burla_r_Page_125.jp2
6814d9f1940ef735527a89d3241de75b
e4755d3b2c3e767e5d74f7355958d501e804d627
F20101204_AAALAA burla_r_Page_138.tif
60d113cdbd49c6b4e67481f4dd04f08ddf6ca821
4316 F20101204_AAALXJ burla_r_Page_148thm.jpg
8fdb946810ffdcc0f937b5845d483707f5be8e00
15982 F20101204_AAALWU burla_r_Page_126.QC.jpg
7353ccd12f26a6eda3bb3061c106f098
021d664ea3616e0f436e65223c47cf17ebfaa1f3
1038738 F20101204_AAAKUH burla_r_Page_126.jp2
146dc40bdb915eabd074b6378f0c6078
109574 F20101204_AAAKTS burla_r_Page_109.jp2
091f2dec14820768bf2a1e29a9bf604a167e07a3
F20101204_AAALAB burla_r_Page_140.tif
099b2d04d6bc5aafbbdabf9e2ae683e7
d7c319f8931943efc41dfcd9862bf0e413140436
17632 F20101204_AAALXK burla_r_Page_150.QC.jpg
1ea9cdf3ab0ece50b50df43d62eea0b7
70c8e0ba159592e06df7b8c6ce94360fb8ddf8a6
4792 F20101204_AAALWV burla_r_Page_126thm.jpg
767459 F20101204_AAAKUI burla_r_Page_127.jp2
aa9d3a8da1de81c90e61a1cc5fba7cbf
85804 F20101204_AAAKTT burla_r_Page_110.jp2
0ab1e773f9897c73ab9cb561754a033b
cc8773d1e81e9eef929328f3d2101b0d17ae985b
F20101204_AAALAC burla_r_Page_141.tif
ed7b53600a8e8f1f12f2ee9d179732d5
b205cbb0f5ccd8a15eafa1236149abbc10f48036
3316 F20101204_AAALXL burla_r_Page_153thm.jpg
b554112c4928b490905b88525c958ed3a32b948b
13546 F20101204_AAALWW burla_r_Page_128.QC.jpg
640928 F20101204_AAAKUJ burla_r_Page_129.jp2
134abba42b554152846121580016a7a6
b0fcba393528bbe699aee83a94c20029c9aab787
95556 F20101204_AAAKTU burla_r_Page_111.jp2
ee50e34f371f5ab89031959f56f2f931
74fd8895f674875920908a367b6ca2b00df368f2
f62095d4ea8ff9540ddca7a99bb70ec4
4837b8657cab3a3175ed67f2d142c7f8e1f4538b
7258 F20101204_AAALYA burla_r_Page_173thm.jpg
b68c397f35a6e262c4167d7e956dd4fd
2ec31ebb2ebd6984fe1dac396d6d1dda9a3f57ec
21253 F20101204_AAALXM burla_r_Page_155.QC.jpg
68e00f3648e359fd74de25c707e157a0b9f911a7
13699 F20101204_AAALWX burla_r_Page_129.QC.jpg
eb9e87247b7bfc11d6996290bea865ca
ee24177f5e1bebe60c3ef4b20b02e1ca83632762
24643 F20101204_AAAKUK burla_r_Page_132.jp2
b858f735a7ebd2feed4e7a670490fd3b62b0c19c
91777 F20101204_AAAKTV burla_r_Page_112.jp2
8b7511efe3ddd368b6d5b21c0851599a
42cdc425524e906258e78b026df4914521247215
F20101204_AAALAE burla_r_Page_143.tif
e865ffef1979037c91403e87032a87d7
440d829555cc89819b72269a22e719fd38fc1e46
8873 F20101204_AAALYB burla_r_Page_174.QC.jpg
91fc28628780afcd893f30b42322cb56
6255 F20101204_AAALXN burla_r_Page_156thm.jpg
93a12e86d84b945373152b9974c04293
7f56eb09ce0f736187d2ab9024849a777591cef5
9263 F20101204_AAALWY burla_r_Page_130.QC.jpg
8b538561f80a1e60e6d20f6b6c976d55
1051964 F20101204_AAAKUL burla_r_Page_135.jp2
535e1d65f48f3dbc83e0c412d0bd8965
108934 F20101204_AAAKTW burla_r_Page_113.jp2
1c7cd35e367c9f6b580122321ddf1dee
6355778e6c241e3a6e21ba1190f1b3a23c30abfc
F20101204_AAALAF burla_r_Page_144.tif
e24550e2491cca20eaf25a157d881a45
59ebc5bcee09cbf307bf590e7f350e79033be268
11026 F20101204_AAALYC burla_r_Page_175.QC.jpg
7f31e9f9ec472c1a49c2a5fcd08f3862
4e3e04eb892e5087f9784319e72244d63be9e720
4438 F20101204_AAALXO burla_r_Page_157thm.jpg
b5eae1e58e16ff0a5efdcd0e348f0c56
5b57591a9ea751089cf6a9ed2553d4a987377402
2708 F20101204_AAALWZ burla_r_Page_130thm.jpg
a8271e1274cfe99abbcd3f5e6f6220e8
1697884c3220e5bec0450efd9c5ef0a8f556fef5
1051930 F20101204_AAAKVA burla_r_Page_156.jp2
4de8dfd7e6ec319f22fee35ed259746bceb70312
644955 F20101204_AAAKUM burla_r_Page_138.jp2
8a655d8e3c25c3d03422b51db2963db9
112901 F20101204_AAAKTX burla_r_Page_114.jp2
d183bd317c1c76fea64e2057147c3e56
fd18eed890b308e21fe9571e4e71e751cf03c3f1
F20101204_AAALAG burla_r_Page_145.tif
0e47c8933a2964ac8c1a216c985353ab
3368 F20101204_AAALYD burla_r_Page_175thm.jpg
77a115962dd130238252d76c1d977c03
bbc819b02e44773b2e439e2f91370ce57262cabd
11918 F20101204_AAALXP burla_r_Page_160.QC.jpg
78005cd224ef74d18248ec12a0448f01
241430a19f3e02ed4d915378fda2604e20fb8547
1051973 F20101204_AAAKUN burla_r_Page_141.jp2
5d48ca1edb92652203e07e38807f9cc4
104717 F20101204_AAAKTY burla_r_Page_115.jp2
2a36e83fa18259beac207e9b503ac5b8
aa81dde0280d6412b5b992343c68d034e101d1b3
F20101204_AAALAH burla_r_Page_148.tif
2747d277728c143cda3f2bb973be83ca
fa17fa2d6b06295c8cbe2ceebbdd16a7c5520e90
557051 F20101204_AAAKVB burla_r_Page_157.jp2
e737784b4060808ee4e1bfd490b47106
c486be8a9c9a75ccc9d8045f54c8f186b2199b3d
3729 F20101204_AAALXQ burla_r_Page_160thm.jpg
29afb56ebbed61be7697f2ea47d42d1e
bacbcabfa886bfdb732b92116f9f1aa07cb98ac2
1051981 F20101204_AAAKUO burla_r_Page_142.jp2
6588b1703493c61b9f5ec2089968ea9a
e7440e12db52fb4705858d4949df34f8dd511dfc
108158 F20101204_AAAKTZ burla_r_Page_116.jp2
8996b17b8cd7cda19b1d4d63617c90fd
b25afea461edd518955f0614db7ab9d47a2959f8
F20101204_AAALAI burla_r_Page_149.tif
acbe7cabe1932d561985ac77e0d12634
cf2cdac064bf73cc4a039e6ce3a178495506ff9a
566596 F20101204_AAAKVC burla_r_Page_158.jp2
6e2890823170b0ef6470715e0e73cf1c
7445 F20101204_AAALXR burla_r_Page_162.QC.jpg
fd28f3d9fa7b46c8b8e48b863ce1d214
28683d46e6e5385853f5b48e58db456e76eb9bf1
347509 F20101204_AAAKUP burla_r_Page_143.jp2
248165427bdc88cda0df430a0738cb9a
414d5ac996f01231e67292b7a6a7c4cab5876678
F20101204_AAALAJ burla_r_Page_150.tif
0d2e69b8ab9ef506ddae4b931ecd892902d38737
330601 F20101204_AAAKVD burla_r_Page_159.jp2
7f4af9b67a5f829db97050454fbbd98d
ce1981ac7c6b6f8d40b063618bf9f6043a3da679
3002 F20101204_AAALXS burla_r_Page_163thm.jpg
fcbc7a2f645a7d1f96f27b679b962493
1051936 F20101204_AAAKUQ burla_r_Page_144.jp2
3fe0ccba05cbe0f74f567a625d5ca0a6
f2ef6d354c0e12943aac2228db8a8cf459fb5508
F20101204_AAALAK burla_r_Page_151.tif
6bd85cd0199c87b9ed66b5dae4ffed1e
c6fa380c43796ae580408bb215cb1853f21eba03
549872 F20101204_AAAKVE burla_r_Page_160.jp2
12ba94de2f44f143a18f732183bd61f7
2405 F20101204_AAALXT burla_r_Page_164thm.jpg
9d520523282b49cda60b460709a88cd0
6a56f16955c67f5a0e4a9dbfdab984ab9c25bb21
399277 F20101204_AAAKUR burla_r_Page_146.jp2
ae479e7359880653713f9f91577ddb3d
ebf58d462b33ebecc2e3fd4987dfd41065e47458
F20101204_AAALAL burla_r_Page_152.tif
b37260279bdda7b9a21efcc73cd035ea
579224 F20101204_AAAKVF burla_r_Page_161.jp2
c80e7b6a9d044f9150cc97261786dbfb
6438 F20101204_AAALXU burla_r_Page_165thm.jpg
166f3850a6584e5713b2a6abc9cf6cf5
bd39766ed02af4caee8176aea7f9a8da293b8e08
F20101204_AAAKUS burla_r_Page_147.jp2
7fccbb3e3f9d531f9525c138fc1ac037
b1284bc232591a188cb22e76f57d64cfeba202c4
F20101204_AAALBA burla_r_Page_169.tif
386cb24ff1586394cd5fe6eaebac84c0
b8b6ca94341d71d7a3e1f15876ce680411f3b1aa
F20101204_AAALAM burla_r_Page_154.tif
7be01e1820aaacabd0af4832be706389
cb65d83f8bf8a71f77bb10318c85d4d60bd7d5be
23647 F20101204_AAAKVG burla_r_Page_162.jp2
e6a00ab49272cb77498a0268f6642e9e
d03dca953ba15fcd99e882ff22a17208e0b8765b
F20101204_AAALBB burla_r_Page_170.tif
cf38a9aca79d366f8d26f296eaf2496e
22a9199f398b62b6465a012f0c023da5bea41a9a
F20101204_AAALAN burla_r_Page_155.tif
bfbe173791b7f696e7c9bbcb8431197c
2d929f6069758808f797a17ca185f03c560ced24
33732 F20101204_AAAKVH burla_r_Page_163.jp2
0f0db6dfbf0801833be05fbf1940d7f1
35964a3d239d23fabf21759785eee69363aa70c2
23607 F20101204_AAALXV burla_r_Page_166.QC.jpg
53b880da09c4aea024b865029cfe20ea
d9e9d736d0013ee5f5ee8b97cc67e3f3e7d53597
995749 F20101204_AAAKUT burla_r_Page_149.jp2
3b7b72bd7a9039fc2394800d923b5212
9ab4fa4f94d5b3c07a48bd25f5fc1c3a2a97a2c5
F20101204_AAALBC burla_r_Page_171.tif
fee2f79a4a3fbf6cdc8259dfe9186487
5c1f208da5c89c8a47a671d5ef6e592fecd238d1
F20101204_AAALAO burla_r_Page_156.tif
31b682dfa6d66852c1291a4b08aee1af
438f4b70dce5589c9b718ab997baf8f0a9c185f9
22740 F20101204_AAAKVI burla_r_Page_164.jp2
ceeca6b3acb4321438a1995fe031db08
300132b9393aeed6b8ba2b89e240a7dd52aa46be
6968 F20101204_AAALXW burla_r_Page_168.QC.jpg
99869f24804a808db01a20f7e3076c2d
253eedd2e4ae1bedab43b82eeb54a88e125175b8
F20101204_AAAKUU burla_r_Page_150.jp2
b8e056f50d7bbfb52cc73c111570d774
F20101204_AAALBD burla_r_Page_172.tif
2a4df431bf55c501fe1565ebf06cedd4
4b19868979a4bdd7404327b5b185f358e4fd31b0
F20101204_AAALAP burla_r_Page_157.tif
4bdffe7e2d8a9c6a213d604b800ccb2d
107505 F20101204_AAAKVJ burla_r_Page_165.jp2
770e848db66db01af087460ce61c76db
861d3e69639408df687e8f5f0ecbe7818ab4c150
7328 F20101204_AAALXX burla_r_Page_169thm.jpg
e97dbcc062d265a6092e9cd3cc6f1cc9
01ac980f75c6c90297e6b262e2547ce5d9f06118
1051837 F20101204_AAAKUV burla_r_Page_151.jp2
fc676c98761055b9461ff6fbbf7bec39
384bd42537bc8dd74b9e080fa0e9bb0531b7e398
F20101204_AAALBE burla_r_Page_173.tif
1a96f0914a23e3343cd409d835076ff1
f1aff0a6c13af282be9f663b23bbd112efe7d286
F20101204_AAALAQ burla_r_Page_158.tif
c89cc481fee97cf8cb1b50e28c8e25d7
095e5771d1651259b87b417432e8ef3205866fcb
113113 F20101204_AAAKVK burla_r_Page_167.jp2
e93600bc5c5f0519430dbaa2696a420febe518d1
26288 F20101204_AAALXY burla_r_Page_171.QC.jpg
7941547904f68dd9188524dff0df1e3a8e16b1c8
1051949 F20101204_AAAKUW burla_r_Page_152.jp2
483b30abd7b92262e6e08c3c633c561a
F20101204_AAALBF burla_r_Page_174.tif
4c57b7c4d2d7789cbc5e672874eea435
5c5abc870893eee331f4c46f0392739dfe2ebf1d
F20101204_AAALAR burla_r_Page_159.tif
f635c2c191aab4e80da55b5f85d37490761231d0
24888 F20101204_AAAKVL burla_r_Page_168.jp2
197a0ab41e620fd472c468b9a3d3409e
0659b8fce38561c02e7966eb1c9cc2538cdf385b
28616 F20101204_AAALXZ burla_r_Page_173.QC.jpg
e028be45f753d65d0e23a0fc737abec3
68b6db720b0d408392a8b96571569fa92a42fdf5
34554 F20101204_AAAKUX burla_r_Page_153.jp2
7ae1b12f6523c2932775c525521a61e9
98f7b050d681c549faaa3053cbfd2c91629c624c
8479 F20101204_AAALBG burla_r_Page_001.pro
6314d22df97204fd23e9fb309bfb4a1b
257709d826b64bcd689e293fd78e9e02459e4baa
F20101204_AAAKWA burla_r_Page_010.tif
febebbd8dc0683b0f22173c48ddc8f9e
bd6275caa77b9e9a5e507b0e6f7e1ed2a44b0139
F20101204_AAALAS burla_r_Page_160.tif
def5057d037bc1641fb21f870d1e1608
096c120b7226ebd0b6d72aeb8f59dbe1412f360f
134734 F20101204_AAAKVM burla_r_Page_169.jp2
2d5a443a12ee09f1d719c84cc010a03b
1fcce2d2ca3c5d0ff1bbc55275d978c40eeb1c8c
F20101204_AAAKUY burla_r_Page_154.jp2
a07bdf84b242c050e11f2600b1681624
885 F20101204_AAALBH burla_r_Page_002.pro
bcdb46e09eb3fa5288560acc61bf977a
99dc98154e18c55b8e881d0bd7183bc81c99ac73
F20101204_AAAKWB burla_r_Page_011.tif
5fe20e80a7772172c1e1aa1877ac3ed6
F20101204_AAALAT burla_r_Page_161.tif
1a811afb159188250b35e7187b50dcba
78958ccc6a228f82b1bc9e8e2684cf45a6ac34a9
133289 F20101204_AAAKVN burla_r_Page_171.jp2
031efcc5d3706f54bacd700109cecd7c
17c4d349c15847935f6ca7bbd266f09120fd0ba4
F20101204_AAAKUZ burla_r_Page_155.jp2
752c0b0280e1aea411735185f3f15931
2b64f326f4658e9eb6414b2ac1c030395ee6219f
29720 F20101204_AAALBI burla_r_Page_004.pro
55abfc171468e86f0a088f042cfb1045
36a294c8b7e9ed7246bd9c6ec7a2b1d812414b84
F20101204_AAAKWC burla_r_Page_012.tif
bcbfa7c50f844d14148abb44804d5cb7
a565abdfaf66de277cc25bd9fa7482aee03d29d2
F20101204_AAALAU burla_r_Page_162.tif
f14114de79820a8b2fedeeab4c859608
4b311b089f8652a3e52ebfcefdfb6eb577341ffb
124769 F20101204_AAAKVO burla_r_Page_172.jp2
95102bdc627b091aa249c05603c9b6b6f2291a69
82150 F20101204_AAALBJ burla_r_Page_005.pro
222b6e85c9355da0af86ceae1b2685d9
F20101204_AAAKWD burla_r_Page_013.tif
7b5237f60107a29634c9bf4c208912ff
478eefb5268e8a743b3efd8cb90c219537d216d2
F20101204_AAALAV burla_r_Page_163.tif
45fc3c824508ca0f35571951c6c2c72d51c41407
140861 F20101204_AAAKVP burla_r_Page_173.jp2
3814c03373ded680469df56b7a41d952
1edd360e90ed31d3a7c704c20c0298d27d853f5a
113475 F20101204_AAALBK burla_r_Page_006.pro
d8d87a333b9485a660b8dd77ddc43349
425a4df767eb69f984768577abef64ecd2bec4cb
F20101204_AAAKWE burla_r_Page_015.tif
f220be14250eede589558068dfb8ab1c
96347293ebcf579f78ca02ba46678bc314f69725
F20101204_AAALAW burla_r_Page_164.tif
d1c77f67e36bc00173227ab319687d2c
d6ee1eed3307324a31d4b92781fb49966a8568f9
36380 F20101204_AAAKVQ burla_r_Page_174.jp2
4b83c7476f1140ce0657d4f73dc17c84
e4d692ebb598297c8a5d0d8ae2ebc93b052fb469
7546 F20101204_AAALBL burla_r_Page_007.pro
2f827c0142e852a341a367d85ce5dee1
970527f9260339feef688bd0b65362d4f7045c2b
F20101204_AAAKWF burla_r_Page_016.tif
ce2a029f2c20292f07c29cdb1d2d1e53
5ab237f0df2cc3bf061269485b8a657dcae90d2f
F20101204_AAALAX burla_r_Page_165.tif
c618a17b84aa4aa2c8202e69b94aa42a179d7afb
46169 F20101204_AAAKVR burla_r_Page_175.jp2
2ef80ea4f07481ea83384a7078347f6e
7b30bacaac3409f955a7627ae527b61ccd83527c
12667 F20101204_AAALBM burla_r_Page_008.pro
8b771f1d34112991679c70e85a10b3d54bf536c0
F20101204_AAAKWG burla_r_Page_017.tif
ffef64dfb8c2856be5d18f26fc3ac264
69f8a121fb6364dd4b5eed494dc05c597050ca85
F20101204_AAALAY burla_r_Page_166.tif
75072205a8fc028d3688a83eb96f491c
F20101204_AAAKVS burla_r_Page_002.tif
3e73305442135385fbc0abae50ec8f5f
77066e453e0dc156c3a3c6426a223061b40dab94
50713 F20101204_AAALCA burla_r_Page_023.pro
9f4a997a21eec25555f36a242ecdbd82
0249c50f1d25f5648f8a7a0e2b951f7b0a175083
68088 F20101204_AAALBN burla_r_Page_009.pro
601f072346bb78164e54e59a3abcf2ed
4d83a14756e16d82d167bbfe16e258e6d12fa885
F20101204_AAAKWH burla_r_Page_018.tif
dc21df7a6b0f2819cf5aa63ba4be4905
906fd17cf3a8df19c4acbd51ae0b56d15f449a89
F20101204_AAALAZ burla_r_Page_168.tif
4d3141db0a70246430029164d15d4309
F20101204_AAAKVT burla_r_Page_003.tif
53965 F20101204_AAALCB burla_r_Page_024.pro
6b2d61e8b0de5420976f16bfc06a76a8
69451 F20101204_AAALBO burla_r_Page_010.pro
2b8f7f5507c58c1bba761224428a8a1d
933af662150ff075c6414399b8c8762a9d20690a
F20101204_AAAKWI burla_r_Page_021.tif
b2df390332ffed82e2ccf143286202f5
c86fd09a858866c93bfaa1e531b11f0a6a1e8d06
52222 F20101204_AAALCC burla_r_Page_025.pro
28d6ea935247f28f074b8683ebc691f0
8682296ab30bc45458a1cf6597b5c5b9265a5c9f
70539 F20101204_AAALBP burla_r_Page_011.pro
6f11df32075a6d6a4d800c1685c1594e
cd4c8b0f175ddd30c4b8d1da726ca7d150bd6129
F20101204_AAAKWJ burla_r_Page_022.tif
9876c1ce0e03f615973d0c67a1d4b54c
F20101204_AAAKVU burla_r_Page_004.tif
7fbaef7a41bd602490d3cbe2dc4a9054
96427d6a3020dce4784dc40fdfc574b34d4ef835
56193 F20101204_AAALCD burla_r_Page_026.pro
903b0ce3637a7d59306af1cfeedd9a46f1bf519a
72092 F20101204_AAALBQ burla_r_Page_012.pro
a55251d6eb9e261f2475384ba0d035cb
4ae1a9b41095031bbd64d13e2d099cbc0f90f4d3
F20101204_AAAKWK burla_r_Page_023.tif
41f62d2f2495162e1996baa6aeea8159
208b7ca463d1fd0c7774e3f583b80cf7d24b6d21
F20101204_AAAKVV burla_r_Page_005.tif
9fe118f397103dfc749d3a7caf04f1ca
82fbd4d4fa430235222eaa037d3e23ea3c89d3a2
56865 F20101204_AAALCE burla_r_Page_027.pro
22213a1e5b21d8e0367d3ba75c25e61f
772f8dca3bce650f3ac20fde430d47c8eac024c3
43582 F20101204_AAALBR burla_r_Page_013.pro
ca7b0c1dc7a0a766d91759460ef2ff39
cd015a7ff1a5def57e390e15fe5f6e8ed14621ec
F20101204_AAAKWL burla_r_Page_024.tif
755078a9839892a331e1143de5ccde49
a0e6b70487c1cdff097f8628b8d4bb60e9a19730
F20101204_AAAKVW burla_r_Page_006.tif
047c36878e3358a2b2411275db255a0b
1ee92bc8b18ce81963fca4c3a94b8e482f4c42b5
55846 F20101204_AAALCF burla_r_Page_028.pro
564a6147c33b698755cbe5b3823d76b3
c9bdb064230164a355d5251d81a65fd25f04d92a
F20101204_AAAKXA burla_r_Page_040.tif
5a2c2ac8f44c1e76635aeb8f80d28401
5a3bfba3e5bab475e1bfda1ca0aae6f10840e121
46142 F20101204_AAALBS burla_r_Page_014.pro
c349ef07639c92ae1e29af0e9c9f998f
F20101204_AAAKWM burla_r_Page_025.tif
e98d639875dc8b325a2d10a1b8a383de
4a8c359fc80c7b72159ed8d9a6e0d6c79b4394e0
F20101204_AAAKVX burla_r_Page_007.tif
8f064e8232eb4b8ec477dd7ff0cd4bd3
000c0c055e966ece89a61640d71bebaefb7f9900
27361 F20101204_AAALCG burla_r_Page_029.pro
eeb21c9eb0b2c1b3bbb1a0a05991b8d0
909c3a624c5f51dd960d5acf095290c282d05d2f
F20101204_AAAKXB burla_r_Page_041.tif
cd7bf31b633f9a745b0778bc3fd2da6b
461f97abf7c9fea1a6116205780026900f42f363
22063 F20101204_AAALBT burla_r_Page_015.pro
4c40e6410fc7385571e7d867a9db52b7
a6a1183ee23ba7bbff6c8c2c50667e13fb2d874e
F20101204_AAAKWN burla_r_Page_026.tif
00d3ef96538fed599f901fb00f1b5e1f
e25f5cff634c344391140b1e0d927905a2c19ba0
F20101204_AAAKVY burla_r_Page_008.tif
00c9695f3134ac24b62413d75bb9cc55
51896 F20101204_AAALCH burla_r_Page_030.pro
6b2fd3f51af94548f4c0b2d35477f6ba
F20101204_AAAKXC burla_r_Page_042.tif
57d3ffc2b6fe8e4f12712e7c3bbfd6402b27cdb9
52310 F20101204_AAALBU burla_r_Page_016.pro
9e9827dddfcf3fdb7739aff4df62a8d0
ddabd0591da92f10975c6def795b5f0d63cdeaeb
F20101204_AAAKWO burla_r_Page_027.tif
b034c45e262b6179dfd067e024b22565
F20101204_AAAKVZ burla_r_Page_009.tif
51886 F20101204_AAALCI burla_r_Page_031.pro
76a107912c6a3ca01406934fd5abea42
98345bccc1976e08e78a7c5762f2c583728eeb6b
F20101204_AAAKXD burla_r_Page_043.tif
3b43bc613bd95be43e4feba845e81de3
6fa7e1c90b298c46dc68ec5dbfc554d18bea5123
55014 F20101204_AAALBV burla_r_Page_017.pro
5d397e9fecacfa5ab65a56da97d0848f
296269136fd113bc5dab1acf0fb60c2d548bd64a
F20101204_AAAKWP burla_r_Page_029.tif
912fcf04748e4631e7b2f079d904a414
1d1617ff0bc5fc3268dfe533441eb347f66ed9cd
41580 F20101204_AAALCJ burla_r_Page_032.pro
5751edeee7cee2a0d1a501ee0fec12d9
6f05bf3d0af6037a5283e92e8e17e8567790b657
F20101204_AAAKXE burla_r_Page_044.tif
2d51193f0040e8964a894a8710425ce7
1d7cb55b9f31a8753c585d9fd1cde0b6a7c1d3cb
51822 F20101204_AAALBW burla_r_Page_018.pro
fa3039c866f655d5b6e3a0a81f7548b7
51f7252404e76aa3aee2c54e656bd406caa66efb
F20101204_AAAKWQ burla_r_Page_030.tif
eacfed89b4400cacb0719805915452a4
d7ffeba858e661b2c486e10c099a0a695df32eb8
38463 F20101204_AAALCK burla_r_Page_033.pro
F20101204_AAAKXF burla_r_Page_045.tif
a30b4540e261c292d72b48d3f8e19bfa
1c434c986d3fe4265c3de1510dc132661fb478a1
52987 F20101204_AAALBX burla_r_Page_019.pro
e303089b42f4895b0a7cea50a5d8b729
F20101204_AAAKWR burla_r_Page_031.tif
362143defd5ba0e98562d2e432716a38
4ff6830c91314bfbb11f87036b21d22b1f86f5ff
5817 F20101204_AAALDA burla_r_Page_054.pro
dc8efc36b897863bd19becda870bcaec
45608 F20101204_AAALCL burla_r_Page_034.pro
1fd3402b4ce0a5be8fc28758c8b18ba9
e46fe5411a6042be7f40a48ffd0320becf154247
F20101204_AAAKXG burla_r_Page_047.tif
4b4419cd424117b85200a61859998faa
34077 F20101204_AAALBY burla_r_Page_020.pro
1c945767d1e9e76ed18330883b643f66
daee3d9d4ab9b797becce329c7c2d3d6ec30a856
F20101204_AAAKWS burla_r_Page_032.tif
0b33108de1d10271fb29b93710ab3612
afec2dcae466c73f670582b09324a02dfe5f74fd
49104 F20101204_AAALCM burla_r_Page_036.pro
d685aec8a6d8877d994d6e952ff38405
F20101204_AAAKXH burla_r_Page_049.tif
37d7c2e920e17e03e27e8700e9038647
20c64813ea0d055b315c38d58c4567943c76bf31
50260 F20101204_AAALBZ burla_r_Page_021.pro
c9c14cb35e6f0048fe99a8340f7bd178f4196d9e
F20101204_AAAKWT burla_r_Page_033.tif
fd5aff14665c4e21225a6c16d27100e9
f7d0530758d26f3b17a81b779026ce0d46ddcf3a
6802 F20101204_AAALDB burla_r_Page_055.pro
9ec5b4fc07378329e686d0f90d355d36
16aecbb300f003275e28d072bfe207140ae1bc2b
41942 F20101204_AAALCN burla_r_Page_037.pro
947fe5466e0140ff11b93d0cdf753e09
f8f2ca543cf41a8814121a182b1b6241eafcabfa
F20101204_AAAKXI burla_r_Page_050.tif
c8ac0dc07beb5d442973d7b0f40b01b1
ac1c96d47a428d83484493497e048384265f1896
F20101204_AAAKWU burla_r_Page_034.tif
16bf3d56f033056cf7978e2975294b1c
ac95049faa3c812fe04b0ce71e75493f9f7cce74
5350 F20101204_AAALDC burla_r_Page_056.pro
fb5e8983d76047653819a476ace958ce
df1ea57ab69e93e44d3149ee771120386fae3a2a
54959 F20101204_AAALCO burla_r_Page_039.pro
F20101204_AAAKXJ burla_r_Page_052.tif
46ce8b14b5571ed6e70e1136ea3cb478
f2d637c8e73fa9ec47cfcc226d2ab2333c8296d8
51321 F20101204_AAALDD burla_r_Page_057.pro
1a73d625ee02446fd1162aac80f2f7dd
24f9d91970c0fa609aea9b5be4b8d2c6ce0e81d9
54761 F20101204_AAALCP burla_r_Page_040.pro
ca5191c6337bdba654ba9c7f8ec3800b
F20101204_AAAKXK burla_r_Page_053.tif
a5fce9af90bcccb2d86e53ec525c9082
a3f7d2165de7a01c31c57b9783fcdc38a6a5d6f6
F20101204_AAAKWV burla_r_Page_035.tif
128ee68fe6ff5c8bc8fa076ab961c3c9
e5bacdbba9ae048a95529121c32e48b39cbe21c8
51459 F20101204_AAALDE burla_r_Page_058.pro
af9aa3def5f3828c120c3993e63332ea
47066 F20101204_AAALCQ burla_r_Page_041.pro
4b6f8aac63cc8d00c373caf124eaaef2
f0d3f8abce11e5ec45d794158ff5983821bef915
F20101204_AAAKXL burla_r_Page_055.tif
3663b035b28dc42844a6a57b6dae283b
f8c7890dea67a04002be5c09ba0f575621f228b8
F20101204_AAAKWW burla_r_Page_036.tif
3f449c728d5e4c92081ec4c9a7078c98
219e5f8d3e3f648cf5378ccaa978f61d53aa1b8a
42311 F20101204_AAALDF burla_r_Page_059.pro
25f8a28c3299907109e81ff71c1c34e8
a6af79c4cb913277121b3893e4d41c7b9df9bd36
47984 F20101204_AAALCR burla_r_Page_042.pro
5812fd11dc1bcc73c5cf99f9602d7edb
F20101204_AAAKXM burla_r_Page_056.tif
67f7eac56a9e103311d0ec8f3e41bcec
7f355ccab38e87c9bf0a7e8c3662092b2a6aa1b3
F20101204_AAAKWX burla_r_Page_037.tif
a8f2a97062d0a93dd27008a251168c25
d115e0156a8074c36fc5ddb2c93be05ba9b4e83f
50596 F20101204_AAALDG burla_r_Page_060.pro
56bdb476b845bc3aa1fdf27e00c20444
66ff90567572fe7388c4122d769985df09e93d5b
F20101204_AAAKYA burla_r_Page_072.tif
47db140a1e4d4a6d4be723404124f3f7
0fd062e654c4f57d3098c3fb53760268b344f849
37214 F20101204_AAALCS burla_r_Page_043.pro
afe2d6f56c48ba2feb8de2c6291771a7
08e197a86cbfdb23a80f7dc1d472bacbd785e5a4
F20101204_AAAKXN burla_r_Page_057.tif
b7edff6631d4b38882b69327693ea24b
80158d03a9799622257cb176a811c800364088ff
F20101204_AAAKWY burla_r_Page_038.tif
0574ab2be543e432556c7147e1e46d5e
a4b60382c062a27e3e701c91b21e313ff01e8b04
51154 F20101204_AAALDH burla_r_Page_061.pro
1d10bd2acf47328f76d9fd82ea602c7f
F20101204_AAAKYB burla_r_Page_073.tif
0e0cf8d8871a49a456d2b5cbdc1b214b
52337 F20101204_AAALCT burla_r_Page_044.pro
210eb26ba92a1b46f460599e1622d857
bd1bc161ee16b64b413dae2924c1bf89691467c5
F20101204_AAAKXO burla_r_Page_058.tif
0f0907dbefa13a5ef2fbe62e501f45e9
63effbaa240a2a1db288f694267f460ff82b374c
F20101204_AAAKWZ burla_r_Page_039.tif
56e3c3150ca56df345703bbb3be1beb4
46262 F20101204_AAALDI burla_r_Page_062.pro
6de918a0989bdffe56424b1107a02b9d
4911e822a0d667d41e3d853d28699a9539c7c573
F20101204_AAAKYC burla_r_Page_074.tif
e51f198fd4c9580332f9d0478c06bea9
d641f33632afe5266873d6302df7b2aabab2a9f5
45189 F20101204_AAALCU burla_r_Page_046.pro
ea020d71923de806c6f0f3d85a699bc5
F20101204_AAAKXP burla_r_Page_059.tif
034bb333056a16109126713a9975f14d
17853 F20101204_AAALDJ burla_r_Page_063.pro
685d4ea44a123eeea52a5274fff52d48
dd75a294e06e70d5f3698dbae346f34261886345
F20101204_AAAKYD burla_r_Page_075.tif
ec5e45858b9d9dbbc0b847a91e8b1a32
c514c8d21dc828429c12ce5b4815644bf2b1446d
38215 F20101204_AAALCV burla_r_Page_047.pro
c5cb8b8f2288cfc98ce941af5c9542d3
afd4deceac4b22520bc2a0a507c4d7c41f1ef2d8
F20101204_AAAKXQ burla_r_Page_060.tif
a95da062a0cd1e6ae18df0583b93e8c5
7e526be5f770aeeacd9fa740a53cd450dbd8042e
15958 F20101204_AAALDK burla_r_Page_064.pro
7aa553041fcc7f37cb1c27ebc3577693
ce7559fa240aafc4dbe6ca95dc84daa197c2a25f
F20101204_AAAKYE burla_r_Page_076.tif
a924efabeb3f8c255547b5f318581233
eeaf3b14e21722f39d26f917b148112c76a18c00
30450 F20101204_AAALCW burla_r_Page_048.pro
23288b2096a6b9cc39479120e852b9b4e2165233
F20101204_AAAKXR burla_r_Page_062.tif
d1342672bc6216866097fbaac29648c8f1c96123
35981 F20101204_AAALEA burla_r_Page_087.pro
32989d0966354525c37e270da5d78807
14032 F20101204_AAALDL burla_r_Page_065.pro
a28493f9bfa49a28edf57021105777a6
641143c23ac826e3b4b6143336a46cab6352acc5
F20101204_AAAKYF burla_r_Page_077.tif
b8dc0492c870c050c6543fc506b460c7
ec1467a1f699bb5f846d8462b945ffb9a641208c
6570 F20101204_AAALCX burla_r_Page_049.pro
c3dbcd18375f73a450f6b19b22f32092
F20101204_AAAKXS burla_r_Page_063.tif
30f0c702f8c1b9fe4a1a1088db2968e2
298a754e82c247d58c471937d5d25d5e1899a856
35065 F20101204_AAALEB burla_r_Page_089.pro
359dbf39fc6ca4a01a7be9cd7d28ea27
6842a57fcc0dd9853741f7c62d1afea8a4a65e9f
24484 F20101204_AAALDM burla_r_Page_066.pro
f6b2c8ca9825007c90b87ff9c701f72a
F20101204_AAAKYG burla_r_Page_078.tif
863877f3ed58c2f1482ffc9d3a663135
16344 F20101204_AAALCY burla_r_Page_052.pro
a246e9475089a6fbf41db511d1c179a4
c6d7b89691010aac56943e58022ee0d389e321fa
F20101204_AAAKXT burla_r_Page_064.tif
68142fe93e0f0d2f0477ba8f62b6da6e
a8924f4c601db7156e71146a93191c8ae094de89
43577 F20101204_AAALDN burla_r_Page_067.pro
de6a0c00540de3c11f210c37cf22c440
F20101204_AAAKYH burla_r_Page_079.tif
2153313443a5c2501001441d5245e7f4
54492486b3721dae74e97c2a96c8ee2575b53628
9814 F20101204_AAALCZ burla_r_Page_053.pro
1faac6742b50e44d43a2acb03d8ef0b910d0d41a
F20101204_AAAKXU burla_r_Page_066.tif
b01b05ca4437a233712a86458697dfaf
45565 F20101204_AAALEC burla_r_Page_090.pro
46b9ab85484e4ec1d98713cb032d757a
eb34c3eea1457f450ecfda49b08c8d862a17cca0
49987 F20101204_AAALDO burla_r_Page_069.pro
736ca2b410c992f7a1bbc82068413bc8
22dc157db9f874e4c10eaec021fb837d784ae0bd
F20101204_AAAKYI burla_r_Page_082.tif
fd10b0a15d09580142d87f0624ea58671939ed1b
F20101204_AAAKXV burla_r_Page_067.tif
7f167e3fd4ccb9949a91040da937ed98
38241 F20101204_AAALED burla_r_Page_091.pro
cbc5d0d64df59daca531a8f64e761fa4
27ccdc5f56ca0fe6374b08d0dc0faafcb5fd4b62
33184 F20101204_AAALDP burla_r_Page_070.pro
0a52b6bd4595e416d05736fe314b7a8a
39b84e6c905701477029f97c0d785c507ee34c3a
F20101204_AAAKYJ burla_r_Page_083.tif
2cd23418cc85efb51894d5e39688fd20
f542a86d3707f01feb9f717c1c984460d19b5e5e
39342 F20101204_AAALEE burla_r_Page_092.pro
0088022420f7b740fd9b00183a3d4d09
aa37b4570216be449bd126ffcbe2b5d7a9abb011
48119 F20101204_AAALDQ burla_r_Page_071.pro
d8c3d575d9a8a1b4ff3f912d58847b2a
d81c7cdbef80f0b08329ece3dbfd1c8416325427
F20101204_AAAKYK burla_r_Page_084.tif
31fcc0cb6d34629171885b760e5ce047
233f3829a754dac541977df3917a2e0802fe3078
F20101204_AAAKXW burla_r_Page_068.tif
5dd8c4007901e134676b3f0b26b7b57c
f2d5dd3515c93166f3d75e18e778ce290206371d
41198 F20101204_AAALEF burla_r_Page_093.pro
868dea969a5277d7dbf875f5b203f9a1
6ea9f330d14c117461212978ffbf322423f79317
44412 F20101204_AAALDR burla_r_Page_072.pro
606990fc36d376c4b6976a97ba6ab2d0
5c55bec033ae6afd2aa4a4883ed20faab9b786c8
F20101204_AAAKYL burla_r_Page_085.tif
231df83d6fbecfcba66a4555c498262e
76e7e480100ab7bcf63eb21ab33d1dba676f3c9e
F20101204_AAAKXX burla_r_Page_069.tif
9bca8d9118fe70e072959ed98c4fdacf
43632 F20101204_AAALEG burla_r_Page_094.pro
13be690735576ef266561d8047864b06
9583cca2f0eab478f8e9974fc69768a29ddea690
F20101204_AAAKZA burla_r_Page_103.tif
65267dd5358659e6695e4b224e03a4ba
9f0a3e5f5b8cb3cda8138104fea332d3776358eb
47168 F20101204_AAALDS burla_r_Page_073.pro
9f3cbb1fab7cdef4313eafcf1e88cb13
F20101204_AAAKYM burla_r_Page_086.tif
e1fd7cd944be8c5b3188007fb5dd655a
88bbe01a624134509e25c88904fe4369f012e1d3
F20101204_AAAKXY burla_r_Page_070.tif
e6b68f4421a302f5abf8794fb10d9e78
b5c3706d1e1f4e91f32eec87a6596c558219f1b3
24262 F20101204_AAALEH burla_r_Page_095.pro
2bc5464c0f387e2c71aac3bcb0b3250b
9b7f077c7b727b86452ab10438322e1001040c94
F20101204_AAAKZB burla_r_Page_104.tif
930955c3250fd6e9a07df00ace26933c
9f10c4e0c1d9b0d35f03856f55d93cf43eb0ebd3
48767 F20101204_AAALDT burla_r_Page_074.pro
5cc45b2d6b06a18eaca758c05af9a7ee
3ba79a7377dfa1f4162e5bea73661a410cb86147
F20101204_AAAKYN burla_r_Page_087.tif
88924eb1d240bd60130278b520623b38
3282 F20101204_AAALEI burla_r_Page_096.pro
23384fd5fc44404631ce82a4d24996b91695dd3f
F20101204_AAAKZC burla_r_Page_105.tif
cffeef3936000d0c942b8b2d331307f8
04e89865cbaaba1b921cdee4e96efabbe486480a
36451 F20101204_AAALDU burla_r_Page_075.pro
f07520943f18ca03b1761fb266350fa3
c1d5bf182860aed721cf196e80fd885afcb1bc04
F20101204_AAAKYO burla_r_Page_088.tif
520109616e44c20012b8cf41eaacbe01
10060120757019c0550ce8c8fd455af47f490d1a
F20101204_AAAKXZ burla_r_Page_071.tif
179b4ce814b924abfb99d7845502e9de
7004f4fcc47c1788fb868879e169cd4258c49d56
5999 F20101204_AAALEJ burla_r_Page_097.pro
a90a079e495cd9e9c529d6514d827fb3
720141dc91b6fd5a8f0cd9def417a28f2cbeb563
F20101204_AAAKZD burla_r_Page_106.tif
852d23ae0ae966a282024182e2354351
7ce0c38dd103d0b594e42e23ee8097a610fe0e08
29707 F20101204_AAALDV burla_r_Page_077.pro
8ef8fc0738ce27d3db828e981af5068d
1098bcd14ce22ede1b59bfc6b773a2302c7134f7
F20101204_AAAKYP burla_r_Page_089.tif
7e8f9b44e2e4f47b0680bc9d344d2012
0b888b77439fa6ea0cb7c22d1c33f9745bc6c9f6
9374 F20101204_AAALEK burla_r_Page_098.pro
6c9155bf2118d939fb00af486a11881b
7a8cd6439a22a2684798dc4a414111437cea0ce6
F20101204_AAAKZE burla_r_Page_107.tif
e58e6681851588579b505139709339e0
f8198552a919a4fe372da390c93c182ee46a518c
32980 F20101204_AAALDW burla_r_Page_078.pro
b125e9d0c6a49d01153f78f09c1b6de4
580113a5e8ee995c716dc2bd30f100a3822f109d
F20101204_AAAKYQ burla_r_Page_090.tif
3eb522a77b0c988958e0c7e7dfb2fa23
c04f9eec47180385a79f4b0607f75b864d9d8d22
13568 F20101204_AAALEL burla_r_Page_100.pro
98e776cb214a58e1fd84b64f8f7b135f
10f0b41175152cafbfa919b80f1fbfecda0d9080
F20101204_AAAKZF burla_r_Page_108.tif
7d43e37637ecd57f494ea3065bee0254
bc75c84ebf433ff862ffd728e86b5bc532741cd7
42360 F20101204_AAALDX burla_r_Page_079.pro
c2096fdd96cb8d5c956071ea84d81899
f13edfccfa053a0eac22b75b6c58f7f57fac6eb6
F20101204_AAAKYR burla_r_Page_091.tif
5334cf1b880038dacfff55becdf684d6
b92b9bf3cc4436aa812f96c2be40c6e03681217a
48697 F20101204_AAALFA burla_r_Page_115.pro
e69fb5a8e3703a292bdb6fa2d2be5267
74e11ac8b879e21f82f35bd891cef106100d1d78
8210 F20101204_AAALEM burla_r_Page_101.pro
f353c047cfe688ae9412fb5737e31a9b
0bab01f19d22e9903d4f855e69a57e842428b182
F20101204_AAAKZG burla_r_Page_109.tif
e8dfdb49b1999fef44c86f0c77db6cf2
82d3bbd07c2c9f38f7eb458b6c94847fd2b35923
47629 F20101204_AAALDY burla_r_Page_082.pro
1a8b9aaffef7c2204ba55b960df88f6f
a8ef462028486ba365bee4672f32e9c47ff45f54
F20101204_AAAKYS burla_r_Page_092.tif
c4abd8ffdd85b7d17a375823b8628832
e2bc2223c4298abc5eb08e7f899ea5e9e4c4bc45
50185 F20101204_AAALFB burla_r_Page_116.pro
9443ac34a06aa09d7d5e7210927bdd3e
56f1e460cd190a28bbc73efa3ed9d173fec4b39c
21729 F20101204_AAALEN burla_r_Page_102.pro
2347e7e217380b22ab9b158195ca4d0e
d124c2059566df0140de84d49970f4f5ca56aa1a
F20101204_AAAKZH burla_r_Page_110.tif
2c30454c370feb6e610bd650fe2f63a6264ceef3
38395 F20101204_AAALDZ burla_r_Page_085.pro
3d1ee9d08ef62fce03142c07ab653dc1
74eebfe99f7bc289667474b1f6dd0dbb42e3c8a9
F20101204_AAAKYT burla_r_Page_094.tif
aab4e05bbdf4c8b4f0126c8022af1d2a4250bff5
51100 F20101204_AAALFC burla_r_Page_117.pro
06c928424c2bf36bf3bbda709bd5864d
3d689754ed2ecb7fe24a55a352edb0721b38779c
50486 F20101204_AAALEO burla_r_Page_103.pro
d730f61a1b516a45c82e5ed99cbde3cb
1961d7a2b3b06407baf50033fba6f8f6e7cd1d6c
F20101204_AAAKZI burla_r_Page_111.tif
ae0f963ab67d556377c977a96fdee3a1
6b954db90d4699561103fce8a38a61f9b053f244
F20101204_AAAKYU burla_r_Page_096.tif
c92828519a556c57d9e053b863c82bf5
674ca3cf91baff6f042956d25c5c286bcb2ec9c3
47901 F20101204_AAALEP burla_r_Page_104.pro
b695663a6980a40d168b98ab5c8cff8a
a5a5a2bbdfcfabca103b5df30452b1b1cef228e7
F20101204_AAAKZJ burla_r_Page_112.tif
f1468f9d39798df420a5bd6f4b419c55
6f9fe19b8f9d82e0d251b531bf579d8f28ba4d57
F20101204_AAAKYV burla_r_Page_097.tif
22402b4c6a2fe7c9b2e945fc5345e860
01b5604218f1982518b89126bdfd077d1754fb1f
53930 F20101204_AAALFD burla_r_Page_118.pro
13e7dc7b5a3898a496f9aa721af9886d
93210d9c658890dbdcc197b8c10cee0eda1bb348
52041 F20101204_AAALEQ burla_r_Page_105.pro
6f17ce636337305b200b5edfa93eae44
F20101204_AAAKZK burla_r_Page_113.tif
13fe160b42bd1deea63f479fbe08da15
4b7ac41508434527bd1e1493e1417d66ea3d83b7
F20101204_AAAKYW burla_r_Page_098.tif
22aa886382db65f500c25621b546cbe6
1fab1d352f6583f2cf39cf6d73b0e4bb8d91169b
50995 F20101204_AAALFE burla_r_Page_119.pro
897e3cd287ae9b66aef3a1263531f856
54939 F20101204_AAALER burla_r_Page_106.pro
54cf8c0586993e3844754fb3a6a2c2de
156a3cf350680df1e6721fd0d874ac9d11381c43
F20101204_AAAKZL burla_r_Page_118.tif
90df3be072397d7589cd1cbd588684d1
20901287d7eca3f946ec764f8105538b7cc5d6cd
41178 F20101204_AAALFF burla_r_Page_120.pro
7f041ef4c6517302f66d944d9cffb9a1
48956 F20101204_AAALES burla_r_Page_107.pro
174d9f963b4c8d3b1210de4338cce888
d53a15c6217c10b348b6713e3b1f8c341dbe18bd
F20101204_AAAKZM burla_r_Page_120.tif
79d786425b1605f9aeaf3e2113628ac5
e41803c224264c798f45368c9d0b98835d018978
F20101204_AAAKYX burla_r_Page_099.tif
76509b3f9aee87d6599b6957dec4e861
c786d641b84be9f0fbd838d84cab562e6bcc2bf4
6172 F20101204_AAALFG burla_r_Page_121.pro
55b4281e01d31388cb6c43d3de7929bb
8714ece1c9944b981cb2a8bbea9362e0ff1a9626
41502 F20101204_AAALET burla_r_Page_108.pro
93130e1a1585a0b047b268cdfce68087
F20101204_AAAKZN burla_r_Page_121.tif
56161c8ce3ebd1ff9d081b12e8ccaa37
3114c08a8dc9f34c7fe79fa532b83cbf99d414c0
F20101204_AAAKYY burla_r_Page_100.tif
6a122ba9d0e75874efe4b90bbc632e1fc2daf468
3220 F20101204_AAALFH burla_r_Page_122.pro
23c1caa5a48da8a331b370f920c770e8
6171ce1e94fdecfe7cde5496b51b26e9917d8b34
50755 F20101204_AAALEU burla_r_Page_109.pro
ab8e6636979ee49fb8aa702a4721d163
85c2eb43a05ebb6ebfd407ddd5b25a43ab56a2a8
F20101204_AAAKZO burla_r_Page_123.tif
b2227b2ab858254a12c5dc84f8d20dc3
4c94573114d552781c7b40f138bd6ce73fd69d60
F20101204_AAAKYZ burla_r_Page_102.tif
c05a5f5e0185f76b4daf05bf8bcc6466
6bbd5b120db3fc199384d01f7d710a870e792bf0
11629 F20101204_AAALFI burla_r_Page_123.pro
da9c9733baef6b22a480c609770f42582ca19b50
47050 F20101204_AAALEV burla_r_Page_110.pro
5d5d27d92e5fb1b4d11e1c03696ea792
F20101204_AAAKZP burla_r_Page_125.tif
8157 F20101204_AAALFJ burla_r_Page_124.pro
4a20460dc9cf0363614847f68c017b6e
aba9991a5c05f552302f1b50e75a68811de78084
45642 F20101204_AAALEW burla_r_Page_111.pro
86e6bf9d5dc868bd9df8e0b53d6dce42
94e33b7ac35dd00b3186e168f90b81a2dbe4005a
F20101204_AAAKZQ burla_r_Page_126.tif
2403b85016eb1e8e557904c1e7ff1556f7dd902d
12005 F20101204_AAALFK burla_r_Page_126.pro
2a93c1508d9a8428eb4503d8f794c0ac9aa41012
F20101204_AAAKZR burla_r_Page_127.tif
bed65776403bae6a49c34708d44af2eb
c875e024047a2eedfa874261edc5e86b9db5e7cb
13317 F20101204_AAALGA burla_r_Page_147.pro
d2e5eacb449fe03ba818099322b78e01
2f990585bf32912b9690b23395094d6718b3e610
7973 F20101204_AAALFL burla_r_Page_129.pro
d2875f7b78869cdceca8250b070c88d5
7678a8c0bae004f88735346c92b0835d2baccd4f
43425 F20101204_AAALEX burla_r_Page_112.pro
785ce359ccd48d459bdb1e651357d391
2af6f1ee7abf2bb0b6bd4848548d61a26e29b888
F20101204_AAAKZS burla_r_Page_128.tif
e34a515b0651b7fb593af0b95df5f144
81ef7163b37a34e874b4914298767030bb531eac
11788 F20101204_AAALGB burla_r_Page_149.pro
1f8ba849d2fa617cdba23d602003b814e3cf11e4
9508 F20101204_AAALFM burla_r_Page_130.pro
243a19a385c5233b76b7a60577d9a02a
1502a177e49eef608c13c717831e37556bc47a97
50905 F20101204_AAALEY burla_r_Page_113.pro
7fa65159c4c4fa8c74268244fb494b6e
0a56d71cfc0d39cabb280b1340788b5248d0bbee
F20101204_AAAKZT burla_r_Page_129.tif
c17e32981b427b8c8bd0e626c712808027af6aed
F20101204_AAAKDA burla_r_Page_124thm.jpg
82aa795fdf1526a4c8ddf74ac39628b6
e0b71f417590b521bff566600502120259066626
12266 F20101204_AAALGC burla_r_Page_150.pro
a8ee7c8a5d5fd12554619c92232c5a6e
5fa0a948c61e2e5aa61c9f6d426b9485f7faeb3b
5746 F20101204_AAALFN burla_r_Page_131.pro
1633a24028dc1319354f2fcc18968551
52537 F20101204_AAALEZ burla_r_Page_114.pro
04e9608b943aaf2d9e411c980a3d6724
2db849f0a63360ab23f351da34a2f9654df6e132
F20101204_AAAKZU burla_r_Page_130.tif
a78c57ddd3b339abb0ebc71936386393
ac9936bc72b6f46582305e72bbec2783759d6ef7
283740 F20101204_AAAKDB burla_r_Page_055.jp2
8e98991704d58052dc39fc555ec5738f
364de43511d75d3cf7b1f54c029ef87a436ed1d5
13152 F20101204_AAALGD burla_r_Page_152.pro
24bdda88f8fd4be10562bce71a5eca0f
0cda5fe337575e4e6cf11aaa51e5bf202f1eea19
9963 F20101204_AAALFO burla_r_Page_132.pro
b21107cf47ca6120e9e25346121c9631
3f038c23345ff227e6fcf77a75b6ca7b8931a47a
F20101204_AAAKZV burla_r_Page_131.tif
0630eb47b423d76c153d1a2b505c1e9c
1f037d4375ed6d6ffcd51f781935728e4a860684
5825 F20101204_AAALFP burla_r_Page_134.pro
3860cd6db8772fa42c52842af7f8e2a3
c2aa74cf471866c5a8466a43b01540f83bd258b2
F20101204_AAAKZW burla_r_Page_132.tif
5bb02e6d1d55b518003acdcbc09295a8
b414b97ea005540e15bec34ecbac0d253e3ffea8
77292 F20101204_AAAKDC burla_r_Page_093.jp2
dae2ddd2150cc386225bf58cd2b0b2e5
20965 F20101204_AAALGE burla_r_Page_153.pro
27a0dcfdaf8dc6324bb756b35e45011a
89db359e78483d4c8cb3641e37021f3f069ae03b
9872 F20101204_AAALFQ burla_r_Page_136.pro
b9dc40e3e2394ce99f463076ce1c633d
237c69e3f9817340d8646e1fa3cbdde169f103d2
F20101204_AAAKZX burla_r_Page_133.tif
f1dec90f34b53deb363ba5a90a6b72eb33018e49
47617 F20101204_AAAKDD burla_r_Page_139.jp2
237432f6a3ed525709788c14655d08a3
06e8d54dddf1eea0f29d4640f098f70964816c90
5883 F20101204_AAALGF burla_r_Page_154.pro
4cf4ffd44bafaee3558d5ab8dec73c20
d3207802378c4cced9d20d621773b6f4dbd3653e
7198 F20101204_AAALFR burla_r_Page_137.pro
9db9e86222e3ba631eb061430c26a345
74755db35185a07b32b1b2d8f060636b528ccebb
102021 F20101204_AAAKDE burla_r_Page_084.jp2
54178d763a69e297759b596dbc33b2ab
e0c5e5d640fb68b55de3da325ce424c139693883
21570 F20101204_AAALGG burla_r_Page_156.pro
e3d007df398a229c61b6caa7bfbbb753
2b806fe709d0a058647c1429c05f56fde49e24e2
8627 F20101204_AAALFS burla_r_Page_138.pro
d035e5544c7c14158af47bb1a0ba6d1ae06a2aef
F20101204_AAAKZY burla_r_Page_135.tif
552b370d6d84d5ccd2f464cb29d9ca33
3163707d5db8d6c7a985a56711c7ed8dd3a25870
2229 F20101204_AAAKDF burla_r_Page_019.txt
79c191693d2f1f1a2b876b4138551189
3530d504bfdae815cb2137b9dd9aa0ac41dff96b
7152 F20101204_AAALGH burla_r_Page_159.pro
42f6de14d33437380a0c08934cf09401
12499 F20101204_AAALFT burla_r_Page_139.pro
f9fd2f0ce2789f7c3c83b5cb005e19c0
7995af653ecbb5dd106817f2f7f3de9af02f1347
F20101204_AAAKZZ burla_r_Page_137.tif
2db45c658466bed98a31312fb7b31301
43f00c4731e20028e7188252b742ff1246541ffd
1861 F20101204_AAAKDG burla_r_Page_080.txt
9c846754dce01d3b0756157eda539702
d81a5d0b96171a823d4453d55e6c74d725af8fbb
10665 F20101204_AAALGI burla_r_Page_160.pro
cdda46befa9c14b27edd29e4dfb7b06626d9b2df
8063 F20101204_AAALFU burla_r_Page_140.pro
59aeee369a5ef81a3588c39b9374da2c
8465821a748046e04eeb930bdcaca5cce143667b
5609 F20101204_AAAKDH burla_r_Page_068thm.jpg
746fea46c30cef4eae14a9e60b31756a
bb2a05e7151958e80285d45388a1a03041bef126
11376 F20101204_AAALGJ burla_r_Page_161.pro
151504439d8a28494e6f467b7a589d17
2fc0ddf1d9f649884b749d82ebe988d51055dbb5
9465 F20101204_AAALFV burla_r_Page_141.pro
b1b37f93bbcb5d09b6d2db9bbbec8c74
8d23e9a8c6127874bdb31f5b2a2a71bb6d9ddc36
69882 F20101204_AAAKDI burla_r_Page_071.jpg
13763 F20101204_AAALGK burla_r_Page_163.pro
6ebb7194d4c609242f96805caf621b40
4a32f81b46bfe7a04f1a28d82930edca51a37efe
5567 F20101204_AAALFW burla_r_Page_143.pro
bd822588c9b7617f513eb6c10870afae
e253643afff68dccd1b07cdd591db03c36c95cfa
6737 F20101204_AAAKDJ burla_r_Page_154.QC.jpg
9a1943e66e356fef9c5abf2951567530b9346824
9895 F20101204_AAALGL burla_r_Page_164.pro
fde9a75f76d2a986012c0d3ca5da4361
2a4feb73e756d09a3da0464e437940aed9c703c5
10363 F20101204_AAALFX burla_r_Page_144.pro
224689c2cbcb9d4cf21f94bb638201f1
b7550798ae2ccf97abcd6be2b0299af7fa857035
4629 F20101204_AAALHA burla_r_Page_006.txt
49634941a06a830d1813f1067fe5bff3
71f1d29dcf898c427f80b894d0ddb589a8f38868
F20101204_AAAKDK burla_r_Page_146.tif
7cf85dac2b0bd247e31705074ec7e5a2
a30587665aa21a2c7af431c0f626461c9ac3ff77
49571 F20101204_AAALGM burla_r_Page_165.pro
d7b8b6f5e1f5fef965cc42b632ecde032c0702bb
2760 F20101204_AAAKCW burla_r_Page_101thm.jpg
78604bd091d08c06fbf4591677ded82c
def3a79e64a9ded32dd4a8913b48dd060df86997
10641 F20101204_AAALFY burla_r_Page_145.pro
b8b53f5542fdecef0dd01a2fd6d506eb
92df20c2c9650bf30875b25edb5bba42f8aa7ef2
299 F20101204_AAALHB burla_r_Page_007.txt
cec58aa542449d2fbfe06fe85ebbb768
549 F20101204_AAALHC burla_r_Page_008.txt
a212a5be2236a0372aaa246440fb85d1
953a7e8aa6fcf84356cf06a057830fc93f6eaf58
11665 F20101204_AAAKDL burla_r_Page_162.pro
89750b84a6c9bdffa8d7c5436efef961
d493fc9b8f37f3be81b23da2bb0491ee42d368c0
51334 F20101204_AAALGN burla_r_Page_166.pro
a336440d73e6481a37f376799383de8a
5dd31d5fc6cbf1f9bed86c24501956dc33aebc78
20180 F20101204_AAAKCX burla_r_Page_092.QC.jpg
25da85832bc6ca864274c908a1b34d00
d8174b72e5a5d5c91036e29f5dda4e0a31be98fe
7253 F20101204_AAALFZ burla_r_Page_146.pro
1aeb53d957d4064e015c83f37e5d7855
5318 F20101204_AAAKEA burla_r_Page_150thm.jpg
7f5ac0b2fd46e6f1b86ed051a55bbcea44b1cc59
2702 F20101204_AAALHD burla_r_Page_009.txt
a1b451f0548af7dfac64d1ddb5f2136b
2275 F20101204_AAAKDM burla_r_Page_088.txt
93c348dc84f9b82ccde7931a91458123
7aec850697f0d4dfbfa4b4ce20ccae2912edcd01
52161 F20101204_AAALGO burla_r_Page_167.pro
3ed6cd565539d6819b59ede4c079fe0f
104889 F20101204_AAAKCY burla_r_Page_083.jp2
78becdc40dc14ccc380db1dfe0c852f0
bc386cd5e58c0737cb06a55ff4e65e8d8a1c7db8
9457 F20101204_AAAKEB burla_r_Page_003.jpg
af395b99b20f4f297eee9c67863fd76cffc8e34f
2727 F20101204_AAALHE burla_r_Page_010.txt
14419890dbd50d8c69d15d49259663a2
f19c071df78b4835541ebddccd578a8f2dc6c3f6
F20101204_AAAKDN burla_r_Page_020.tif
210e4c5d856d1539537d810193c6d804
000e938d5763de7b15ac2f4d3d94f565427cf0a0
F20101204_AAALGP burla_r_Page_168.pro
8a4c1626e2123f4de30c1634a0d66ac361908078
3560 F20101204_AAAKCZ burla_r_Page_136thm.jpg
07e218f68a7921fb3f25451924d2a0d04d9bb8cc
3034 F20101204_AAAKEC burla_r_Page_063thm.jpg
bfc63955af2cf18a45c4a3a6f0b75ac5
4a077748bb72c10c544469c0cdb9a45c3a4d519a
1827 F20101204_AAAKDO burla_r_Page_108.txt
4218e9083d2a5e4e6c94f8a18a9e8de4
2852a44581bcf46d9c405211161ca37ceb99e3cf
63645 F20101204_AAALGQ burla_r_Page_169.pro
e14d6bca92f0d13ae5e2f73514ac2c9f
342754c192b5e2fda425795efa4310641f48c000
2788 F20101204_AAALHF burla_r_Page_011.txt
91b70f2e14a3876fa966cec2076511ea
66062 F20101204_AAAKDP burla_r_Page_048.jp2
6285f8781165c82f00cf29c284e86f3b
c1c672aab3106f3b5745084dfb53a7e05f42f033
63076 F20101204_AAALGR burla_r_Page_171.pro
5f64054a9969256cab5dca5d1512b1ff
5cd65781953fd2a43ed79a3cee154c6c85120c1b
1352 F20101204_AAAKED burla_r_Page_153.txt
688d56b162dae979a90ee86e722f7e71
9e4620d099598ffc46e6589e31494907fa72217a
2832 F20101204_AAALHG burla_r_Page_012.txt
03008b45cb25962085d97ccc6e9ae9c5
a64f1b5ed97d4ebde664d168c362c000af10fbd2
11302 F20101204_AAAKDQ burla_r_Page_139.QC.jpg
ed19e3ef3a02bbedd4799f80b8620ff6
ccdcea2cae60e890fa82dc17934e8a7cdd31f23e
66661 F20101204_AAALGS burla_r_Page_173.pro
ecc2130014d028976073599e51574713
113599 F20101204_AAAKEE burla_r_Page_039.jp2
d698f5e5712e4f767d380141d7005838
711dc7d7c5798a79761c39a94d8b79441bb0aa78
1728 F20101204_AAALHH burla_r_Page_013.txt
e497d14dc4048973566fe2ae750b793b
5e5feda0f1ba8a81d1d898a5c3f874dedd889ce1
111176 F20101204_AAAKDR burla_r_Page_166.jp2
036960e8d8bca1020bcfc280f7b6866a
2032a8226e836aef238a7c3b83cdcd3a77ccbf9a
15782 F20101204_AAALGT burla_r_Page_174.pro
5bb98dbbda8280ec3a1fbf6d889ff5bc
4c0b8abc07353e92864cdcc3e93aed1049fecc6e
31117 F20101204_AAAKEF burla_r_Page_130.jp2
a9d52a4e8dcde60250b3a6ed325c5c21d3e1cddd
F20101204_AAALHI burla_r_Page_014.txt
24c47d1c783cccaa745657535e73e346
5d8efe17b2858f3e3e457c6894b8ff2771a3f8e9
52846 F20101204_AAAKDS burla_r_Page_093.jpg
02c5465579fd5cc00ecd69988039042f
4fedb3892201a7a60d36c6c13b763b66d56c0a52
19281 F20101204_AAALGU burla_r_Page_175.pro
567f3f7f7464ec08769abd7df5cf2db9
05e12ee896f3c5856c44d3e8e0dcf9f6e609ca10
25845 F20101204_AAAKEG burla_r_Page_102.jpg
aa8a3c5dcb215090ee26c8f814fa01c7
a3fc10ab9641fb18425a58f85ff7fca2ecbab69c
2169 F20101204_AAALHJ burla_r_Page_016.txt
f1ff35c9fdd9e93bc4e962f74a008c37
9a553f7a6ed8cd7344e0d28f0b3c16edd918e6f6
F20101204_AAAKDT burla_r_Page_061.tif
41f75436f1c38c21a928617be56cf3a1
7340260cf98921aa121d23ff8fa3a9652a2e18a9
493 F20101204_AAALGV burla_r_Page_001.txt
7cc73aebb97a2866b92cc7a62ddd0290
F20101204_AAAKEH burla_r_Page_101.tif
a0e8d95ee2f6ecfb8926a57fac0deecf
1bbecff3a88e8501530fd35046afc93924ca76db
2159 F20101204_AAALHK burla_r_Page_017.txt
22c13f948631567771622d6ca704fb27
338718b32faea984e6b2d54db58b4ddca13d088d
35592 F20101204_AAAKDU burla_r_Page_053.jpg
5087ab72d1b617f87663119030b55833
84f13061ae03047d7ba4067712ca59eddb8123bd
92 F20101204_AAALGW burla_r_Page_002.txt
3fef43e643474bf6ab83c6107ed2696f
b10637f690757fd89a0552f393d79dbcdb7c4b4f
39393 F20101204_AAAKEI burla_r_Page_081.pro
f00c5311df8360592d0f06a614f4b834
ed8ac8dfd5539c6c1e56756c88bdfee2bff545da
1757 F20101204_AAALIA burla_r_Page_035.txt
b94606365f7ceaa4631ea04f5c6ccd01
c2c9a881c253e0cffe5eaa12d03b047bed188a00
2046 F20101204_AAALHL burla_r_Page_018.txt
4a3c16d3b62faa7707f707653b64c1d7
66f446811da7ed3f2d06b33da340d226861bd921
76007 F20101204_AAAKDV burla_r_Page_114.jpg
a9bdca9f47970db1bd396d988238de37
01258421bfddd7e949edf9b554e69cdc39c6d562
82 F20101204_AAALGX burla_r_Page_003.txt
c0f82858de8a262db7ff805ef75df187
6ae7301f65cb9a1997bb0c25fd0a47e57186d869
4460 F20101204_AAAKEJ burla_r_Page_144thm.jpg
6b3bce1e3742760dbe53802a3e509ee2
aabc3a03d0835cc01085655086f736d67525b0f9
1813 F20101204_AAALIB burla_r_Page_037.txt
b53bf5753b9d116aeab640cb48481207
12667a1b1e8a02830e82f23240d7169d224fb18a
1377 F20101204_AAALHM burla_r_Page_020.txt
929cbe547c90bd6012af317e80b4817f
57007 F20101204_AAAKDW burla_r_Page_059.jpg
c96f0b78f8a613c38e4efd5b67348046
f446b332bcf6a5721edbe275abf2145720d736e3
1235 F20101204_AAALGY burla_r_Page_004.txt
6b041ba13722e7b106d142a0bf43cc44
d7054b96043ed647baf4dbb55cb488dc1836405e
59467 F20101204_AAAKEK burla_r_Page_037.jpg
c2e0035b36658e2504777f5fc26f776a
1997 F20101204_AAALIC burla_r_Page_038.txt
cf922e021a1f543b0801f73189d10753
409b8a9c08559261aacf4fc916eaa13ff56163a3
2067 F20101204_AAALHN burla_r_Page_021.txt
a7bfa34ba7779929916fcf42583cc98d
cf32ec05087a4cf03873fd4598508b116e5b9705
F20101204_AAAKDX burla_r_Page_065.tif
daabaf64fd543bca68f9281c93fc3209
bc881cac21e8da6ff9a4fa8bd8e002499d9cb719
3399 F20101204_AAALGZ burla_r_Page_005.txt
fcd149f1ff582385985af5ddd2b3b158
84948e02ba20946594d595a6f8532f197c4192f1
8459 F20101204_AAAKFA burla_r_Page_135.pro
e77a45f72ae0b0876077cdba34b7f6f7
55b8229d1bcc3ac9c659bb1dd3eb9790279cbb37
F20101204_AAAKEL burla_r_Page_117.tif
15a4b4d2d94ab3da1dfb662c29f679ff
96c6b7fc21e24f6f010e6876de7847d0bc9968eb
2193 F20101204_AAALID burla_r_Page_039.txt
b15b9295938c0ab145ba85c6690813b0
2153 F20101204_AAALHO burla_r_Page_022.txt
3263fe405334d21cda1824be8dbf711f8f749b1a
6486 F20101204_AAAKDY burla_r_Page_074thm.jpg
9fd260186c542a39db8063e76ddc2fbf
524e073d9f072155788e9d7115db83c28ba030d1
F20101204_AAAKFB burla_r_Page_081.tif
9b08c135fb12ed70fd13e4c7010358a1
6e68f4dea418c29b61bdcd7d788c427e56b6ddbb
72044 F20101204_AAAKEM burla_r_Page_165.jpg
3da4426e32bd76880c4c9ba8eb8e87dcb4c9b7a7
2181 F20101204_AAALIE burla_r_Page_040.txt
84c4cd018db74dba5381a17338ca8ac6
27e57be360cf8df5cb09cf0d9a0530a0a7fb46a0
2162 F20101204_AAALHP burla_r_Page_024.txt
b9a2710b8ed90b01e2348e74f40f5d5c
5e90f7a62e0a4944535b7abc38e4eaabbaaabaaa
23178 F20101204_AAAKDZ burla_r_Page_061.QC.jpg
287dd69fcb056ff823ca71874ba5649bd0c40d65
1933 F20101204_AAAKFC burla_r_Page_062.txt
cbe1d1c12f650824a9e45f826fb9a08483f906f4
7087 F20101204_AAAKEN burla_r_Page_132.QC.jpg
a7f787fb76055351acf8d0e0f3ce8654
c4d3329e01aa656a4bb3eb40aed0f92dee0bfb64
1942 F20101204_AAALIF burla_r_Page_041.txt
447d192c5bbd5243f3b2077a696567bb
cb6f336b998ca7ea99e8c022ceb6dbc0afe6018a
2052 F20101204_AAALHQ burla_r_Page_025.txt
4222e4ca85518ce1be9c91c9abddb865
d3eacdf5525e3523cf0017e80c42b4c5ec0e5c3e
13292 F20101204_AAAKFD burla_r_Page_007.jpg
31e9c573bac52f337333ebbc91251597
f3787ca8568f5dc7f380b32c3ce145979313c351
2331 F20101204_AAAKEO burla_r_Page_154thm.jpg
d3e389ec200e9d323dbea232543509e4
ebc9a48983a0bb49a77b4e4ac77f0067eaff07d3
2205 F20101204_AAALHR burla_r_Page_026.txt
e3ab6d7b0e3b169a528010507ccdfdfd
0b9bf5efaf733e733dded6362c3ecd8ef13a917a
7622 F20101204_AAAKEP burla_r_Page_050.pro
02d3a1ef58a0b85b9c61f4d98b5d5524
c8cde6c9cfdc27e0f7d5ef4ae05a02073aa1a7df
2083 F20101204_AAALIG burla_r_Page_044.txt
33c1e4db779d56bebbcb0c52c12a2085
11b81053b96619986ca8023b87d2d80862c13367
2233 F20101204_AAALHS burla_r_Page_027.txt
234d92517f4d7dac4a6951aef8af58cb
119eeeddf59dee16e286000dc55d537e14f29c0e
40317 F20101204_AAAKFE burla_r_Page_076.pro
b57d9c8dbe33d203c12bef0ef4a7aa4b
05dbae164d791695ef656a68dfa4dc25623154c8
F20101204_AAAKEQ burla_r_Page_116.tif
4fdd551186326beb56897075637ef24b
a7c0be00fdcd6889d6fbe6883024c8e68d5f63f9
F20101204_AAALIH burla_r_Page_045.txt
69a7d4b60106fb8c9bf1ec7ee8a4ef4e
73258ed5f237cacfff4b289d7e47c4a6558b68f7
2214 F20101204_AAALHT burla_r_Page_028.txt
06d8b01387c80067bb2363d5f458f396
98998e457f246d4b54437f2eee40bcf5cf99baa4
2015 F20101204_AAAKFF burla_r_Page_069.txt
6432eeac07e7a81a9a5b852d0feb8ccb
77360056c71a46964442673916d03a023fbc138e
25719 F20101204_AAAKER burla_r_Page_117.QC.jpg
3f0aa75522c0bba2debf6f80aee3943b
6038df325d951a61de72f31d1fc45dac1e8e3064
1915 F20101204_AAALII burla_r_Page_046.txt
ac50b86b5d8bca9d5ebe12a66d615ae2
a4d78daac3f6754d48fb32be407d7f4f0095a178
1091 F20101204_AAALHU burla_r_Page_029.txt
a20be862e0b9e3674a21c7e73da0a4fc
b162e6a8e0f52566b798da73a136f5c8cfa5130a
4922 F20101204_AAAKFG burla_r_Page_020thm.jpg
83a20cf5e594273ba958df44bf380b07
858cbb68844956e22625402edbb1843756bbedac
51298 F20101204_AAAKES burla_r_Page_088.pro
f6e49f425be528ac2cbfb069a003d01a
5ba6cb16982de8dc1582cb6092c6f87a7065bc80
1619 F20101204_AAALIJ burla_r_Page_047.txt
ca1d49407e1f2892f39fdc4cabbe6d8c
4f675893e36ed463d660e54ec54901cac393379c
2101 F20101204_AAALHV burla_r_Page_030.txt
1d5b44b1723a036a5823ef8d4582578d23dd376c
6169 F20101204_AAAKFH burla_r_Page_009thm.jpg
db690f28effe66444a5cbfaea63f6f7c
119162 F20101204_AAAKET burla_r_Page_006.jpg
583cf3b33e22d9a020e8172c3d600547
49114aa0f9044515f95a3e1ff90a9f821a7483c4
1475 F20101204_AAALIK burla_r_Page_048.txt
a1cc068d7ee132ed2e6a96f7e3069217
4fdc1cc79066e5b446c85ae6ce9284ac2d9a873d
2085 F20101204_AAALHW burla_r_Page_031.txt
0b876987b7eb25f012bff44eff16ba12
301743959fe8c1715be731f6c924fa74e7796fa5
4589 F20101204_AAAKFI burla_r_Page_135thm.jpg
b6444e578516d8f77dbc604c41939dde
e2fdaa49d4a0475a35b0e511c8a74ccd4848f980
42741 F20101204_AAAKEU burla_r_Page_133.jpg
863e96d2b48224d9f33d6ec2fac3edf6
fdee0e4bc86c358c99d095c7298db6feb6bb83fe
963 F20101204_AAALJA burla_r_Page_065.txt
ccf14669c12a5c89ce4394c502ee0879
0234afa6a62192325942053ea5ca4665b5f8b893
394 F20101204_AAALIL burla_r_Page_049.txt
0d75329ed51d342009431ae922004f35
fa22a3378f750dd1d98a8bcca739a184f978e3b6
1775 F20101204_AAALHX burla_r_Page_032.txt
0ce8dc136ccd981e86ee4d8342216b2c
04fd965365c3d1b653f015db89cce1806450e98b
73326 F20101204_AAAKFJ burla_r_Page_058.jpg
a26629653da57e94b67180fe1d97ee2d
6992 F20101204_AAAKEV burla_r_Page_106thm.jpg
4aceaae4f9f52d983943b1bf3053898c
42167024016c478a981dda2af41c2d274fd3f183
1489 F20101204_AAALJB burla_r_Page_066.txt
967078126eab145961bd6e9048119866
9855cb03282d0e0ea5dda5c00d304d6084d99559
339 F20101204_AAALIM burla_r_Page_050.txt
9fa2113e79abe70559a2fb25d9180800
3fb030bb7fec1c2aaef70255e99bd61b943218af
1744 F20101204_AAALHY burla_r_Page_033.txt
13661f9c26a4b57e89f34516735a80e8
6d5b734c1e4fcf45d300ca11bef26cb46a3900d3
100948 F20101204_AAAKFK burla_r_Page_082.jp2
73d3a2aa74f6c33387c4435029a851eb
70db869792f389e85598b818166142986f7a38f9
4564 F20101204_AAAKEW burla_r_Page_128thm.jpg
6b6e6d5b9433b2212989e8d1b50fa5a4
78985c24cfa73fb3b14f2aa842c2f93530dd1154
1697 F20101204_AAALJC burla_r_Page_068.txt
5422202ce4ebd869342846898f1e8ebc
200 F20101204_AAALIN burla_r_Page_051.txt
86341524be8d28367a7c0398848cd165
8bdb7275e335d533ea3699d4b95dfde237d0b0af
1937 F20101204_AAALHZ burla_r_Page_034.txt
6b7fc3a78cc45cda564576cc08ebb236
9c1e55722e2d107d9331f4ef5e224099d9107e62
76887 F20101204_AAAKGA burla_r_Page_118.jpg
75418610f517ff5a258cc47a30b7c770
bcf37867e2da9e4fb00d33e081d0906fb93dd789
9580 F20101204_AAAKFL burla_r_Page_151.pro
e25af8aaf8de4643126d588255045685f21e26d7
F20101204_AAAKEX burla_r_Page_147.tif
1945 F20101204_AAALJD burla_r_Page_070.txt
30ccc3720e08e1169224e24f57028015
0308399171169657faa539ef398dc05c9d7991f6
766 F20101204_AAALIO burla_r_Page_052.txt
80aaec05c94f19370f78e73cdb0e115d
c57ac2f332d157816d5c8c71b1cdac3f914f6bcc
18821 F20101204_AAAKGB burla_r_Page_056.jpg
1234cafedc64756e4d2bbdef2f8eab11
0844f67fcf9af33503c358f43a6c13cf69f7ac1f
15883 F20101204_AAAKFM burla_r_Page_078.QC.jpg
db83576a009ab084d94ac394f5cbc257
39010 F20101204_AAAKEY burla_r_Page_068.pro
802c48fbd401488edd278f8bac557a84
e0a909c1ae48084b2ed50a0691d84d1003216265
2044 F20101204_AAALJE burla_r_Page_071.txt
dc6660abffa01f7275d2e873f8b8c215
650 F20101204_AAALIP burla_r_Page_053.txt
08e8e380c62bca569b3d82c95e4c4999
24af06312587b7e4d4e9e1c9dba0ee7ea525d9f2
F20101204_AAAKGC burla_r_Page_093.tif
4147d49c786318bbb55c96e9f73c56a4
2a9bdcdbd7d3d1f2447f98a88c3b44a1e9baf350
4314 F20101204_AAAKFN burla_r_Page_140thm.jpg
bb8a1186f73112bd50c2e2a02de7c96fd868e09c
76264 F20101204_AAAKEZ burla_r_Page_040.jpg
eddd29b585398570e5f6ec794eb0fd8a
cfd20cd51c25b67d6203e30057925caa1b0dda57
2128 F20101204_AAALJF burla_r_Page_072.txt
318 F20101204_AAALIQ burla_r_Page_054.txt
a958fb7b70d2e083818e14ca8b9740fc
dc56705b93f1db4cbb6326aea03dcc0d56242a41
6767 F20101204_AAAKGD burla_r_Page_056.QC.jpg
378f63796160434026ab47f91a7f0090
1794 F20101204_AAAKFO burla_r_Page_043.txt
04dcca6db2cdae8abbc05e7110301d10
fea12e923abe5b0b2e2788727eed1f57d3e63fa0
1804 F20101204_AAALJG burla_r_Page_075.txt
9921eb2c71cc8791fcf10d5dc892ceb895c7fe15
354 F20101204_AAALIR burla_r_Page_055.txt
d0b9695fcd0db85fe3a797dd8fce5548
2657 F20101204_AAAKGE burla_r_Page_132thm.jpg
ae70d692e812ef315c7e1747457096ea
ef03f7578cab57e9c7eeacbe70acd2491d7e8a8e
55107 F20101204_AAAKFP burla_r_Page_043.jpg
e48bfdeb72c8b47c9137ef950ba8f6c1
fa56fd7593cfeac93dde4b0857189efa32fdb847
F20101204_AAALIS burla_r_Page_056.txt
db067c5edde421051322c243fc1a6d39
6204095459165c41039446a9d81742aba763c36b
73448 F20101204_AAAKFQ burla_r_Page_119.jpg
ceced7b2c92bd69021e00da974aced2820c22690
1994 F20101204_AAALJH burla_r_Page_076.txt
861bc8ac40e39d6613f9c4fa9cefe91c
F20101204_AAALIT burla_r_Page_057.txt
ccfe903e7a445c060853a7417974f210
ea7257e99552f37f93990b0580f8fac9d6c9e56b
38744 F20101204_AAAKGF burla_r_Page_080.pro
a800239277741220e15e049ac70649bf
e4f91fae23a956aea401ba39094847a2f0928a5a
85318 F20101204_AAAKFR burla_r_Page_012.jpg
2b1135ec335714669334c036cb658bedaec9fd19
1360 F20101204_AAALJI burla_r_Page_077.txt
b21d25ddfba575ccc3270d64c10b31f7
F20101204_AAALIU burla_r_Page_058.txt
ce5b3c4083b6aed71c92cb490cfc89a7
82e1bdf8bc3506e34cbe51d380a64741e9636473
3901 F20101204_AAAKGG burla_r_Page_051.pro
2e20bb24e8fc372223789159e312e170
e5b603c647b4fcc0011bb44967bac6127e131c0b
2360 F20101204_AAAKFS burla_r_Page_172.txt
cf9b49da9b4c201506edc7ab68ba8061
ca7b031b98edd3b36b837c46f7c232d61c2f07e0
1906 F20101204_AAALJJ burla_r_Page_078.txt
bdf451f166110bb6952efce93a10a1b1
caacdba036889be14873ec7bb5e53ce44b1766fb
F20101204_AAALIV burla_r_Page_059.txt
b0df9d5120feacc68d80b82822118b74
0ab897d7940c96168ab9c4d5a43e577c3d8b6f0f
5407 F20101204_AAAKGH burla_r_Page_127.pro
bd5cddd930db03515394ce484068af877655669b
F20101204_AAAKFT burla_r_Page_153.tif
90a89b186dbcb12dce7b640a03edde11
a3ba0d004d88a1f9b4b4fcb36885a53b885b822e
1821 F20101204_AAALJK burla_r_Page_079.txt
40531a6893921e6b563bd273691561d9
2266 F20101204_AAALIW burla_r_Page_060.txt
767486 F20101204_AAAKGI burla_r_Page_140.jp2
cca4ee15b8fbd08592d34bc3b2c62068
9093b31c0d1b93c4e60c1a3b20029b67f5a18523
F20101204_AAAKFU burla_r_Page_134.tif
00102cec126831235cae2f0fd2096f0e
78f638f95ab4eba741c3451e090929b00d36f7da
499 F20101204_AAALKA burla_r_Page_098.txt
0da4e17580a49839abe8a38e0c8b1001
b73fa04775232fe6960b7890a6091ef6ff5516b3
1888 F20101204_AAALJL burla_r_Page_081.txt
481166bab6942e28ea0cc8b8312c097b
e5272a1e93724a0447fc9fc59acfd4ea8791d3f7
2093 F20101204_AAALIX burla_r_Page_061.txt
38aa6db57b0a266b10bf1b73d9e05fd0
5d5fc80b167c4b190fa5efa392fd60bd45c0bf06
8145 F20101204_AAAKFV burla_r_Page_128.pro
7baaec887c3f3912d41637133761603e
39318 F20101204_AAAKGJ burla_r_Page_063.jp2
c5e066d75905d2a42a8d7e0362ff3e7d
bf5169035361cb6824e34ab44d741f3a6e3d3d82
482 F20101204_AAALKB burla_r_Page_099.txt
e21f6deffc156254751790724651314d
b76f2309b9ce3bcc09f50e83b6cdcfcf802cdeb8
1948 F20101204_AAALJM burla_r_Page_082.txt
786df783ff0ee9f92f1ba00d1d5a8eb7
dbaefb49cc0d7474c7337b765c0c3b46d494d1e1
746 F20101204_AAALIY burla_r_Page_063.txt
2c205e7dde96a9173b67d088a8945262
38a247c9162f1bbf0af5fc0e9713ea34950bcee1
176646 F20101204_AAAKFW burla_r_Page_122.jp2
c957e6b32333be00bbcfa2f0697ace11
a8de27239f2446267f50fee786cf8276d9588f15
F20101204_AAAKGK burla_r_Page_139.tif
c98d41bb7791dd1b587d5a02130d81d707d32741
420 F20101204_AAALKC burla_r_Page_101.txt
4060a2059a884423d0d186edfe1bd74c
F20101204_AAALJN burla_r_Page_083.txt
1f3c64941d99f514e8241eb500b82ff6
a1506820ccd93d5720d4ff7209f827250cbe2f21
1229 F20101204_AAALIZ burla_r_Page_064.txt
9de7a32cc63578b7ae8ed3427b743de5
e52d7e4186faf9fbe65f2e6ce8af48c89a2362ef
60629 F20101204_AAAKFX burla_r_Page_032.jpg
d96cf047a8b8b80f5d02f6d7e5f27f41b685c500
5474 F20101204_AAAKHA burla_r_Page_033thm.jpg
eca41210fd1b48a75fbac9bed5485442
4959c8f46531be94d943c39cc4c712634e74658a
1842 F20101204_AAAKGL burla_r_Page_112.txt
a935b2f6c792af8d7b47e491e7cb3727
5f8a2378c02e3e6a5793ec69f047501efa1a8c46
999 F20101204_AAALKD burla_r_Page_102.txt
2014ded90f0e9ed0c2e05d192615be0e
de977b98a06e595781cd13867f02973bca1a6a78
1864 F20101204_AAALJO burla_r_Page_084.txt
c36411df60577f75abc34d7d0e44b53e
e5b2eebfe3ef829dc5c425af94b1919cb144435e
22091 F20101204_AAAKFY burla_r_Page_079.QC.jpg
fdf652251661888038ce3db2c71e8885
c4ab65250f0df5d5c1efa3c2d53b101972e5d532
F20101204_AAAKHB burla_r_Page_080.tif
d40f77665cb363f6c8889947e775c2cb
351afe9439813dab64bb8465dd6815576151cff9
63013 F20101204_AAAKGM burla_r_Page_112.jpg
c00287fd59f432e6cbb621af88dffb4f
2079 F20101204_AAALKE burla_r_Page_103.txt
7bf5df2f05fc597a7e11ca4256a05391
95dc7d1f78236f9f0ffa40ec673c1ef6cb62c208
2006 F20101204_AAALJP burla_r_Page_085.txt
f2b126205ff55c620177c6395b73be6c
26f28448ce8de45873ecf55470376fee1e20914a
616416 F20101204_AAAKFZ burla_r_Page_134.jp2
4709 F20101204_AAAKHC burla_r_Page_050thm.jpg
26c2f97f48510543be4acac77072883a
dc864b79a04847fe13d6661ed9525a33f24eaeed
F20101204_AAAKGN burla_r_Page_074.txt
caa0668755778fedcebb761459a34c2b
3b687b145239a05f2a552e3ab5f515a07d64c66f
1990 F20101204_AAALKF burla_r_Page_104.txt
ef7d5078d5e1cfbd649c3fa305400443
7705b163f5f3be207b7baa19bbf52efc2a16ba2f
1767 F20101204_AAALJQ burla_r_Page_086.txt
a7f494373460636b983061005d5142e2
7232f3e717fc18f8554c8c0100a7e572f1128555
F20101204_AAAKHD burla_r_Page_002.jp2
d72a696ec2814ca1911d9046f02806abc3829026
3906 F20101204_AAAKGO burla_r_Page_065thm.jpg
8d21ebf5e9f5b932e1194581d2bc120f
fe1f7623f59b88323c7ffc82f52271f744e70ccf
2151 F20101204_AAALKG burla_r_Page_105.txt
38d094864071abcf79e64ea98e026b7e
F20101204_AAALJR burla_r_Page_087.txt
0876991cefdd6741e500d21a7ddaef71
d2bb480cc533926eb04e6b989acb0dc5e6d647c1
44883 F20101204_AAAKHE burla_r_Page_155.pro
045037d282416e1c1a2c68b16450fbdc1c58619f
1494 F20101204_AAAKGP burla_r_Page_142.txt
f00d1db01f116127bc6565b888bc0097
243b212a456d41a9d3e0d48cd1b4b2917d9278bb
F20101204_AAALKH burla_r_Page_106.txt
5c670ce438cec16c00cc8cc6c0be0b32
41a8b9dec367271fa89eeaed5a661240e20d649f
1748 F20101204_AAALJS burla_r_Page_089.txt
43acbdc0bdae169cf1792b7aa0a5cca997b35b93
25868 F20101204_AAAKHF burla_r_Page_131.jp2
1df5955500a2170b689eebd0738e51b5
3219 F20101204_AAAKGQ burla_r_Page_121thm.jpg
d23f73c25e9dabe7613a76cf188a1060
803a5196ff049ac04461967f58c0a5d85ac7c851
F20101204_AAALJT burla_r_Page_090.txt
2ccbbae48a0d86e6cf8d89d0f1b5ccb7
6af3b6dbed0510347101840d34c532803e988c76
11077 F20101204_AAAKGR burla_r_Page_099.pro
36dda2799e53b17333b9ce4f7bcc9eff
bd2c2f34a3102db32656a98d86dc69388662b047
1991 F20101204_AAALKI burla_r_Page_107.txt
242b257739fb4354a5c582bd42e60d01
48f2d2ff00f86eab0fa491000060b1cd21632872
1776 F20101204_AAALJU burla_r_Page_092.txt
08eeafd2e550475e765ae330fe824190
2f335b7f1c898a196d52ac872bf4ec438a094227
87622 F20101204_AAAKHG burla_r_Page_108.jp2
8c33a2f837c3fde4d19ba6bb92b59ac2
381dd7b9f95afa032fff08d43583f950b5996c73
53987 F20101204_AAAKGS burla_r_Page_022.pro
35066bde0a800766450b86bfbd817386
d414b66ca8cd83371fa39b045928d7339bb57ff4