Citation
Quantifying the Effects of Interference through Use of an Alternative Method of Productivity Estimation

Material Information

Title:
Quantifying the Effects of Interference through Use of an Alternative Method of Productivity Estimation
Creator:
CHOI, JAEHYUN ( Author, Primary )
Copyright Date:
2008

Subjects

Subjects / Keywords:
Case studies ( jstor )
Modeling ( jstor )
Pavements ( jstor )
Productivity ( jstor )
Rain ( jstor )
Simulations ( jstor )
Total parenteral nutrition ( jstor )
Trucks ( jstor )
Turning lanes ( jstor )
Weather ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Jaehyun Choi. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
2/28/2007
Resource Identifier:
649814532 ( OCLC )

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

QUANTIFYING THE EFFECTS OF INTERFERENCE THROUGH USE OF AN ALTERNATIVE METHOD OF PR ODUCTIVITY ESTIMATION By JAEHYUN CHOI A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006

PAGE 2

Copyright 2006 by Jaehyun Choi

PAGE 3

Dedicated to my parents and my wonderful wife, Hyo Kum Ryu.

PAGE 4

iv ACKNOWLEDGMENTS The completion of this dissertation re quired the encouragement, support and guidance of research committee members, fa mily, and friends. I would like to acknowledge the contributions of the many indi viduals who made this project possible. First, I would like to extend my heartfelt th anks to my research committee. As my graduate advisor and research chair, Dr. R. Edward Mi nchin provided the motivation, direction, and insight necessary to complete the project. Through his incredible work ethic and dedication to his numerous respons ibilities, Dr. Minchin has been a very positive role model not only in my scholarly endeavors but also in every aspect of my life. I have enjoyed being his student and l ook forward to being his colleague. Dr. Ralph D. Ellis, as a cochair of the committee, demonstrated excellence in scholarship and mentorship. Dr. EllisÂ’s systematic approach to research excellence and mentorship will serve as a model for al l of my future endeavors. Dr. Zohar J. Herbsman served as a constant source of encouragement and an exampl e of scientific excellence. He has been thoughtful and caring as we worked toward the completion of this research project. Dr. Reynaldo Roque, as a research mentor and co mmittee member, played a critical role in my understanding of the subtle ties of roadway design and c onstruction. Finally, I am extremely thankful to have th e support of Dr. A. Alexandr e Trindade, who challenged me to understand statistical research methodologi es, and to use them as powerful tools for analysis.

PAGE 5

v On a personal level, I have been partic ularly fortunate to have Dr. Charles R. Glagola as a long-time mentor . Dr. Glagola provided me with support and confidence whenever I was searching for direction in my gr aduate study. I am forever grateful to Dr. Ok-Kyu Kim and Dr. Hong-Sub Ahn. They have been an integral source of support and encouragement since I started my graduate st udy in the University of Florida. I also would like to acknowledge and thank Drs. H. Randolph Thomas, William J. OÂ’Brien, R. Raymond Issa, Mang Tia, Chang-Hak Kim, Chang-Ho Choi, Chan-Sik Park, and EulBum Lee for their insightful gui dance during this process. And above all, I extend my most sincere appreciation to my parents and family members. My parents have been an inspiration for me throughout my life. They are unwavering in their support, a nd I love and respect them more than I can ever express. They always remind me to focus my work in a manner that will positively influence the lives of those I serve. Most importantly, I would like to extend my deepest gratitude and love to my wife Hyo Kum Ryu for being my primary source of support, encouragement, and love throughout my doctoral program and the completion of this project.

PAGE 6

vi TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...............................................................................................................x LIST OF FIGURES.........................................................................................................xiii ABSTRACT....................................................................................................................xvi i CHAPTER 1 INTRODUCTION...................................................................................................1 1.1 Research Background........................................................................................1 1.2 Research Objective............................................................................................3 1.3 Chapter Organization.........................................................................................4 1.4 Research Scope..................................................................................................5 2 LITERATURE REVI EW DISCUSSION................................................................7 2.1 Production and Productivity Variability............................................................7 2.1.1 Work Flow Variability vs . Productivity Variability...........................8 2.1.2 Productivity Measurement..................................................................9 2.1.2.1 Activity sampling...............................................................10 2.1.2.2 Field rating.........................................................................10 2.1.3 Methods of Estimating Productivity Loss.........................................11 2.1.3.1 Measured mile....................................................................11 2.1.3.2 Control charts.....................................................................11 2.1.4 Labor-intensive vs. Equipment-intensive.........................................12 2.2 Definitions........................................................................................................13 2.2.1 Unit Completed Method (UCM).......................................................13 2.2.2 Baseline Productivity (BP)...............................................................14 2.2.3 Cumulative Productivity (CP)...........................................................14 2.2.4 Project Management Index (PMI)....................................................15 2.2.5 Conversion Factors (CF)...................................................................15 2.3 Computer Simulation.......................................................................................16 2.3.1 General Aspects of Computer Simulation........................................16 2.3.2 Process Simulation Models...............................................................18 2.3.2.1 General purpose models.....................................................18

PAGE 7

vii 2.3.2.2 CYCLONE.........................................................................18 2.3.2.3 Petri-Nets...........................................................................20 2.3.2.4 Process model for construction process.............................22 2.4 Simplified Modeling Efforts............................................................................24 2.4.1 Framework........................................................................................24 2.4.2 Resource-based Modeling (RBM)....................................................26 2.4.3 Activity-based Modeling (ABM)......................................................27 2.4.4 Simplified Discrete Event Simulation Approach (SDESA).............28 2.5 Transient-Effect (Non-steady) Model..............................................................29 2.6 Summary of Literature Discussion..................................................................33 3 RESEARCH METHODOLOGY...........................................................................34 3.1 Overview of Experimental Methodology........................................................34 3.2 Construction Process for Asphalt Paving........................................................37 3.3 Data Collection for Productivity Measurement...............................................38 3.3.1 Productivity Output and Conversion Factor.....................................39 3.3.2 Productivity Input.............................................................................41 3.4 Productivity Parameters...................................................................................42 3.4.1 Weather Interference.........................................................................43 3.4.2 Management Interference.................................................................43 3.4.3 Work Content Interference...............................................................44 3.5 Methods for Statistical Anal yses and Productivity Estimation........................45 3.6 Summary of Methodology...............................................................................45 4 PROCESS PRODUCTIVITY ANALYSES..........................................................46 4.1 Introduction......................................................................................................46 4.2 Productivity Analyses for Pavement Operation...............................................47 4.3 Project 1, SR-20 (Alachua)..............................................................................48 4.3.1 Project Description............................................................................48 4.3.2 Work Flow Management..................................................................50 4.3.3 Work Content....................................................................................51 4.3.4 Weather.............................................................................................51 4.4 Project 2, SR-102 (Duval)................................................................................51 4.4.1 Project Description............................................................................51 4.4.2 Work Flow Management..................................................................54 4.4.3 Work Content....................................................................................55 4.4.4 Weather.............................................................................................55 4.5 SR-20 (Palatka, Putnam County).....................................................................55 4.5.1 Project Description............................................................................55 4.5.2 Work Flow Management..................................................................57 4.5.3 Work Content....................................................................................58 4.5.4 Weather.............................................................................................58 4.6 I-10 (Pensacola, Escambia County).................................................................58 4.6.1 Project Description............................................................................58 4.6.2 Work Flow Management..................................................................61

PAGE 8

viii 4.6.3 Work Content....................................................................................61 4.6.4 Weather.............................................................................................62 4.7 Summary of Production Analyses and Conclusion.........................................62 5 EFFECTS OF INTERFERFERENCE...................................................................64 5.1 Introduction......................................................................................................64 5.2 Multiple Mean Comparisons with Transformed Productivity.........................67 5.2.1 Further Analysis by TukeyÂ’s Test.....................................................68 5.2.2 Further Analysis by Least Si gnificant Difference (LSD) Test.........69 5.3 Productivity Comparison by Interference Factor.............................................69 5.4 Productivity Comparison for Management Interference.................................72 5.5 Correlation Analyses........................................................................................73 5.5.1 Effects by the Monthly Amount of Rainfall.....................................74 5.5.2 Effects by the Monthly Number of Rain Days.................................76 5.5.3 Effects of Rainfall without Outliers..................................................77 5.6 Effects of Weather and Prerequisite Work......................................................79 5.6.1 Initial Design.....................................................................................81 5.6.2 Effects of Weather on Probabilit y of Management Interference Occurrence.....................................................................................83 5.6.3 Effects of Weather Interference on Productivity..............................85 5.6.4 Effects of Prerequisite Work on Management Interference Occurrence.....................................................................................86 5.6.5 Effects of Prerequisite Work on Productivity...................................87 5.7 Summary..........................................................................................................88 6 ALTERNATIVE METHOD FOR PR ODUCTIVITY ESTIMATION.................90 6.1 Introduction......................................................................................................90 6.2 Simulation Method for Productivity Estimation..............................................90 6.2.1 Simulation of Pavement Production Process....................................91 6.2.1 Result of Simulation using Initial Time Duration.............................94 6.2.2 Quantification of Interference for Alternative Time Duration.......101 6.2.2.1 Further categorizati on of interference..............................101 6.2.2.2. The order of magnitude for interference.........................102 6.2.2.3 Example result of inte rference quantification..................105 6.2.3 Development of Interference M odel by Timed Petri-Net (TPN)....105 6.2.4 Result of TPN for Alte rnative Time Duration................................107 6.2.5 Validation of Alternative Time Duration from TPN......................109 6.3 Simulation of Pavement Process with Alternative Time Duration................111 6.4 Case Study 1: SR-20 (Alachua).....................................................................111 6.5 Case Study 2: SR-102 (Duval).......................................................................117 6.6 Case Study 3: SR-20 (Putnam)......................................................................121 6.7 Case Study 4: I-10 (Escambia)......................................................................125 6.8 Result Summary.............................................................................................129

PAGE 9

ix 7 VERIFICATION AND PROD UCTIVITY PREDICTION.................................132 7.1 Project Description.........................................................................................132 7.2 Production Data Collected.............................................................................133 7.3 Estimation of Alternative Time Duration for SR-26.....................................136 7.4 Productivity Estimation for SR-26.................................................................139 7.5 Modeling Uncertainty in Time Duration and Daily Shift Effect...................141 7.5.1 Improved Result using the Simulation Method..............................143 7.5.1 Limitation of using St atistical Distribution.....................................146 7.6 Estimation of Construction Duration.............................................................146 8 SUMMARY.........................................................................................................149 8.1 Conclusion.....................................................................................................149 8.1.1 Productivity Analyses for Pavement Construction Operation........149 8.1.2 Causes-and-Effects of Interference on Productivity.......................150 8.1.3 Method of Productivity Estimation Developed..............................152 8.2 Future Work and Research Limitation...........................................................153 8.2.1 Statistical Analyses.........................................................................153 8.2.2 Integrated System for Modeling Interference.................................154 8.2.3. Development of Project Productivity Factor.................................155 APPENDIX A DATA COLLECTION FORMS..........................................................................156 B FURTHER ANAYSES FOR PROJEC T PRODUCTIVITY COMPARISON...160 C FURTHER ANAYSES FOR IN TERFERENCE COMPARISON.....................161 D THE RESULT OF FURTHER ANAYSES FOR DESIGN 1.............................163 E THE RESULT OF FURTHER ANAYSES FOR DESIGN 2.............................164 F INTERFERENCE QUANTIFICATION.............................................................165 G RESULT OF TPN SIMULATION......................................................................173 H RESULTS OF PRODUCTIVITY ESTIMATION..............................................191 LIST OF REFERENCES.................................................................................................195 BIOGRAPHICAL SKETCH...........................................................................................200

PAGE 10

x LIST OF TABLES Table page 2-1 Classes of variability..................................................................................................8 2-2 Basic modeling elements of CYCLONE.................................................................19 2-3 Factors related to productivity transients.................................................................29 3-1 Case study projects and project category.................................................................38 3-2 Production rates a nd conversion factor....................................................................41 3-3 Example of productivity calculation........................................................................42 4-1 Project statistics........................................................................................................4 7 4-2 Disruptions occurred in project SR-20 (Alachua)....................................................49 4-3 Disruptions occurred in project SR-102...................................................................53 4-4 Disruptions occurred in project SR-20 (Putnam).....................................................56 4-5 Disruptions occurred in project I-10........................................................................60 5-1 Mean and standard deviation for four projects.........................................................68 5-2 Result of F-test for mean comparison......................................................................68 5-3 Mean and standard deviation for interference factors..............................................71 5-4 Result of F-test for comparison of interference factors............................................71 5-5 Mean values with management interference............................................................72 5-6 Result of mean comparison for management interference.......................................72 5-7 Correlation test result between the amounts of rainfall and NLA............................76 5-8 Correlation test result between th e number of rain days and NLA..........................77 5-9 Correlation of test re sult without outliers.................................................................78

PAGE 11

xi 5-10 Design of correlation test between causes and effects.............................................80 5-11 The model with crosse d and nested effects..............................................................82 5-12 Design for 2-factor experiment................................................................................83 5-13 GLM test result for Design 1....................................................................................84 5-14 2-factor experiment result for Design 1...................................................................84 5-15 2-factor experiment result for Design 2...................................................................85 5-16 2-factor experiment result for Design 3...................................................................87 5-17 2-factor experiment result for Design 4...................................................................88 6-1 Deterministic method for work task time duration..................................................93 6-2 Results of survey for time duration..........................................................................93 6-3 Hauling times and number of trucks........................................................................94 6-4 Comparison of simulation results.............................................................................99 6-5 Categories and causes of interference on sub-tasks...............................................102 6-6 Method of quantifying ATPL on work day............................................................103 6-7 Method of quantifying NLA and AT on sub-tasks................................................104 6-8 Loading example for quantifying interference.......................................................105 6-9 Result summary fo r TPN simulation......................................................................109 6-10 Validation of TPN model for interference categories............................................110 6-11 Validation of TPN mode l for interference causes..................................................110 6-12 Probability and NLA of interf erence for loading (Case Study 1)..........................112 6-13 Probability and NLA of interfer ence for spreading (Case Study 1).......................112 6-14 Probability and NLA of interfer ence for compacting (case Study 1)....................112 6-15 Probability and NLA of interf erence for loading (Case Study 2)..........................118 6-16 Probability and NLA of interfer ence for spreading (Case Study 2).......................118 6-17 Probability and NLA of interfer ence for compacting (Case Study 2)....................118

PAGE 12

xii 6-18 Probability and NLA of interf erence for loading (Case Study 3)..........................121 6-19 Probability and NLA of interfer ence for spreading (Case Study 3).......................122 6-20 Probability and NLA of interfer ence for compacting (Case Study 3)....................122 6-21 Probability and NLA of interf erence for loading (Case Study 4)..........................125 6-22 Probability and NLA of interfer ence for spreading (Case Study 4).......................126 6-23 Probability and NLA of interfer ence for compacting (Case Study 4)....................126 7-1 SR-26 General informati on for pavement structure...............................................133 7-2 Production data for SR-26......................................................................................134 7-3 Work areas associated with work content (APUB)................................................134 7-4 Work areas associated with work content (FDAP)................................................135 7-5 Initial time duration for SR-26...............................................................................136 7-6 Production prediction results for the SR-26 projects.............................................141 7-7 Production prediction results for th e SR-26 projects after improvement...............145

PAGE 13

xiii LIST OF FIGURES Figure page 2-1 Simplified method of measured mile.......................................................................11 2-2 Basic structure of control charts...............................................................................12 2-3 Example of NORMAL entity...................................................................................19 2-4 Example of COMBI element....................................................................................20 2-5 General structures of PNs fo r construction process modeling.................................21 2-6 Input and output system...........................................................................................23 2-7 Example of earthwork process model......................................................................24 2-8 General system architecture of SPS tools................................................................25 2-9 Atomic model (l oading) example.............................................................................26 2-10 Atomic model library example.................................................................................26 2-11 Bar example in CYCLONE......................................................................................28 2-12 Bar example in ABC................................................................................................28 2-13 Example of transient effect of productivity..............................................................30 2-14 In-progress inventories between tasks......................................................................30 2-15 Work task for repair-of-equipment..........................................................................31 2-16 The result of simu lation by stopping rules...............................................................32 3-1 Overall process of the research................................................................................35 4-1 Daily productivity plot for SR-20 (Hawthorne).......................................................48 4-2 Daily productivity plot for SR-102..........................................................................52 4-3 Daily productivity plot for SR-20 (Putnam)............................................................57

PAGE 14

xiv 4-4 Daily productivity plot for I-10................................................................................59 5-1 Probability plot of populat ion before transformation...............................................65 5-2 Probability plot of popul ation after transformation..................................................66 5-3 Equal variance test among four projects..................................................................67 5-4 Box plot for mean comparison among projects.......................................................67 5-5 Equal variance test amon g interference factors........................................................70 5-6 Box plot for mean comparison among interference factors.....................................70 5-7 Box plot for the mean comparison with management interference.........................73 5-8 Monthly time series pl ot for rainfall and NLA........................................................75 5-9 Monthly time series plot for the number of rain days and NLA..............................77 6-1 General process of asphalt pavement.......................................................................91 6-2 Example of MicroCYLONE input file.....................................................................92 6-3 Initial simulation resu lt of SR-20 (Alachua)............................................................96 6-4 Initial simulation result of SR-102 (Duval)..............................................................96 6-5 Initial simulation resu lt of SR-20 (Putnam).............................................................97 6-6 Initial simulation result of I-10.................................................................................97 6-7 Productivity comparison for no in terference in SR-20 (Alachua)...........................98 6-8 Productivity comparison for no interference in SR-102 (Duval).............................98 6-9 Productivity comparison for no interference in SR20 (Putnam)..............................98 6-10 Productivity comparison for no interference in I-10................................................99 6-11 Conceptual model of TPN for alternative time duration........................................106 6-12 Example of TPN model for sub-task interference..................................................107 6-13 TPN model for loading (Case Study 1)..................................................................113 6-14 TPN model for spreading (Case Study 1)..............................................................113 6-15 TPN model for compacting (Case Study 1)...........................................................114

PAGE 15

xv 6-16 Productivity simulation result from MicroCYLCONE (Case Study 1).................115 6-17 Productivity simulation result from DISCO (Case Study 1)..................................115 6-18 Productivity analysis for SR-20 (Alachua)............................................................117 6-19 TPN model for loading (Case Study 1)..................................................................118 6-20 TPN model for spreading (Case Study 1)..............................................................119 6-21 TPN model for compacting (Case Study 1)...........................................................119 6-22 Productivity simulation result from MicroCYLCONE (Case Study 2).................120 6-23 Productivity simulation result from DISCO (Case Study 2)..................................120 6-24 Productivity analysis for SR-102...........................................................................121 6-25 TPN model for loading (Case Study 3)..................................................................122 6-26 TPN model for spreading (Case Study 3)..............................................................123 6-27 TPN model for compacting (Case Study 3)...........................................................123 6-28 Productivity simulation result from MicroCYLCONE (Case Study 3).................124 6-29 Productivity simulation result from DISCO (Case Study 3)..................................124 6-30 Productivity analysis for SR-20 (Putnam).............................................................125 6-31 TPN model for loading (Case Study 4)..................................................................126 6-32 TPN model for spreading (Case Study 4)..............................................................127 6-33 TPN model for compacting (Case Study 4)...........................................................127 6-34 Productivity simulation results from MicroCYLCONE (Case Study 4)................128 6-35 Productivity simulation results from DISCO (Case Study 4)................................128 6-36 Productivity analysis for I-10.................................................................................129 6-37 Overall productivity estimation resu lts from the four case studies........................130 7-1 Typical section for the APUB method (SR-26).....................................................133 7-2 Typical section for the FDAP method (SR-26)......................................................133 7-3 Result of MicroCYCLONE simula tion for the SR-26 APUB project...................139

PAGE 16

xvi 7-4 Result of DISCO simulation for the SR-26 APUB project....................................139 7-5 Result of MicroCYCLONE simula tion for the SR-26 FDAP project....................140 7-6 Result of DISCO simulation for the SR-26 FDAP project....................................140 7-7 Frequency histogram of producti vity for the normal distribution..........................143 7-8 Frequency histogram of productiv ity for the lognormal distribution.....................143 7-9 Probability plot of productiv ity for the normal distribution...................................143 7-10 Lognormal distribution for spreading task.............................................................144 7-11 Improved result of productivity by MicroCYCLONE (SR-26, APUB)................144 7-12 Improved result of productiv ity by DISCO (SR-26, APUB).................................144 7-13 Improved result of productivity by MicroCYCLONE (SR-26, FDAP).................145 7-14 Improved result of productiv ity by DISCO (SR-26, FDAP).................................145 7-15 Summary of prediction results...............................................................................148

PAGE 17

xvii Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy QUANTIFYING THE EFFECTS OF INTERFERENCE THROUGH USE OF AN ALTERNATIVE METHOD OF PR ODUCTIVITY ESTIMATION By Jaehyun Choi August 2006 Chair: R. Edward Minchin Cochair: Ralph D. Ellis Major Department: Civil and Coastal Engineering In order to determine the expected co mpletion time of the highway pavement projects, the Department of Transportation (DOT) needs to be able to estimate the durations of individual activities required for the projects. The estimation for the activities on the critical path is even mo re important when calculating total project duration, and the asphalt pavement operation is one of those activities in the highway construction schedule. The objectives of this research were to de velop, evaluate, and verify an alternative method to estimate construction productivity of the asphalt pavement operation. The researcher measured various productivity pa rameters from four pavement research projects. Each project fell into a different category when classified by geographic feature and the pavement structure. The researcher identified various inte rference factors and categorized them by their types such as weat her, management, and work content. The factors were further categorized by their causes and association with each sub-task. The

PAGE 18

xviii likely magnitude of each sub-task was quantif ied by the number of loads affected by each factor. Then the alternative time durations of each sub-task from each project were estimated using the Timed Petri-net (TPN) simulation model developed. Finally, the durations were entered to an existing pa vement process model and productivity of pavement operations was estimated. The re sults showed that the accuracy of the productivity estimation was improved in all four research projects from different category. The estimation method developed was ve rified through two other pavement projects. For the weather and management in terference, the researcher used the same magnitude estimated from one of the research projects that belongs to the same project category. However, for the work content interference, the actual magnitude was calculated because this information is avai lable before production begins. The accuracy of productivity prediction was 27% and 31% respectively from the actual productivity. The researcher also applied st atistical distribution to a sub-task durat ion and daily shift effect, and the accuracy was significantly im proved to 1% and 9% respectively of the actual productivity. By using the alternative method proposed by this research, DOT can estimate construction productivity for th e pavement operation more accurately.

PAGE 19

1 CHAPTER 1 INTRODUCTION 1.1 Research Background Construction operations utilize various re sources and technology and convert them into the physical components that make up a project. In order to estimate the project duration, management allocates the require d resources and schedules the required individual operations before production be gins. Once production begins, management monitors the performance to ensure that it conforms to the original schedule. The duration of each operation is determined based on the quantity and construction productivity of the operation. The construction productivity is usually derived from historical data taken from past projects or publi shed reference materials. However, construction operations involve s ophisticated resource interactions and are performed in the field where different w eather and site conditions may influence production. Therefore, construc tion productivity may vary fr om project to project or from workday to workday. Construction productivity is commonly measured as the ratio between work hours of input and construction goods or services of output. For example, Thomas and Sanvido (2000) measured the labor productivity of concrete formwork and placement operations on a daily basis. They monitored the fluctu ation of daily productiv ity for all workdays and identified factors that negative ly affected the daily productivity. One of the primary advantages of using this measurement is that it provides management with an accurate feedback for the performance during and after the

PAGE 20

2 operation. The daily amounts of input and outpu t are derived along with the factors that affect the production of the operation of inte rest. Then, the produc tivity fluctuation on each workday can be easily verified. Manage ment also can quantify each factor that causes variation in the c onstruction productivity. However, the construction productivity m easured by this method is difficult to apply to some projects due to the uniquene ss of each construction project. Variables— geographical features, weather conditions, size of projects, size of the crew, and number of crews employed for the operation—are likely to be different from one project to another. Thus, the productivity rate also vari es from project to project and from workday to workday even though the same types of operations are monitored. Simulation of construction processes is another proven method of estimating construction productivity. It has been widely used to analyze the uncertain and dynamic nature of the construction operations. According to Halpin and Riggs (1992), a simulation model in general is a representation of a real-w orld situation and provides a framework with which a given situation can be analyzed. With the help of computer technology, simulation has expedited the m odeling, analysis, and optimization of construction operations. A common met hod of estimating productivity by using simulation is to build a model for an ope ration, conduct a simulation experiment by the iteration of random numbers, and estimate an average measure of performance. The primary advantage of using a simulati on model for construction productivity is that it can apply various resour ce utilizations such as work hours of the crew, size of the crew, and number of crews. Besides, most simulation models can assign uncertain nature of work task durations in their system. De pending on the time durations associated with

PAGE 21

3 each work task, the transition of simulation flow is delayed to represent work task duration, and the transition can also be assigne d by stochastic time estimate. Therefore, management can utilize various st atistical distributions to estimate the work task duration. By utilizing the uncertainty of time durations, the construction productivity may be predicted before actual production begins so that management can calculate the duration of the operation by using productivity rates. As a result, the constr uction productivity of an operation can be generally used on other projects by changing the resource utilizations and work task durations accordingly. However, in order to estimate the productivity more accurately using simulation models, users should find the statistical distri bution of each work task to be used through measurements of actual time durations. This task requires huge management efforts, and is not always feasible. Also, due to the nature of computer simulation, most existing simulation tools have difficulty responding to delays and inte rruptions once the simulation begins. Huang and Halpin (1995) discussed that steady-state environment has been assumed in most simulation research even though it is seldom achieved in realworld practice. Therefore, it is recognized that construction produc tivity can be more accurately estimated by integrating the adva ntages of the two productivity estimation methods introduced. 1.2 Research Objective The objectives of this research were to de velop, evaluate, and verify an alternative method to estimate construction productivity of asphalt pavement operation. Existing simulation methods have the ability of appl ying various resource ut ilizations and the uncertainty in work task durations. Howe ver, the simulation results for productivity

PAGE 22

4 estimation become more reliable only when th e effects of various in terference factors in the process of production are ap plied to the simulation. In order to determe project durations for the highway pavement projects, the Department of Transportation (DOT) needs to be able to estima te the durations of individual activities required for the projec ts. The estimation for the activities on the critical path is even more important wh en calculating total pr oject duration and the asphalt pavement operation is one of thos e activities in the highway construction schedule. By using the alternative method proposed by this research, DOT can estimate construction productivity for the pavement operation more accurately, leading to more accurate prediction of project duration. 1.3 Chapter Organization The overall process employed to develop the alternative method is described in Chapter 3, titled Research Methodology, and the details are discussed from Chapter 4 to Chapter 6. Production data are collected from four highway research projects for the purpose of analyzing productivity for the pavement ope rations from different types of project. The production data are used to calculate various productiv ity parameters and identify factors that affect a pavement crewÂ’s efficien cy. The loss of efficiency by each factor is quantified and evaluated by using the productiv ity analysis method presented in Chapter 4. In order to validate and further analyze the results of the pr oductivity study in Chapter 4, the researcher performed statis tical analyses involving mean productivity comparisons and effects of interference factors. The primary purpose of mean comparison is to identify and rank interf erence factors by their adverse effects on

PAGE 23

5 productivity. The researcher tested the eff ects of weather and prerequisite work on the crewsÂ’ performances represented by probability of interference occurrence and productivity variation. The pro cess and results of the analyses are presented in Chapter 5. In Chapter 6, the method developed to estimate the productivity of pavement operations with various interfer ence factors is discussed. The types of interference are further categorized and quantif ied by their causes and by thei r association with each subtask of the operation. Then a simulation tool was developed to captu re the interference quantified and estimate alternative time durati ons of the sub-tasks. The results of the productivity estimation were evaluated thr ough four research proj ects investigated, presented in Chapter 6. In Chapter 7, the ability of the deve loped methodology was verified by applying the method to the new projects, in order to predict the productiv ity and construction duration of their pavement operations. For the verification purpose, the uncertainty in the task durations and daily shif t effect were also applied to the simulation model. 1.4 Research Scope The scope of this research mainly includes the following: To collect production data for the paveme nt operations from various types of highway research projects. To analyze the productivity of the op erations based on the production data collected. To identify and categorize th e interference factors caused by various disruptions in the process of the operations. To identify and rank interf erence factors based on thei r effects on the productivity rates of the projects investigated. To quantify the effects of interference f actors on each sub-task for different types of projects.

PAGE 24

6 To develop a simulation model to estimat e alternative time duration for each subtask using the quantifie d effects on the tasks. To simulate construction process of pave ment operation by applying the effects of interference factors and estimate productivity more accurately. To evaluate the results by comparing th e results of productivity estimation with other productivity parameters. To verify the developed method to new pr ojects to predict their productivity and construction durations. The methodology for accomplishing the scope of this research is described in Chapter 3 in more detail.

PAGE 25

7 CHAPTER 2 LITERATURE REVIEW DISCUSSION This chapter discusses theories and issues in the areas of construction productivity and process simulation models. Overall, c onstruction productivity issues involve the methodology to evaluate construction cr ews’ performance by measuring their productivity with valid methods. The proce ss simulation models serve as a tool to validate the results of productivity measurements under various circumstances. 2.1 Production and Productivity Variability High uncertainty and variability are part of the unique nature of the construction industry. This is especially true in terms of controlling the production and productivity of operations. Ballard and Howe ll (1997) reported that cons truction planning is the production of budgets, schedules, and specifica tions to be followed. Once production begins, project management “controls” performa nce to ensure that it conforms to the budget, schedule, and specifications. Ho wever, in manufacturing, management “controls” the physical production pr ocess at a more detailed leve l. This process consists of material and information flow (Horma n and Kenley 1998; Ballard and Howell 1998). In other words, the manufacturing industr y has applied the c oncept of “production variability control” in orde r to improve the efficiency of the production line. In construction, the processes of design and construction are interrelated and performed by multiple contract ors. Construction site c onditions are more unpredictable than those of a factory production line, so th e concept of uncertainty and variability in production is defined more conservatively in manufacturing than in construction (Hopp

PAGE 26

8 and Spearman 2000). Table 2-1 presents the class of variability defined in manufacturing. The coefficient of variation (c) is defined as the standard deviation (s) divided by the mean ( ) of the sample. According to the definition of variability class in Table 2-1, variability in construction produc tion is likely to fall into the range of “moderate” or “high” because the produc tion process of construction is more sophisticated. Table 2-1. Classes of variability Variability Class Coeffi cient of Variation (c) Low (LV) c < 0.75 Moderate (MV) 0.75 c < 1.33 High (HV) c 1.33 2.1.1 Work Flow Variability vs. Productivity Variability Howell and Ballard (1994) defined workflow variability as the quantity difference between “planned to be completed” and “actually completed” in a given time period (usually a week). Thomas (2002), however, measured the variability of work flow productivity (the ratio of input to output) on a daily basis and argued that the variability of workflow productivity is also an importa nt indicator of performance because of the conceptual similarities between “work flow variability” and “work flow productivity variability” Gulezian and Samelian (2003) distinguish ed the source of va riation in daily productivity by natural variati on and variation due to an assignable cause. Natural variation is either a variati on that cannot be explained or that is caused by chance or common cause. For example, a crew can perf orm at the different productivity level even though they perform a repetitive work and no ex ternal interference fa ctors affected their work. Assignable causes of variation are ones that are not attributable to chance, and

PAGE 27

9 they usually are subject to the control of the contractor(s), the owner, or inclement weather conditions that affect the performance of the contractor. Damrianant and Wakefield ( 2000) stated that “interfere nce in construction process is considered composing of two components, delay and interruption. Delay is the state that slows the rate of an activity execution, a nd lengthens the norma l activity duration while interruption is the temporary or pe rmanent stopping of execution”. Even though both delay and interruption affect cons truction productivity, they were modeled differently due to their different effects on th e execution of activity. On the other hand, the term “Disruption” typically refers to a loss of productivity caused by a change in a contractor’s working conditions, reso urces, or methods (Finke 1998). In this research, “interference” was used as a general term to refe r to the factors that affected productivity of pavement operation studied. 2.1.2 Productivity Measurement Alfeld (1988) defined performance as a ratio of “Accomplishment” to “Method”. Accomplishment represents finished work of value while method is the cost spent to accomplish the value. Performance is improved by reducing the cost of the methods required to accomplish a given work. He also recognized the difference between performance and productivity by stating that “productivity m easures only one performance dimension” (Alfeld 1988, pp.19). Produc tivity is defined as a ratio of output to input with output corresponding to accomplishment, and input to methods in performance; however, output usually measur es a single work accomplished relative to a single input (e.g. labor man-hours) . It is concluded that good productivity can lead to high performance, but it is not the only c ontribution to job site performance.

PAGE 28

10 Drewin (1982) provided the general proce dure of productivity measurement for the purpose of productivity improvement. The pr ocedure involved five steps in general: activity selection, activity r ecording, data analysis, new method development, and new method installation. Among them, activity reco rding requires to use such methods as activity sampling and field rating to collect data. 2.1.2.1 Activity sampling Activity sampling is performed for the pur pose of determining the proportion of working time for a crew of interest. It re quires the obser vation of all crew members employed for the activity and rating their wo rk time by predetermined category of their working status. The simplest rating category is “working” and “not working”. By randomly sampling the crew member’s working status several times per day (e.g. 4 times per day), management can calculate the effi ciency of them. Balance chart has been commonly used to record the work status. An activity of interest is observed and recorded in the form of a balance chart with its sub-tasks a nd their corresponding time duration to analyze the process. Once the new process of th e activity has been established, the new balance chart is compared to the existing one in order to measure the improvement. It has been shown that produc tivity improves by utilizing this method at the activity level and making the process more efficient (Oglesby et al. 1989). 2.1.2.2 Field rating A general process of field rating is to obser ve a crew for a short period of time (e.g. several minutes) and to note the ratio of ine fficient hours to total time observed. If the inefficient hours exceed 50 percent of the tim e, the time period is noted “inefficient”. Otherwise, the time period is noted “worki ng”. Management can calculate the field

PAGE 29

11 rating of the crew by multiplying the sum of workers to the times noted as “working” divided by the total observation time (Alfeld 1988). 2.1.3 Methods of Estimating Productivity Loss 2.1.3.1 Measured mile Measured mile has been prevalently used to estimate the loss of productivity in construction claims (Schwarzkopf 1995; Jones 2001). The objective of using measured mile method is to identify unimpacted peri od by disruption for a construction activity by using the linear relation between cumulative input (hours) and output (percent complete). The productivity of unimpacted period is th en projected to entir e project including impacted period to estimate the amount of productivity loss by the hours (Zink 1986). This method is more reliable for the activitie s that have distinct unimpacted and impacted periods in their process. Figure 2-1 shows the simplified method of measured mile (Schwarzkopf 1995; Jones 2001). Figure 2-1. Simplified method of measured mile 2.1.3.2 Control charts Control chats were originally used to control the quality of a manufacturing production by applying a statistical method. Figure 2-2 shows the basic structure of

PAGE 30

12 control charts when the chart is used to an alyze the variation of daily productivity rates (Schwarzkopf 1995; Jones 2001). Figure 2-2. Basic structure of control charts The basic structure of control chart has th ree horizontal lines, namely center line (CL), upper control limit (UCL), and lower control limit (LCL), as seen in Figure 2-2. CL corresponds to the mean value of th e daily productivity rates. UCL and LCL typically correspond to three times of the sta ndard deviation of the set from the CL to both upper and lower directions. The producti vity rates of workdays 42, 44, and 45 are outside of the UCL, and they are truncated in the first step. This process is repeated until all productivity rates stay with in the range of UCL and LCL, and when it occurs, the CL is determined as the baseline productivity. 2.1.4 Labor-intensive vs. Equipment-intensive The primary difference between labor-inten sive and equipment-intensive activities is that in equipment-intensive activities, part icular pieces of equipment play a major role in the production that brings revenue to the contractor. In contra st, labor provides these major contributions to production in labor-inten sive activities. For example, an asphalt paver and rollers are the ma jor contributors to the produc tion of roadway pavement

PAGE 31

13 operations with the help of laborers, while la borers fabricate reinforcing steel with the help of fabricating equipment or tools. Most published research has applied productivity measur ement techniques to laborintensive operations (Thomas et al. 1999; Thomas and Sanvido 2000; Thomas 2000). Thomas et al. (2003a) quantified ineffective control of workflow, which led to a labor inefficiency of 51% in their case studies and demonstrated that effective workflow management improves labor productivity. Highway construction projects are classified as linear and repetitive construction operations due to their geometric layout. Cr ews repeat the same task in all sections (Vorster et al. 1992; Hassanei n and Moselhi 2004). However, the daily productivity at the crew level tends to have some degree of variability even though the crew size and their work hours are consistent (Thomas et al. 2003b; Choi and Minchin, In press). Vorster and De La Garza (1990) emphasized that maintaining crew work continuity is important for highway construction because the equipment used is typically heavy and expensive. Pavement operations are one of the critical activ ities in the highway construction schedule (Lee 2000, Hassanein and Moselhi 2004). 2.2 Definitions The following terminologies and concepts, related to highway paving and productivity, are used. 2.2.1 Unit Completed Method (UCM) The principle methods for measuring quantit y installed for a part icular activity are unit completed, percent complete, level of effort, incremental milestones, and start/finish percentages (Thomas and Zavrski 1999). Among those, UCM determines output by simply measuring the units completed. This method can be used where the activity scope

PAGE 32

14 is well defined and where output can be dete rmined easily. Its advantage is that it measures quantity the most accurately and objectively by not assigning any different level of credit for the sub-tasks of the activit y. It is, then, best for measuring asphalt pavement activity because the completed quant ity of the pavement activity is easily measured by truck loads, and although thr ee major sub-tasks are involved (hauling, placing, compacting), no sub-task is le ft unfinished at the end of the day. 2.2.2 Baseline Productivity (BP) Baseline productivity is theoretically the best productivity that a crew can achieve for the activity of interest. Gulezian a nd Samelian (2003) stated three conditions to determine baseline productivity as listed. 1. Individual productivity valu e rather than cumulative output should be used to establish a baseline. 2. Baseline productivity should not necessarily be dependent on separately identified consecutive periods of unimp acted and impacted production. 3. Variability in productivity should be considered. Thomas and Zavrski (1999) suggested that in order to measure baseline productivity, the baseline subset—the 10 percen t of work days (minimum of five) with the largest output—be verified and that the pr oductivity of each of these days be noted. The baseline productivity is defined as the median value of the subset. 2.2.3 Cumulative Productivity (CP) Cumulative productivity is a compilation of the total quantity installed divided by total work hours spent at any given time duri ng the paving, as shown in Equation [2-1]. The reason for using cumulative productivity is to monitor work progress as a whole and to predict the final pr oductivity rate upon completion of the activity (Thomas and Zavrski 1999). The values of cumulative productivity are different from those of the mean

PAGE 33

15 productivity because the mean value averages all daily productivity rates as shown in Equation [2-2]. [2-1] Cumulative Productivity = n i i n i iCH DQ1 1 [2-2] Mean Productivity = nn Day n Day 2 Day 2 Day 1 Day 1 Day CH DQ CH DQ CH DQ Where, CH = crew hours, DQ= daily quantity, and n = total work days. 2.2.4 Project Management Index (PMI) PMI represents the amount of disruption caused by management factors. It is measured by the difference between cumulativ e productivity and baseline productivity divided by baseline productivity as shown in Equation [2-3]. Because cumulative productivity contains disrup tions stemming from both mana gement and design factors and baseline productivity is only associated with the design effects, the PMI represents the effects of management. A higher value means a higher number of disruptions caused by management (Thomas and Zavrski 1999). [2-3] PMI = BP CP BP 2.2.5 Conversion Factors (CF) Conversion factors (CF) can be utilized when the crew worked on various work items on a single work day, and those items requ ire different level of effort. The level of effort is calculated by the unit rate, which translates to the work hours required to finish a unit of the item in question. The conversion fa ctor is the unit rate for the item in question

PAGE 34

16 divided by that for the standard item as shown in Equation [2-4] (Thomas and Zavrski 1999). [2-4] Conversion Factor = item standard for the rate Unit question in item for the rate Unit 2.3 Computer Simulation According to Halpin and Riggs (1992), a m odel in general is a representation of a real-world situation and usua lly provides a framework with which a given situation can be analyzed. In construction, mathematical models and discrete models are widely used for modeling a given problem. Mathemati cal models usually address problems of planning and scheduling. Network models fo r scheduling and cash flow models for cost control are examples of mathematical mode ls. Discrete models are utilized when mathematical modeling is not possible. Discre te models view a model as a set of events and transitions (Hajjar and AbouRizk 2000). En tities in discrete m odels represent the active elements of the model and they trav el throughout the event network and trigger transformations. Construction operation pr ocesses are usually modeled by discrete models. 2.3.1 General Aspects of Computer Simulation Shi (1999) stated that co mputer simulation is th e process of designing a mathematical-logic model of a real syst em and experimenting with the model on a computer. Computer simulation has been ut ilized in construction for the past three decades because it expedites the modeling, analysis, and optimization of a construction operation with the help of computer technology. Construction modeling requires represen ting a process using modeling elements, and a simulation enables portrayal of the ch anging state of construction process over

PAGE 35

17 time. Since Halpin (1977) developed CYCLONE (CYCLic Operation Network), a general-purpose system for simulating constr uction operations, numerous features were built into the various implementations of CYCLONE in order to improve simulation capability. Among the general-purpose si mulation languages are CYCLONE (Halpin 1977), RESQUE (Chang 1986), COOPS (Liu 1991), CIPROS (Odeh 1992), STEPS (McCahill and Bernold 1993), DISCO (Hua ng and Halpin 1995), and STROBOSCOPE (Martinez 1996). Sawhney et al. (1998) di scussed the enhancement of CYCLONE to allow the development of individual models for all processes that constitute a project and link them to simulate at the project level. The latest research endeavor s to simplify methods of simulation for users. Shi and AbouRizk (1997) developed resource-base d modeling (RBM), in which operating processes of active resources are defined as atomic models. Shi (1999) developed the activity-based construction (ABC) simulation with one single “activity” element. Hajjar and AbouRizk (2000) stated that “effective tr ansfer of computer simulation knowledge to the construction industry is best done thr ough specialization and customization of the modeling, known as special purpose simulation (SPS).” Lu (2003) developed SDESA (Simplified Discrete-Event Simulation Approach) to extract the construction features from existing events and, by using SDESA he extracted productivity rates of construction activities. Huang and Halpin (1995) stated that “construction operations usually involve complex resource interactions and are pe rformed in the field and influenced by uncontrollable exogenous factors such as in clement weather conditions and equipment breakdowns”. Therefore, the steady-state en vironment of simulation systems, which is

PAGE 36

18 usually assumed in the models, is not realistic. The transien t-effect of productivity needs to be simulated (Tavakoli 1983, Bernold 1989, Huang and Halpin 1995). 2.3.2 Process Simulation Models 2.3.2.1 General purpose models Martinez and Ioannou (1999) defined ge neral-purpose simulation languages as tools that target a very broad domain and can be used to model almost any type of operation. On the other hand, special-purpose simulations (SPS) are tools that target a narrow domain (e. g. pipe installation). Th ey also discussed that modeling paradigm (strategy) and flexibility char acterize models developed. Th e main paradigms used today for modeling construction process are process in teraction (PI) and activity scanning (AS). Most of the manufacturing ope rations are modeled by PI strategies because they are written from the point of view of entities that flow through a system. Typically, those entities “arrive, undergo some processing where they capture and release scarce resources, then exit.” (Martinez and Ioannou 1999) By contrast, AS models focused on identifying various activities and the conditi ons under which they take place; therefore, AS models are adept at modeling such system s that various activities are interdependent according to highly dynamic rules. Martinez and Ioannou (1999) concluded that in construction, AS is a more natural and effective strategy than PI. 2.3.2.2 CYCLONE CYCLONE (Halpin 1977) is th e most widely used general-purpose simulation tool in the construction domain. It uses AS where activity cycling diagram (ACD) is the simulation model. It has been used to model many construction operations including concrete batch plants, tunneling, and asphalt pa ving. In order to represent a work state

PAGE 37

19 and entity flow, CYCLONE uses a graphical modeling format as basic elements. Table 2-2 shows the basic modeling elements. Table 2-2. Basic modeling elements of CYCLONE Modeling Element Name of Element Description NORMAL The normal work task modeling element, which is unconstrained in its starting logic and indicates active processing of resource entities. COMBI The constrained work task modeling element, which is l ogically constrained in its starting logic, otherwise similar to the normal work task modeling element. Q NODE The idle state of a resource entity symbolically representing a queuing up or waiting for use of passive state of resources. ARROW The resource entity directional flow modeling element. COUNTER The number of times a simulation unit passes the point is counted so that the system can be stopped at a certain time and production can be measured. As explained in Table 2-2, the arrival of a resource entity in a NORMAL network commences the work task because there is no queue built up before a NORMAL work task. As shown in Figure 2-3, the work task can be processed when either resource entity A or B arrives. In contrast, all required resources should arrive before a COMBI work task is initialized. In Figure 2-4, the “loading truck with dirt using a loader” work task can start only when the three resources (truck , dirt, and loader) have arrived. Work Task Lable Resource entity A Resource entity B Resource entity exit Figure 2-3. Example of NORMAL entity

PAGE 38

20 Load truck with dirt using loader Truck Loader Dirt Loaded truck available Front end loader available Figure 2-4. Example of COMBI element The idle or passive state of a resource is modeled by a QUE node. The QUE node acts as a storage location for resources and releases the resources to the following COMBI node whenever the logic of the COMBI node is satisfied. The ARROW modeling element has no time parameters affec ting resource entity flow, since it serves purely as a mean of indicating resource di rectional flow logic among various nodes of NORMAL, COMBI, and QUEUE (Halpin and Riggs 1992). The COUNTER element is used to control the number of simulation cycles before the simulation stops. A modeler can assign th e number of cycles to the COUNTER, and once the number of units has reached the elements, the simulation experiment is terminated. COUNTER is also used to measure production. For example, if a multiplier, which goes through COUNTER, is a truck with the capacity of 15 yd3, the multiplier increases by 15 yd3 each time a truck entity passes the COUNTER element. 2.3.2.3 Petri-Nets Since C. A. Petri introduced his networks in 1962, Petri-Nets (PNs) have advanced in both theory and application as a general purpose simulation tool. PNs, as a general purpose modeling tool, have been widely applie d in the areas of manufacturing, computer software and hardware, and queuing systems to model their system. Wakefield and Sears (1997, pp. 107) stated that PNs function supe rior compared queuing theory or discrete event simulation when simulate the systems with concurrency, where “several state

PAGE 39

21 changes happen simultaneously and where even t-driven characteristics are present”. The basic mathematical definition of PNs is shown with a four t uple (Peterson 1981). PN = < P, T, IN, OUT > Where, P = { p1, p2, p3,…., pn} for the set of n places T = { t1, t2, t3,…., tm} for the set of m transitions IN = {P X T} N is an input function that defines directed arcs from places to transitions OUT = {T X P} N is an output function of directed arcs from transitions to places Figure 2-5 illustrates the ge neral structures of PNs th at are commonly used to model construction process (Damrianant and Wakefield 2000; Wakefield and Sears 1997, pp. 107). Figure 2-5. General structures of PNs for construction process modeling “Sequential execution (a)” models pr ecedence dependencies in construction activities; for example, the transition t2 can fires the token only after t1 fires the token entered in p1. “Conflict (b)” models res ource sharing among multiple activities. For

PAGE 40

22 example, if a crane is shared by three activities modeled t1, t2, and t3 in the figure, the token travels either one of three transitions based on the priority or probability assigned to each transition. Even though all transiti ons are enabled, firing of token to any transition disables the other two transitions. “Concurrency (c)” is used to model multiple activities performed in parall el. “Synchronization (d)” mode ls an activity that requires multiple types of resources. Transition t1 can fire the token only when the token arrives in previous places. “Merging (e)” is used to model several materials that are needed for an activity; for example, the constituent materials of hot mixed asphalt arrive at the asphalt plant for a mixing activity. “Confusion (f)” models when conflict and concurrency occur simultaneously. The ge neral structures of PNs are thoroughly explained from the literatures such as Peterson (1981) and Resig (1985). Among the typical structures introduced, the “Conflict” function is particularly important in this research because the researcher assigned pr obability with the relevant transitions to model various types of interference and estima te alternative time duration. The detailed method of using these functions is described in Chapter 6. Timed Petri-Nets (TPN) is another adva nced modeling technique, and it is very useful to model construction activity, associated with time duration. When all input conditions are met, the firing of the token w ill be delayed during the assigned time before it is deposited to the output place. 2.3.2.4 Process model for construction process The construction process can be modeled as an input and output system as shown in Figure 2-6 (Blanchard and Fabrychy 1981). I nput has seven categories: material, tools, equipment, labor, management, time, and c ondition. Bernold (1989) stated that the efficient use of the input factor s is related to the effectivene ss of a model and that the user

PAGE 41

23 of the model should be able to input proba bilistic elements to reflect real-world situations. Bernold (1989) also stated that the production rates of complex systems cannot simply be decided because the relatio nship between input and output variables is usually unknown; therefore, simulation models are very useful to handle the probabilistic nature of the real world. Figure 2-6. Input and output system The objective of process mode ls is to provide the user s with a decision process relating to the operation. Ha lpin and Riggs (1992) mentioned that the process models must focus on the decision variables pertinent to the construction operation itself and that the decision variables should be availabl e to the users for manipulation in the management of the construction operation. Sawhney and AbouRizk (1996) found that co nstruction processes are modeled at either a process level or a project level. In the process level method, the CYCLONE model can be linked to each pr ocess. By using external li nkage of the CPM network, the model has a capability to represent process interdependencies. The limitation of this method is that resource allocation is done indi vidually for each activity and the modeling of the process independency is limite d by CPM. Sawhney and AbouRizk (1996) proposed the project-level simulation by perf orming the following four steps: “(1)

PAGE 42

24 identify all the processes that are to be m odeled; (2) identify and define the resources required for the processes identified in step (1); (3) develop process models using the CYCLONE method; (4) perform simultaneous simulation by using the resource pool. The project-level simulation enhanced the capability of the process-level model by establishing process interdependencies and comm on resources at the project level. Figure 2-7 shows the example of an earthwork process model with multiple successors. In Figure 2-7, the availability of resources su ch as backhoes and trucks are checked and allocated (AR) to the activity (COMBI). Th en, they are freed by th e element of FR. The earthwork process is a predecessor of both the “piling process” of pier-1 and the “excavation process” in pier-2. Figure 2-7. Example of earthwork process model 2.4 Simplified Modeling Efforts 2.4.1 Framework Computer simulation requires a relatively large investment of time and money, and this hinders effective application of simula tion; therefore, simulation researchers have attempted different ways to simplify the m odeling process. Hajjar and AbouRizk (2000) emphasized that the development of a framew ork that includes the reuse of common code and common design, can help in overcoming these obstacles. Once the domain of the

PAGE 43

25 application is determined, the common features (frozen spots) and va riables (hot spots) are identified. “Frozen spot s are framework components that the developers have no control over while hot spots ar e areas that they can custom ize for specific requirement.” (Hajjar and AbouRizk 2000) Figur e 2-8 illustrates the general system architecture of the framework. Figure 2-8. General system architecture of SPS tools In Figure 2-8, “modeling elements” are th e building blocks of the model, which distinguish a given model from another. A “simulation engine” is a processor and it continues its simulation until all events have been processed or until simulation time reaches a maximum specified by the user. A “random number generator” and an “external database” supply i nput values. Random number generators produce common transformations such as be ta; exponentially; normally; tr iangularly; and uniformlydistributed values. “Statistical collection and analysis” involves the collection of observed values during simulation execution. The benefit of applying the framework approach is reducing the development effo rt of construction simulation tools.

PAGE 44

26 2.4.2 Resource-based Modeling (RBM) RBM, developed by Shi and AbouRizk (1997), uses resources as the basic building blocks in order to build construction simu lation models. They mentioned that “even though each construction process is unique, th e operation processes of its component resources are somewhat generic”. The major f eature of RBM is that it provides flexible atomic models as compared to the fixed mode ls that have to use a fixed library. An atomic model is defined “a basic and unique description of a particular process”. A model library consists of a collection of the atomic models (Shi and AbouRizk 1997). Figure 2-9. Atomic model (loading) example In RBM, a separate model library should be designed for each type of construction process with corresponding resources. Figure 2-9 shows an example of the atomic model for a loading operation and Figure 2-10 shows an example of an atomic model library that contains the loading atomic model. An atomic model can interact with other atomic models through communication ports (e.g . LKI: input and LKO: output). Figure 2-10. Atomic model library example

PAGE 45

27 2.4.3 Activity-based Modeling (ABM) Because a construction process can be de fined a collection of activities and an activity consumes both time and resources, activity and resources constitute the two fundamental components of a construction ope ration (Shi 1999). Shi (1999) found that various modeling elements in existing simulation systems are all activity-related. Also, activity can be started when such conditions as logical dependencies and resource requirements are met. Therefore, an activity is in its “active” stat e when it meets all conditions (NORMAL or COMBI in CYCLONE); otherwise, it is in an idle state (QUEUE in CYCLONE). Figure 2-11 shows the example of the “Bar” model written in CYCLONE and Figure 2-12 shows the simplified “Bar” model wr itten in ABC. The queue elements in CYCLONE are included in relevant activities as attributes, which means that resources can wait at activities instead of queue nodes. Shi (1999) divided the simulation entity into a resource entity and a processing entity. The resource entity can be at an activity or resource pool. A resource at an activity is only available to that activity; however, a resource at a resource pool is available to all ac tivities. Shi (1999) c oncluded that “it is practical to use one single element (i.e. act ivity) to model a constr uction process without sacrificing functionality”.

PAGE 46

28 Figure 2-11. Bar example in CYCLONE Figure 2-12. Bar example in ABC 2.4.4 Simplified Discrete Event Simulation Approach (SDESA) Lu (2003) discussed the importance of discre te-event simulation. The event in the context of a discrete-e vent is defined as an instance of time at which a significant state change occurs in the system (Pidd, 1988). SDESA focuses on “extracting the constructive features from the existing even t/activity based a generic simulation method; therefore, the algorithm and th e model structure of simulatio n are streamlined such that simulating construction systems is made as easy as applying the critical path method (CPM).” (Lu 2003)

PAGE 47

29 2.5 Transient-Effect (Non-steady) Model Bernold (1989) discussed that computer simulation primarily concerns steady-state performance, while construction processes ar e usually interrupted by disruption events (e.g. equipment breakdown, delays, and weathe r conditions.) He al so categorized the factors related to productivity tran sients as shown in Table 2-3. A productivity curve should reflect the transien t effect as illustrated in Figure 2-13. In Figure 2-13, the productivity is defined as the rate of production based on cumulative units produced for the total time elapsed since the start of the process (Huang and Halpin 1995). Table 2-3. Factors related to productivity transients Category Type Examples Unbalanced system Resource allo cation Not enough trucks to keep loader busy Resource unavailability Material Tools Equipment Labor Management Insufficient inventories Left in shop Breakdown Coffee break Missing plans Changing conditions Environmental Site Work place Experience Weather Soil conditions Move to another floor Learning curve The operation usually requires “warm-up” before it reaches steady-state as illustrated between t0 and t1. Usually, mobiliz ation or start-up causes this effect. The other circumstance for the productivity transien ts is the operation delay in between time interval t2 and t3. Huang and Halpin (1995) found that operation delays included inclement weather conditions, an equipmen t breakdown, and a management delay; therefore, operation delays can occur rega rdless of whether the operation reaches steadystate or not.

PAGE 48

30 Figure 2-13. Example of transi ent effect of productivity In order to capture the function of productiv ity transients in the simulation, Bernold (1989) developed a systematic framework based on the enhancement of the CYCLONE model. This framework handles two causes of disruption events: In -progress inventories between tasks and equipment breakdowns. As shown in Figure 2-14, the second task cannot start until th e queue holds inprogress inventory. If the maximum capacity of the queue is limited, this storage limitation combined with an interruption of a chained work task will force the operation to stop. This is called the “repercussion eff ect”. One good example of this chain of work is a pallet of bricks on a scaffold. A laborer supplies a mason with br icks on a pallet on a scaffold. The scaffold is designed to hol d only a certain amount of pallets before it collapses due to a structural failure. Figure 2-14. In-progress i nventories between tasks For simulating an equipment breakdown, the simulator randomly chooses a breakdown within the given time frame by putting one unit into a work task called “repair of equipment”, as shown in Figure 2-15.

PAGE 49

31 Figure 2-15. Work task for repair-of-equipment Huang (1995) pointed out the impact of shif t effects at the end of each working day by modeling the typical ending states of cons truction operations on a daily basis. He defined “shift effects” as the effect that a certain work schedule ha s on what work can be started. He developed a simulation mechanis m that is capable of stopping simulation at the end of each working day and resuming it the next day. This mechanism consists of four rules: 10-minute-mark, 0-minute-mark, back trace, and start-from-scratch. In the 10-minute-mark rule, all the work ta sks that are still in progress and can be finished in 10 minutes after the end of the da y can be carried to completion. Otherwise, these tasks are not started. The 0-minute-ma rk rule is more restrictive than the 10minute-mark rule since tasks cannot be starte d unless they can be finished by the end of each working day. The back trace rule impos es even more restrictive stopping criteria than 0-minute-mark. It traces back through the sequence of work tasks and controls the feasibility of the start of the work task. Fo r example, if a batch of concrete cannot be finished before the end of working day, it s hould not be spread, it should not be poured, and it should not be mixed or transported. The start-from-scratch rule would stop a work

PAGE 50

32 item in mid-task. This rule assumes that work tasks that are left unf inished at the end of work day are resumed the next day, with resources never having been delivered. Figure 2-16 shows the result of simulati ng an asphalt-paving operation using each stoppage rule. Even though the 0-minute-mark rule is more restrictive than the 10minute-mark rule and the back trace rule is mo re restrictive than the start-from-scratch rule, the result shows that the 0-minute-mark ru le and the start-from-scratch rule are more optimistic than the 10-minute-mark rule and the back trace rule. Huang (1995) explained that the model formation combined with ha ving multiple flow entities in the operation causes this result. Figure 2-16. The result of simulation by stopping rules Huang and Halpin (1995) developed a program called DISCO (Dynamic Interface for Simulation of Construction Operation) based on the CYCLONE schematic model. DISCO allows breaks or interruptions by st opping the simulation clock so that the transient states of resources can be determin ed. Thus, with the cap ability of interrupting the simulation run, DISCO captures the pr oductivity transient effects on time and productivity calculation by using an y of the four stopping rules.

PAGE 51

33 2.6 Summary of Literature Discussion It is recognized that the simulation model should have a capability of reflecting a real-life situation because the lack of reality limits the use of the simulation model. The construction process has a dynamic nature of in teraction with resource s and techniques in order to convert them to physical com ponents (Halpin and Riggs 1992). Also, the processes are interrelated in the activity nexus . Huang and Halpin ( 1995) described that the steady-state performance, which is comm only assumed in the construction simulation, is rarely achieved in real life construction practice. However, most simulation models were developed based on limited production data from construction projects. Thus, the productivity variability caused by various disrup tion events is neglected in the models. In the next section, the theories related to production and productiv ity issues are discussed. Various simulation models and their features are introduced in this chapter. As explained, the primary objective of simulation is to model real-life s ituations and analyze operation processes. In order to accomplis h this objective, the simulation model should respond to the transient-effects of productivity because the steady-state situation that is assumed in most models is impractical. Hu ang and Halpin (1995) analyzed the transienteffects of resources and work tasks in a process by interrupting the simulation run and stopping the simulation time clock. Howe ver, their simulation did not apply the transient-effects resulting from delays such as equipment breakdown and inclement weather conditions. It is rec ognized that those delays are di fficult to quantify due to the limited production data from the construction projects. Thus, the quantitative analyses for the management delays are major parts of th is research and it is described in Chapters 4 and 5.

PAGE 52

34 CHAPTER 3 RESEARCH METHODOLOGY In this chapter, the researcher descri bed the overall processes of the research methodology. The researcher selected highw ay pavement construction operations in order to analyze an equipment intensive ope ration, develop an alternative method to estimate productivity based on the results of analyses, and apply the method. 3.1 Overview of Experimental Methodology Figure 3-1 shows the overall process of th e experiment for this research. The researcher selected four hi ghway construction projects a nd collected production data from their pavement operations. This da ta collection proce ss included numerous interviews, weekly project visits, and project document s acquisitions. The general construction process of the pavement operations was also surveyed before and during the data collection process. From the production data collected, daily productivity rates were calculated by taking the ratio between output (quantity installe d) and input (efforts spent) factors. The two factors should be pre-defined in any pr oductivity study because they can vary by the purpose of the study. Collecting reasonably a ccurate data for input and output factors provides the framework for measuring the produc tivity of an activity of interest, so the units of the factors should be readily measurable. From the daily productivity calculated, productivity parameters such as baselin e, mean, and cumulative are calculated.

PAGE 53

35 Chapter4:Data Collection Process Chapter4:ProductivityAnalysis Chapter5:StatisticalAnalysis Chapter6:DevelopmentofProductivtyEstimationMethod Chapter7:VerificationofProductivityEstimationMethod ConstructionProcess ThreesubtasksCrewcompositionWorkcharacteristic Projects FourcasestudyprojectsDatacollectionprotocolProjectinputandoutput ConversionFactor MeasureoutputbyunitrateVariouspavementcourses DailyProductivity EfficiencyofperformanceIdentifying&Quantifydisruption events CorrelationAnalysis1 RainfallvsweatherinferferenceOutlierstruncated InterferenceCategorization CategorizationbythreegroupsinefficienthourscalculationPMI MeanComparion DatatransformationAmongprojectproductivityAmonginterferenceAmongmanagementcauses ProductivityParameters BaselineMeanCumulative CorrelationAnalysis2 WeathervsProb.OfManagementinferferenceWeathervsProductivityvariationPrerequisitevs.Prob.OfmanagementinterferencePrerequisitevs.Productivityvariation ProcessSimulation1 DiscreteeventsimulationInitialtimedurationsurveyOptimisticproductivty AlternativeTimeDuration InterferenceCategorizationAssociationwithsubtasksTimedPetri-Netsimulation ProcessSimulation2 AlternativetimedurationCaptureinteferencesMorerealisticresults Figure 3-1. Overall process of the research Daily productivity is also useful in mon itoring the efficiency of the construction performance by identifying disruption events that adversely affect the activity and quantifying the adverse effect s more accurately (Thomas a nd Zavrski 1999; Choi and Minchin, In press). When co llecting production data, the factor s that contribute to costly, inefficient work hours were identified and grouped in order to quantify the loss of productivity caused by each factor. The resear cher categorized thos e interference factors into three major groups: management, work c ontent, and weather interference. Then the causes of each interference fact or were described in more detail. The results of the productivity analyses are presented w ith four case studies in Chapter 4. From the result of productivity analyses, statistical analyses were performed in order to validate and further analyze the result s. The analyses consisted of two parts:

PAGE 54

36 mean productivity comparisons and correlati ons among interference factors. For the purpose of mean comparisons, productivity va lues obtained from f our case study projects (population) were grouped by the project from which the values were collected and by interference (or no interf erence) with which the daily produc tivity values were associated. The result of mean productivity comparison by their association with interference shows which factor is most detrimental to pr oductivity. The mean of daily productivity associated with poor management was also co mpared to find which factor contributed to reduce productivity the most within manageme nt interference. Correlation tests were then performed between causes of effects. The correlation tests include the effects of: rainfall in various project locations on the probability of weather interference occurrence, weather interference on the probability of management interference occurrence, weather interference on the productivity variation, defects of prerequisite work on the probability of management interference occurrence, and defects of prerequisite work on the productivity variation. The results of statistical analys es are presented in Chapter 5. Finally, the researcher discussed the met hod developed to estimate the productivity of the pavement operation with various combina tions of interference. The results of the productivity estimation were validated through four case study projec ts investigated, as presented in Chapter 6. Finally, the pr oductivity estimation method was verified by applying the method to different projects. The productivity and project durations were predicted. The research verifi cation was shown in Chapter 7.

PAGE 55

37 3.2 Construction Process for Asphalt Paving The asphalt pavement operation for HMA base and pavement courses involves three sub-tasks: delivering HMA material, la ying the material down, and compacting it. Once the subgrade is ready, material deliver y vehicles (usually dump trucks) deliver HMA material from a designated asphalt plant to the site. No material storage on the site is required because the materials are dumped di rectly into the asphalt paver. When the vehicle approaches to paver, it should not m ove back into the pave r because it can push the paver back and cause ridges or bumps in the mat. Thus, the paver should move forward to until its roller bars make firm contact with the truck’s rear wheel (FDOT 2002d). Jiang (2003) discussed that management needs to balance the utilization of the different types of equipment used to ach ieve maximum production. If the delivery capacity of the dump trucks exceeds that of the asphalt paver, then the dump trucks queue, and wait until the paver is re ady. On the other hand, if th e material delivery rate is slower than the paving rate, then the asphalt paver is sitting idle. This is not costeffective for the contractor because the cost of the asphalt paver is more than the cost of the dump trucks. The asphalt paver places the material on top of the subgrade for the base construction or on the base layer for structural course construction. At least two skilled operators are needed for operating one paver, with one controlling the paver itself and the other controlling a screed. Sometimes, a moto r grader substitutes for the asphalt paver when placing the first lift of the HMA base when a “leveling” course is required. Then “break-down” rollers and a finish-roll er follow the paver, compacting the lift of material placed. Rolling should begin as soon as the mat will car ry the roller with displacing the mix. Breakdown rolling is us ually done with a steel wheel roller. The

PAGE 56

38 required number of rollers and passes vari es depending upon the compacting method and density requirement. When a roller gets onto newly laid mix adjacent to an existing pavement, it should be driven onto the old pavement. The final rolling is usually done with a tandem static roller. If a vibratory roll er is used, the vibrations are turned off, and the roller makes a static pass to insure a smooth surface (FDOT 2002d). The paving operation is sequential in th at the rollers shoul d compact all HMA material placed and, at the end of the wor kday, no work should be left unfinished. The paving crew consists of a foreman, equipment operators (skilled or semi-skilled), laborers (common or trainee), and sometimes a superintendent. 3.3 Data Collection for Productivity Measurement Over 15 months were spent to gather data on pavement construction production from four highway construction projects in Florida: SR-20 (Palat ka, Putnam County), SR-20 (Hawthorne, Alachua County), I-10 (Pensacola, Escambia County), and SR-102 (Jacksonville, Duval County), as shown in Table 3-1. Table 3-1. Case study proj ects and project category Pavement Structure (HMA Usage) Project Category Pavement Only (APUB) Base and Pavement (FDAP) Rural SR-20 (Alachua) SR-20 (Putnam) Geographic Feature Urban SR-102 (Duval) I-10 (Escambia) As highway construction projec ts are classified, all four projects studied fall into the “added-capacity” category of projects b ecause the project scopes include adding new lanes (Daniels et al. 1999). The projects were further categorized by geog raphical feature and the pavement structure for HMA usage. Each of the projects studied falls into a different geographic/material category, as s hown in Table 3-1. The SR-20 (Alachua) and SR-102 projects used aggregat e base (limerock base); th erefore, only the pavement

PAGE 57

39 course, consisting of a structural course and a friction course was monitored. The SR-20 (Putnam) and I-10 projects had a full-depth pavement structure, and HMA material was placed in the base course as well as a paveme nt course, but each course had a different HMA mix design (FDOT 2004). More detailed pr oject statistics are sh own in Table 4-1. In order to expedite data gathering, project inspectors were given a production measurement form, as seen Appendix A. Th e form contained space for information more relevant to the research than the FDOT Daily Diaries or Asphalt Reports routinely filled out by project inspectors and was completed da ily with the proper information by project personnel. When completed, this form contains such information as date of operation, lift constructed, daily quantity, work crew a nd hours, equipment number and hours, and details of any incident that might affect productivity. Daily diar ies and asphalt reports also supplemented the information from the production measurement form, when appropriate. The sample forms of daily diarie s and asphalt reports we re also presented in Appendix A. The detailed incidents (inter ference) were categorized into three major groups of factors (management, work content, and weathe r), as described in Chapter 2. The data gathered all pertained only to the work item of interest and consisted of production of truck loads (output) and crew hours (input), and informati on regarding any event or occurrence that could have affected productivity. 3.3.1 Productivity Output and Conversion Factor The output of the production was readily measured by the actual number of truck loads per day. Paving contractors are supposed to supply the inspector with a ticket for each truck load, which determines how much the contractor delivered, placed, and compacted on the site each working day. Th e number of truck loads was also used as a

PAGE 58

40 flow unit in the process simulation model, whic h will be described in Chapter 6. Halpin and Riggs (1992) stated that identifying the fl ow unit is the first step of the modeling a process because it determines the degree of modeling detail. In simulation models, depending on the type of process and the appr oach of the modeling effort, the flow units can be equipment, labor, or material. In any case, the units ar e discrete quantities relevant to the productivity measurement of the process. The daily quantity of truck loads was c onverted by using conversion factor (CF) since the paving crew worked various pavement courses during a single work day, that is, base, structural, and friction courses for SR20 (Putnam) and I-10 proj ects and structural and friction courses for SR-20 (Alachua) and SR-102 projects. The definition of CF is described in Section 2.2.5 of Chapter 2. As an indication of productivity, RS M eans Manual (2004) pr ovides the production rates of various pavement courses when a st andard crew worked 8 hours per day, shown in Table 3-2. The quantity measured by th e surface area completed on each work day decreases as the thickness of the course incr eases. When the quantity is measured by the volume, however, the thicker the course, th e larger the quantity of daily output. The volume is converted to the weight (tons) by multiplying the average density of the asphalt material by the volume. Then, the weight of the daily output is converted to the truck loads by dividing it by 17 MT/load. The number of loads is then revers ed to calculate the unit rate (hr/load). The unit rate indicates the level of effort as the thickness of the course changes. A 100-mm-thick base course wa s determined as a standard item, and conversion factors for other work items were calculated by the ratio, shown in Equation [2-4].

PAGE 59

41 Table 3-2. Production rates and conversion factor Course Thickness (mm) Area (m2) Volume (m3) Tonnage (MT) Load hr/load CF 100 3800 380.00 885.40 52 0.15 1.00 150 3094 464.10 1081.35 64 0.13 0.82 200 2508 501.60 1168.73 69 0.12 0.76 Base 250 2128 532.00 1239.56 73 0.11 0.71 40 6459 258.36 601.98 35 0.23 1.47 50 5305 265.25 618.03 36 0.22 1.43 75 4101 307.58 716.65 42 0.19 1.24 100 3461 346.10 806.41 47 0.17 1.10 Structural 130 2850 370.50 863.27 51 0.16 1.03 25 8842 221.05 515.05 30 0.26 1.72 40 6459 258.36 601.98 35 0.23 1.47 50 5305 265.25 618.03 36 0.22 1.43 65 4582 297.83 693.94 41 0.20 1.28 Friction 75 4097 307.28 715.95 42 0.19 1.24 3.3.2 Productivity Input As an input value, only crew hours that the pavement crew spent on site were counted. Thomas and Zavrski (1999) menti oned that as a general rule, a crew only performs a single category of work, and crew hours are not divided in to more detail than the category of work; therefore, the daily crew hours that the paving crew spent is the relevant number. As mentioned, the pa vement operation involve s three sub-tasks, namely delivering asphalt material to the site, spreading the material, and compacting it. Each sub-task is performed mainly by specific equipment. There was only one crew for the operation most of time, and their hours included equipment operation hours. Even though the crew composition (crew size) slightly varies through working days within a project, the variance is usually only one worker and does not have a significant effect on the production rate, unlike la bor-intensive activities.

PAGE 60

42 3.4 Productivity Parameters The researcher focused on validating the effect of work flow management in equipment-intensive tasks by conducting case studies. Each case study project was analyzed in order to measure productivity pa rameters such as daily, mean, cumulative, and baseline productivity and PM I, as defined in Chapter 2. Table 3-3 shows an example of the produc tivity calculation method. Column (C) identifies the pavement layer installed; for example, the friction course requires only one lift of pavement layer, whereas the structural course requires two lifts. By using 1.72 and 1.03 as conversion factors for the friction and structural courses, the actual number of truck loads finished by the end of each work da y (D) is converted and recorded in column (E). By dividing the number of convert ed loads by crew hour s spent (G), daily productivity (F) is calculated. Upon the completion of data collection, baseline productivity is calculated by using the met hod described in Section 2.2.2. Mean and cumulative productivity rates are also calcu lated by using Equations [2-1] and [2-2]. Table 3-3. Example of productivity calculation Work day (A) Date (B) Lift (C) No. of Loads (D) C. Loads (E) Daily Productivity (F) CH (G) CH by baseline (H) Wasted (I) Category (J) Cause (K) 1 5/22/03 1/1 12.00 13.39 1.67 8 2.40 5.60 Work content turning lane 3 8/14/03 1/2 27.00 27.81 3.97 7 4.99 2.01 Weather Rain Each work day contains information a bout whether any part icular type of interference occurred, whether it affected the performance of the pavement crew, and how long it delayed the execution of work. Ca tegory (column J) and more detailed cause (column K) identify the types of interferen ce that occurred. By projecting baseline productivity on those days when any interfer ence occurred, crew hours that would have spent were calculated and recorded in column (H). Wasted hours (column J) is the crew

PAGE 61

43 hour difference between actual crew hours spent (column (G) and crew hours by baseline (column H). It is important to note that the researcher categori zed interference by three main typesweather, management, and work content. 3.4.1 Weather Interference Any unfavorable weather conditions, such as rainfall and very low temperatures, that affect the performance of the pavement crew is monitored. HMA cannot simply be laid in the rain, and FDOT (FDOT 2004) specifies the temperature requirement for various thicknesses of layer placed as s hown below. El-Rayes and Moselhi (2001) asserted that heavy rainfall is a significant factor in the delay of highway construction because it leads to saturated and unworkable soil conditions. Temperature: Spread the mixture only when the air temperature in the shade and away from artificial h eat is at least 40ºF [4ºC] for layers greater than 1 inch (100 lb/yd2) [25 mm (55 kg/m2)] in thickness and at least 45 ºF [7ºC] for layers 1 inch (100 lb/yd2) [25 mm (55 kg/m2)] or less in thickness (this includes leveling courses). The minimum temperature requirement for le veling courses with a spread rate of 50 lb/yd2 [25 kg/m2] or less is 50ºF [10ºC]. 3.4.2 Management Interference Management interference can be broken down into problems with prerequisite work (including re-work caused by defects of prerequisite work), out-of-sequence work, work conflict, work area, material shorta ge, and equipment breakdown. “Prerequisite work” means that some deficiency from prere quisite work has a negative effect on the performance of the pavement crew. “Out-ofsequence work” means that the crew has to wait or slow down because the prerequisite wo rk has slowed down or stopped. If the crew spends time correcting previous work (re-work), they get no credit for the quantity involved in the rework regardless of th e causes (Thomas and Zavrski 1999). “Work

PAGE 62

44 area” means that the crew could not obtain e nough work space to efficiently perform their work or they had to make an unplanned move to another location. 3.4.3 Work Content Interference Work content is a rating of the complexi ty of the work to be performed; for example, turning lanes, intersections, ra mp, and the area around manhole and curbs tend to require more work hours than the main ro adway. It is necessary to do handwork in those areas listed because the paver cannot re ach. For example, a paver will place mix as close to a manhole as possible. The area ar ound the manhole must then be filled in by hand. Fresh mix is shoveled into the area and carefully placed. The rollers then compact the mix as close as they can to the manhole. The area not compacted by the rollers must be hand tamped. Handwork is also necessary on crossovers and tur nouts. The paver will cover as much of the area as possible, but the curved shape of crossovers and turnouts makes it almost impossible for the paver to cove r the total area of each. As a result, the crew sometimes must stop and resume the work multiple times in those areas if they have a small amount of work to perform in each area. The described areas have the entirely di fferent work processes from those of mainline areas. If the crew hours spent onl y on those areas were measured on each work day, the daily production output could be measured by assigning higher conversion factors to those work areas than mainline ar eas, based on their unit rates as explained in Section 2.2.5 titled Conversion Factors. Howe ver, this type of information can only be measured through field observation. Theref ore, the paving items on those areas were defined as a type of interference, as sugge sted by other productiv ity research (Thomas and Zavrski 1999).

PAGE 63

45 3.5 Methods for Statistical Analyses and Productivity Estimation The methods employed for statistical anal yses and the develo pment of productivity estimation are described in Chapter 5 and Chap ter 6 in greater detail. The two chapters stand alone in terms of their structures. 3.6 Summary of Methodology The research methodology discussed in this chapter includes th e development of a methodology to estimate construction produc tivity for the highway construction operations. The method development requires an extensive data co llection process from the highway research projects. The resear cher conducted an extensive data collection process with the help of construction proj ect personnel, designers, and contractors gathering data from various sources includ ing interviews, project visits, and project documents. The input and output data collected from the projects are used to estimate how many crew hours can be spent inefficiently. The researcher identified and quantified disruption events that cause th e inefficient work hours. The loss of work hours then are applied to the TPN simulation model developed to calculated alternative time durations. The durations were entered to existing simulation models to estimate construction productivity. The model estimates the c onstruction productivity for the pavement operation. Then, the researcher discussed about the application of the model. The application provides a framework to the highway practitioners to estimate the productivity rates of asphalt pavement operation.

PAGE 64

46 CHAPTER 4 PROCESS PRODUCTIVITY ANALYSES 4.1 Introduction The objective productivity analyses presente d in this chapter is to measure the productivity rate of highway pavement operations , to identify factors th at adversely affect performance, and to quantify the loss of pr oductivity caused by each factor. Traditionally, productivity for equipment-intens ive work has been defined by charts and tables provided by manufacturers in equipment handbooks with no regard for factors other than those acknowledged by the handbook. Even t hough highway pavement operations are considered highly equipment-intensive a nd repetitive (Harris and Ioannou 1998), the productivity of the operations contains signifi cant variability, some of which cannot be accounted for simply by using the methods pr ovided by the equipment manufacturers. The measurement of performance helps to iden tify the causes for uncertainty in the work flow of equipment-intensive tasks. Again, the primary difference between the two types of tasks is that in the equipment-intensive tasks, particular pieces of equipment play the largest role in the production w ith laborers providing support. In labor-intensive tasks, the reverse is true. The research hypothesis of productivity analys es is that controlling uncertainty and variability is a significant factor in the effective management of production for equipment-intensive operations. In order to prove the hypothesis, the researcher studied highway pavement operations, which are repe titive throughout the work process. Pavement operations are a critical part of the highway construction schedule, and sub-

PAGE 65

47 tasks are related by finish-to-start depende ncy (Hassanein and Moselhi 2004). This analyses found variability in the equipm ent-intensive area of asphalt pavement construction and identified the primary causes of the variability, whic h should lead to an improvement of the reliability of work fl ow in highway pavement operations. This chapter also focuses on valida ting the effect of work flow management in equipmentintensive tasks by conducting case studies for highway pavement operations. 4.2 Productivity Analyses for Pavement Operation Table 4-1 shows the statistics for each projec t. The data covers a total of 253 work days, 2,398 crew hours, and 6509.22 trucks load s for all projects. As seen, all four projects had a large quantity of asphalt pavement. Total paved areas for the four projects varied from 176,157.55 m2 to 448,732.88 m2, based on the number of work days collected. Total work days and cr ew hours spent were also varied fr om project to project. Table 4-1. Project statistics Project Data SR-20 (Alachua) SR-102 (Duval) SR-20 (Putnam) I-10 (Escambia) Date from 5/22/2003 5/13/2003 9/12/2003 3/30/2003 Date to 4/06/2004 3/25/2004 5/11/2004 4/09/2004 Total Work Days 42 50 64 97 Total Crew Hours (CH) 389.00 445.00 646.00 918.00 Total C. Truck Loads (C.loads) 851.85 1365.15 1409.28 2882.94 Total Area Paved (m2) 176,157.55 245,430.63 242,351.56 448,732.88 Mean Productivity (C.loads/hr) 2.166 2.995 2.165 3.120 Standard Deviation 1.533 1.849 1.280 1.723 Cumulative Productivity 2.190 3.068 2.182 3.140 Baseline Productivity 5.574 6.440 5.05 5.89 PMI 0.6071 0.5236 0.5679 0.4669 Inefficient CH 177.37 186.55 205.32 288.51 Inefficient CH (%) 45.6 41.9 31.8 31.4 In addition to cumulative productivity, the mean productivity and standard deviation of the daily productivity rates were calculated for each proj ect. As described,

PAGE 66

48 the mean productivity values are different from those of cumulative productivity. Note that the number of truck loads is conver ted to the equivale nt numbers by using conversion factor when the crew worked various courses. 4.3 Project 1, SR-20 (Alachua) 4.3.1 Project Description This project involved converti ng an existing two-lane road into a four-lane divided highway. During the entire project, the pavi ng crew consisted of one charge hand (either superintendent or foreman), sometimes two, two skilled operators (one paver operator and one that controlled the screed elevation) , up to four semi-skilled operators (roller operators), and up to four laborers. 0.00 1.00 2.00 3.00 4.00 5.00 6.00 1357911131517192123252729313335373941 Work dayProductivity (C.load/hr) Figure 4-1. Daily productivity plot for SR-20 (Hawthorne) Since this project used limerock material for its base course , production data for installing only structural and friction courses were collected from May 22, 2003, to April 6, 2004, for 42 work days, as shown in Table 4-1. A structural course averaging 80 mm

PAGE 67

49 (3.2 inches) thick and a friction course aver aging 20 mm (0.8 inches) thick were placed on top of the base course. Figure 4-1 shows the daily productivity ra te of the project. The baseline productivity is marked by the dotted line, and the workdays on baseline subsets are marked as enlarged points. The figure implie s that there was some variability in the daily productivity during the entire pr oject with low productivity rates on days 15 and 20. Table 4-2. Disruptions occurred in project SR-20 (Alachua) Work Day Date lift C. Loads C.Load/ hr CH CH by baseline Wasted Category Cause 1 5/22/03 1/1 13.39 1.67 8 2.40 5.60 Work content turning lane 3 8/14/03 1/2 27.81 3.97 7 4.99 2.01 Weather Rain :stopped 6 9/19/03 1/2 23.69 2.79 8.5 4.25 4.25 Work content turning lane 7 10/20/03 1/1 20.60 3.43 6 3.70 2.30 Management lack of finish grading LR 9 10/22/03 2/2 38.11 3.46 11 6.84 4.16 Ma nagement Prerequisite work Management Equipment breakdown 11 10/24/03 1/1 14.42 1.44 10 2.59 7.41 Work content turning lane 12 10/31/03 2/2 13.39 1.34 10 2.40 7.60 Work Content Shoulders in various area 13 11/3/03 1/1 18.54 1.85 10 3.33 6.67 Work Content Turn lane Work Congestion 14 11/5/03 1/1 9.27 1.03 9 2.31 6.69 Management Prerequisite work 15 11/6/03 1/1 1.03 0.34 3 0.18 2.82 Work Content Ramp 17 12/11/03 1/1 6.18 0.62 10 1.54 8.46 Ma nagement Prerequisite work Management Prerequisite work 18 12/12/03 1/2 5.15 0.52 10 1.28 8.72 Work Content Ramp 20 12/17/03 2/2 2.06 0.26 8 0.37 7.63 Work Content Ramp 23 12/30/03 1/2 13.39 1.34 10 2.40 7.60 Work Content Turning lane 24 12/31/03 1/2 8.24 1.18 7 1.48 5.52 Weather Rain 27 1/20/04 1/2 7.21 0.72 10 1.29 8.71 Work Content Turning lane Management Plant breakdown 28 1/21/04 1/2 9.27 0.93 10 1.66 8.34 Weather lower 40 31 1/26/04 2/2 25.75 2.58 10 6.41 3. 59 Work Content Crossover 34 1/29/04 1/1 20.64 1.88 11 3.71 7.29 Work Content Turning lane 35 1/30/04 1/2 19.57 1.96 10 3.51 6.49 Work Content Turning lane 36 2/18/04 1/2 22.66 2.27 10 4.07 5.93 Management Rework 37 3/19/04 1/2 5.15 0.52 10 0.92 9.08 Work Content Turning lane 39 3/23/04 1/2 4.12 0.82 5 0.74 4.26 Work Content Turning lane 40 3/30/04 1/2 21.63 2.16 10 3.88 6.12 Ma nagement Prerequisite work

PAGE 68

50 Table 4-2 shows the workdays when various disruptions interrupted the pavement crew during operations. The di sruption factors and their cause s are also indicated. The inefficient crew hours are the difference betw een the actual crew hours the paving crew spent and the crew hours the cr ew would have spent if they worked at the baseline productivity. If the baseline productivity were projected to those workdays listed in Table 4-1, then the crew hour would have been 177.37 less than the total crew hour spent (389 hours). Thus, the inefficient crew hours by di sruptions were 45.6 pe rcent of total crew hours. 4.3.2 Work Flow Management Poor work flow management resulted in 64.82 inefficient crew hours. This value was calculated by adding inefficient crew hour s caused by management factors as shown in Table 3. This translates into 36.5 percen t of the total ineffici ent crew hours (177.37 hours) on the project. The PMI of this pr oject is 0.6071, which is the highest value among the four projects studied. This indicates that the proj ect has a relatively high level of inefficiency caused by management factors. A notable observation from Days 7, 9, 14, 15, 17, and 18 is that when the paving crew placed the first lift of structural course, the crew had to place overage material due to the finishing grade for the limerock base not being at the proper elevation. Having to adjust the lift thickness duri ng the operation resulted in a loss of productivity. On Day 14, the crew had a work conflict with the base crew when the base crew placed the subgrade and ba se material adjacent to the paving area. By not having enough work space, the crew worked at inefficient productivity. On Day 20, the paving crew faced a variable depth of embankment and installed HMA base in lieu of subgrade

PAGE 69

51 where no limerock could be used in order to level the embankment. This unplanned activity resulted in low productivity for that da y. On Days 27 and 37, the crew worked at various locations; thus, they lost productivity as they moved their equipment from one location to another. 4.3.3 Work Content Work content caused 96.69 inefficient cr ew hours when adding inefficient crew hours caused by work content factors as shown in Table 4-2. This translates into 54.5 percent of the total inefficient crew hours on the project. On Days 23, 34, and 39, the crew worked on turning lanes, and on Day 35, they worked on a ramp area. Because of the complexity and limited amount of work to perform, relatively low productivity was measured on those days. 4.3.4 Weather Unfavorable weather conditions caused 15. 86 inefficient crew hours when adding inefficient crew hours caused by weather factors as shown in Table 4-2. This translates into 8.9 percent of the total inefficient crew hours on the proj ect. Rain on Day 3 and cold weather on Days 24 and 28 disrupted the performance of the crew. 4.4 Project 2, SR-102 (Duval) 4.4.1 Project Description This project mainly involved constructi ng a new four-lane access road to the Jacksonville Airport. The to tal length of the roadway was 2683.62 meters (10,734 lane meters). The scope of work by the pavi ng contractor also in cluded resurfacing the existing road for 356 meters (1,424 lane meters ). Limerock material was installed as a base course. HMA material was then pl aced with a 130-mm-thick (5 1/8 inches) structural course and 19-mm-thick (3/4 of an inch) friction course. Typical paving crew

PAGE 70

52 composition for this project was one charge hand (either superinte ndent or foreman), sometimes two, two skilled operators (one paver operator and one that controlled the screed elevation), two semi-skilled operators (roller operators), and five laborers. Because this project used a limerock materi al for its base course, production data for installing only structural a nd friction courses were coll ected from May 13, 2003, to March 25, 2004, for 50 work days, as shown in Table 4-3. Projec t statistics are also listed in Table 4-1. Table 4-3 showed the workdays when disruptions occurred during operations. Figure 4-2 shows the daily productivity rate of the project. It implies that there were some variations and non productive work days in the workflow, especially from Day 30 to 39. The total crew hours would have been 186.55 less th an the total crew hours spent (445 hours) if the crew had perf ormed the paving operation at the baseline productivity rate on those days when disruptions occurred. In all, the inefficient crew hours amounted to 41.9 percen t of total crew hours. 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 161116212631364146Work dayC.Load/CH Figure 4-2. Daily produc tivity plot for SR-102

PAGE 71

53 Table 4-3. Disruptions occurred in project SR-102 Work day Date lift C. Loads C.Load /hr CH CH by baseline Wasted Category Cause 1 5/13/03 1/3 15.45 3.09 5 2.28 2.72 Management Prerequisite work 2 5/14/03 1/3 52.53 5.25 10 7.74 2.26 Management Prerequisite work 2 5/14/03 1/3 52.53 5.25 10 7.74 2.26 Management Material rejected 2 5/14/03 1/3 52.53 5.25 10 7.74 2.26 Work content turning lane 4 5/20/03 3/3 28.84 2.88 10 4.25 5.75 Work content turning lane 5 5/21/03 2/2 19.57 2.45 8 2.88 5.12 Work content turning lane 6 6/6/03 NA 2.06 0.52 4 0.30 3.70 Management Prerequisite work 7 6/9/03 NA 5.15 0.52 10 0.76 9.24 Management Prerequisite (Rework) 8 6/12/03 1/3 29.87 3.73 8 4.40 3.60 Weather Rain Management Problem at plant Management Material failure 14 9/20/03 1/3 14.42 2.40 6 2.12 3.88 Work content turning lane 15 11/12/03 1/3 12.36 1.55 8 1.82 6.18 Work content turning lane 16 11/13/03 2/3 17.51 2.19 8 2.58 5.42 Management Problem at plant 18 11/15/03 1/3 45.32 4.53 10 6.68 3.32 Work content tapered area & turn lane 19 11/22/03 2/3 49.44 4.94 10 7.28 2.72 Management Material delay 20 11/25/03 2/3 28.84 2.88 10 4.25 5.75 Work content turning lane 21 11/26/03 1/3 56.65 5.67 10 8.35 1.65 Work content crossover 22 12/2/03 1/3 15.45 2.21 7 2.28 4.72 Work content turning lane Weather temperature 23 12/3/03 3/3 20.60 4.12 5 3.03 1.97 Weather rain 24 12/4/03 3/3 23.69 3.95 6 3.49 2.51 Weather rain 27 12/10/03 1/2 3.09 1.03 3 0.46 2.54 Weather rain 0.00 Work content turning lane 30 1/16/04 1/3 25.75 3.22 8 3.79 4.21 Work content turning lane 32 1/21/04 1/3 14.42 1.44 10 2.12 7.88 Management Rework 33 1/22/04 3/3 46.35 4.21 11 6.83 4.17 Management Rework 34 1/23/04 2/2 35.02 3.50 10 5.16 4.84 Management Rework 35 1/26/04 1/1 2.06 0.34 6 0.30 5.70 Management Equipment breakdown 36 1/27/04 1/1 15.45 1.55 10 2.28 7.72 Management Rework 39 1/30/04 1/1 1.03 0.10 10 0.15 9.85 Management Rework 40 2/3/04 1/1 6.18 0.77 8 0.91 7.09 Management Previous work 42 3/2/04 2/3 7.21 0.72 10 1.06 8.94 Management Previous work 43 3/3/04 1/1 10.30 1.03 10 1.52 8.48 Management Previous work 44 3/4/04 2/2 8.24 0.82 10 1.21 8.79 Management Previous work 46 3/9/04 3/3 7.21 0.72 10 1.06 8.94 Management Rework 48 3/19/04 1/1 41.28 3.75 11 3.64 7.36 Management Problem at plant

PAGE 72

54 4.4.2 Work Flow Management A combination of poor work flow manage ment and mechanical problems resulted in 125.82 inefficient crew hours on this project, which translates into 67.4 percent of the total inefficient crew hours on the project. Th ere were also some de ficient reactions to the mechanical problems. On Days 1 and 2, “pumping” from the limerock base (prerequisite work) occurred, and the crew had to stabilize it prior to paving. On Day 6, due to the wet condition of the limerock ma terial, the scheduled paving operation was suspended. On Day 7, the constructed widt h of the roadway was less than the planned width; therefore, additional ma terial was required in order to maintain proper width. On Days 14, 16, and 48, a portion of asphalt mate rial delivered could not be placed because it was out of the allowable temperature rang e, and the contractor stopped paving at his discretion due to problems at the plant. FDOT requires for the project inspectors to periodically check the temperature of the mix to be sure that it is within 6.7C(20F) of the approved job mix formula. For example, if a load of mix is 300 F at the plant, the temperature should not be less than 280 F at the job site (FDOT 2002). On Day 33, the crew had to mill the deficient pavement section. On Days 34, 36, and 46, because previously placed HMA mate rial did not meet the density requirement, the crew had to perform re-work. Obvious ly, this rework severely impacted the productivity rate. Also, the asphalt paver ha d a mechanical problem on Day 35, and the crew had to stop working until it is fixed. Th e paver is required to be checked and the screed properly adjusted and crowned befo re spreading the hot mix (FDOT 2002). On Days 40, 42, 43, and 44, the crew milled th e existing asphalt struct ure on the resurfacing area before paving. The PMI of this project is 0.5236. This indicates that this project had

PAGE 73

55 more disruptions caused by management fact ors than the I-10 project that used HMA base. 4.4.3 Work Content Work content caused 45.33 inefficient cr ew hours, which translates into 24.3 percent of the total inefficient crew hours on th e project. On days 2, 4, 5, 15, 22, 27 and 30, the crew worked on turning lanes, mis cellaneous asphalt placem ent, and intersection areas that require more crew hours per m2 of production than the mainline area. 4.4.4 Weather Severe weather conditions resulted in 15.34 inefficient crew hours, which translates into 8.2 percent of the total inefficient crew hours on the proj ect. The cause of disruption by weather conditions was rainfall on days 23, 24, and 27. 4.5 SR-20 (Palatka, Putnam County) 4.5.1 Project Description This project involved adding two lanes to an existing two-lane roadway. The paving crew of this project consisted of one charge hand (superintendent or foreman), two skilled operators (one pa ver operator and one that cont rolled the screed elevation), three semi-skilled operators (roller operators), and up to five laborers. As shown in Table 2, pertinent data were collected from September 12, 2003, to May 11, 2004, for 64 working days. Base Option Group 1 (100 mm, 4 inches) and Base Option Group 15 (170 mm, 9 inches) were required for the shoulder ar ea and mainline base course, respectively. Each lift of the base course was either 50 mm thick (2 inches) or 75 mm thick (3 inches) because FDOT limits the maximum thickness pe r lift to 75 mm (3 inches) (FDOT 2002). On top of the base course, 75 mm (3 inches) of structural course and 37 mm (1.5 inches) of friction course were installed. A total of 646 crew hours were spent to install HMA on

PAGE 74

56 an area of 242,351.56 m2. Other project statistics are shown in Table 4-1. Table 4-4 shows the workdays when disruptions occurred during operations. Table 4-4. Disruptions occurred in project SR-20 (Putnam) Work day Date lift C. Loads C. Load/hr CH CH by baseline Wasted Category Cause 3 9/15/03 1/3 34.04 3.09 11 6.75 4.25 Management Rework 5 10/2/03 1/3 6.66 0.56 12 1.32 10.68 Management Prerequisite work 5 10/2/03 1/3 6.66 0.56 12 1.81 10.19 Work content Turning lane 6 10/3/03 1/3 18.50 1.95 9.5 3.67 5.83 Management Prerequisite work 8 10/16/03 1/3 26.64 2.54 10.5 5.28 5.22 Management Material delivery 9 10/27/03 1/3 19.98 2.22 9 3.96 5.04 Weather Rain 10 10/28/03 1/3 8.14 0.90 9 1.61 7.39 Management Prerequisite work 11 10/29/03 1/3 11.10 1.23 9 2.20 6.80 Weather Rain 12 10/30/03 1/3 12.58 1.40 9 2.49 6.51 Work content Turning lane 15 12/2/03 1/3 8.88 0.89 10 1.76 8.24 Management Material delivery 16 12/18/03 1/3 23.68 2.63 9 4.69 4.31 Management Material delivery 17 12/19/03 1/3 17.76 1.97 9 3.52 5.48 Management Out of Sequence 20 1/7/04 1/3 30.34 3.37 9 6.01 2.99 Weather Temperature 22 1/16/04 1/3 8.88 0.99 9 1.76 7.24 Work content Turning lane 24 1/20/04 1/3 8.14 0.74 11 1.61 9.39 Weather Temperature 25 1/21/04 1/3 19.24 1.75 11 3.81 7.19 Management Work Conflict 26 1/22/04 1/3 19.98 1.82 11 3.96 7.04 Work content Turning lane 28 1/26/04 1/3 14.80 1.41 10.5 2.93 7.57 Management Equipment breakdown 30 1/29/00 1/3 1.48 0.11 13 0.29 12.71 Management Work Conflict 34 2/13/04 1/3 17.76 1.78 10 3.52 6.48 Work content Turning lane 35 2/19/04 1/3 7.40 0.82 9 1.47 7.53 Work content Intersection 37 2/26/04 1/3 11.84 1.18 10 2.35 7.65 Work content Turning lane 45 4/3/04 1/3 8.88 0.89 10 1.76 8.24 Management Plant breakdown 47 4/6/04 1/3 11.10 1.01 11 2.20 8.80 Management Prerequisite work 52 4/14/04 4/4 16.28 1.36 12 3.23 8.77 Work content Driveway 54 4/17/04 2/4 8.14 0.90 9 1.61 7.39 Management Out of Sequence 55 4/19/04 3/4 17.76 1.48 12 3.52 8.48 Management Plant breakdown 62 5/5/04 1/3 15.54 1.41 11 3.08 7.92 Management Prerequisite work Figure 4-3 shows the daily productivity rate of the project and implies that there were some variations in the workflow with the lowest productivit y on Day 30. If the baseline productivity were proj ected to those days when disruptions occurred, then the crew hours would have been 205.32 less than the total crew hour s spent (646 hours). Thus, the inefficient crew hours were 31.8 percen t of the total crew hours on the project.

PAGE 75

57 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 14710131619222528313437404346495255586164 Work daysC. load/CH Figure 4-3. Daily productivity plot for SR-20 (Putnam) 4.5.2 Work Flow Management Poor workflow management resulted in a total of 119.69 inefficient crew hours, which translates into 58.3 per cent of the total inefficient cr ew hours on the project. On Day 3, HMA material placed in the mainline was thicker than planned, and excessive material was placed outside of the planned widt h. The contractor had to re-do the work in order to correct it. FDOT specifies th e plan thickness (compacted thickness), and the contractor is responsible fo r obtaining the compacted th ickness shown in the plans. On Days 5, 6, and 47, the subgrade (prere quisite work) was not ready, and thus it had to be either stabilized by more compacti ng or at least checked. On Days 10, the paving crew cleaned up the surface of a previ ous lift where another HMA layer would be placed. This work was not profit-generating work and was attributed to poor management. On Days 25 and 30, the paving co ntractor had a conflict with the inspector of the project. The contr actor was advised that he should use a plate tamp around manholes where the roller was inaccessible, but failed to do so. This resulted in a workday with lower quality and productivity than was necessary. The factors that

PAGE 76

58 contribute to diminished crew performance are interrelated, and this type of poor management can also be attributed to work content (work complexity). On Day 52, the contractor had to remove and replace driveway s before paving (out of sequence work). The PMI for this project is 0.5679, lower than the other two projects of SR-102 and I-10, located on urban areas. 4.5.3 Work Content Work content caused a total of 61.42 ineffi cient crew hours, which translates into 29.9 percent of the total inefficient hours on the project. On Days 12, 22, and 26, the contractor worked on turning lanes and in tersections, not ideal production work. On Days 34, 35, and 37, the paving crew had to lay a base wedge down, and the complexity of the work led to low produc tivity. It is not known whet her or not poor management contributed to this situation. 4.5.4 Weather Poor weather conditions caused a total of 24.21 inefficient crew hours, which translates into 11.8 percent of the total in efficient hours on the proj ect. Typically, the contractor chose not to work if rain was expected, but on Da ys 9 and 11, he decided to proceed and the rain diminished the quality of work. On Days 20 and 24, with the temperature below 4C (40F ), the contractor could not lay the HMA base, as specified in the FDOT Flexible Pavement Design manual (FDOT 2002). 4.6 I-10 (Pensacola, Escambia County) 4.6.1 Project Description This project involved adding la nes to and resurfacing an ur ban section of Interstates 10 and 110. Base Option Group 11 (180 mm, 7 inches) and Base Option Group 4 (100 mm, 4 inches) were used for adding lanes on the mainline and shoulder base,

PAGE 77

59 respectively; then, the pavement structure (str uctural and friction courses) was installed. For resurfacing, the top 50 mm (2 inches) of existing pavement structure was milled and repaved. In order to install the required 180-mm-thick base c ourse (7 inches), three lifts were used because FDOT limits the maximum thickness per lift to 75 mm (3 inches), as described (FDOT 2002). Each lift of the base course was either 50 mm (2 inches) or 75 mm (3 inches) thick, but the thickness vari ed depending upon the area. On top of the base course, 75 mm (3 inches ) of structural course and 37 mm (1.5 inches) of friction course were laid. A total of 918 crew hours we re spent to install pa vement structure in a total area of 242,351.56 m2. 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00 16111621263136414651566166717681869196 Work daysC. load/CH Figure 4-4. Daily productivity plot for I-10 The project statistics are shown in Table 4-1. If the baseline productivity were projected to those days when disruptions occurred, then th e crew hours would have been 288.51 less than the total crew hours spent (918 hours). Thus, the inefficient crew hours were 31.4 percent of total crew hours. Tabl e 4-5 shows the workdays when disruptions occurred during the process.

PAGE 78

60 Table 4-5. Disruptions occurred in project I-10 Work day Date lift C. Load C. Load /hr CH CH by baseline Wasted Category Cause 4 4/7/2003 1/3 3.90 0.98 4.00 0.66 3.34 Management Prerequisite work 6 4/26/2003 1/2 16.38 2.34 7.00 2.78 4.22 Management Material delivery 7 4/28/2003 1/2 39.42 3.94 10.00 6. 69 3.31 Management Congestion 9 4/30/2003 2/2 35.54 2.96 12.00 6.03 5.97 Work content Work zone 10 5/1/2003 1/1 47.13 3.93 12.00 8.00 4.00 Work content Work zone 15 5/12/2003 3/3 29.64 2.96 10.00 5.03 4. 97 Management material failure 16 5/13/2003 1/1 4.12 0.82 5.00 0.70 4. 30 Management Material delivery 17 5/20/2003 1/1 30.90 3.09 10.00 5.25 4. 75 Work content Turning lane 18 5/21/2003 1/2 15.00 3.00 5.00 2.55 2.45 Work content Shoulder area 19 5/22/2003 1/2 20.12 2.52 8.00 3.42 4.58 Work content Shoulder area 24 6/4/2003 1/1 7.21 0.80 9.00 1.22 7.78 Management Prerequisite work 26 6/23/2003 1/1 18.54 1.85 10.00 3.15 6.85 Work content Intersection 27 7/14/2003 1/1 30.90 3.86 8.00 5.25 2.75 Management material failure 29 7/16/2003 1/1 18.54 2.32 8.00 3.15 4.85 Weather Rain 33 7/22/2003 1/1 25.75 3.22 8.00 4.37 3.63 Weather Rain 42 8/27/2003 1/3 42.12 3.83 11.00 7.15 3.85 Management Equipment breakdown 43 9/2/2003 1/3 24.96 2.50 10.00 4.24 5.76 Work content Intersection 45 9/5/2003 3/3 24.96 2.50 10.00 4.24 5.76 Management Prerequisite work 48 9/11/2003 1/2 30.00 3.00 10.00 5.09 4. 91 Management Material delivery 49 9/19/2003 1/3 19.50 1.95 10.00 3.31 6.69 Management Prerequisite work 51 9/29/2003 4/4 17.16 1.63 10.50 2.91 7.59 Management Prerequisite work 52 9/30/2003 2/3 24.96 2.27 11.00 4.24 6.76 Work content Ramp 54 10/6/2003 1/2 14.00 1.12 12.50 2.38 10. 12 Management Material delivery 55 10/9/2003 2/2 27.81 2.78 10.00 4.72 5.28 Work content Ramp 56 10/14/2003 4/4 10.92 1.37 8.00 1.85 6.15 Management Prerequisite work 57 10/16/2003 4/4 30.42 3.04 10.00 5.16 4.84 Management Prerequisite work 58 10/17/2003 4/4 21.84 1.99 11.00 3.71 7. 29 Management Material delivery 62 10/22/2003 3/3 8.58 0.78 11.00 1.46 9.54 Weather Rain 63 10/23/2003 1/2 20.60 1.72 12.00 3.50 8.50 Management Prerequisite work 64 11/10/2003 1/3 31.20 2.60 12.00 5.30 6.70 Work content Ramp 67 12/18/2004 1/3 10.92 0.99 11.00 1.85 9.15 Management Prerequisite work 68 12/19/2004 1/3 45.24 4.11 11.00 7.68 3.32 Management Equipment breakdown 69 1/8/2004 3/3 14.82 1.85 8.00 2.52 5. 48 Work content Turning lane 72 1/20/2004 1/2 14.00 1.27 11.00 2.38 8.62 Management Congestion 74 1/22/2004 2/2 6.18 0.62 10.00 1.05 8.95 Management Prerequisite 76 1/28/2004 1/2 14.00 1.40 10.00 2.38 7.62 Weather Rain 77 2/3/2004 1/1 18.54 2.32 8.00 3.15 4. 85 Work content Turning lane 79 2/5/2004 1/2 18.00 1.64 11.00 3.06 7. 94 Work content Turning lane 80 2/9/2004 1/2 24.00 2.18 11.00 4.07 6.93 Management Prerequisite work 81 2/26/2004 2/2 27.81 2.53 11.00 4.72 6.28 Management Prerequisite work 82 2/27/2004 1/2 26.78 2.98 9.00 4.55 4.45 Management Prerequisite work 85 3/1/2004 1/2 6.00 0.60 10.00 1.02 8.98 Management Rework 86 3/4/2004 1/1 10.30 2.06 5.00 1.75 3.25 Weather Rain 88 3/8/2004 1/2 5.15 0.52 10.00 0.87 9. 13 Management Material delivery 91 3/29/2004 2/2 19.53 3.91 5.00 3.32 1.68 Work content Median 92 3/31/2004 2/2 23.58 2.14 11.00 4.00 7.00 Management Rework 95 4/6/2004 1/3 13.26 1.21 11.00 2.25 8. 75 Management Material delivery 96 4/8/2004 2/3 2.34 0.26 9.00 0.40 8.60 Work content Ramp

PAGE 79

61 4.6.2 Work Flow Management Poor work flow management caused 177.92 inefficient crew hours on the project, which translates into 61.7 percent of the total inefficient hours on the project. On Days 4, 45 and 49, the area planned for paving expe rienced “pumping”, and a subgrade failure delayed the paving. On Days 6, 16, 48, 54, and 58, the material delivery trucks were late, delaying the paving. Again, coordinating the capacity of asphalt plant with the speed of paver is important factor for the eff ective production of pavement, and the number of trucks should be determined to optimize the coordi nation. On Days 42 and 68, the paver was broken, and the crewmembers remained idle un til the paver was fixed. On Days 51, 56, and 57, the crew had to level the subgrade in the area where the base course was to be placed. Because the pre-paved area did not meet the density requirement, re-work was performed on Day 85. The PMI of this proj ect is 0.4669, the lowest value among the four projects studied. This result indicates that this project was the best-managed of the projects analyzed. 4.6.3 Work Content Work content caused 81.69 inefficient cr ew hours, which translates into 28.3 percent of the total inefficient hours on the proj ect. On Days 9 and 10, access to the work zone was restricted by traffic, making work more difficult. On Days 18 and 19, the crew worked inside the shoulder, and on Day 43, an intersection was paved. On Day 26, the crew worked on the resurfacing area, where they had to wait for the top pavement layer to be milled before paving. It is not know n whether poor management contributed to this situation. On Days 52, 55, and 64, the crew worked on a ramp, which requires much handwork, lowering productivity.

PAGE 80

62 4.6.4 Weather Poor weather conditions cause d a total of 28.9 inefficient crew hours on the project, which translates into 10.0 percent of the tota l inefficient hours on the project. Rainfall on Days 29, 33, 62, and 86 caused delays on the project. 4.7 Summary of Production Analyses and Conclusion The pavement operation, one of the most equipment-intensive of all construction activities, has significant produc tivity variability due to disr uption events in the workflow caused by such factors as poor management , work content, and severe weather conditions. The loss of crew hours caused by poor management ranged from 40% to 62% of the total inefficient crew hours on the four projects. One of the primary factors that contributed to poor management was out-of-sequence work and deficiencies in prerequisite work. Project SR-20 (Alachua ) and Project SR-102 repeatedly required rework in order to correct the deficiencies of the limerock base when the crew placed the first lift of the structural course. Compensa ting for this prerequisite activity diminished the paving crewsÂ’ productivity . It has been confirmed that the quantitative and qualitative uncertainty in pr oduction output caused adverse e ffects, extending them to succeeding activities in the activity nexus. Th e PMI values of the two projects were higher than the other two pr ojects that used HMA base. In addition, equipment breakdown and material shortages were occa sionally observed dur ing the process of paving. The loss of crew hours caused by work cont ent was from 21% to 48% of the total inefficient crew hours on the four projects. Ev en if the loss by work content is inevitable, the impact of work content can be mitigated with minimum impact on the project by planning ahead for the most frequent causes and improving predictability of workflow.

PAGE 81

63 Efficiency losses due to work content in this research mostly related to use of hand work while working in areas such as ramps, tu rning lanes, intersec tions, and crossovers; however, management and work content are some times interrelated, i.e., the deficiencies of previous work not only diminish the producti vity of later work but also make the work more complicated. The effect of severe weather conditions wa s not very significant, ranging from 6% to 17% of the total inefficien t crew hours of the four pr ojects. Even though severe weather prolonged the projects, the delay was not observed in this research because the contractors usually chose not to work when the weather was expected to be unfavorable and productivity was only measured on days when the contractors worked. Finally, the measurement of performance on a daily basis can provide information about the primary causes of productivity va riability in paving operations. This information, in the hands of competent mana gers, can lead to more effective and more reliable work flow and better quality contro l during the operation. Statistical analyses were performed to validate the re sults and presented in Chapter 5.

PAGE 82

64 CHAPTER 5 EFFECTS OF INTERFERFERENCE 5.1 Introduction In Chapter 5, the researcher described stat istical analyses that were performed to further analyze the results in Chapter 4. Th e analyses consist of two major parts: mean productivity comparisons and correlations among factors. For the purpose of mean comparisons, productivity values obtained fr om four case study projects (population) were grouped by the project from which the va lues were collected and by interference (or no interference) with which the values were as sociated (samples). Mean values of each sample were compared and ranked among four projects and among interference factors. The correlation analyses among interference f actors will follow in Section 5.5, titled Correlation Analyses. Before performing mean comparisons, tw o general requirements should be met1. The population should be normally distributed. The variances of each sample should be the same. The null hypothesis (Ho) for the normality test provided by the Minitab Software (Version 14) is that all data from the populat ion are normally distri buted. As seen in Figure 5-1, when the probability of the population was plotted, Howas rejected at the 5% level (p-value less than 0.05). It was c oncluded that the population was not normally distributed. 1 The samples were assumed to be independent (no correlation).

PAGE 83

65 Daily ProductivityPercent 12 10 8 6 4 2 0 -2 -4 99.9 99 95 90 80 70 60 50 40 30 20 10 5 1 0.1 Mean2.695 StDev1.671 N253 AD2.549 P-Value<0.005Probability Plot of Daily ProductivityNormal Figure 5-1. Probability plot of population before transformation As is shown in Figure 5-1, the data were distributed asymmetric ally with a long tail of high values. Such a distribution is said to be “skewed to the right,” and instead of the points plotting in a straight line, they fall along a smooth curve, which is convex upward (Hart 2002). The Productivity rate, defined as the output quantity divided by input quantity of each working day, cannot be less than zero, and they can be higher values further from the mean values as seen in th e figures from 4-1 to 4-4. Such data are typically severely skewed to the right. The population, therefore, needs to be tran sformed to meet the first condition. A transformation of the sample data is defi ned as a systematic process in which the measurements on the original scale are convert ed to a new scale (Ott 2001). Possibilities for converting the original variable y are y2, y, log y, or some other transformed variables. Most often, the conversion may be accomplished by choosing a particular power of the data, and the fam ily of transformations may be expressed as transf(y) = yp, where y and transf(y) are the original and tr ansformed data, respectively, and p is the power to which y is raised. The more severely that y is skewed to the right, the lower the value of p is required to obtain a near-norma l transformation. A suitable transformation

PAGE 84

66 is found by trial and error. Because the popul ation was skewed to the right somewhat severely, the original data (y ) were converted by taking the sq uare root (transf(y) and p = 0.5). After re-testing the normality with transformed data, the result shows that Ho cannot be rejected at the 5% le vel with the p-value of 0.225, as shown in Figure 5-2. The population was concluded to be normally distributed. C3Percent 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 99.9 99 95 90 80 70 60 50 40 30 20 10 5 1 0.1 Mean1.558 StDev0.5187 N253 AD0.486 P-Value0.225Probability Plot of C3Normal Figure 5-2. Probability plot of population after transformation Equal variance, the second condition for the mean comparison, was then tested. Hoof the test was that all variances from each sample group were the same. As shown in Figure 5-3, the p-value of LeveneÂ’s Test was 0. 131, which is insignificant at the 5% level. Hocannot be rejected, and the variances from each project were conclusively the same at the 5% level. This result from the equal variance test meets the second condition for the mean comparison.

PAGE 85

67 C195% Bonferroni Confidence Intervals for StDevs SR-20 Palatka SR-20 Hawthorne SR-102 I-10 0.8 0.7 0.6 0.5 0.4 0.3 Test Statistic5.13 P-Value0.162 Test Statistic1.89 P-Value0.131 Bartlett's Test Levene's Test Test for Equal Variances for C3 Figure 5-3. Equal variance te st among four projects 5.2 Multiple Mean Comparisons with Transformed Productivity Once the two conditions were met, the m ean values of transformed productivity among four projects were comp ared by using the F-test. Hoof the test was that all of the mean values were the same, and by rejecting Ho, it was concluded that the means are not the same. The box plot, mean and standard de viation, and the result of the F-test are shown in Figure 5-4, Table 5-1, a nd Table 5-2, respectively. ProjectsTrans. I-10 SR-102 SR-20 Hawthorne SR-20 Palatka 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 Boxplot of Trans. vs Projects Figure 5-4. Box plot for mean comparison among projects

PAGE 86

68 Table 5-1. Mean and standard deviation for four projects Sample Number of data Mean Standard deviation Rank I-10 (Escambia) 97 1.6980 0.4884 1 SR-102 (Duval) 50 1.6316 0.5830 2 SR-20 (Palatka) 64 1.4342 0.4342 3 SR-20 (Alachua) 42 1.3768 0.5262 4 Table 5-2. Result of F-test for mean comparison Source Degree of Freedom (DF) Sum of Square (SS) Mean Square (MS) F-value P-value Projects 3 5.014 1.671 6.63 0.001 Error 249 62.787 0.252 Total 252 67.802 Since all of the mean values of the four pr ojects turned out to be different (p-value less than 0.05), as is shown in Table 5-2, pair-wise mean comparisons were performed. The least significant difference (LSD) test a nd TukeyÂ’s test are gene rally used for this purpose. The two tests are used only after the F-test has been shown to be significant. Although similar, TukeyÂ’s test has a smaller e rror range and is more conservative than LSD. Therefore, TukeyÂ’s test concludes significant difference between sample means fewer than LSD does. 5.2.1 Further Analysis by TukeyÂ’s Test Once the mean values of the four projects we re listed in order from the highest to the lowest, as seen in Table 5-1, 95% confid ence intervals for the difference between the two means were generated for each project m ean value. For example, the I-10 project has the highest mean value among the four proj ects. The 95% confid ence interval for the mean difference between I-10 and the other projects (2 1 ) was calculated. If the confidence interval contains zero, then the tw o mean values are not significantly different at the 5% level. The output of pair-wise m ean comparisons can be found in Appendix C.

PAGE 87

69 As a result of Tukey’s test, the mean of the I-10 project is higher than the SR-20 (Alachua) and SR-20 (Putnam) projects even though the means from the I-10 and SR-102 projects were not significantly different. 5.2.2 Further Analysis by Least Significant Difference (LSD) Test Unlike the result of Tukey’s test, the result of the LSD test shows that projects SR102 and I-10 have better productivity than proj ects SR-20 (Alachua) and SR-20 (Putnam). Again, the result of the LSD test is less c onservative than Tukey’s test, and the sample means between SR-102 (second highest) a nd SR-20 (Alachua, third highest) are concluded to be different. It is very likely that the pave ment crews for the two projects with higher productivity than the other two projects were encouraged to be more productive because those two pr ojects were located in urban areas with AADT of more than 12,000 (FDOT 2002c). The results of the LSD test can also be found in Appendix C. Next, the population was gr ouped by five major interfer ence factors: multiple, management, work content, weather, and no interference. The mean values of each sample group were compared in order to ra nk the factors by their magnitude of adverse effects on production rates. 5.3 Productivity Comparison by Interference Factor The F-test again was performed to compare the mean values from five sample groups associated with the o ccurrence of interference. If there were more than two incidents of interference on a single work day, then the productivity on that day was defined as “multiple.” Since the population of all of the sample groups was normally distributed, only equal variance was tested for the mean comparison. As shown in Figure

PAGE 88

70 5-5, the test concluded that the variances of four sample groups are the same at the 5% level because the p-value of LeveneÂ’s test is 0.824 in favor of Ho(equal variance). Interference95% Bonferroni Confidence Intervals for StDevs No interference Weather Work Content Management Multiple 1.2 1.0 0.8 0.6 0.4 0.2 Bartlett's Test 0.824 Test Statistic2.51 P-Value0.643 Levene's Test Test Statistic0.38 P-ValueTest for Equal Variances for T. Prod. Figure 5-5. Equal variance test among interference factors Figure 5-6, Table 5-3, and Ta ble 5-4 show the output of the box plot, the mean and standard deviation, and the results of the F-te st, respectively. As shown in Figure 5-6 and Table 5-3, the productivity means are ranked in the order of multiple, management, work content, weather, and no interference from the lowest to the highest. The result of the Ftest in Table 5-4 shows that the means associat ed with five interference factors are not the same at the 5% level (p-value less than 0.05). InterferenceT. Prod. No interference Weather Work Content Management Multiple 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 Boxplot of T. Prod. vs Interference Figure 5-6. Box plot for mean co mparison among interference factors

PAGE 89

71 Table 5-3. Mean and standard de viation for interference factors Sample Number of data Mean Standard deviation Rank Multiple Interference 9 1.21890.4954 1 Management 65 1.29750.4456 2 Work Content 42 1.38710.4196 3 Weather 14 1.50930.4273 4 No Interference 123 1.78260.4998 5 Table 5-4. Result of F-test for comparison of interference factors Source DF SS MS F-value P-value Projects 4 12.909 3.227 14.62 0.001 Error 248 54.740 0.221 Total 252 67.648 Again, further analyses were performed becau se the result of the F-test in Table 5-4 shows that all the means are not the same at the 5% level. The result can be found in Appendix D. The productivity means were ranked by multiple, management, work content, weather, and no interference from the lowest to the highest base d on their association with types of interference. When multiple types of interference occurred, the mean productivity rate was the lowest. For exam ple, pavement crew experienced the most sever interference when they worked on tu rning lanes, ramps, driveways, and around manholes (work content), while other interf erence factor caused by management or weather occurred as well. The mean values of productivity associat ed with any type of interference are significantly lower than the one with no interf erence, as confirmed by the result of the LSD test. The result of TukeyÂ’s test, however, shows that there was no significant difference between the mean value of w eather interference (second highest) and no interference (the highest). Again, TukeyÂ’s te st is more conservative than the LSD test, and concludes the difference between two mean values fewer than the LSD test does.

PAGE 90

72 Inclement weather was not a si gnificant factor in productiv ity because contractors chose not to work when inclement weather was expected. Even though the differences are not signifi cant at the 5% level, management, as a single factor, has not only the most frequent occurrence (Table 5-3) but also the most adverse effects on productivity . Therefore, the mean comparison was expanded within management interference, as is de scribed in the next section. 5.4 Productivity Comparison for Management Interference Interference caused by the management f actor was broken down into four subcategories: defects in prerequisite work, material-related interference (material delay or failure), equipment breakdown, and work c ongestion. The mean values of four management sub-categories were then compar ed to rank productivity rates associated with each management interference sub-cate gory. The results of the comparison are shown in Table 5-5, Table 5-6, and Figure 5-7. Table 5-5. Mean values with management interference Level N Mean StDev Rank Work congestion 7 1.16290.502 1 Prerequisite 421.24050.44252 Equipment 111.36910.46893 Material 151.49070.47224 Table 5-6. Result of mean comparison for management interference Source DFSS MS F-Value P-Value Management interference 3 0.878 0.2931.4 0.25 Error 71 14.8610.209 Total 74 15.738

PAGE 91

73 Mgt interT.Prod.R Material Equipment Prerequisite Work congestion 2.5 2.0 1.5 1.0 0.5 Boxplot of T.Prod.R vs Mgt inter Figure 5-7. Box plot for the mean co mparison with management interference As is shown in Table 5-5, a total of 75 working days had management-related interference. The mean producti vity rates were ranked as wo rk congestion, prerequisite work, equipment, and material in order from the lowest to the highest even though the differences among them were not significant at the 10% level (p-value of 0.25). The mean value of the productivity rate s associated with the defects from prerequisite work was the second lowest among the four sub-categories, and the prerequisite interference occu rred on 42 working days out of total 75 working days when management interference occurred, as is show n in Table 5-5. Therefore, the researcher focused on the prerequisite interference for the correlation analyses along with weather interference from the following section. 5.5 Correlation Analyses The correlation analyses in this section cons ist of two parts. The first part deals with the correlation between th e amount of rainfall in project areas and the effects of it on project productivity. The s econd part includes more in-depth correlation analyses between causes of weather a nd prerequisite work and th eir effects on the crewsÂ’ performances represented by probability of interference occurrence and productivity variation.

PAGE 92

74 5.5.1 Effects by the Monthly Amount of Rainfall Even though the mean productivity associated with the weather interference factor was ranked the highest out of four interf erence sample groups, excluding the group of no interference, the researcher performed th e correlation analyses between the cause variables (X: the monthly amounts of rainfall and the number of rain days in project locations) and the effect vari ables (Y: the number of work days that production was influenced by the rainfall and NLA). Because the researcher collected productivity data that included inefficient work hours caused by ra infall, the likelihood of the effects of the rainfall were also quantified by multiplying the inefficient work hours by the baseline productivity of each project. Thus, the or der of magnitude was represented by the number of truck loads affected (NLA) by rainfall. The mean values of the monthly rainfa ll on the areas of Jacksonville (SR-102), Pensacola (I-10), and Gainesville (SR-20 Alachua and SR-20 Putnam) were obtained2. Each value is the average of the monthly rainfall for the 30 years from 1975 to 2005. Then the number of work days and the NLA by the rainfall were plotted on a monthly time series plot, seen in Figure 5-8. The un its of the second Y-axis—as seen on the right side of the chart—were number of days, hours, and loads, respectively. NLA were divided by 10 to plot them on a similar scal e. Due to the geographical proximity among the three areas, the shapes of the three monthly rainfall plots are very similar, with a small spike in March and high amounts of rainfall between June and September. 2 Source: National Climate Data Center (NCDC) (http://lwf.ncdc.noaa.gov/oa/ncdc.html)

PAGE 93

75 0 1 2 3 4 5 6 7 8 9 123456789101112 MonthInches0 2 4 6 8 10 12Effects Gainesville Jacksonville Pensacola Affe. Days Ineff. Hours NLA (/10) ` Figure 5-8. Monthly time series plot for rainfall and NLA The Pearson correlation test was then perfor med for the five factors. The Pearson correlation measures the strength of the lin ear relationship between the X and Y variables on a probability plot. The correlation between the two variables lies between 0 and 1, with higher values indicating a better fitting distribution. The results of the test are shown in Table 5-7. Again, the monthly am ounts of rainfall in the three locations were very close, and this similarity was confirmed by the test results. The Pearson correlation coefficient (r-value, appears on top) between the amounts of rainfall in the Jacksonville and Gainesville areas is 0.759, meaning that 75.9 % of variation in the amount of rainfall in one area can be explained by that of the ot her area. The p-value is 0.004 (appears on bottom in parenthesis), and it confirms the si gnificant correlation (significant at the 5% level). The p-values for the amounts of rain fall for the three areas are 0.001, 0.004, and 0.009, and they are all significant at the 5 % level. In addition, the number of work days affected by rain and the NLA are highly corr elated with the r-value of 0.917 and the pvalue of less than 0.001.

PAGE 94

76 Table 5-7. Correlation test result be tween the amounts of rainfall and NLA GainesvilleJacksonvillePensacola Work days affected 0.759 Jacksonville (0.004) 0.884 0.712 Pensacola (0.001) (0.009) -0.054 -0.109 -0.018 Work days affected (0.868) (0.737) (0.955) -0.059 -0.127 -0.085 0.917 NLA (/10) (0.855) (0.695) (0.793) (0.001) 5.5.2 Effects by the Monthly Number of Rain Days Next, the monthly number of rain days in the year 2003 was obtained on the areas of Jacksonville (SR-102), Pensacola (I-10), and Gainesville (SR-20 Alachua and SR-20 Putnam)3. The number of work days and NLA by the rainfall was also plotted on a monthly time series plot, seen in Figure 5-9. As seen in Figure 5-8, the shapes of the three monthly number of rain days are very si milar, with more than 15 days in March and June. Note that the Pensacola area had 29 rain days in June, and the areas of Pensacola and Jacksonville had 23 rain days in July, 2003 The results of the correlation test are show n in Table 5-8. The monthly number of rain days in the three locations was close related. The r-values between GainesvilleJacksonville, Gainesville-Pens acola, and Jacksonville-Pensacola were 0.712, 0.791, and 0.792, respectively. The p-values for the numbe r of rain days for the three areas were all smaller than 0.01, and they are all significant at the 5 % level. Again, the number of work days and the number of loads affect ed by rain were highly correlated. 3 Source: National Climate Data Center (NCDC) (http://lwf.ncdc.noaa.gov/oa/ncdc.html)

PAGE 95

77 0 5 10 15 20 25 30 35 123456789101112 MonthRain days0 2 4 6 8 10 12Effects Gainesville Jacksonville Pensacola Affe. Days Ineff. Hours NLA (/10) ` Figure 5-9. Monthly time se ries plot for the number of rain days and NLA Table 5-8. Correlation test result between the number of rain days and NLA GainesvilleJacksonvillePensacola Work days affected 0.712 Jacksonville (0.009) 0.791 0.792 Pensacola (0.002) (0.002) -0.18 0.042 0.000 Work days affected (0.576) (0.897) (1.000) -0.253 -0.138 -0.035 0.917 NLA (/10) (0.428) (0.669) (0.915) (0.001) 5.5.3 Effects of Rainfall without Outliers As seen in Tables 5-7 and 5-8, the amount of rainfall and the number of rain days in the three areas had little or no influen ce on either NLA or the number of work days that had weather interference. The r-values were negative (boldfaced), meaning that the more rainfall or rain days the pavement cr ew had, the fewer work days and the smaller number of loads were affected by rainfall. The correlations (r-values) were not significant with the p-values of over 0.428, the highest correlati on, seen between the number of rain days and NLA by rain in Gainesville.

PAGE 96

78 There were two reasons that explain this re sult. First, total number of work days that had rain interference wa s only 14 days as seen in Table 5-3. Due to the limited number of dataset, the result was not very reliable. Second, the dataset included outliers that deviated considerably from the shape of rainfall. As seen in Figures 5-8 and 5-9, the shape of number of work days and NLA obtai ned in September and December were very different from that of the rainfall. The researcher attempted to test the correlation after truncating the data for September and December by defining them as m ild outliers. The objectiv e of this test is to show the possible correlation when exclus ive volume of data is available. Mild outliers are defined as the values that do not fall within the range of the inter-quartile range (IQR) multiplied by 1.5 from both lower and upper quartiles (Ott 2001). For example, IQR of NLA was 2.53 calculated by th e first quartile (0) subtracted from the third quartile (2.53). When multiplying 1.5 by the upper quartile (2.53) , it yields 3.795, which is the upper limit for mild outliers. The NLA in December was 9.46, making it a mild outlier. The result after removing those two outliers shows that the correlation increased, seen in Table 5-9. Table 5-9. Correlation of te st result wit hout outliers GainesvilleJacksonvillePensacolaMean No. of Work days 0.923 Jacksonville (0.001) 0.873 0.828 Pensacola (0.001) (0.003) 0.977 0.958 0.936 Mean (0.001) (0.001) (0.001) 0.232 0.452 0.348 0.351 No. of Work days (0.520) (0.190) (0.325) (0.319) 0.464 0.714 0.517 0.583 0.917 NLA (/10) (0.177) (0.020) (0.126) (0.077) (0.001)

PAGE 97

79 Since the amounts of rainfall in the three areas were highly correlated, the mean value of the areas was added in Table 5-9. The correlation coefficient between the number of work days and the amount of rain fall in the three areas was increased even though the p-values were not si gnificant at the 10% level. The coefficient between the mean value of the rainfall and the number of work days is 0.351 with the p-value of 0.319. The correlation between NLA and the ra infall increased more significantly. The amounts of rainfall in Gainesville, Jacksonvill e, and Pensacola explain the 46.4%, 71.4%, and 51.7% of variability in NLA by rainfall. The mean value of the rainfall explains the 58.3% of variation in NLA, and the correla tion between the mean rainfall and NLA is significant at the 10 % level w ith the p-value of 0.077. As explained, production data were colle cted only when the paving crew worked, and the crew usually chose not to work wh en inclement weather was expected. Thus, only 14 work days were affected by rain. This partly explains why the mean productivity rate when weather interference occurred was hi gher than when other types of interference occurred. Even though the amount of rainfall or the number or rain days have little effect on NLA or the number of work days affected by rain, the correlation is expected to be more significant if the production data is co llected on the calendar day scale instead of working day scale. 5.6 Effects of Weather and Prerequisite Work The researcher performed more experiments to test the effects of interference factors related to weather and prerequisite work. The experiments were performed with limited number of data collected because the whole purpose was to show possible effects of the two factors on performance. The cause s include inefficient hours by weather and prerequisite work interference. The ineffici ent hours were categoriz ed by their severity;

PAGE 98

80 if the inefficient hours on a work day were le ss than 3 hours, then it was considered to be “Modest,” while over 3 hours were considered to be “Severe.” The effects are the probability of occurre nce by management inte rference and daily productivity rates when a cause factor occurred , as shown in Table 510. It is important to note that the probability of management interference occurr ence, as a hypothetical rule, increases linearly over time until the next interference occurs. For example, if the crew worked a certain number of truck loads until management interference occurred, then the probability was calculated by 1 divided by the number of truck loads until the interference occurred. Then the probability was set back to zero until the next interference occurred. Table 5-10. Design of correlation test between causes and effects Design Causes (severe, modest) Effects Correlation 1 Inefficient hours by weather interference Probability of occurrence by management interference (1st, 2nd, 3rd, and 4th week) 0.683 (0.007) Significant 2 Inefficient hours by weather interference Production rate (Same day, 1st, 2nd, 3rd, and 4th week) -0.531 (0.042) Significant 3 Inefficient hours by prerequisite work Probability of occurrence by other management interference (1st, 2nd, 3rd, 4th week) 0.228 (0.363) Not significant 4 Inefficient hours by prerequisite work Production rate (Same day, 1st, 2nd, 3rd, and 4th week) -0.391 (0.015) Significant at 20 % The effects of Design 1 and Design 3 are measured by the probability of management interference occu rrence after weather interference (Design 1) or defects from prerequisite work (Design 3) occu rred. They consist of four levels—1st, 2nd, 3rd, and 4th week—which are based on the time elapse afte r the cause factor occurred. The effects of Design 2 and Design 4 are measured by produ ctivity rates after weather (Design 2) or

PAGE 99

81 prerequisite (Design 4) interf erence occurred. The effects of Designs 2 and 4 have the same four levels as Designs 1 and 3 had, but in addition to the four levels, they include another level: the same day th at the cause occurred. The Pearson correlations were also tested to measure how the causes of each design influenced their effects. First, the research er performed the test for Designs 1 and 3 on a daily basis. However, little or no corr elation could be found between the actual inefficient hours of the cause factor (X) and the probability of management interference occurrence of the effect factor (Y). For example, after ra in delay occurred, there was no interference recorded by any interference factor for the following two or three weeks because the pavement crews worked irregularl y. Therefore, the correlation tests for Designs 1 and 3 were performed between the in efficient hours of the cause factor and the probability of management interference occurrence of the effect factor for four weeks of the time period after the cause factor occurr ed, assuming ripple effects of interference. For Designs 2 and 4, however, the correla tion between the actual inefficient hours by the cause factor (X) and th e productivity rate on the same day when the interference occurred (Y) was tested. The result of e xperiment for each design is described from Sections 5.6.2 to 5.6.5. 5.6.1 Initial Design The model with crossed and nested effects was considered as an initial model, as shown in Table 5-11. Again, measurement responses are the probability of management interference occurrence for Design 1 and Desi gn 3 while they are the productivity rates for Design and Design 4. The severity of cau ses is categorized by either modest (less than or equal to 3 hours of inefficient hours) or severe (more than 3 hours).

PAGE 100

82 Table 5-11. The model with crossed and nested effects Project 1 Project 2 Project 3 Project 4 Modest Severe Modest Severe Modest Severe Modest Severe Week 1 111 121 Week 2 Week 3 Week 4 444 The model with crossed and nested eff ects is represented in Equation [5-1]. [5-1] ijkl ijk ik k ij i ijklE y ) ( ) ( Where, A: Project, B: Seve rity, and C: Week i : The effect of level i of A (i=1, 2, 3, 4) ij : The effect of level j of B (j=1.5, 4.5) nested within ith level of A. k : The effect of level k of C (k=1, 2, 3, 4) ik) ( : Interaction between project and weeks ijk) ( : Interaction between severity and weeks ijklE: Error term The model tests the interaction effects, as well as the main effect (effects by individual factors). For Design 1, as an example, those effects include the following. 1. Whether each project has a different proba bility of management interference after weather interference occurred (i ). 2. Whether the probabilities of management in terference are different within a project depending on the severity of weather interference (ij ). 3. Whether the probabilities of management in terference are different over four weeks of time after weather interference occurred (k ). 4. Whether each project has a different proba bility of management interference in each week for four weeks after weather interference occurred (ik) ( ). 5. The severity of weather interference infl uences the probabili ty of management interference for four weeks of time periodijk) ( .

PAGE 101

83 The same analyses can be performed for th e other three designs. Due to the lack and suitability of data collected, however, th e initial design was changed to a 2-factor experiment model, seen in Equation [5-2]. [5-2] ijk ij j i ijkE y ) ( . Where, A: Project (4-level) and B: Severity (2-level) The researcher believes that the initia l design is useful when proper data are available. To test the 2factor experiment model, th e General Linear Method (GLM) procedure was applied because this procedure can be used when the dataset is unbalanced (different number of measurements in each category). The format of the 2-factor experiment model is shown in Table 5-12. The measurements in each category in cludes the probability of management interference occurrence for f our weeks of time combined for Design 1 and Design 3, while they include productivity rates for the sa me day when interference occurred and for four weeks for Design 2 and 4. Each test result will be described in more detail. Table 5-12. Design for 2-factor experiment Project 1 Project 2 Project 3 Project 4 Modest 11 21 31 41 Severity 21 22 32 42 5.6.2 Effects of Weather on Probability of Management Interference Occurrence Design 1 in Table 5-10 shows the correla tion between the seve rity of weather interference and the probability of management interference occurrence over a four-week time period after weather interf erence occurred. In the SR-102 (Duval) project for example, the scheduled paving operation was suspended due to the wet condition of the limerock material. The correlation between the two factors was cal culated by using the Pearson correlation test again. Even though the cause (inefficient hours) was categorized

PAGE 102

84 by severity, the correlation test shown in Table 5-10 was performed quantitatively, using the actual number of ineffi cient hours. The r-value was 0.683, and the correlation was significant at the 5% level. This result shows that as the inefficient hours by weather interference increases, the probability of management interference occurrence tends to increase and that the inefficient hours explai ns the 68.3% of variab ility in occurrence of management interference. The result of the 2-factor experiment is shown in Tables 5-13 and 5-14, with the result of further analyses in Appendix E. The result in Table 5-13 shows that the mean values in each category (total of 8) are not significantly di fferent in favor of the null hypothesis of equal mean (p-value: 0.3614). Table 5-13. GLM test result for Design 1 Source DF SS MS F-value P-value Model 7 0.00036053 0.0000515 1.36 0.3614 Error 6 0.00022698 0.00003783 Corrected Total 13 0.00058751 Table 5-14. 2-factor experi ment result for Design 1 Source DF Type III SS MS F-value P-value A 3 0.00005965 0.00001988 0.53 0.6807 B 1 0.00008573 0.00008573 2.27 0.1829 A*B 3 0.00006458 0.00002153 0.57 0.6556 Table 5-14 shows that the main effects of factors A and B were not significant at the 10% level. This result is conclusive because the interaction effect between two factors is insignificant. In other words, the probability occurrence of management interference remains the same for the four pr ojects after different severities of rain interference occurred. This result can also be confirmed by further analyses, which is shown in Appendix D. Only Project A ha s a different occurrence of interference by

PAGE 103

85 management caused by different severity of th e weather interference. All four projects have a similar probability of occurrence when the severity is the same (that is, fixed). 5.6.3 Effects of Weather Interference on Productivity Design 2 was targeted to test the effect s of weather interference on productivity. The PearsonÂ’s correlation test in Table 5-10 showed that the inefficient hours caused by weather interference had a signifi cant effect on the productivity variation at the 5% level (r-value: -0.531, p-value: 0.042). The negative r-value shows that as the inefficient hours increase, the productivity decrea ses. The result of the 2-f actor experiment is shown in Table 5-15. Table 5-15. 2-factor experi ment result for Design 2 Source DF Type III SS MS F-value P-value A 3 2.8652725 0.955090836.02 0.0011 B 1 1.039700971.039700976.55 0.0128 A*B 3 0.088745730.029581910.19 0.9052 Based on the result, the inter action effect between factors A and B is insignificant, meaning that the productivity difference among four projects was not affected by different levels of inefficient hours by w eather. The main effect of factor A was significant, so productivity ra tes after weather interferen ce occurred were significantly different from project to project. Since factor A is significant at th e fixed level of B, as shown in Appendix E, it also can be conclude d that even though a similar severity of weather interference occurred, th e effect of the in terference on productivity varied from project to project among the four projects investigated. This result is conclusive because no significant interaction effect between factors A and B were observed.

PAGE 104

86 The main effect of factor B was not signi ficant at the 10% level, meaning that the severity of weather in terference does not have a significan t effect on the variability of productivity. This result was also conclusive. 5.6.4 Effects of Prerequisite Work on Management Interference Occurrence As an engineering assumption, the defects of prerequisite work have adverse effects on the probability of other management inte rference occurrences. Details of possible defects from prerequisite work were discussed in Chapter 4, and one of the most frequent findings during pavement opera tion was of defects on the pa vement layer constructed previously. Examples are pumping from th e limerock base before the crew had constructed the first lift of the structural cour se and the low (or high) profile of the layer not conforming to the specificati on. In the SR-102 project, the constructed width of the roadway was less than the planne d width; therefore, additional materi al was required in order to maintain proper width. Also, the cr ew had to mill the deficient pavement section because previously placed HMA material did not meet the density requirement. Thus, the objective of Design 3 was to test the hypothesis that as the defects in the prerequisite work increase, the probabili ty of management interference occurrence increases. The order of magnitude for prer equisite work was measured by the hours that the pavement crew had to spend to fix th e problem, and those hours were specifically recorded in the production data collected. The results of the Pearson correlation test, shown in Table 5-10, show that the correlation between the caus e (defects of prerequisite work) and the effect factor (probability of management interference occurrence) is not significant at the 10% le vel with the r-value of 0.228 and the p-value of 0.363. The result of the 2-factor experiment is shown in Table 5-16.

PAGE 105

87 Table 5-16. 2-factor experi ment result for Design 3 Source DF Type III SS MS F-value P-value A 3 0.024098620.008032873.86 0.0382 B 1 0.014873840.014873847.14 0.0203 A*B 1 0.013203190.013203196.34 0.0270 The main effects of factors A and B were significant at the 5 % level; however, the interaction between factors A and B (A*B) was also significant at the 5% level with the p-value of 0.0270. The existence of the interaction effect can obscure the effect of factor A or B. In other words, the change in the mean response under one factor will depend on the level of the other factor. However, si nce the main effects of A and B were all significant at the 5% level in this experiment, it is conclusive that th e probability of other management interference occurrences caused by defects in the prerequisite work is significantly different among the f our projects. In addition, th e severity of defects has a significant effect on the occurr ence of management interferen ce because the mean of the probability of occurrence under modest (crew spen t less than 3 hours to fix it) severity of defects is not the same as that of the pr obability of occurrence under severe (crew spent over 3 hours) severity. 5.6.5 Effects of Prerequisite Work on Productivity Design 4 involves testing the effects of prerequisite work on the variability of productivity. Again, in cases of defects from prerequisite work, the pavement crew had to correct the problems before starting to cons truct a new layer of pa vement. When they had to work on something unplanned, or when they had to work to fix problematic end results, the productivity suffere d, as was discussed in Chapter 4. For example, when the paving crew placed the first lift of structural course in th e SR-20 (Alachua) project, the crew had to place overage material due to the finishing grade for the base not being at the

PAGE 106

88 proper elevation. Having to adjust the lif t thickness during the operation resulted in a loss of productivity. The result of the Pearson correlation test in Table 5-10 shows that the hours they spent fixing the problem has no significant correlation to th e daily productivity rates at the 10% level. However, the correlation was still very high with the r-value of -0.391 and p-value of 0.015 (significant at the 20% level). The r-value can be translated into meaning that the work hours they spent fixing defects of prerequisite work explains the 39.1% of variation of the same day productivity. The results for Design 4 in Table 5-17 s how that the interac tion effect between factors A and B is not significant. The main ef fect of factor A is si gnificant (conclusive), meaning that productivity rates after prereq uisite work interference occurred were significantly different from proj ect to project. The main e ffect of factor B is also significant (conclusive). It means that th e mean productivity rates were different according to the severity of th e interference and that the seve rity of prerequisite work interference was a significant factor in the va riability of productivity after prerequisite work occurred. Table 5-17. 2-factor experi ment result for Design 4 Source DF Type III SS MS F-value P-value A 3 1.829001730.609667242.55 0.0585 B 1 4.717048344.7170483419.77 <.0001 A*B 3 0.4682427 0.1560809 0.65 0.5819 5.7 Summary Statistical analyses discussed in this chapter showed the following results. Two pavement projects located in urban ar eas had higher produc tivity, compared to the other two projects in rural areas.

PAGE 107

89 The productivity associated with any interf erence is significan tly lower than one with no interference. The productivity means are ranked as multiple, management, work content, weather, and no interference in order from the lowest to the highest when they are compared by their association with types of interference. The mean productivity value when weather interference occurred was the highest among others because the crew usually chos e not to work when inclement weather was expected. Thus, only 14 work days were affected by rain. The mean value of management interferen ce was the lowest as a single interference factor, and poor management was the most frequent interference among others (65 out of 253 work days). Within management interference, the m ean productivity rates are ranked as work congestion, prerequisite work , equipment, and material in order from the lowest to the highest even though the differences among them were not significant at the 10% level. The mean value of defects from prerequis ite work has the most frequent occurrence among other types of management inte rference (41 out of 65 work days). The amount of rainfall in proj ect areas has little or no in fluence on either number of work days or NLA influenced by rain fall; however, the correlation increased significantly after truncating mild outliers. As the inefficient hours by weather in terference increases, the probability of management interference occurr ence tends to increase as well. The inefficient hours caused by weather inte rference have significant effects on the productivity variation, a nd as the inefficient hours increase, the productivity decreases. The correlation between the defects of prer equisite work and the probability of management interference occu rrence is not significant. The correlation between the hours the paving crew spent fixing the defects and the variability of daily productivity rates was significant at the 20% level.

PAGE 108

90 CHAPTER 6 ALTERNATIVE METHOD FOR PR ODUCTIVITY ESTIMATION 6.1 Introduction Computer simulation methods were devel oped in the early 1960s and have been commonly used as analytical t ools of management science. One of the most essential objectives of employing simulation methods is finding the optimal solution for a production system. To accomplish the objectiv e, the simulation methods are used as a vehicle for experimentation, often by trial a nd error in order to demonstrate the likely effect of various factors in th e system. In this chapter, th e method developed to estimate the productivity of pavement ope rations with various types of interference is discussed. The results of the productivity estimation were validated through four case study projects investigated. 6.2 Simulation Method for Productivity Estimation Productivity estimation using discrete event simulation involves three steps in general. First, a model of the production system of interest is built; ei ther flow charts or a set of instructions that disp lays the logic of the producti on process can be employed. Second, computer programs ar e written that “embody the mode l to imitate the system’s behavior when subject to a variety of operating policies su ch as resource allocation and utilization” (Pidd 1988). Last, the simulation is repeatedly implemented with the various factors in order to observe the effects of various combinations of the factors.

PAGE 109

91 6.2.1 Simulation of Pavement Production Process The main objective of this chapter is to develop a methodology to quantify various interference factors and their ca uses and apply them to the pr oductivity estimation. Thus, the existing simulation tools were employed to model the proce ss of the pavement operation instead of writing a nother computer program to re present the same process. Figure 6-1 illustrates the simulation model written in CYCLONE for representing the general process of the asphalt paving operation (Halpin and Rigg 1992). Figure 6-1. General proce ss of asphalt pavement MicroCYCLONE and DISCO support the form of input file as shown in Figure 6-2. The input file is the means by which the user translates a graphical model into a problemoriented language (POL), which can be unde rstood by MicroCYCLONE. The input file consisted of general information, network input, duration input, and resource input. General information, as shown in the first li ne of the file, contains the name of the network and program parameters such as the length of simulation run and the number of cycles. The network input segment is used to enter the actual pr ocess network. Each

PAGE 110

92 statement of this segment specifies a networ k element, its attri butes, and its logical relationship to other elemen ts in the network. Figure 6-2. Example of MicroCYLONE input file Duration input segment defines the duration t ype of each task represented in either COMBI or NORMAL elements and the parame ters of the distribution. The time durations should be assigned by e ither a deterministic method or a statistical distribution. The deterministic method assigns a fixed value for each work task, as shown in Table 6-1. This method is used when the duration of th e task is subject to small variations. A broadly defined work task that is subject to large variation should be broken down into a set of sub-tasks that can be evaluated according to the deterministic method. The subtasks of the asphalt pavement operation are good candidates for this method because they are equipment-intensive with short time variation.

PAGE 111

93 Using statistical distributions for s ub-task durations requires tremendous investigation of time and money. In order to use the dist ribution for the paving operation for example, each cycle time duration of each sub-task should be measured. The measurement requires field observations with at least four people at the same time with a stop watch—one in the plant, one beside the paver, one beside brea kdown roller, and one beside the finishing roller. This method was not feasible for this research. The researcher, therefore, surveyed actual work task durations from four projects investigated with the help of project personnel. Each proj ect inspector from each project provided the median value for those time duratio ns as accurate as possible. The survey form was presented in Appendix A, and the resu lt of the survey is shown in Table 6-2 and 6-3. Table 6-1. Deterministic method for work task time duration Node No. Work Task Duration (min) 2 Load at plant 5 3 Travel to job 10 5 Reposition for new pass 22 6 Return 9 9 Dump to spreader 2 10 Spread 12 15 Compact asphalt section 35 18 Finish section 20 Table 6-2. Results of survey for time duration Node No. Unit SR-20 (Alachua) SR-102 SR-20 (Putnam) I-10 Median 2 min 5 5 5 6 5 5 min 25 15 30 20 23 9 min 12 8 8 10 9 10 min 10 8 10 12 10 11 load 20 20 20 20 20 13 load 2 2 2 2 2 15 min 10 10 12 5 10 18 min 10 10 8 5 9

PAGE 112

94 Time durations for nodes 3 and 6 in Fi gure 6-1 involve hauling times by the delivery vehicles; the time varied based on the hauling distance between the designated asphalt plant and the site and traffic volume of each projec t. The maximum hauling times of all four projects were set to 45 minutes, FDOTÂ’s maximum hauling time. In addition to the hauling times, the researcher also surv eyed the average number of trucks used in each project, shown in Table 6-3. The tr iangular distribution (minimum, mode, and maximum hauling times) was used for nodes 3 and 6 because the minimum, mean, and maximum time durations were available. Table 6-3. Hauling times and number of trucks Projects Hauling Distance (mile) Hauling Time (min) Min, Mode, Max (min) No. of Trucks used SR-20 (Alachua) 15-17 45 40,45,45 8 SR-102 12.2 18 15,18,45 8 SR-20 (Putnam) 18 40 35,40,45 10 I-10 5 25 20,25,45 6 Finally, the resource input determines the number of units of each resource type to be used in the network. Each resource type includes equipment, labor, or materials, and the resources utilized at each task are entered at corres ponding QUE nodes, as discussed in Chapter 2. To initialize the simulation, the QUE node that will be the starting point for the simulation needs to be defined at the firs t line of the resource input segment. For example, 8 trucks, 1 spreader, 1 breakdown rolle r, and 1 finish roller were used in the input file shown in Figure 6-2. 6.2.1 Result of Simulation us ing Initial Time Duration The time duration of each sub-task is assume d to have small variation if little or no interference occurs during the production pr ocess. Based on this assumption, the

PAGE 113

95 pavement operation process was simulated using MicroCYCLONE for the four case study projects. The researcher entered the initial time durations shown in Table 6-2 to the corresponding nodes of sub-tasks. The actual number of de livery trucks employed was also entered to node 7 in each project. Th erefore, the deterministic method of time duration was used for all subtasks except th e hauling sub-tasks, which were modeled by triangular distribution in nodes 3 and 6. Then, the process models for each proj ect were simulated for 600 minutes in simulation time to estimate the daily producti vity assuming 10-hour work days. Due to the steady-state environment of the model, this simulation result shows the productivity rates when no interference and daily shift e ffect occur. In other words, the simulation result include the daily shift effect because the simulation was stopped after 10 simulation hours to model just one work day. Shift effects—defined as the effect that a certain work schedule has on what work can be started—reduces productivity by some degree. Previous research suggested that, to estimate the daily shif t effects for an 8-hour working day, the simulation run be stopped every 480 minutes in simulation time. The resources remaining in the simulation netw ork at the end of a work day (every 480 minutes of simulation time) should be then reassigned to their in itial nodes. This approach simulates the “start-from-scratch” rule, by disregarding all the tasks left unfinished on the previous working day. Th e result of this approach showed that the productivity of asphalt pavement operation wa s reduced up to 17% (Huang and Halpin 1995). The daily shift effect usi ng this rule is app lied to the new projects, and the results are evaluated in Chapter 7. The simulation resu lts are shown in Figures 6-3, 6-4, 6-5, and 6-6.

PAGE 114

96 The productivity rates in Fi gures from 6-3 to 6-6 were calculated as shown in Equation [6-5]. The number of units produced was multiplied by 5 because in the process model, each flow unit (truck load ) was accumulated by 5 at the COUNTER node (node 20 in Figure 6-1, and row 20 in Figure 6-2). Thus, each unit produced included 5 truck loads. The simulation time was then di vided by 60 in order to convert it to an hourly rate. [6-5] System Productivity = 60 / 5 Time Simulation Total x produced Units Figure 6-3. Initial simulation result of SR-20 (Alachua) Figure 6-4. Initial simulati on result of SR-102 (Duval)

PAGE 115

97 Figure 6-5. Initial simulation result of SR-20 (Putnam) Figure 6-6. Initial simu lation result of I-10 The simulation results calculated were 5.48, 7.47, 6.45, 6.00 loads per hour for the SR-20 (Alachua), SR-102, SR-20 (Putnam), a nd I-10 projects, respec tively. Since the productivity results obtained by the simulation models indicate ideal productivity when no interference or daily shift-e ffect occurred, the results were compared to the actual daily productivity results of the days wh en no interference was recorded. The comparisons of each project are presented from Figures 6-7 to 6-10. The horizontal lines show the simulation results, baseline productiv ity (dotted line), the third quartile, median, and the first quartile of actual productivity from top to bottom. Table 6-4 shows the actual values for each project. The fluctuation of actual productivity is al so shown for each project. As discussed in Chapter 2, the natural variation can be th e source of variation in daily productivity. Natural variation is either a variation that cannot be explai ned or that is caused by chance or common cause. For example, a crew can pe rform at a different productivity level even

PAGE 116

98 though they perform a repetitive work task and no external interfer ence factors affected their work. 0.00 1.00 2.00 3.00 4.00 5.00 6.00 123456789101112131415161718 No interference daysProductivity Figure 6-7. Productivity comparison for no interference in SR-20 (Alachua) 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 12345678910111213141516171819 No interference daysProductivity Figure 6-8. Productivity comparison fo r no interference in SR-102 (Duval) 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 16111621263136 No interference daysProductivity Figure 6-9. Productivity comparison fo r no interference in SR20 (Putnam)

PAGE 117

99 0.00 2.00 4.00 6.00 8.00 10.00 12.00 1112131415161718191101111121 No interference daysProductivity Figure 6-10. Productivity comparis on for no interference in I-10 Table 6-4. Comparison of simulation results Project Min. 1st Qua. Median 3rd Qua. Max. IQR Upper limit Base. Sim. Results SR-20 (Alachua) 2.84 1.82 2.53 3.96 5.57 2.15 7.18 5.57 5.48 SR-102 3.88 2.50 3.35 5.24 6.66 2.75 9.36 6.44 7.47 SR-20 (Putnam) 2.65 1.72 2.72 3.31 6.15 1.60 5.70 5.05 6.45 I-10 2.73 1.18 2.53 3.66 10.82 2.49 7.40 5.89 6.00 Unit: loads per hour For all four projects, the simulation results were higher (optimistic) than the values of the third quartile of the act ual productivity rates. As expl ained in Chapter 5, IQR is calculated by the difference between the values of 3rd and 1st Quartile. The value of upper limit was calculated by multiplying 1.5 to IQR and adding the calculated value to the 3rd Quartile value. Only the simulated resu lt of the SR-20 (Putnam) project is outside of the upper limit. This is because the pr oject has the smallest variation of actual productivity rates among the four projects; IQ R of the project was 1. 60. Figure 6-9 also shows the result. The simulated productivity of the SR-102 pr oject was outside of the range of actual rates because the time durations obtained from the SR-102 project we re shorter than the ones from other projects. For example, the dur ations for spreader reposition (node 5) and

PAGE 118

100 spreading (node 10) took 15 and 8 minutes ev en though the median values of those tasks for the four projects were 23 and 10 minutes, as seen in Table 6-2. The SR-102 project also had the shortest haul time from the plant to the site among four projects, as seen in Table 6-3. There are two main reasons that the estimated results by simulation were more optimistic than the actual ones even though no interference was recorded on those days. Firstly, the natural variation can be the s ource of variation in da ily productivity. As explained, natural variation is either a varia tion that cannot be explai ned or that is caused by chance or common cause. A crew can perf orm at a different productivity level even though they perform a repetitive work task and no external interfer ence factors affected their work. Research showed that regardless of the causes of varia tion, it causes a longer cycle time of each task and reduce the effici ency of performance (Ballard and Howell 1997; Horman and Kenley 1998) . Secondly, the surveyed time durations were more optimistic than the actual duration; even though the data we re as close to the median tim e duration as possible, they do not include all auxiliary times actually spent for each sub-task. For example, when the researcher made site visits, it was observed that the paver was waiting for the material delivered for less than 20 minutes. Since it wa s not a major delay as far as the inspectors were concerned, that incident was not pr ovided to the researcher. Therefore, the additional time associated with each sub-task should be included in its time duration in order to make the deterministic method more reliable. Especially when any type of interference influences the performance of a sub-task and delayed it, the additional time should be included in the duration of each task.

PAGE 119

101 6.2.2 Quantification of Interference for Alternative Time Duration As was discussed in Chapter 4, the res earcher (1) collected production data from four pavement projects, (2) calculated daily prod uction rates by using number of truck loads converted by conversion factors as th e output and crew hours as the input, (3) calculated mean, cumulative, and baseline pr oductivity for each project, (4) categorized types of interference by three major factors of management, weather, and work content, with the description of causes for the interf erence, and (5) calculated inefficient hours by projecting baseline productivity on those days when interference occurred. 6.2.2.1 Further categorization of interference The interference factors are further analyzed in this ch apter to quantif y the order of magnitude to which each factor affects each sub-task of operation. Table 6-5 shows the three sub-tasks, the major categories of type s of interference possibly associated with each sub-task (CUpper case), and th e possible causes (cLower case). Loading sub-task, for example, can be affected by weather or management interference (C1 or C2) because work conten t (C3) was defined as the work areas in which the crew had to spend more time than in mainline areas such as turning lanes, driveways, and ramps. Working on those areas has little or no effect on loading material into the paver. Also, material delay (c24) , material failure (c25), and work congestion (c27) occur only on the site, and they can not affect on loading task. For example, if there was a material delay or failure on the site, those types of interference affected spreading and compacting tasks. Because loading is the predecessor of the other two tasks, the interference that occurred before the paving operation began a ffected only the loading task most of the time unless the interference influenced the pr oductivity of the other two succeeding tasks.

PAGE 120

102 For example, if the rain (C 1-c11) interrupted the produc tion before the crew began working (crew had to stop working), and it also delayed the pr oduction of spreading (crew had to slow down), the rain interference affected the spreading and compacting tasks. Table 6-5. Categories and cause s of interference on sub-tasks Sub-task Category (C) Cause (c) Weather C1 Rainfall c11 Temperature c12 Management C2 Prerequisite work (before production begins) c21 Equipment (breakdown) c26 Loading Material failure (plant problem) c25 Weather C1 Rainfall c11 Temperature c12 Management C2 Prerequisite work c21 Out-of-sequence c22 Operator/ Labor c23 Material delay c24 Material failure c25 Equipment (breakdown) c26 Work Congestion c27 Accident c28 etc. c29 Spreading Work Content C3 Piecemeal work (Turning lane, driveway, etc.) c31 Weather C1 Rainfall c11 Management C2 Prerequisite work (Rework) c21 Out-of-sequence c22 Operator/ Labor c23 Material failure c25 Equipment (breakdown) c26 Work Congestion c27 Accident c28 etc. c29 Compacting Work Content C3 Piecemeal work c31 6.2.2.2. The order of magnitude for interference The method of quantifying types of interfer ence is shown in Table 6-6 and Table

PAGE 121

103 6-7. Columns (A) and (B) in Table 6-5 show the dates when interference occurred. Columns (C) and (D) record the output and input amount for the production of the day. Column (E) was the calculat ed value, meaning that wh en baseline productivity would have been achieved on that day, the crew could have spent 0.87 hours less than 6 hours, shown in column (D). Column (F), additional time per load (ATPL) was calculated by wasted time (E) divided by output of the day (C ). Column (G) represents total number of loads affected (TNLA) across three sub-tasks and is calculated by adding columns (2), (6), and (10) from Table 6-7. The order of magnitude for each type of interference was measured by either hours affected or number of loads affected (NLA ). The researcher recorded actual hours affected by each interference in columns (1), (5), and (9) in Table 6-7 when the actual hours were provided by the proj ect inspectors. Then baseline productivity of the corresponding project (loads/hr.) was multiplie d by hours affected in order to calculate NLA in columns (2), (6), and (10) of Table 6-7. The actual NLA were provided when certain types of interference occurred. For example, when material was rejected by the in spector on the job sites, or when the crew worked on the areas defined as work cont ent, the actual NLA were provided by the inspectors. Table 6-6. Method of quantifying ATPL on work day Work day (A) Date (B) C. Loads (C) Crew hour (D) Wasted (E) ATPL (F) TNLA (G) Category (H) Cause (I) A B/BB/BB 20.60 6 0.87 0.0042 40 C2 c21

PAGE 122

104 Table 6-7. Method of quantif ying NLA and AT on sub-tasks Loading Spreading Compacting Hours affected (1) NLA (2) AT (3) Cause (4) Hours affected (5) NLA (6) AT (7) Cause (8) Hours affected (9) NLA (10) AT (11) Cause (12) The additional times (AT) associated with each sub-task, shown in columns (3), (7), and (11), were calculated by prorating ATPL by the proporti on of NLA from TNLA. The calculations are summarized by Equations [6-1] to [6-4]. [6-1] ATPL= Loads C hours Wasted . [6-2] TNLA= NLA (Loading) + NLA (Spreading) + NLA (Compacting) [6-3] NLA= Hours affected x baseline productivity [6-4] ATi subtask = ATPL x TNLA NLAi subtask After recording the forms shown in Tabl e 6-6 and Table 6-7 for each project, NLA in columns (2), (6), and (10) in Table 6-7 were categorized by each sub-task and by the interference cause. By doing so , the effects of interferen ce were quantified by NLA on each sub-task. In short, the magnitude of inteference wa s either calculated or actual number of load affected (NLA). For example, when th e crew had to stop working for three hours by an interruption factor, NLA was calculate d by multiplying three hours to the baseline productivity of the project. When the actua l NLA was provided, for instance, number of loads that had material failure or number of loads worked on work content areas, the actual NLA was used.

PAGE 123

105 6.2.2.3 Example result of interference quantification The researcher obtained the result shown in Table 6-8 by categorizing types of interference by their a ssociation with the sub-task and quantifying them by NLA. Table 6-8 shows only the loading task from one of the case study projects. NLA for each category (B) and cause (E) was calculated. Total number of loads completed for the project was 1,365, and the probability of each category and cause was also calculated by the proportion of NLA. For example, NLA by C1 (weather) was 77.41, and it took 5.7% of the total number of loads affected. Ou t of 77.41 NLA by C1, c11 (rain) affected 58.09 loads (75%), and c12 (temperature) affected 19.32 loads (25%). Times in column (F) were calculated by taking the mean value of AT in column (3) from Table 6-7. Table 6-8. Loading example for quantifying interference Category (A) NLA (B) Probability (C) Cause (D) NLA (E) Probability (F) Time (hr) (F) Time (min.) (G) C1 77.41 0.057 c11 58.09 0.75 0.19 11.34 c12 19.32 0.25 0.22 12.97 C2 469.97 0.344 c21 395.91 0.84 1.44 86.29 c26 74.06 0.16 0.22 13.30 C3 0.00 0.000 c31 0.00 0.00 0.00 0.00 6.2.3 Development of Interference Model by Timed Petri-Net (TPN) General aspects of Petri-Net were discusse d in Chapter 2. As discussed, Timed Petri-Net (TPN) has been widely used becau se it has the capabili ty of delaying firing tokens in transition to represent task durati on. A transition can delay firing by either deterministic or statistical distribution, as CYCLONE does. Be sides, TPN provides a function of probabilistic arc. Probabilistic arcs utilize the function of probabilistic occurrence of events to mode l uncertainty in the activity nexus. For example, the equipment breakdown that occurre d 2% of the time in an ac tivity can be modeled by the

PAGE 124

106 arc by assigning the probability. These functio ns of TPN incorporate risk and uncertainty factors in work task durati on and project scheduling. Based on the results of cate gorization and quantification of types of interference, the researcher developed an interference simulation model by using TPN. The model, shown in Figure 6-11, includes three layers to represent category, cause, and AT, respectively. Figure 6-11. Conceptual model of TPN for alternative time duration A flow unit of the simulation model is a token in the “start” place, representing each truck load of a sub-task. T1 delays the firing of the token for the initial time duration of each sub-task obtained from the su rvey shown in Table 6-2. Three places (C1, C2, and C3) in the first layer (first dotted line from the left) represent three categories of interference, followed by each cau se in the second layer (second dotted line from the left). AT for each interference cause is included in the transitions within the dotted box. The model uses probabilistic arcs between places and transitions in the first and second layers to represent probabilistic distribution of interference occurrence by

PAGE 125

107 using “conflict” function of TPN models. Again, “Conflict” function means that when a token travels from an input place to multiple output transitions, firi ng of the token to any transition disables the other two transitions even though all transitions are enabled. By assigning probabilities calculated in Table 6-9 to each transition, the token travels through three transitions according to the pr obability to model three categories of interference. This confliction function is al so used to model each cause of interference. By building and implementing the model for eac h sub-task, alternative time durations can be calculated by adding additional time of each interference cause to the initial time duration. 6.2.4 Result of TPN for Alternative Time Duration The researcher used Visual Simnet (version 1.37) to build and implement the interference model for each sub-ta sk from four pavement proj ects. Figure 6-12 shows the screenshot of modeling interference for loading the sub-task. The token initially assigned in the first place (P13) travels through the ne t according to the given rules of delay (work task duration) and probabilistic arc. Unlik e the initial model shown in Figure 6-11, Figure 6-12 has four categorie s of interference because it includes “No category of interference (NC)” on top. Figure 6-12. Example of TPN mode l for sub-task interference

PAGE 126

108 The firing of the token is de layed from the first transition (T1) for initial time duration of the task given in Table 6-2. Once the token passes T1, it travels through either one of C1, C2, C3, or NC (no inte rference). The number of passes among four categories depends on the probabilistic dist ribution in the arcs, assigned by the calculation from the previous step (Table 6-9). For example, the total number of loads completed was 1,365. The total number of loads should match the number of events shown at the bottom of Figure 6-12 multiplied by 12 (number of events for one cycle of the model). Total number of loads is the same as total number of cycles that the simulation has to be run. Out of 1,365 load s (cycles), the token will theoretically travel 77.41 times through C1, 469.97 times through C2, zero through C3, and the rest of the time (817.62) through NC, as shown in Table 6-9. In the second layer, each category of in terference is broken down by its cause, based on the probability distribution. The toke n travels by the same distribution rule. No time delay was entered into the transitions of first and second layers because they only work as a function of probabilistic distributi on. Once the token reaches the third layer for AT (T10 though T14 and T20), the transitions de lay the firing of the token for a given duration. The token comes back to P13, where it began traveling, and this is considered to be one cycle. Again, the token travels 1,365 cycles—equivalent to the total number of truck loads completed—and the number of tr uck loads determines the simulation time, shown in the bottom of Figure 6-12. The simulation model needs to be built a nd implemented for each sub-task, and as the result, the alternative time duration of each sub-task is calculated. The alternative

PAGE 127

109 time duration is basically the initial time duration of a sub-task added by additional time duration caused by each in terference factor. 6.2.5 Validation of Alternative Time Duration from TPN The summary result of the simulation mode l in Figure 6-12 is presented in Table 610, and the full version of the result can be found in Appendix H. The simulation result shows that the total time elapsed to run 1, 365 cycles was 40,140 (min). Column (A) in Table 6-10 shows the categories and causes in the model, and column (B) shows total number of times that the token passed each category and cause. Column (C) shows the average number of passes per unit time. For example, the token passed C1 (Weather interference) 77 times for the time period of 40,140. Seventy-seven divided by 40,140 yields 0.00192. Service distance in column (D) is the average time duration per pass. The service distance in t17 yiel ds the alternative time duration for the sub-task because the token must pass through t17 after passing the category and cause in each cycle. Thus, the service sum of t17 is the same as the tota l number of cycles, as shown in column (B) in Table 6-9 Table 6-9. Result summary for TPN simulation Transition Name (A) Service Sum (B) Service/ Time (C) Service Distance (D) C1 77 0.00192 520 C2 478 0.0119 83.6 C3 2 5.85E-05 402 c11 57 0.00142 697 c12 20 0.000499 1765 c21 403 0.01 98.9 c26 75 0.00186 524 t17 1365 0.034 29.4 NI 808 0.0201 49.6 Note: Total time: 40,140 and Event: 16,380

PAGE 128

110 Table 6-11 and Table 6-12 present the results of the simulation with the quantification data. Columns (A), (B), and (C ) come from the actual data as presented in Table 6-9, and columns (D) and (E) show the result of TPN simulation. Also, the simulation result shows (Appendix H) that th e token passed through the transition NI 808 times, and the probability of no interference was 0.591. The results from the simulation model in columns (D) and (E) are very clos e to those assigned from the results of quantification in columns (B) and (C). Therefore, the result of this simulation is concluded to be valid. Based on the validity of the simulation model developed, 29.4 minutes per cycle is used as the alterna tive time duration for th e sub-task when the paving crew experienced the give n amount of interference. The researcher also built and implemente d TPN models for the other two sub-tasks to calculate the alternative time duration of each sub-task. Then, the process was repeated for all the projects investigated. Table 6-10. Validation of TPN m odel for interference categories Category (A) NLA (B) Probability (C) NLA passed through categories (D) Probability in Simulation (E) C1 77.41 0.057 77 0.056 C2 469.970.344 478 0.350 C3 0.00 0.000 0.00 0.000 Table 6-11. Validation of TPN model for interference causes Cause (A) NLA (B) Probability (C) NLA passed through causes (D) Probability in Simulation (E) c11 58.09 0.750 57 0.740 c12 19.32 0.250 20 0.260 c21 395.91 0.840 403 0.843 c26 74.06 0.160 75 15.69 c31 0.00 0.000 0.00 0.00

PAGE 129

111 6.3 Simulation of Pavement Process with Alternative Time Duration Once alternative time durations were calcula ted as the results of TPN simulations, the durations were applied to the pavement pr ocess model as discussed in Section 6.2.1. Again, DISCO (Huang 1995) a nd MicroCYCLONE (Halpin a nd Riggs 1992) were used as a simulation tool because DISCO uses MicroCYC LONE to generate its input files. The result of the pavement process simulati on by using alternative time durations for each sub-task is the productivity rate of each case study project. Once the productivity rates were obtained from each project, the results were analyzed for further validation. The resear ch hypothesis for validation was that if the productivity results, which incorporate inte rference by using altern ative time durations, fall in the range between baseli ne and cumulative productivity rates, then the results are valid. As discussed in Chapter 4, base line productivity is theo retically the best productivity that the pavement crew can achieve when little or no management interference occurs. Cumulativ e productivity, on th e other hand, is th e productivity that the crew performed with a ll occurrences of in terference included. Even though the productivity obtained from the process simu lation captured all inte rference categorized and quantified from the production data collected, it may not be the same as the cumulative productivity. The results of pr ocess simulation and their validations will follow in the next section by reporting results fr om four pavement project case studies. 6.4 Case Study 1: SR-20 (Alachua) The results of interferen ce quantified from the SR20 (Alachua) project are summarized in Tables 6-12 to 14 for each sub-task. The original production data that contain interference factors ar e also shown in Appendix F. The production data followed the methods described; for example, the method for calculating AT for each cause of

PAGE 130

112 interference was described with the examples in Table 6-6 and Table 6-7 and Equations [6-1] to [6-4]. As the product of this met hod, the probabilities of each category and cause and AT were calculated. Table 6-12. Probability a nd NLA of interference for loading (Case Study 1) Category NLA % Cause NLA % Time (hr) Time (min.) C1 52.92 0.066 c11 38.99 0.740.36 21.67 c12 13.93 0.260.89 53.44 C2 177.81 0.223 c21 105.270.590.59 35.28 c26 52.45 0.290.65 39.25 c27 20.09 0.110.19 11.41 C3 0.00 0.000 c31 0 0.000.00 0.00 NI 566.54 0.711 Table 6-13. Probability a nd NLA of interference for spreading (Case Study 1) Category NLA % CauseNLA % Time (hr) Time (min.) C1 8.36 0.010 c11 8.36 1.000.02 1.18 C2 79.02 0.099 c21 79.02 1.000.15 8.89 C3 109.00 0.137 c31 109.001.000.47 27.97 NI 600.89 0.754 Table 6-14. Probability a nd NLA of interference for compacting (case Study 1) Category NLA % CauseNLA % Time (hr) Time (min) C1 0.00 0.000 c11 0.00 0.000.00 0.00 C2 68.16 0.085 c21 68.16 1.000.10 6.14 C3 109.00 0.137 c31 97.00 1.000.48 28.63 NI 620.10 0.778 The results of the proba bilities and AT were pl ugged into the TPN model developed for each sub-task in order to estimate the alternative time duration. The TPN models for loading, spreading, and comp acting sub-tasks of the SR-20 (Hawthorne) project are shown Figure 6-13, Figure 6-14, and Figure 6-15, respectively. The TPN model in Figure 6-12 incorporates the resu lts shown in Table 6-12 to estimate the alternative time for loading. The initial time duration for lo ading, which can be found in Table 6-2 as the result of a survey, is entered into the transition of t1, and all interference

PAGE 131

113 categories and causes are modeled by their pr obabilities and AT. Note that the total number of events is 9,576, which simulate s the total number of 798 cycles (9,576/12). Total number of cycles in the TPN model is equivalent to the total number of loads completed. Figure 6-13. TPN model fo r loading (Case Study 1) Figure 6-14. TPN model for spreading (Case Study 1)

PAGE 132

114 Figure 6-15. TPN model for compacting (Case Study 1) The alternative time durations for loading, spreading, and compacting were 11.4, 15.1, and 15 minutes, as shown in Appendix G. Compared to the initial time durations (5, 10, and 10 minutes) for each sub-task, seen in Table 6-2, the alternative time durations were increased by 2.28, 1.51, and 1.5 times , respectively. Once the alternative time durations for each sub-task were obtained from the results of TPN simulation, the durations were entered into the DISCO/Mi croCYLONE simulation model in order to estimate the productivity of pavement operati ons when the given amount of interference occurred in the process. The simulation results are shown in Figure 6-16 and Figure 617. The detailed productivity result is shown in Appendix H, and it is calculated to be 5.31 loads per hour by Equation [6-5]. The productivity obtained by simulating in DISCO can be calculated by the same equation as in MicroCYCLONE. Total numbe r of loads (shown in the COUNTER node in Figure 6-17) was multiplied by 5 and then divided by the simulation time converted to an hourly rate. It returned the same result of productivity (5.31 load s per hr.) as the one estimated by MicroCYCLONE.

PAGE 133

115 Figure 6-16. Productivity simulation result from MicroCYLCONE (Case Study 1) Figure 6-17. Productivity simulation re sult from DISCO (Case Study 1) Again, alternative time durations for load ing, spreading, and compacting sub-tasks were entered into COMBI node 2, NORMA L node 10, and COMBI nodes 15 and 16, seen in Figure 6-17. The number of material delivery trucks utilized in this project was entered in the QUE node 7 and is shown in Ta ble 6-3. The hauling times of the trucks were entered in NORMAL nodes 3 (deliver materi al) and 6 (return). The researcher also obtained the average number of truck loads th at the spreader repositioned to make a new pass parallel to the just-com pleted pass and the average number of truck loads that the

PAGE 134

116 spreader section released to the roller. Th ey were entered into function nodes 11 and 12, respectively. Finally, the productivity estimated by the simulation was compared to the productivity parameters such as cumulative a nd baseline productivity in order to validate the result. The simulation result has two valu es: the value when the initial time durations were used and the value when alternative time durations were substituted for the initial ones. Intuitively, the productivity result that included initial time durations is assumed to be more optimistic because it was run w ithout any regard for interference. Daily productivity in Figure 4-1 from Chapter 4 shows the fluctuation of productivity. The cumulative productivity, wh en calculated by Equation [2-1], was 2.91 loads per hour. The baseline productivity was 5.57 loads per hour. Figure 6-18 shows the productivity plot for those parameters. Cumulative productivity was calculated for every working day as the time elapsed, and it showed that there was some degree of a learning curve at the beginning of the pr oduction. As shown in Figure 6-18, the simulation results with initial time durations were very optimistic, even higher than the baseline productivity; the productivity results with a lternative time durations, however, fall between baseline and cumulative productiv ity. Based on the research hypothesis, the productivity result simulated with alternative ti me durations is concl uded to be valid, and the method of categorizing and quantifying type s of interference was effective when the method was used for the purpose of simulation.

PAGE 135

117 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 1611162126313641 Work dayProductivity (load/hr) Project sim. Baseline productivity Cumulative productivity Daily productivity Improved Figure 6-18. Productivity anal ysis for SR-20 (Alachua) 6.5 Case Study 2: SR-102 (Duval) The results of interferen ce quantified from the SR -102 (Duval) project are summarized in Tables 6-15 to 6-17 for each sub-task. The original production data that contain interference factors are also shown in Appendix F. The results of the proba bilities and AT were pl ugged into the TPN model developed for each sub-task in order to esti mate alternative time duration. The TPN models for loading, spreading, and comp acting sub-tasks of the SR-20 (Hawthorne) project are shown Figure 6-19, Figure 6-20, and Figure 6-21, respectively. The total number of events was 16,380, which simu lates the total number of 1,365 truck loads completed.

PAGE 136

118 Table 6-15. Probability a nd NLA of interference for loading (Case Study 2) Category NLA % CauseNLA % Time (hr) Time (min.) C1 77.41 0.057 c11 58.09 0.75 0.19 11.34 c12 19.32 0.25 0.22 12.97 C2 469.97 0.344 c21 395.91 0.84 1.44 86.29 c26 74.06 0.16 0.22 13.30 C3 0.00 0.000 c31 0 0.00 0.00 0.00 NI 817.62 0.599 Table 6-16. Probability a nd NLA of interference for spreading (Case Study 2) Category NLA % CauseNLA % Time (hr) Time (min.) C1 0.00 0.000c11 0.00 0.00 0.00 0.00 c12 0.00 0.00 0.00 0.00 C2 42.76 0.031c24 10.00 0.23 0.05 3.30 c25 7.00 0.16 0.01 0.72 c26 25.76 0.60 2.77 165.92 C3 84.00 0.062c31 84.00 1.00 0.08 4.59 NI 1238.24 0.907 Table 6-17. Probability a nd NLA of interference for compacting (Case Study 2) Category NLA % CauseNLA % Time (hr) Time (min.) C1 0.00 0.000c11 0.00 0.00 0.00 0.00 c12 0.00 0.00 0.00 0.00 C2 0.00 0.000c11 0.00 0.00 0.00 0.00 C3 84.00 0.062c31 84.00 1.00 0.09 5.42 NI 1281.00 0.938 Figure 6-19. TPN model fo r loading (Case Study 1)

PAGE 137

119 Figure 6-20. TPN model for spreading (Case Study 1) Figure 6-21. TPN model for compacting (Case Study 1) The alternative time durations for loading, spreading, and compacting were 29.4, 13.6, and 10.3 minutes, as shown in Appendix G. Compared to the initial time durations (5, 8, and 10 minutes) for each sub-task, seen in Table 6-2, the alternative time durations were increased by 5.88, 1.7, and 1.3 times, respectively. The simulation results by MicroCYCLONE and DISCO are shown in Figure 6-22 and Figure 6-23. The detailed productivity results are shown in Appendix H. The productivity rates from both simulation models we re calculated to be 5.08 loads per hour. The productivity obtained by simulating in DISCO can be calculated by the same equation as in MicroCYCLONE. Total numbe r of loads (shown in the COUNTER node in Figure 6-23) was multiplied by 5 and then divided by the simulation time converted to

PAGE 138

120 an hourly rate. It returned the same result of productivity (5.08 lo ads per hour) as the one estimated by MicroCYCLONE. Figure 6-22. Productivity simulation result from MicroCYLCONE (Case Study 2) Figure 6-23. Productivity simulation re sult from DISCO (Case Study 2) The productivity obtained from both the si mulation with initial time durations and with alternative time durations were compared with cumulative and baseline productivity. The baseline productivity was 6.59 loads pe r hour. Figure 6-24 shows the productivity plot for those parameters. The simulation resu lts with initial time durations were very optimistic, whereas the productiv ity results with a lternative time dura tions fall between the baseline productivity and cumulative productiv ity, as happened in Case Study 1. The

PAGE 139

121 simulation result with alternative time durations is concluded to be valid again in Case Study 2. 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00 161116212631364146 Daily productivity Cumulative productivity Project sim. Improved Baseline productivity Figure 6-24. Productivity analysis for SR-102 6.6 Case Study 3: SR-20 (Putnam) The results of interference for each sub-ta sk quantified from the SR-20 (Putnam) project are summarized in Tables 6-18 to 6-20. The original production data for those days when interference occurred ar e also shown in Appendix F. Table 6-18. Probability a nd NLA of interference for loading (Case Study 3) Category NLA % Cause NLA % Time (hr) Time (min.) C1 61.55 0.05 c11 21.19 0.340.59 35.61 c12 40.36 0.660.86 51.53 C2 396.40 0.29 c21 199.06 0.500.97 57.96 c22 40.26 0.100.90 54.05 c26 56.71 0.141.02 61.16 c27 100.37 0.256.26 375.68 C3 0.00 0.00 c31 0.00 0.000.00 0.00 NI 908.04 0.66

PAGE 140

122 Table 6-19. Probability a nd NLA of interference for spreading (Case Study 3) Category NLA % Cause NLA % Time (hr) Time (min.) C1 0.00 0.00 c11 0.00 1.000.00 0.00 C2 127.82 0.09 c24 89.64 0.700.37 22.35 c26 38.18 0.300.41 24.60 C3 76.00 0.06 c31 76.00 1.000.30 17.81 NI 1162.18 0.85 Table 6-20. Probability a nd NLA of interference for compacting (Case Study 3) Category NLA % Cause NLA % Time (hr) Time (min.) C1 0.00 0.00 c11 0.00 1.000.00 0.00 C2 73.07 0.05 c24 49.26 0.670.18 10.84 c26 23.81 0.330.22 13.12 C3 76.00 0.06 c31 76.00 1.000.26 15.87 NI 1216.93 0.89 The results of probabilities and AT we re plugged into the TPN model developed for each sub-task, and alternative time durat ions of each were estimated. The TPN models for loading, spreading, and compacti ng sub-tasks of the SR-20 (Putnam) project are shown Figure 6-25, Figure 626, and Figure 6-27, respectively. The total number of events was 16,392, which simulates the tota l number of 1,409 truck loads completed. Figure 6-25. TPN model fo r loading (Case Study 3)

PAGE 141

123 Figure 6-26. TPN model for spreading (Case Study 3) Figure 6-27. TPN model for compacting (Case Study 3) The alternative time durations for loading, spreading, and compacting were 40.5, 13.1, and 15.2 minutes, as shown in Appendix G. Compared to the initial time durations (5, 10, and 12 minutes) for each sub-task, seen in Table 6-2, the alternative time durations were increased by 8.1, 1.31, and 1.27 times, respectively. The simulation results by MicroCYCLONE and DISCO are shown in Figure 6-28 and Figure 6-29. The detailed productivity re sults are shown in Appendix H. The productivity rates from both simulation models we re calculated to be 3.68 loads per hour.

PAGE 142

124 Figure 6-28. Productivity simulation result from MicroCYLCONE (Case Study 3) Figure 6-29. Productivity simulation re sult from DISCO (Case Study 3) The productivity obtained by the simulation with the initial ti me duration was 7.65 loads per hour, which is higher than the baseli ne productivity of the project (5.05). The productivity with alternative time durations was plotted, as seen in Figure 6-30, to compare it with other productivity parameters . Figure 6-30 shows that the simulation results with initial time durations were very optimistic, whereas the productivity results with alternative time durations fall between baseline and cumulative productivity, as in Case Studies 1 and 2. The simulation result with alternative time durations is concluded to be valid.

PAGE 143

125 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 161116212631364146515661 Work dayC. Loads/hr Initial duration Baseline Interferences Cumulative Daily Figure 6-30. Productivity anal ysis for SR-20 (Putnam) 6.7 Case Study 4: I-10 (Escambia) The results of interference for each sub-ta sk quantified from the I-10 project are summarized in Tables 6-21 to 6-23. The or iginal production data for those days when interference occurred are also shown in Appendix F. Table 6-21. Probability a nd NLA of interference for loading (Case Study 4) Category NLA Prob. CauseNLA Prob. Time (hr) Time (min.) C1 72.46 0.03 c11 72.46 1.00 0.48 28.50 C2 586.74 0.20 c21 539.61 0.92 0.55 32.88 c26 47.13 0.08 0.08 4.94 C3 0.00 0.00 c31 0 0.00 0.00 0.00 NI 2222.80 0.77 The results of probabilities and AT for each sub-task were used by the TPN model, and alternative time durations of each were estimated. The TPN models for loading, spreading, and compacting sub-tasks of the I-10 project are shown Figure 6-31, Figure 6-

PAGE 144

126 32, and Figure 6-33, respectively. The total number of events was 34,584, which simulates the total number of 2, 882 truck loads completed. Table 6-22. Probability a nd NLA of interference for spreading (Case Study 4) Category NLA Prob. CauseNLA Prob. Time (hr) Time (min.) C1 0.00 0.00 c11 0.00 1.00 0.00 0.00 C2 191.06 0.07 c21 30.00 0.16 0.29 17.60 c24 102.88 0.54 0.35 21.23 c25 21.00 0.11 0.06 3.85 c27 37.17 0.19 0.17 10.50 C3 204.00 0.07 c31 204.00 1.00 0.23 13.82 NI 2486.94 0.863 Table 6-23. Probability a nd NLA of interference for compacting (Case Study 4) Category NLA Prob. CauseNLA Prob. Time (hr) Time (min.) C1 0.00 0.00 c11 0.00 1.00 0.00 0.00 C2 191.06 0.07 c21 30.00 0.16 0.29 17.60 c24 102.88 0.54 0.35 21.23 c25 21.00 0.11 0.06 3.85 c27 37.17 0.19 0.17 10.50 C3 204 0.07 c31 204.00 1.00 0.23 13.82 NI 2486.94 0.86 Figure 6-31. TPN model fo r loading (Case Study 4)

PAGE 145

127 Figure 6-32. TPN model for spreading (Case Study 4) Figure 6-33. TPN model for compacting (Case Study 4) The alternative time durations for loadi ng, spreading, and compacting were 13, 16.6, and 9.63 minutes, as show n in Appendix G. Compared to the initial time durations (6, 12, and 5 minutes) for each sub-task, s hown in Table 6-2, the alternative time durations were increased by 2.17, 1.38, a nd 1.93 times, respectively. The simulation results by MicroCYCLONE and DISCO are s hown in Figure 6-34 and Figure 6-35. The detailed productivity results are shown in A ppendix H. The productivity rates from both simulation models are calculated to be 5.41 loads per hour.

PAGE 146

128 Figure 6-34. Productivity simulation result s from MicroCYLCONE (Case Study 4) Figure 6-35. Productivity simulation re sults from DISCO (Case Study 4) The productivity obtained by the simulation with the initial ti me duration was 6.49 loads per hour, which is higher than the baseli ne productivity of the project (5.89). The productivity with alternative time durations was plotted in Figure 6-36 to compare it to other productivity parameters. As shown in Figure 6-36, the simulation results with initial time durations were very optimisti c, whereas the productivity results with alternative time durations fa ll between baseline and cumula tive productivity, as was true in previous case studies. The simulation result with alternative time durations is concluded to be valid.

PAGE 147

129 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00 16111621263136414651566166717681869196 Work daysC. Load/hr Project Baseline productivity Cumulative productivity Daily productivity Improved Figure 6-36. Productivity analysis for I-10 6.8 Result Summary Figure 6-37 shows the overall productivity estimation resu lts for the four case study projects. The productivity estim ated by using alternate time dur ations all fall in the range between baseline and cumulativ e productivity. Therefore, the method developed to categorize and quantify interference is valid when it is used to estimate construction productivity by computer simulation.

PAGE 148

130 0 1 2 3 4 5 6 7 8 9 10 SR-20 (Alachua) SR-102SR-20 (Putnam) I-10 Initial Time Duration. Baseline Alternative Time Duration Cumulative Figure 6-37. Overall produc tivity estimation results from the four case studies In Chapter 5, the mean productivity rates of the four projects were ranked I-10, SR102, SR-20 (Alachua), and SR-20 (Palatka), in order from the highest to the lowest; however, when the simulation method was a pplied, the productivity of the SR-20 (Alachua) project was higher than that of the SR-102 project. In addition, the amount of productivity reduction from when initial time durations were used to when alternative time durations were used varies based on th e frequency of occurrence and the likely magnitude of interference, recorded in the production measurement form. The initial time durations were the median values that the inspectors provided, and they are subject to some vari ation. Even though the data co llected were as accurate as possible, not all of the interference factors we re recorded and provided to the researcher, and this affected the productivity differences between the simulation result using alternative time durations a nd cumulative productivity. As explained, the impact of shift effect s at the end of each working day was not considered in the simulation models when estimating the productivity rates. Shift effects—defined as the effect that a cert ain work schedule has on what work can be

PAGE 149

131 started—reduces productivity by some degree. For example, to estimate the daily shift effects for 8-hour working day, the simulati on run was stopped in every 480 minutes in simulation time. The resources remained in the simulation network at the end of work day (every 480 simulation time) were reassigne d in their initial nodes. This approach simulates the “start-from-scratch” rule, by disr egarding all the tasks left unfinished on the previous working day. The result of this approach showed that the productivity of asphalt pavement operation was reduced up to 17% (Huang and Halpin 1995).

PAGE 150

132 CHAPTER 7 VERIFICATION AND PRODUCTIVITY PREDICTION 7.1 Project Description The researcher selected the SR-26 project to verify the ability of the model to predict the productivity of a pavement ope ration using the developed methodology. The researcher then used the productivity result to estimate the project duration. The SR-26 project involved adding two lanes to an ex isting two-lane road through rural Alachua County into the city of Newbe rry. The total project length was 8,097 meters (5.03 miles). The low bidder for the project chose to us e both the APUB method with 100 mm (Base Option Group 1) and 260 mm (Base Option Group 9) limerock base in the rural part of the project and the FDAP method with 200 mm (Base Option Group13) HMA base for the urban section, making it an extremely va luable project for applying the productivity estimation method developed. The researcher separated the project according to which of the two pavement types was used, as if there were two separate projects. Base Option Group 1 was used for shoul der base only. Also, 75-mm Type-SP structural course and 20-mm FC-5 friction course were used on top of the base course. Figures 7-1 and 7-2 show the typical sections of the two pavement methods used on the project. General information regarding the pavement method s, including the actual unit price bid by the contractor , is shown in Table 7-1 .

PAGE 151

133 Figure 7-1. Typical section for the APUB method (SR-26) Figure 7-2. Typical section fo r the FDAP method (SR-26) Table 7-1. SR-26 General information for pavement structure Item Description Thickness Plan Qty Unit Unit Price Base Option 1 (Limerock) 100 mm 28,286.00 m2 $3.90 Base Option 9 (Limerock) 260 mm 69,424.00 m2 $7.20 Base Option 13 (HMA) 200 mm 45,418.00 m2 $18.60 Structural Course (Super-pave Asphalt Conc.) 75 mm 30,080.50 MT $40.95 Friction Course (FC-5) 20 mm 5,552.80 MT $67.86 7.2 Production Data Collected The pavement production information obtai ned during the data collection period was shown in Table 7-2. In the secti ons where the APUB method was used (SR-26 APUB Project), the crew inst alled 27,981.69 MT of the st ructural course on top of limerock base constructed, and in the sect ions where the FDAP method was used (SR-26 FDAP Project), 25,001.18 MT for the base course and 2,609.14 MT for the structural course were installed during th e data collection period.

PAGE 152

134 The total number of truck loads was calculated by divi ding the total tonnage of asphalt material by the 17 MT-capacity per tr uck. The number of truck loads was then converted by using the conversion factors as seen in Table 3-2 in Section 3.3.1 titled Productivity Output and Convers ion Factor. The SR-26 APUB and FDAP projects used 260 mm limerock base (Base Option 9) and 200 mm HMA base (B ase Option Group 13), respectively, for their base material, but bot h projects used 75 mm st ructural courses. The conversion factors for the structural co urse and HMA base were 1.24 and 0.76, respectively, as seen in Table 3-2. Table 7-2. Production data for SR-26 Data APUB FDAP Total Quantity (MT) Structural: 27981.69 Base: 25001.18 Structural:2609.14 Number of Truck loads 1645.98 1624.14 C. Truck loads 2041.02 1308.01 Crew hours (CH) 372.83 440.91 Productivity (C. Truck loads/CH) 5.47 2.97 Table 7-3. Work areas associat ed with work content (APUB) Length (m) Width (m) Thick. (m) Volume (m3) MT No. of loads Note 164.00 3.65 0.075 44.90 104.616.15 Turning lane 235.00 3.65 0.075 64.33 149.898.82 Turning lane 12.00 3.65 0.075 3.29 7.65 0.45 Turning lane 65.00 3.65 0.075 17.79 41.46 2.44 Cross-over 54.00 3.65 0.075 14.78 34.44 2.03 Center turning lane 112.00 3.65 0.075 30.66 71.44 4.20 Center turning lane 807.00 3.65 0.075 220.92 514.7330.28 Turning lane 33.00 3.65 0.075 9.03 21.05 1.24 3rd, Turnout (Rt) 187.00 3.65 0.075 51.19 119.287.02 3rd, Turnout (Rt) Total number of loads 62.62 The CH spent for each method was also obtai ned from the project. The pavement crew usually installed the sa me category of pavement cour se on a single work day (e.g. base course on the SR-26 FDAP project or structural course on the SR-26 APUB project).

PAGE 153

135 If the crew constructed the structural courses on both AP UB and FDAP projects on a single work day, the daily crew hours were pror ated based on the quantity completed. By dividing the converted truck loads by CH spent, the cumulative productivity of each method was calculated. Table 7-4. Work areas associat ed with work content (FDAP) Length (m) Width (m) Thick. (m) Volume (m3) MT No. of loads Note 4.00 3.65 0.075 1.09 2.55 0.15 Turn Lane 29.00 3.65 0.075 7.94 18.50 1.09 Turn Lane 35.00 3.65 0.075 9.58 22.32 1.31 3rd, Turnout (Rt) 34.00 3.65 0.075 9.31 21.69 1.28 3rd, Turnout (Rt) 332.00 3.65 0.075 90.89 211.76 12.46 Turn Lane 30.00 3.65 0.075 8.21 19.14 1.13 Crossover 39.00 3.65 0.075 10.68 24.88 1.46 Crossover 153.00 3.65 0.075 41.88 97.59 5.74 Turn Lane 42.00 3.65 0.075 11.50 26.79 1.58 Turn out 27.00 3.65 0.075 7.39 17.22 1.01 Turn out 29.00 3.65 0.075 7.94 18.50 1.09 Turn out 41.00 3.65 0.075 11.22 26.15 1.54 Intersection CR-235 15.00 3.65 0.075 4.11 9.57 0.56 Intersection CR-235 35.00 3.65 0.075 9.58 22.32 1.31 Turn out 5.00 3.65 0.075 1.37 3.19 0.19 Turnout 131.00 3.65 0.075 35.86 83.56 4.92 Turn Lane Total number of loads 36.81 The researcher obtained the actual quantity of work areas that were associated with work content from each project. As disc ussed, those areas include turning lanes, intersections, and crossover pavements. This type of information is readily obtained from the project documents such as plans and draw ings (e.g. typical section) before actual production begins so the actual quantity can be used in the TPN model to estimate the alternative time duration for each sub-task . Tables 7-3 and 7-4 show the quantity calculated from each project during the data collection process. For the verification purpose, the researcher used same width (3 .65 meter) and thickness (0.075 meter) for all

PAGE 154

136 lifts constructed. The volume of each lift was converted to MT by multiplying the average density of asphalt material (2.33 MT per m3) to the volume. The tonnage of the material then was converted to the number of truck loads by divi ding the tonnage by the average MT per each truck load (17 MT per load). 7.3 Estimation of Alternative Time Duration for SR-26 The researcher followed the procedure developed in order to estimate the alternative time durations of each sub-task for the SR-26 APUB and FDAP projects. The project inspector measured the median time duration of each sub-task and made them available to the researcher. The same survey form used fo r the four case study projects was used (Appendix A). Table 7-5 shows th e result of initial time durations for the project. Table 7-5. Initial ti me duration for SR-26 Node No. Work Task Unit Duration 2 Load at plant min 4 3 Travel to job min 6 5 Reposition for new pass load 20 6 Return min 10 9 Dump to spreader load 8 10 Spread min 10 15 Compact asphalt section min 8 18 Finish section min 8 Since the SR-26 project was located in a rural area of Alachua County, the SR-26 APUB project was assumed to have the same probability and magn itude of C1 (weather) and C2 (management) interference as the SR-2 0 (Alachua) project had. For the C2 (work content) interference, the act ual number of truck loads re quired in the SR-26 APUB project was used. In other words, the sa me magnitudes of the C1 and C2 types of interference were assumed for the projects that fall into the same category, as seen in Table 3-1; however, for the C3 type of inte rference, actual number of loads can be used

PAGE 155

137 for each project to predict the productivity of the pavement operation when using the methodology developed. Therefore, the same TPN models develope d to simulate interference for each subtask of the SR-20 (Alachua) project were us ed for the SR-26 APUB project with its own initial time durations and proba bility of C2 interference. The TPN models of loading, spreading, and compacting sub-tasks were seen in Figure 613, 6-14, and 6-15, respectively. For the loading task as seen in Table 6-12, the probabilities of C1 and C2 interference for the SR-20 (Al achua) project were 6.6 % and 22.3 % of the total number of loads. The proportions were estimated by the ratio between the NLA by each interference type and total numbe r of loads. No C3 interference was associated with the loading task since loading the asphalt materi al to the delivery truc ks at the plant has nothing to do with those work areas. Also, th e probability of the in terference causes were also found in Table 6-12, such as c11 (rain), c12 (low temperatur e), c21 (prerequisite work), and c26 (equipment breakdown). The pr obability of each category and cause and the addition times corresponding to each cause were used to simulate for the loading task of the SR-26 project, except that the initial time duration of the SR-26 APUB project was used. The total number of simulation events for the SR-26 APUB project was 24,492 to simulate 2,041 truck loads (24,492/12). It wa s described that one cycle consisted of 12 events in the TPN model developed. For estimating the alternative time durat ions for the spreading and compacting tasks, the probability of C3 interference were calculated by dividing 62.62 (actual number of loads) by 2,041 (total number of loads) , yielding 3.06 %. The probability of C3

PAGE 156

138 interference occurred in th e spreading and compacting tasks of the SR-20 (Alachua) project was 0.137. This number was repla ced by 0.0306. Also, the probabilities of NI (No interference) increased by the differen ce of 0.137 and 0.0306 (0.107). For the SR-26 APUB project, the probabilities of NI b ecame 0.861 (86.1%) for spreading and 0.885 (88.5%) for compacting. After replacing the pr obability of C3 interference for spreading and compacting subtasks, the TPN models were simulated to estimate the alternative time durations. The alternative time durations fo r loading, spreading, and compacting subtasks were estimated to be 19.1, 12.5, and 11.0 minutes, respectively, and the results of TPN simulation are provided in Appendix G. Similarly, the C1 and C2 types of inte rference that occurred in the SR-26 FDAP project was assumed to be the same as those of the SR-20 (Putnam) project because those two projects had the same pavement met hods (FDAP) and the geographical proximity. The TPN models developed for the SR-20 (Putnam) project we re also used for the SR-26 (Alachua) project to estimate the alternative time durations for its sub-tasks. The TPN models were seen in Figure 6-25, 6-26, and 6-27 for loading, spreading, and compacting subtasks, respectively. In the models fo r the spreading and compacting tasks, the probability for C3 was recalculated by dividing 36.81 (actual number of loads for C3) by 1,396, yielding 0.03. Since the probability of C3 for the spreading and compacting tasks in the project SR-20 (Palatka) was 0.06 (see Tables 6-19 and 6-20), the probability was replaced by 0.03. Then the probabilities for NI of spreading and compacting were replaced by 0.88 and 0.92 (increased by 0.03) in stead of 0.85 and 0.89, as seen in Tables 6-19 and 6-20. The total number of simula tion events for the SR -26 FDAP project was 16,752 to simulate 1,396 converted truck loads (16,752/12). The alternative time

PAGE 157

139 durations for loading, spreading, and compac ting of the project were 38.1, 12.6, and 10.6 minutes, respectively, as shown in Appendix G. 7.4 Productivity Estimation for SR-26 The project-specific time dur ations, resources utilized, and the alternative time durations for the three subtasks were then entered in the DISCO/ MicroCYCLONE simulation model to predict the productivity of the pavement operation for both SR-26 APUB and FDAP projects. Figur es 7-3 and 7-4 show the resu lts of the simulation of the SR-26 APUB project. The productivity rate calculated by Equation [6-5] was 6.98 loads per hour. The same procedure was followed to predict the productivity for the SR-26 FDAP project. Figure 7-5 and 7-6 showed th e results of the simu lation, yielding 3.90 loads per hour. Figure 7-3. Result of MicroCYCLONE si mulation for the SR-26 APUB project Figure 7-4. Result of DISCO simula tion for the SR-26 APUB project

PAGE 158

140 Figure 7-5. Result of MicroCYCLONE si mulation for the SR-26 FDAP project Figure 7-6. Result of DISCO simula tion for the SR-26 FDAP project As seen in Table 7-2, the actual produc tivity rates for the SR-26 APUB and FDAP projects were 5.47 and 2.97 loads per hour, so the result of pr oductivity predictions obtained by using the devel oped methodology were 27% and 31% higher than the actual rates. The summary result for productivity prediction is shown in Table 7-6. This result confirmed that, as seen in the four case study projects, the produc tivity estimation method developed yielded more optimistic productiv ity results than th e actual cumulative productivity. This result was consistent when the method was applied to predict the productivity of the pavement operation for bot h the SR-26 APUB an FDAP projects. Two main reasons that lead to this result were discussed in Chapter 6. First, the researcher used the deterministic method (median value) for modeling work task

PAGE 159

141 durations; however, the durations are likely to vary from cycle to cycle in reality. Even though the additional times caused by interfer ence factors were included in the time durations to estimate the actua l time durations as accurate as possible, the time durations were still subject to vary. Second, daily work shift effect was not applied in the simulation method even though it could influence the actual productivity. Therefore, the researcher applied (1) a statis tical distribution for work ta sk duration and (2) the daily shift effect to the simulation model. Table 7-6. Production prediction re sults for the SR-26 projects Productivity (C. Loads/CH) APUB FDAP Actual value 5.47 2.97 Predicted value 7.75 3.92 % error in predicted value 28.5 % 31.3% 7.5 Modeling Uncertainty in Time Duration and Daily Shift Effect In Chapter 4, it was shown that the pavement operation, one of the most equipment-intensive of all cons truction activities, had signifi cant productivity variability primary due to disruption events in the workflow caused by such factors as poor management, work content, and severe weather conditions. In Chapter 6, the productivity rates that were not associated with any type of interference also had some degree of variation due to th e natural causes. However, the productivity variances from each project were conclusively the same at the 5% level as shown in Chapter 5. The task of spreading asphalt material on th e site each work day serves as a major function of daily productivity measurement because the material spread on the site should be compacted and finished at the end of each wo rk day. Therefore, it is assumed that the work time duration of the spreading task follo ws the same distribution as that of daily

PAGE 160

142 productivity rates. The resear cher tested the pr oductivity results from four case study projects to find a probability distribution. As shown in Chapter 5, the productivity rates from the four projects were not normally distributed as seen in Figure 5-1. The rates were transformed to compare their mean values. Figure 7-7 shows the frequency histogram for the productivity rates, and it confirms the abnormality of the data. Since th e data were skewed to right with a long tail (see Figure 5-1), the distribut ion of the data was similar to lognormal distribution as shown in Figure 7-8. By definition, a dist ribution of random variab le X is lognormal if its natural logarithm, Y= log (X) is norm al. As shown in Figure 7-9, however, the distribution of natural logarith m of productivity rates was not normal at the 5% level (pvalue less than 0.05). The researcher tested th e distribution fit of productivity rates with other types of distributions su ch as logistic, gamma, beta, weibull, and exponential, but the distribution of the data was similar to lognormal distributio n the most with the location value of 1.0 and the scale value of 0.6. The effect of the location parameter is to translate the graph to the right or left on the horizontal axis. For example, relative to the standard normal di stribution, 10 units of a location parameter shift the graph 10 units to the right on the horizontal axis. The effect of a scale parameter greater th an one is to stretch the gra ph. The greater the magnitude, the greater the stretching. The productivity rates of the SR-26 APUB and FDAP projects were then estimated, assuming that the durati on of their spreading tasks have lognormal distribution.

PAGE 161

143 Prod. rateFrequency 10 8 6 4 2 0 35 30 25 20 15 10 5 0 Mean2.695 StDev1.672 N253Histogram of Prod. rateNormal Figure 7-7. Frequency hist ogram of productivity fo r the normal distribution Prod. rateFrequency 10.5 9.0 7.5 6.0 4.5 3.0 1.5 0.0 40 30 20 10 0 Loc1 Scale0.6 N253Histogram of Prod. rateLognormal Figure 7-8. Frequency hist ogram of productivity for the lognormal distribution C1Probability 1.5 1.0 0.5 0.0 -0.5 -1.0 0.999 0.99 0.95 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.05 0.01 0.001 Mean <0.005 0.3285 StDev0.3329 N253 AD3.746 P-ValueProbability Plot of C1Normal 95% CI Figure 7-9. Probability plot of productivity fo r the normal distribution 7.5.1 Improved Result using the Simulation Method Figure 7-10 shows the alterna tive time duration entered to the spreading task for the SR-26 APUB project in its si mulation model. The lognormal distribution of the duration was entered with the parameters of its locati on and scale. As shown in Figures 7-11 and

PAGE 162

144 7-12, the simulation was run only for 600 minut es in simulation time to apply the daily work shift effect. As a result, the productivity was calculated to be 5.97 loads per. Similarly, the alternative time duration of the task for the SR-26 FDAP project was entered with the lognormal dist ribution. The results of the simulation are estimated to be 2.99 loads per hour, as shown in Figures 7-13 an d 7-14. The results of the simulation are summarized in Table 7-7 to compare them with the actual pr oductivity rates. Figure 7-10. Lognormal distri bution for spreading task Figure 7-11. Improved result of produc tivity by MicroCYCLO NE (SR-26, APUB) Figure 7-12. Improved result of pr oductivity by DISCO (SR-26, APUB)

PAGE 163

145 Figure 7-13. Improved result of produc tivity by MicroCYCL ONE (SR-26, FDAP) Figure 7-14. Improved result of pr oductivity by DISCO (SR-26, FDAP) The productivity estimation was improved when applying: (1 ) alternative time durations for the three sub-tasks to model th e interference factors (Predicted value 1 in Table 7-7) and (2) statistical distribution to model uncertainty in the time duration and daily shift effects (Predicted value 2). When the alternative time durations were used, the estimated productivity rates were within 29 % (APUB) and 31% (FDAP) range from the actual values. The predicted values for the SR-26 APUB and FDAP projects were within 10% and 1% error range from th e actual productivity rates. Table 7-7. Production prediction results fo r the SR-26 projects after improvement Productivity (Loads / ho ur) APUB FDAP Actual value 5.47 2.97 Predicted value 1 (Actual hours, Deterministic distribution) 6.98 (27 %) 3.90 (31%) Predicted value 2 (10 hours, Lognormal distribution) 5.97 (9.0%) 2.99 (0.67%)

PAGE 164

146 7.5.1 Limitation of using Statistical Distribution The researcher planned to simulate the pr ocess as many times as the actual number of truck loads for both project s, as seen in Table 7-2. For example, by simulating 10 simulation hours for each work day, the simu lation for the SR-26 APUB project requires 38 iterations to simulate actua l 372.83 crew hours. Theoreti cally, the productivity result for each simulation should be different, sinc e the time duration of spreading sub-task follows the lognormal distribution. However, the result were the same in the simulation because the time duration of only three subtasks (including hauling and returning subtasks) used the statistical distribution while all other sub-tasks used the deterministic method. 7.6 Estimation of Construction Duration The construction duration for the SR-26 AP UB project consisted of the duration of structural and friction course construction. As shown in Ta ble 7-2, the crew finished 27,981.69 MT of structural course on top of the limerock base constructed. The result of productivity prediction for the construction of the st ructural course of the project was 5.97 loads per hour. As seen in Table 7-2, the number of truck loads required was derived by dividing 27,981.69 MT by the average tonnage per each truck load (17 MT per loads), and it yielded 1645. 98 truck loads. Then, 1645.98 truck loads were converted to 2041.02 converted truck lo ads by using the conversion factor of 1.24 as before. By dividing the converted truck load s by the predicted pr oductivity, the CH required were calculated as 341.87 (2041.02 / 5. 97). This CH is less than actual CH measured (372.83), as presented in Table 72 because the productivity predicted (5.73

PAGE 165

147 loads /hr) was higher than the actual productivity (5.47 load s/hr). The CH were then converted to 35 ten-hour work days. Similarly, the result of productivity pr ediction for the SR-26 FDAP project was 2.99 loads per hour. As seen in Table 7-2, the number of truck loads required was derived by dividing 27,610.32 MT (25001.18+ 2609.14) by the average tonnage per each truck load (17 MT per loads) , and it yielded 1624.13 truc k loads. Then, 1624.13 truck loads were converted to 1308.01 converted tr uck loads by using the conversion factors1.24 for the structural course and 0. 76 for the base course as before. By dividing the converted truck load s by the predicted pr oductivity, the CH required were calculated as 437.46 (1308.01 / 3.99) . This CH also is less than actual CH measured (440.91), as presented in Table 72 because the productivity predicted (2.99 loads /hr) was higher than the actual productivity (2.97 load s/hr). The CH were then converted to 44 ten-hour work days. As seen, by using the developed methodology, the productivity of the pavement operation can be predicted more accurately. The predicted productivity will serve as a reliable tool fo r the practitioners to estimate construction duration (work days) of paveme nt operation. Figure 7-15 sh ows the result of prediction for the SR-26 projects as a summary.

PAGE 166

148 335.39 372.83 440.91 292.41 437.46 341.87 34 45 38 30 44 350 50 100 150 200 250 300 350 400 450 500APUBFDAPCrew hours0 10 20 30 40 50 60 70 80 90 100Work days Actual CH Predicted CH (1) Predicted CH (2) Actual wrok day Predicted work day (1) Predicted work day (2) Figure 7-15. Summary of prediction results

PAGE 167

149 CHAPTER 8 SUMMARY 8.1 Conclusion 8.1.1 Productivity Analyses for Pa vement Construction Operation The objective of productivity analyses pres ented in Chapter 4 was to measure the productivity rate of highway pavement opera tions, one of the most equipment-intensive of all construction activities. By measuring calculating daily producti vity, the researcher also identified factors that adversely aff ected performance and quantified the loss of productivity caused by each factor. The result of analyses showed that the pavement operation had significant productivi ty variability due to disrup tion events in the workflow caused by such factors as poor management , work content, and severe weather conditions. The loss of crew hours caused by poor ma nagement ranged from 40% to 62% of the total inefficient crew hour s on the four projects. One of the primary factors that contributed to poor manage ment was out-of-sequence work and deficiencies in prerequisite work. It has been confirmed th at the quantitative and qualitative uncertainty in production output caused adverse effects, extending them to succeeding activities in the activity nexus. The PMI values of the tw o projects were higher than the other two projects that used HMA base. In addition, equipment breakdown a nd material shortages were occasionally observed during the process of paving. The loss of crew hours caused by work cont ent was from 21% to 48% of the total inefficient crew hours on the four projects. Ev en if the loss by work content is inevitable,

PAGE 168

150 the impact of work content can be mitigated with minimum impact on the project by planning ahead for the most frequent causes and improving predictability of workflow. Efficiency losses due to work content in this research mostly related to use of hand work while working in areas such as ramps, tu rning lanes, intersec tions, and crossovers; however, management and work content are some times interrelated, i.e., the deficiencies of previous work not only diminish the producti vity of later work but also make the work more complicated. The effect of severe weather conditions wa s not very significant, ranging from 6% to 17% of the total inefficien t crew hours of the four pr ojects. Even though severe weather prolonged the projects, the delay was not observed in this research because the contractors usually chose not to work when the weather was expected to be unfavorable and productivity was only measured on days when the contractors worked. The measurement of performance on a daily basis can provide information about the primary causes of productivity variability in paving operati ons. This information, in the hands of competent managers, can lead to more effective and more reliable work flow and better quality control dur ing the operation. Other produ ctivity parameters such as cumulative, mean, and baseline productivity served as the source of statistical analyses and productivity estimation by using process simulation models. 8.1.2 Causes-and-Effects of Interference on Productivity The researcher performed quantitative anal yses by using produc tion data collected and productivity parameters calculated. The analyses consisted of a series of mean comparisons and correlations tests. The anal yses yielded the results as following: Two pavement projects located in urban ar eas had higher produc tivity, compared to the other two projects in rural areas. It is assumed that the pave ment crews for the two

PAGE 169

151 projects with higher productivity than the other two projects we re encouraged to be more productive due to the high AADT and RUC. The productivity means are ranked as multiple, management, work content, weather, and no interference in order from the lowest to the highest when they are compared by their association with types of interference. The productivity associated with any interference is significantly lower than one with no inte rference. The mean productivity value when weat her interference occurred wa s the highest among others because the crew usually chose not to work when inclement weather was expected. The mean value of management interferen ce was the lowest as a single interference factor, and poor management was the most fr equent interference am ong others (65 out of 253 work days). Within management interfer ence, the mean productivity rates are ranked as work congestion, prerequisite work, equipm ent, and material in order from the lowest to the highest even though the differences among them were not significant at the 10% level. Also, the mean value of defects from prerequisite work ha s the most frequent occurrence among other types of management interference (41 out of 65 work days). The amount of rainfall and the number of rain days in project areas has little or no influence on either number of work days or NL A influenced by rainfall partly because the dataset analyzed included only 14 work days associated with weat her interference. However, the correlation increased significantly after truncating mild outliers in the dataset. As the inefficient hours by weather in terference increases, the probability of management interference occurr ence tends to increase as well. The inefficient hours also

PAGE 170

152 have significant effects on the productivity variat ion, and as the inefficient hours increase, the productivity decreases. The correlation between the defects of prer equisite work and the probability of management interference occu rrence is not significant; howev er, the correl ation between the hours the paving crew spent fixing the de fects and the daily productivity rates was significant at the 20% level. 8.1.3 Method of Productivity Estimation Developed The researcher developed, implemented, and evaluated the method of productivity estimation by applying quantified interference factors to process simulation model. In quantitative analyses, the mean productivity rates of the four case study projects were ranked I-10, SR-102, SR-20 (Alachua), and SR-2 0 (Palatka), in order from the highest to the lowest; however, when the simulation me thod was applied, the productivity of the SR-20 (Alachua) project was higher than that of the SR-102 project. In addition, the amount of productivity reduction from when in itial time durations were used to when alternative time durations were used varies based on the frequency of occurrence and the likely magnitude of interference, record ed in the production measurement form. The initial time durations were the median values that the inspectors provided. To test the reliability of the initial time durations, the durat ions were entered in the simulation model, and the model was simula ted for 10-simulation hours. The simulation results were then compared to actual da ily productivity rates th at had no interference recorded. For all four projects, the simula tion results were higher (optimistic) than the values of the third quartile of the actual productivity rates. The survey data for time durations provided by the project inspectors were more optimistic than the actual duration. Even though the data were as close to the median time duration as possible, they do not

PAGE 171

153 include auxiliary times. Even though the data collected were as accurate as possible, not all of the interference factor s were recorded and provided to the researcher, and this affected the productivity differences between the simulation result using alternative time durations and cumulative productivity. The researcher quantified interference factors using the developed method and estimated the alternative time durations for each subtask. When the alternative time durations were used, the productivity all fell in the range between baseline and cumulative productivity for four case study pr ojects. The method developed was applied to new projects to evaluate the accuracy for productivity estimation. In addition to using the deterministic method for modeling time durat ions and modeling daily shift effect, the researcher applied a statistical distribution to model the duration of the major sub-task in the process. As a result, th e result of productivity estimati on improved significantly. The productivity rates estimated were used to pr edict the number of work days required. 8.2 Future Work and Research Limitation 8.2.1 Statistical Analyses The researcher attempted to test both main and interaction effects between the severity of interference factors and their e ffects by using the model with crossed and nested effects. Testing of the model require s more exclusive volume of data related to the interference factors occurred in the pro cess of production, and the data can only be collected through on-site obser vations. Even though the data collection process by observation takes more time and money, te sting the method will provide very useful information for the effects of individual factors and interactio ns among interference factors.

PAGE 172

154 More accurate prediction m odel can be developed if the on-site observation is conducted. Developing a prediction model re quires an accurate measurement for each cycle of each sub-task. The measurements incl ude such data as the number of trucks, the volume of trucks per load, paver waiting time, truck load start time, truck load end time, the number of truck loads that the paver repositioned to make a new pass parallel to the just-completed pass, and the average number of truck loads that the spreader section released to the roller. When those data are collected, the statistical distribution of time durations can be identified and applied to the simulation model in order to estimate productivity more accurately. Besides, regressi on analysis can be used to determine the statistical relationship between a response (e.g. productivity rate) and predictors (e.g. cycle times). As explained, the data collection me thod using field observations requires tremendous investigation of time and money, a nd it was not feasible in this research. The researcher used the project documents to estimate average additional time for each cycle of each subtask. Each subtask did not act ually take the altern ative time duration; however, if the task was delayed by an inte rference factor reco rded in the project documents, the delayed time was incl uded in the duration of each task. 8.2.2 Integrated System for Modeling Interference In order to estimate constr uction productivity, the researcher used two simulation tools. First, the interference model was de veloped by using a TPN modeling technique to calculate alternative time durations of each s ub-task based on the frequency of occurrence and the likely magnitude of interference, r ecorded in the production measurement form. The alternative time durations were then en tered in the asphalt pavement simulation model developed in DISCO/MicroCYCLONE e nvironment. PNs, as a general purpose

PAGE 173

155 modeling tool, can also be used to mode l asphalt pavement operation with additional functions. For example, Colored Petri-Nets (CPN), as an extended form of PNs, allow different colors (attributes) to be associated with tokens. This function is used to distinguish different types of re sources (e.g. material, equipmen t, information, etc.) to be modeled. In pavement operations, the activ ity “spread” requires asphalt material to be delivered by trucks and an asphalt paver to spread the material. Each resource can be represented by different color tokens to estimate resource utilizat ion of each resource. Another research plan by using PNs i nvolves modeling construction processes at the project level. PNs also provide hierarchical featur es in its transitions, and a hierarchical transition is utili zed to depict a group of work tasks that are linked to each other. For example, a transition in a higher level model (e.g. site work) can be linked to the lower level model (e.g. excavating, load ing, etc.). Also, the interference model developed can be linked to each sub-task of pavement operation model by using this feature. 8.2.3. Development of Project Productivity Factor If more case studies were conducted, a set of project specific factors can be developed to predict pr oductivity on new or future projects . These factors would serve to calibrate the model to the unique features of the new project. Since the researcher measured productivity-specific parameters from one case study project from each category, more case study projects will be requ ired to develop project productivity factors. Future studies should consider these issu es based on the work by this research.

PAGE 174

156 APPENDIX A DATA COLLECTION FORMS Production Measurement Form Contact info: Volkert Construction Services, Inc. P. O. Box 11428 Pensacola, FL 32524 Office: 850.477.7485 Fax: 850.477.7517 Ed Blackmon, P.E. Mobile: 850.723.3166 Today's Date:10-May-031 of 3 Thickness:~2.9" 1035.85 Width:12 ft 1035.85 6729.33 Begin STA:726+86 753+50 Remarks:P2021, 726+86-753+50 R1, 728+67-752+50 L1, I10, 285-711 (group 11 base) Supt.: Foreman: Skilled: Semi-Skilled: Laborer: Hours 8.0 8.0 0.0 8.0 8.0 8.0 Asphalt Summary Sheets Project 222434-1-52-01, 21557 I10/I110 Interchange Pensacola, Florida Email: eblackmon@volkert.com LIFT CONSTRUCTED: QUANTITIES Quantity Delivered (Tons): Quantity Completed (Tons): Quantity Completed (SY): End STA: WORK CREW No.Hours 18.0 18.0 48.0 48.0 18.0 EQUIPMENT TypeNo. Dual Axle Dump Truck 8 Tri Axle Dump Truck Steel Wheel Roller Ingersoll Rand & Hypac C35002 Traffic Roller Ingersol Rand0 Tack Truck1 Blaw Knox PF172B Asphalt Paver1 Pickup Truck2 INCIDENTS THAT AFFECT PRODUCTION

PAGE 175

157 Survey Form Project Name: ________________ Project Inspector: _____________ 1. Average number of trucks used for paving: ____trucks (e.g. 10) 2. Average hauling time of HMA material from the plant to the site: _____(e.g. 20 minutes.) 3. Time for loading one truck with hot mix asphalt (HMA) from a plant (Do not include travel time): ______min (e.g. 8 min) 4. Time for dumping one truck load of HM A into the spreader (paver) skip: _____min (e.g. 4 min) 5. Average number of truck loads that the spreader is repositioned to make a new pass parallel to the just-completed pass:_____ loads (e.g. 15 loads) 6. Time to reposition the spread er: ______ min (e.g. 1 min) 7. Average number of truck loads that the spreader section is released to the roller:______ loads (e.g. 5 loads) 8. Time for distributing one tr uck load of HMA via the spreader:______ min (1 min) 9. Time for compacting one truck load of HMA with the breakdown roller (initial compacting): ______ min (e.g. 5 min) 10. Time for compacting one truck load of HMA with the finish roller (final compacting) : _______ min (e.g. 1.5 min) Thank you so much!!!

PAGE 176

158 Asphalt Report (Sample)

PAGE 177

159 Daily Diary (Sample)

PAGE 178

160 APPENDIX B FURTHER ANAYSES FOR PROJECT PRODUCTIVITY COMPARISON TukeyÂ’s Test Projects = SR-20 Palatka subtracted from: Projects Lower Center Upper SR-20 Hawthorne -0.2860 -0.0300 0.2260 SR-102 -0.0185 0.2248 0.4681 I-10 0.0836 0.2912 0.4988 Projects ---------+---------+---------+---------+ SR-20 Hawthorne (--------*--------) SR-102 (-------*--------) I-10 (------*------) ---------+---------+---------+---------+ -0.30 0.00 0.30 0.60 Projects = SR-20 Hawthorne subtracted from: Projects Lower Center Upper ---------+---------+---------+---------+ SR-102 -0.0150 0.2548 0.5246 (--------*--------) I-10 0.0831 0.3212 0.5593 (-------*-------) ---------+---------+---------+---------+ -0.30 0.00 0.30 0.60 Projects = SR-102 subtracted from: Projects Lower Center Upper ---------+---------+---------+---------+ I-10 -0.1580 0.0664 0.2908 (------*-------) ---------+---------+---------+---------+ -0.30 0.00 0.30 0.60 LSD Test Projects = SR-20 Palatka subtracted from: Projects Lower Center Upper SR-20 Hawthorne -0.2264 -0.0300 0.1664 SR-102 0.0381 0.2248 0.4115 I-10 0.1319 0.2912 0.4505 Projects -------+---------+---------+---------+-SR-20 Hawthorne (------*------) SR-102 (-----*------) I-10 (-----*----) -------+---------+---------+---------+--0.30 0.00 0.30 0.60 Projects = SR-20 Hawthorne subtracted from: Projects Lower Center Upper -------+---------+---------+---------+-SR-102 0.0478 0.2548 0.4618 (-----*------) I-10 0.1385 0.3212 0.5039 (-----*-----) -------+---------+---------+---------+--0.30 0.00 0.30 0.60 Projects = SR-102 subtracted from: Projects Lower Center Upper -------+---------+---------+---------+-I-10 -0.1057 0.0664 0.2386 (-----*-----) -------+---------+---------+---------+--0.30 0.00 0.30 0.60

PAGE 179

161 APPENDIX C FURTHER ANAYSES FOR INTERFERENCE COMPARISON TukeyÂ’s Test Interference = Multiple subtracted from: Interference Lower Center Upper Management -0.3774 0.0786 0.5347 Work Content -0.3028 0.1683 0.6393 Weather -0.2575 0.2904 0.8383 No interference 0.1209 0.5637 1.0065 Interference ----+---------+---------+---------+----Management (---------*--------) Work Content (--------*---------) Weather (----------*----------) No interference (--------*--------) ----+---------+---------+---------+-----0.50 0.00 0.50 1.00 Interference = Management subtracted from: Interference Lower Center Upper Work Content -0.1643 0.0896 0.3435 Weather -0.1661 0.2117 0.5896 No interference 0.2884 0.4851 0.6817 Interference ----+---------+---------+---------+----Work Content (----*----) Weather (------*-------) No interference (---*---) ----+---------+---------+---------+-----0.50 0.00 0.50 1.00 Interference = Work Content subtracted from: Interference Lower Center Upper Weather -0.2736 0.1221 0.5179 No interference 0.1663 0.3955 0.6246 Interference ----+---------+---------+---------+----Weather (------*-------) No interference (----*---) ----+---------+---------+---------+-----0.50 0.00 0.50 1.00 Interference = Weather subtracted from: Interference Lower Center Upper No interference -0.0884 0.2733 0.6350 Interference ----+---------+---------+---------+----No interference (------*-------) ----+---------+---------+---------+-----0.50 0.00 0.50 1.00

PAGE 180

162 FURTHER ANAYSES FOR INTERFEREN CE COMPARISON (continued) LSD Test Interference = Multiple subtracted from: Interference Lower Center Upper Management -0.2505 0.0786 0.4078 Work Content -0.1716 0.1683 0.5081 Weather -0.1049 0.2904 0.6857 No interference 0.2442 0.5637 0.8832 Interference ------+---------+---------+---------+--Management (-------*-------) Work Content (-------*--------) Weather (---------*---------) No interference (-------*-------) ------+---------+---------+---------+---0.40 0.00 0.40 0.80 Interference = Management subtracted from: Interference Lower Center Upper Work Content -0.0936 0.0896 0.2728 Weather -0.0609 0.2117 0.4844 No interference 0.3432 0.4851 0.6270 Interference ------+---------+---------+---------+--Work Content (---*----) Weather (------*------) No interference (--*---) ------+---------+---------+---------+---0.40 0.00 0.40 0.80 Interference = Work Content subtracted from: Interference Lower Center Upper Weather -0.1634 0.1221 0.4077 No interference 0.2301 0.3955 0.5608 Interference ------+---------+---------+---------+--Weather (------*------) No interference (---*---) ------+---------+---------+---------+---0.40 0.00 0.40 0.80 Interference = Weather subtracted from: Interference Lower Center Upper No interference 0.0123 0.2733 0.5343 Interference ------+---------+---------+---------+--No interference (------*-----) ------+---------+---------+---------+---0.40 0.00 0.40 0.80

PAGE 181

163 APPENDIX D THE RESULT OF FURTHER ANAYSES FOR DESIGN 1 A=1 Source DFSS MS F-value P-value Model 1 5.40E-07 5.40E-07 Infty <.0001 Error 1 0 0 Corrected Total 2 5.40E-07 A=2 Source DFSS MS F-value P-value Model 1 1.04E-06 1.04E-06 0.93 0.5122 Error 1 1.13E-06 1.13E-06 Corrected Total 2 2.17E-06 A=3 Source DFSS MS F-value P-value Model 1 0.000117810.000117811.05 0.4133 Error 2 0.000224490.00011224 Corrected Total 3 0.0003423 A=4 Source DFSS MS F-value P-value Model 1 0.000044890.0000448965.53 0.0149 Error 2 0.000001370.00000068 Corrected Total 3 0.00004626 B=1 Source DFSS MS F-value P-value Model 3 1.85E-06 6.16E-07 1.35 0.4059 Error 3 1.37E-06 4.57E-07 Corrected Total 6 3.22E-06 B=2 Source DFSS MS F-value P-value Model 3 0.000159550.000053180.71 0.6087 Error 3 0.000225610.0000752 Corrected Total 6 0.00038516

PAGE 182

164 APPENDIX E THE RESULT OF FURTHER ANAYSES FOR DESIGN 2 A=1 Source DFSS MS F-value P-value Model 1 0.195607270.195607271.04 0.3245 Error 14 2.626392730.18759948 Corrected Total 15 2.822 A=2 Source DFSS MS F-value P-value Model 1 0.439733570.439733571.85 0.2041 Error 10 2.381891430.23818914 Corrected Total 11 2.821625 A=3 Source DFSS MS F-value P-value Model 1 0.102536890.102536891.04 0.32 Error 22 2.178658950.09902995 Corrected Total 23 2.28119583 A=4 Source DFSS MS F-value P-value Model 1 0.385244790.385244792.35 0.1412 Error 20 3.283832480.16419162 Corrected Total 21 3.66907727 B=1 Source DFSS MS F-value P-value Model 3 1.886498920.628832973.14 0.0389 Error 32 6.417601080.20055003 Corrected Total 35 8.3041 B=2 Source DFSS MS F-value P-value Model 3 1.242196550.414065523.47 0.0265 Error 34 4.0531745 0.11921101 Corrected Total 37 5.29537105

PAGE 183

165 APPENDIX F INTERFERENCE QUANTIFICATION SR-20, Alachua (Case Study 1) Work day Date C. Loads C. Load/hr Crew hour W/ baseline Wasted ATPL TNLA Category Cause 7 10/20/03 20.60 3.43 6 3.70 2.30 0.112 40.00 Management lack of finish grading LR 9 10/22/03 38.11 3.46 11 6.84 4.16 0.109 40.17 Management Prerequisite work 11 10/24/03 14.42 1.44 10 2.59 7.41 0.514 35.63 Management Equipment breakdown 14 11/5/03 9.27 1.03 9 2.31 6.69 0.722 76.24 Management Work Congestion 14 11/5/03 9.27 1.03 9 2.31 6.69 0.722 76. 24 Management Prerequisite work 17 12/11/03 6.18 0.62 10 1.54 8.46 1.369 62.26 Management Prerequisite work 18 12/12/03 5.15 0.52 10 1.28 8.72 1.693 32.09 Management Prerequisite work 28 1/21/04 9.27 0.93 10 1.66 8.34 0.899 14.06 Management Plant breakdown 36 2/18/04 22.66 2.27 10 4.07 5.93 0.262 17.51 Management Rework 40 3/30/04 21.63 2.16 10 3.88 6.12 0.283 18. 54 Management Prerequisite work 3 8/14/03 27.81 3.97 7 4.99 2.01 0. 072 30.64 Weather Rain :stopped 24 12/31/03 8.24 1.18 7 1.48 5.52 0.670 16.71 Weather Rain 28 1/21/04 9.27 0.93 10 1.66 8.34 0.899 14.06 Weather lower 40 1 5/22/03 13.39 1.67 8 2.40 5.60 0.42 4.00 Work content turning lane 11 10/24/03 14.42 1.44 10 3.59 6.41 0.445 35.63 Work content turning lane 18 12/12/03 5.15 0.52 10 1.28 8.72 1.693 32.09 Work Content Ramp 6 9/19/03 23.69 2.79 8.5 4.25 4.25 0.179 30.00 Work content turning lane 12 10/31/03 13.39 1.34 10 2.40 7.60 0.567 26.00 Work Content Shoulders in various area 13 11/3/03 18.54 1.85 10 3.33 6.67 0.360 6.00 Work Content Turn lane 15 11/6/03 1.03 0.34 3 0.18 2.82 2.733 2.00 Work Content Ramp 20 12/17/03 2.06 0.26 8 0.37 7.63 3.704 4.00 Work Content Ramp 23 12/30/03 13.39 1.34 10 2.40 7.60 0.567 26.00 Work Content Turning lane 27 1/20/04 7.21 0.72 10 1.29 8.71 1.207 14.00 Work Content Turning lane 31 1/26/04 25.75 2.58 10 6.41 3.59 0.139 20.00 Work Content Crossover 34 1/29/04 20.64 1.88 11 3.71 7.29 0.353 24.00 Work Content Turning lane 35 1/30/04 19.57 1.96 10 3.51 6.49 0.331 26.00 Work Content Turning lane 37 3/19/04 5.15 0.52 10 0.92 9.08 1.762 10.00 Work Content Turning lane 39 3/23/04 4.12 0.82 5 0.74 4.26 1.034 8.00 Work Content Turning lane

PAGE 184

166 SR-20, Alachua (Case Study 1) (continued) Loading Spreading Compacting Work day Date H-a NLA AT Cause H-a NLA AT Cause H-a NLA AT Cause 7 10/20/03 20.00 0.06 c21 20.00 0.02 c21 9 10/22/03 5.00 20.09 0.02 c21 5.00 20.09 0.02 c21 11 10/24/03 7.41 41.31 0.60 c26 14 11/5/03 6.69 20.09 0.19 c27 14 11/5/03 6.99 38.93 0.37 c21 6.99 28.08 0.27 c21 17 12/11/03 8.46 47.13 1.37 c21 18 12/12/03 8.72 22.09 1.17 c21 28 1/21/04 2.00 11.14 0.71 c26 36 2/18/04 4.36 17.51 0.19 c21 40 3/30/04 4.62 18.54 0.21 c21 3 8/14/03 4.00 22.28 0.05 c11 1.50 8.36 0.02 c11 24 12/31/03 3.00 16.71 0.67 c11 28 1/21/04 2.50 13.93 0.89 c12 1 5/22/03 2.00 0.21 c31 2.00 0.17 c31 11 10/24/03 4.00 0.05 c31 4.00 0.05 c31 18 12/12/03 5.00 0.26 c31 5.00 0.85 c31 6 9/19/03 15.00 0.09 c31 15.00 0.05 c31 12 10/31/03 13.00 0.28 c31 13.00 0.25 c31 13 11/3/03 3.00 0.18 c31 3.00 0.15 c31 15 11/6/03 1.00 1.37 c31 1.00 1.33 c31 20 12/17/03 2.00 1.85 c31 2.00 1.82 c31 23 12/30/03 13.00 0.28 c31 13.00 0.25 c31 27 1/20/04 7.00 0.60 c31 7.00 0.57 c31 31 1/26/04 10.00 0.07 c31 10.00 0.07 c31 34 1/29/04 12.00 0.18 c31 12.00 0.14 c31 35 1/30/04 13.00 0.17 c31 13.00 0.13 c31 37 3/19/04 5.00 0.88 c31 5.00 0.85 c31 39 3/23/04 4.00 0.52 c31 4.00 0.48 c31

PAGE 185

167 SR-102 (Case Study 2) Work day Date C. Loads C. Load/hr Crew hour W/ Baseline Wasted ATPL TNLA Category Cause 1 5/13/03 15.45 3.09 5 2.28 2.72 0.176 17. 54 Management Prerequisite work 2 5/14/03 52.53 5.25 10 7.74 2.26 0.043 38. 46 Management Prerequisite work 2 5/14/03 52.53 5.25 10 7.74 2.26 0.043 38. 46 Management Material rejected 6 6/6/03 2.06 0.52 4 0.30 3.70 1.794 23.81 Management Prerequisite work 7 6/9/03 5.15 0.52 10 0.76 9.24 1.794 25.76 Ma nagement Prerequisite (Rework) 14 9/20/03 14.42 2.40 6 2.12 3.88 0.269 29.32 Management Problem at plant 14 9/20/03 14.42 2.40 6 2.12 3.88 0.269 29. 32 Management Material failure 16 11/13/03 17.51 2.19 8 2.58 5.42 0.310 22.54 Management Problem at plant 19 11/22/03 49.44 4.94 10 7.28 2.72 0.055 10.00 Management Material delay 32 1/21/04 14.42 1.44 10 2.12 7.88 0.546 16.10 Management Rework 33 1/22/04 46.35 4.21 11 6.83 4.17 0.090 19.32 Management Rework 34 1/23/04 35.02 3.50 10 5.16 4.84 0.138 16.10 Management Rework 35 1/26/04 2.06 0.34 6 0.30 5.70 2.765 25.76 Management Equipment breakdown 36 1/27/04 15.45 1.55 10 2.28 7.72 0.500 25.76 Management Rework 39 1/30/04 1.03 0.10 10 0.15 9.85 9.561 9.66 Management Rework 40 2/3/04 6.18 0.77 8 0.91 7.09 1.147 45.66 Management Previous work 42 3/2/04 7.21 0.72 10 1.06 8.94 1.240 57.56 Management Previous work 43 3/3/04 10.30 1.03 10 1.52 8.48 0.824 54.63 Management Previous work 44 3/4/04 8.24 0.82 10 1.21 8.79 1.066 56.58 Management Previous work 46 3/9/04 7.21 0.72 10 1.06 8.94 1.240 12.88 Management Rework 48 3/19/04 41.28 3.75 11 3.64 7.36 0.178 32.20 Management Problem at plant 8 6/12/03 29.87 3.73 8 4.40 3.60 0.121 12.88 Weather rain 22 12/2/03 15.45 2.21 7 2.28 4.72 0.306 27.32 Weather temperature 23 12/3/03 20.60 4.12 5 3.03 1.97 0.095 12.66 Weather rain 24 12/4/03 23.69 3.95 6 3.49 2.51 0.106 16.16 Weather rain 27 12/10/03 3.09 1.03 3 0.46 2.54 0.824 49.08 Weather rain 2 5/14/03 52.53 5.25 10 7.74 2.26 0.043 38.46 Work content turning lane 4 5/20/03 28.84 2.88 10 4.25 5.75 0.199 8.00 Work content turning lane 5 5/21/03 19.57 2.45 8 2.88 5.12 0.261 4.00 Work content turning lane 14 9/20/03 14.42 2.40 6 2.12 3.88 0.269 29.32 Work content turning lane 15 11/12/03 12.36 1.55 8 1.82 6.18 0.500 8.00 Work content turning lane 18 11/15/03 45.32 4.53 10 6.68 3.32 0.073 10.00 Work content tapered area & turn lane 20 11/25/03 28.84 2.88 10 4.25 5.75 0.199 56.00 Work content turning lane 21 11/26/03 56.65 5.67 10 8.35 1.65 0.029 20.00 Work content crossover 22 12/2/03 15.45 2.21 7 2.28 4.72 0.306 27.32 Work content turning lane 27 12/10/03 3.09 1.03 3 0.46 2.54 0.824 49.08 Work content turning lane 30 1/16/04 25.75 3.22 8 3.79 4.21 0.163 18.00 Work content turning lane

PAGE 186

168 SR-102 (Case Study 2) (continued) Loading Spreading Compacting Work day Date H-a NLA AT Cause H-a NLA AT Cause H-a NLA AT Cause 1 5/13/03 2.7 17.54 0.176 c21 2 5/14/03 2.3 14.56 0.016 c21 2 5/14/03 N/A 5.00 0.006 c25 6 6/6/03 3.7 23.81 1.794 c21 7 6/9/03 4.0 25.76 1.794 c21 14 9/20/03 3.0 19.32 0.177 c26 14 9/20/03 2.00 0.018 c25 16 11/13/03 3.5 22.54 0.310 c26 19 11/22/03 10.00 0.055 c24 32 1/21/04 2.5 16.10 0.546 c21 33 1/22/04 3.0 19.32 0.090 c21 34 1/23/04 2.5 16.10 0.138 c21 35 1/26/04 4.0 25.76 2.765 c26 36 1/27/04 4.0 25.76 0.500 c21 39 1/30/04 1.5 9.66 9.561 c21 40 2/3/04 7.1 45.66 1.147 c21 42 3/2/04 8.9 57.56 1.240 c21 43 3/3/04 8.5 54.63 0.824 c21 44 3/4/04 8.8 56.58 1.066 c21 46 3/9/04 2.0 12.88 1.240 c21 48 3/19/04 5.0 32.20 0.178 c26 8 6/12/03 2.0 12.88 0.279 c11 22 12/2/03 3.0 19.32 0.216 c12 23 12/3/03 2.0 12.66 0.095 c11 24 12/4/03 2.5 16.16 0.106 c11 27 12/10/03 2.5 16.39 0.275 c11 2 5/14/03 12.00 0.013 c31 12.00 0.013 c31 4 5/20/03 4.00 0.100 c31 4.00 0.100 c31 5 5/21/03 2.00 0.131 c31 2.00 0.131 c31 14 9/20/03 4.00 0.037 c31 4.00 0.037 c31 15 11/12/03 4.00 0.250 c31 4.00 0.250 c31 18 11/15/03 5.00 0.037 c31 5.00 0.037 c31 20 11/25/03 28.00 0.100 c31 28.00 0.100 c31 21 11/26/03 10.00 0.015 c31 10.00 0.015 c31 22 12/2/03 4.00 0.045 c31 4.00 0.084 c31 27 12/10/03 2.00 0.034 c31 2.00 0.147 c31 30 1/16/04 9.00 0.082 c31 9.00 0.082 c31

PAGE 187

169 SR-20, Putnam (Case Study 3) Work day Date C. Loads C. Load /hr CH CH by baseline Wasted ATPL TNLA Category Cause 47 4/6/04 11.10 1.01 11 2.20 8.80 0.793 29.34 Management Prerequisite work 5 10/2/03 6.66 0.56 12 1.32 10.68 1.604 55.46 Management Prerequisite work 6 10/3/03 18.50 1.95 9.5 3.67 5.83 0.315 16. 43 Management Prerequisite work 8 10/16/03 26.64 2.54 10.5 5.28 5.22 0.196 23.93 Management Material delivery 10 10/28/03 8.14 0.90 9 1.61 7.39 0.907 24.95 Management Prerequisite work 15 12/2/03 8.88 0.89 10 1.76 8.24 0.928 55. 77 Management Material delivery 16 12/18/03 23.68 2.63 9 4.69 4.31 0.182 18.82 Management Material delivery 17 12/19/03 17.76 1.97 9 3.52 5.48 0.309 15.33 Management Out of Sequence 25 1/21/04 19.24 1.75 11 3.81 7.19 0.374 21.20 Management Work Conflict 28 1/26/04 14.80 1.41 10.5 2.93 7.57 0.511 47.61 Management Equipment breakdown 30 1/29/00 1.48 0.11 13 0.29 12.71 8.586 46.32 Management Work Conflict 45 4/3/04 8.88 0.89 10 1.76 8.24 0.928 27. 89 Management Plant breakdown 55 4/19/04 17.76 1.48 12 3.52 8.48 0.477 11.03 Management Plant breakdown 62 5/5/04 15.54 1.41 11 3.08 7.92 0.510 9.19 Management Prerequisite work 3 9/15/03 34.04 3.09 11 6.75 4.25 0.125 6.40 Management Rework 54 4/17/04 8.14 0.90 9 1.61 7.39 0.907 9.19 Management Out of Sequence 9 10/27/03 19.98 2.22 9 3.96 5.04 0.252 4.41 Weather Rain 11 10/29/03 11.10 1.23 9 2.20 6.80 0.613 11.03 Weather Rain 20 1/7/04 30.34 3.37 9 6.01 2.99 0.098 9.19 Weather Temperature 24 1/20/04 8.14 0.74 11 1.61 9.39 1.153 20.22 Weather Temperature 5 10/2/03 6.66 0.56 12 1.81 10.19 1.530 55.46 Work content Turning lane 12 10/30/03 12.58 1.40 9 2.49 6.51 0.517 16.00 Work content Turning lane 22 1/16/04 8.88 0.99 9 1.76 7.24 0.815 12. 00 Work content Turning lane 26 1/22/04 19.98 1.82 11 3.96 7.04 0.352 24. 00 Work content Turning lane 34 2/13/04 17.76 1.78 10 3.52 6.48 0.365 30. 00 Work content Turning lane 35 2/19/04 7.40 0.82 9 1.47 7.53 1.018 20.00 Work content Intersection 37 2/26/04 11.84 1.18 10 2.35 7.65 0.646 12. 00 Work content Turning lane 52 4/14/04 16.28 1.36 12 3.23 8.77 0.539 20.00 Work content Driveway

PAGE 188

170 SR-20, Putnam (Case Study 3) (continued) Loading Spreading Compacting Work day Date H-a NLA AT Cause H-a NLA AT Cause H-a NLA AT Cause 47 4/6/04 8.80 44.40 1.20 c21 5 10/2/03 10.68 53.89 1.56 c21 6 10/3/03 5.83 29.43 0.56 c21 8 10/16/03 5.22 26.34 0.22 c24 3.25 11.97 0.06 c24 10 10/28/03 7.39 37.27 1.36 c21 15 12/2/03 8.24 41.57 0.69 c24 7.58 27.89 0.43 c24 16 12/18/03 4.31 21.73 0.21 c24 2.56 9.41 0.05 c24 17 12/19/03 5.48 27.65 0.56 c22 25 1/21/04 7.19 36.26 0.64 c27 28 1/26/04 7.57 38.18 0.41 c26 6.47 23.81 0.22 c26 30 1/29/00 12.71 64.11 11.88 c27 45 4/3/04 8.24 41.57 1.38 c26 55 4/19/04 3.00 15.14 0.66 c26 62 5/5/04 2.50 12.61 0.70 c21 3 9/15/03 4.25 21.46 0.42 c21 54 4/17/04 2.50 12.61 1.25 c22 9 10/27/03 1.20 6.05 0.35 c11 11 10/29/03 3.00 15.14 0.84 c11 20 1/7/04 2.50 12.61 0.14 c12 24 1/20/04 5.50 27.75 1.58 c12 5 10/2/03 9.00 0.25 c31 9.00 0.25 c31 12 10/30/03 8.00 0.26 c31 8.00 0.22 c31 22 1/16/04 6.00 0.41 c31 6.00 0.37 c31 26 1/22/04 12.00 0.18 c31 12.00 0.14 c31 34 2/13/04 15.00 0.18 c31 15.00 0.15 c31 35 2/19/04 10.00 0.51 c31 10.00 0.47 c31 37 2/26/04 6.00 0.32 c31 6.00 0.29 c31 52 4/14/04 10.00 0.27 c31 10.00 0.83

PAGE 189

171 I-10 (Case Study 4) Work day Date C. Load C. load /hr Crew hours w/ baseline Wasted ATPL TNLA Category Cause 4 4/7/2003 3.90 0.98 4.00 0.66 3.34 0.86 19. 66 Management Prerequisite work 7 4/28/2003 39.42 3.94 10.00 6.69 3.31 0.08 39.00 Management Congestion 15 5/12/2003 29.64 2.96 10.00 5.03 4.97 0.17 18.00 Management material failure 16 5/13/2003 4.12 0.82 5.00 0.70 4.30 1.04 41.24 Management Material delivery 24 6/4/2003 7.21 0.80 9.00 1.22 7.78 1.08 40. 00 Management Prerequisite work 42 8/27/2003 42.12 3.83 11.00 7.15 3.85 0.09 17.67 Management Equipment breakdown 48 9/11/2003 30.00 3.00 10.00 5.09 4.91 0.16 28.28 Management Material delivery 51 9/29/2003 17.16 1.63 10.50 2.91 7.59 0.44 44. 69 Management Prerequisite work 54 10/6/2003 14.00 1.12 12.50 2.38 10.12 0.72 14.00 Management Material delivery 58 10/17/2003 21.84 1.99 11.00 3.71 7.29 0.33 47.13 Management Material delivery 68 12/19/2004 45.24 4.11 11.00 7.68 3.32 0.07 29.45 Management Equipment breakdown 74 1/22/2004 6.18 0.62 10.00 1.05 8.95 1. 45 52.73 Management Prerequisite 85 3/1/2004 6.00 0.60 10.00 1.02 8.98 1.50 52.91 Management Rework 88 3/8/2004 5.15 0.52 10.00 0.87 9.13 1.77 10.00 Management Material delivery 92 3/31/2004 23.58 2.14 11.00 4.00 7.00 0.30 23.56 Management Rework 95 4/6/2004 13.26 1.21 11.00 2.25 8.75 0.66 18.00 Management Material delivery 6 4/26/2003 16.38 2.34 7.00 2.78 4.22 0.26 47. 13 Management Material delivery 27 7/14/2003 30.90 3.86 8.00 5.25 2.75 0.09 24.00 Management material failure 45 9/5/2003 24.96 2.50 10.00 4.24 5.76 0.23 33. 95 Management Prerequisite work 49 9/19/2003 19.50 1.95 10.00 3.31 6.69 0.34 39. 41 Management Prerequisite work 56 10/14/2003 10.92 1.37 8.00 1.85 6.15 0.56 36.21 Management Prerequisite work 57 10/16/2003 30.42 3.04 10.00 5.16 4.84 0.16 28.49 Management Prerequisite work 63 10/23/2003 20.60 1.72 12.00 3.50 8.50 0.41 50.09 Management Prerequisite work 67 12/18/2004 10.92 0.99 11.00 1.85 9.15 0.84 53.88 Management Prerequisite work 72 1/20/2004 14.00 1.27 11.00 2.38 8.62 0.62 35.35 Management Congestion 80 2/9/2004 24.00 2.18 11.00 4.07 6.93 0.29 60. 80 Management Prerequisite work 81 2/26/2004 27.81 2.53 11.00 4.72 6.28 0.23 36. 99 Management Prerequisite work 82 2/27/2004 26.78 2.98 9.00 4.55 4.45 0.17 26. 24 Management Prerequisite work 29 7/16/2003 18.54 2.32 8.00 3.15 4.85 0.26 13.55 Weather Rain 33 7/22/2003 25.75 3.22 8.00 4.37 3.63 0.14 11.78 Weather Rain 62 10/22/2003 8.58 0.78 11.00 1.46 9.54 1.11 17.67 Weather Rain 76 1/28/2004 14.00 1.40 10.00 2.38 7.62 0.54 17.67 Weather Rain 86 3/4/2004 10.30 2.06 5.00 1.75 3.25 0.32 11.78 Weather Rain 9 4/30/2003 35.54 2.96 12.00 6.03 5.97 0.17 22.00 Work content Turning lane 10 5/1/2003 47.13 3.93 12.00 8.00 4.00 0.08 10.00 Work content Turning lane 17 5/20/2003 30.90 3.09 10.00 5.25 4.75 0.15 16.00 Work content Turning lane 18 5/21/2003 15.00 3.00 5.00 2.55 2.45 0.16 20.00 Work content Turning lane 19 5/22/2003 20.12 2.52 8.00 3.42 4.58 0.23 12.00 Work content Turning lane 26 6/23/2003 18.54 1.85 10.00 3.15 6.85 0.37 30.00 Work content Turning lane 43 9/2/2003 24.96 2.50 10.00 4.24 5.76 0.23 20.00 Work content Turning lane 52 9/30/2003 24.96 2.27 11.00 4.24 6.76 0.27 46.00 Work content Ramp 55 10/9/2003 27.81 2.78 10.00 4.72 5.28 0.19 26.00 Work content Ramp 64 11/10/2003 31.20 2.60 12.00 5.30 6.70 0.21 70.00 Work content Ramp 69 1/8/2004 14.82 1.85 8.00 2.52 5.48 0.37 20.00 Work content Turning lane 77 2/3/2004 18.54 2.32 8.00 3.15 4.85 0.26 20.00 Work content Turning lane 79 2/5/2004 18.00 1.64 11.00 3.06 7.94 0.44 26.00 Work content Turning lane 91 3/29/2004 19.53 3.91 5.00 3.32 1.68 0.09 38.00 Work content Median 96 4/8/2004 2.34 0.26 9.00 0.40 8.60 3.68 32.00 Work content Ramp

PAGE 190

172 I-10 (Case Study 4) (continued) Loading Spreading Compacting Work day Date H-a NLA AT Cause H-a NLA AT Cause H-a NLA AT Cause 4 4/7/2003 3.3 19.66 0.86 c21 7 4/28/2003 3.3 19.50 0.04 c27 3.3 19.50 0.04 c27 15 5/12/2003 9.00 0.08 c25 9.00 0.08 c25 16 5/13/2003 3.5 20.62 0.52 c24 3.5 20.62 0.52 c24 24 6/4/2003 20.00 0.54 c21 20.00 0.54 c21 42 8/27/2003 3.0 17.67 0.09 c26 48 9/11/2003 2.4 14.14 0.08 c24 2.4 14.14 0.08 c24 51 9/29/2003 7.6 44.69 0.44 c21 54 10/6/2003 7.00 0.36 c24 7.00 0.36 c24 58 10/17/2003 4.0 23.56 0.17 c24 4.0 23.56 0.17 c24 68 12/19/2004 5.0 29.45 0.07 c26 74 1/22/2004 9.0 52.73 1.45 c21 85 3/1/2004 9.0 52.91 1.50 c21 88 3/8/2004 5.00 0.89 c24 5.00 0.89 c24 92 3/31/2004 4.0 23.56 0.30 c21 95 4/6/2004 9.00 0.33 c24 9.00 0.33 c24 6 4/26/2003 4.0 23.56 0.13 c24 4.0 23.56 0.13 c24 27 7/14/2003 12.00 0.04 c25 12.00 0.04 c25 45 9/5/2003 5.8 33.95 0.23 c21 49 9/19/2003 6.7 39.41 0.34 c21 56 10/14/2003 6.1 36.21 0.56 c21 57 10/16/2003 4.8 28.49 0.16 c21 63 10/23/2003 8.5 50.09 0.41 c21 67 12/18/2004 9.1 53.88 0.84 c21 72 1/20/2004 3.0 17.67 0.31 c27 3.0 17.67 0.31 c27 80 2/9/2004 6.9 40.80 0.19 c21 10.00 0.05 c21 10.00 0.05 c31 81 2/26/2004 6.3 36.99 0.23 c21 82 2/27/2004 4.5 26.24 0.17 c21 29 7/16/2003 2.3 13.55 0.26 c11 33 7/22/2003 2.0 11.78 0.14 c11 62 10/22/2003 3.0 17.67 1.11 c11 76 1/28/2004 3.0 17.67 0.54 c11 86 3/4/2004 2.0 11.78 0.32 c11 9 4/30/2003 11.00 0.08 c31 11.00 0.08 c31 10 5/1/2003 5.00 0.04 c31 5.00 0.04 c31 17 5/20/2003 8.00 0.08 c31 8.00 0.08 c31 18 5/21/2003 10.00 0.08 c31 10.00 0.08 c31 19 5/22/2003 6.00 0.11 c31 6.00 0.11 c31 26 6/23/2003 15.00 0.18 c31 15.00 0.18 c31 43 9/2/2003 10.00 0.12 c31 10.00 0.12 c31 52 9/30/2003 23.00 0.14 c31 23.00 0.14 c31 55 10/9/2003 13.00 0.09 c31 13.00 0.09 c31 64 11/10/2003 35.00 0.11 c31 35.00 0.11 c31 69 1/8/2004 10.00 0.19 c31 10.00 0.19 c31 77 2/3/2004 10.00 0.13 c31 10.00 0.13 c31 79 2/5/2004 13.00 0.22 c31 13.00 0.22 c31 91 3/29/2004 19.00 0.04 c31 19.00 0.04 c31 96 4/8/2004 16.00 1.84 c31 16.00 1.84 c31

PAGE 191

173 APPENDIX G RESULT OF TPN SIMULATION SR-20, Alachua (Case Study 1): Loading

PAGE 192

174 SR-20, Alachua (Case Study 1): Spreading

PAGE 193

175 SR-20, Alachua (Case Study 1): Compacting

PAGE 194

176 SR-102 (Case Study 2): Loading

PAGE 195

177 SR-102 (Case Study 2): Spreading

PAGE 196

178 SR-102 (Case Study 2): Compacting

PAGE 197

179 SR-20, Putnam (Case Study 3): Loading

PAGE 198

180 SR-20, Putnam (Case Study 3): Spreading

PAGE 199

181 SR-20, Putnam (Case Study 3): Compacting

PAGE 200

182 I-10 (Case Study 4): Loading

PAGE 201

183 I-10 (Case Study 4): Spreading

PAGE 202

184 I-10 (Case Study 4): Compacting

PAGE 203

185 SR-26, Alachua (APUB): Loading

PAGE 204

186 SR-26, Alachua (APUB): Spreading

PAGE 205

187 SR-26, Alachua (APUB): Compacting

PAGE 206

188 SR-26, Alachua (FDAP): Loading

PAGE 207

189 SR-26, Alachua (FDAP): Spreading

PAGE 208

190 SR-26, Alachua (FDAP): Compacting

PAGE 209

191 APPENDIX H RESULTS OF PRODUCTIVITY ESTIMATION SR-20 (Alachua, Case Study 1)

PAGE 210

192 SR-102 (Duval, Case Study 2)

PAGE 211

193 SR-20 (Putnam, Case Study 3)

PAGE 212

194 I-10 (Escambia, Case Study 4)

PAGE 213

195 LIST OF REFERENCES Alfeld, L. E. (1988). Construction productivity , McGraw-Hill, New York. Ballard, G., and Howell, G. (1997). “Implemen ting lean construction: stabilizing work flow.” Lean Construction; Proc., 2nd Annu. Meeting of Int. Group for Lean Constr. , A.A. Balkema, Publishers, Rotterdam, The Netherlands. Ballard, G., and Howell, G. (1998). “Shieldi ng production: Essential step in production control.” J. Constr. Engrg. and Mgmt. , 124(1), 11-17. Bernold, L. E.(1989). “Simulation of nonsteady construction processes.” J. Constr. Engrg. and Mgmt. , 115(2), 163–178. Blanchard, B. S., and Fabrycky, W. J. (1981). Systems engineering and analysis , Prentice-Hall, Inc., New Jersey. Chang, D. Y. (1986). “RESQUE: A resource based simulation system for construction process planning.” PhD disse rtation, University of Michigan, Ann Arbor, Mich. Choi, J., and Minchin, R. E. (In press). “W ork flow management a nd productivity control for asphalt pavement operations.” Can. J. Civ. Eng. . Choi, J., Minchin, R. E., and Herbsman, Z. J. (In press). “A decision making model for base material options.” Transp. Res. Rec,. Damrianant, J., and Wakefield, R. R. (2000) , “An alternative appr oach for modeling of interference in discre te-event systems.” J. Civ. Eng. & Envt. Sys., 17 (3), 213-235. Daniels, G., Ellis, D. R., & Stockton, W. R. (1999). Techniques for m anually estimating road user costs associated with construction projects , Texas Transportation Institute, Texas A&M University, College Station. Drewin, F. J. (1982). Construction productivity . Elsevier, New York. El-Rayes, K., and Moselhi, O. (2001). “Impact of rainfall on the produc tivity of highway construction.” J. Constr. Engrg. and Mgmt. , 127(2), 125-131. Finke, M. R. (1998). “A better way to estimate and mitigate disruption.” J. Constr. Engrg. and Mgmt. , 124(6), 490-497.

PAGE 214

196 Florida Department of Transportation. ( 2002a). “Flexible pavement design manual.” FDOT website , (Jan. 10, 2003). Florida Department of Transportation. (2002b). “Statewide CCEI salary rate.” FDOT website , (Sep. 20, 2003). Florida Department of Transportation. ( 2002c). “Annual average daily traffic reports.” FDOT website , (Jun 15, 2003). Florida Department of Transportation. ( 2002d). “Manual of uniform minimum standards for design, construction and mainte nance for streets and highways.” FDOT website , (March 16, 2003). Florida Department of Transportation. ( 2004). “Standard specifications for road and bridge construction.” FDOT website , (Oct. 12, 2003). Goodrum, P. M., and Hass, C. T. (2002). “P artial factor produc tivity and equipment technology change at activity level in U.S. construction industry.” J. Constr. Engrg. and Mgmt. , 128(6), 463-472. Griffiths, D., Gulati, C., a nd Ollis, J. (2003). “Statistical control for road pavements.” Aust. N. Z. Stat. 45(2), 129-140 Gulezian, R., and Samelian, F. (2003). “Bas eline determination in construction labor productivity-loss claim.” J. Constr. Engrg. and Mgmt. , 19(4), 160-165 Hajjar, D., and AbouRizk, S. (2000). “App lication framework fo r development of simulation tools.” J. Comput. Civ. Eng. , 14(3), 160–167. Hajjar, D., and AbouRizk, S. M., ( 2002). “Unified modeling methodology for construction simulation.” J. Constr. Engrg. and Mgmt. , 128(2), 174-185 Halpin, D. W. (1977). “CYCLONE—Methods for modeling job site processes.” J. Constr. Div. , ASCE, 103(3), 489–499. Halpin, D. W., and Riggs, L. S. (1992). Planning and analysis of construction operations. John Wiley and Sons, New York. Harris, R. B., and Ioannou, P. G. (1998). “S cheduling projects with repeating activities.” J. Constr. Engrg. and Mgmt. , 124(4), 269-278. Hassanein, A., and Moselhi, O. (2004). “Pla nning and scheduling highway construction.” J. Constr. Engrg. and Mgmt. , 130(5), 638-646.

PAGE 215

197 Hart, M. and Hart, R. (2002). Statistical process c ontrol for health care , Pacific Grove, California. Hopp, W., and Spearman, M. (2000). Factor y physics: foundations of manufacturing management. 2nd Ed., Irwin McGraw-Hill, Boston Horman, M. J., and Kenley, R. (1998). “Pro cess dynamics; Identifying a strategy for the deployment of buffers in building projects.” Int. J. Logistics; Research and Applications . 1(3), 221-237. Howell, G., and Ballard, G. (1994). “Impleme nting lean construction: Reducing inflow variation.” Proc., 2nd Annual Meeting of the Int. Group fro Lean Construction , A.A. Balkema, Rotterdam, The Netherlands. Huang, R. Y., and Halpin, D. W. (19 95). “Graphical-based method for transient evaluation of construction operations.” J. Constr. Engrg. and Mgmt. , 121(2), 222229. Huang, Y. H (1992). Pavement analysis and design . 2nd Ed., Englewood Cliffs, New Jersey. Jiang, Y. (2003). “The effect of traffic fl ow rates at freeway work zone on asphalt pavement construction productivity.” Transportation Quarterly , 57(3), 83-102. Jones, R. M. (2001). “Lost productivity: Claims for the cumulative impact of multiple change orders.” Public Contract Law J ., 31(1), 1-46. Lee, E. B. (2000). “Constructability and pr oductivity analysis for long life pavement rehabilitation strategies (LLPRS).” PhD thesis, Univ. of California, Berkeley, Calif. Liu, L. Y. (1991). “COOPS: Construction object-oriented simulation system.” PhD dissertation, Univ. of Mi chigan, Ann Arbor, Mich. Lu, M. (2003). “Simplified discrete-event simulation approach for construction simulation.” J. Constr. Engrg. and Mgmt. , 129(5), 537–546. Martinez, J. C. (1996). “STROBOSCOPE: St ate and resource based simulation of construction processes.” PhD dissertation, University of Michigan, Ann Arbor, Mich. Martinez, J. C., and Ioannou, P. G. (1999) , “General purpose systems for effective construction simulation.” J. Constr. Engrg. and Mgmt. ,125(4), 265-276. McCahill, D. F., and Bernold, L. E. (1993) . “Resource-oriented modeling and simulation in construction.” J. Constr. Engrg. and Mgmt. , ASCE, 119(3), 590–606.

PAGE 216

198 Minchin, R.E, Herbsman, Z., and Choi, J. (2003). A n economic evaluation based on total cost of aggregate base vs. asphalt base in the FDOT road construction operation . Florida Department of Transportation. (BC354-72.) Odeh, A. M. (1992). “Construction integr ated planning and simulation model.” PhD dissertation, University of Michigan, Ann Arbor, Mich. Oglesby, C. H., Parker H. W., and Howell. G. A. (1989). Productivity improvement in construction , Irwin/McGraw Hill, New York. Ott, L., and Longnecker, M. (2001). Statistical methods and data analysis , 5th Ed., Wadsworth Group, California. Peterson, J.L. (1981). Petri net theory and the modeling of systems , Prentice-Hall, New Jersey. Pidd, M. (1988). Computer simulation in management sciences , 2nd Ed., Wiley, New York. Resig, W. (1985). Petri nets: an introduction, EACTS monographs on theoretical computer science . Springer-Verlag KG, Berlin, Germany. Sawhney, A., AbouRizk, S. M., and Halp in, D. W. (1998). “Construction project simulation using CYCLONE.” Can. J. Civ. Eng., (25) 16-25. Sawhney, A., and AbouRizk, S. M. (1996) . “Computerized tool for hierarchical simulation modeling.” J. Comput. Civ. Eng. , 10(2), 115–124. Schwartzkopf, W. (1995). Calculating lost labor producti vity in construction claims , Wiley, New York. Seed, S. B. (2004). Flexible pavement de signsummary of the state of the art, Transportation Research Board, < http://gulliver.trb.org/publications/millennium /00040.pdf> (Jun. 2005). Shahin, M. Y. (2005). Pavement management for air ports, roads, and parking lots , 2nd Ed., Plenum, New York. Shi, J. (1999). “Activity-b ased construction (ABC) mode ling and simulation method.” J. Constr. Engrg. and Mgmt. , 125(5), 354–360. Shi, J., and AbouRizk, S. M. (1997). “R esource-based modeling for construction simulation.” J. Constr. Engrg. and Mgmt. , 123(1), 26–33. Smith, S. D. (1999). “Earthmoving producti vity estimation usi ng linear regression techniques.” J. Constr. Engrg. and Mgmt. , 125(3), 133-140.

PAGE 217

199 Tavakoli, A. (1983). “ Productivity analysis of he avy construction operations.” PhD thesis, Georgia Ins. of Tech., Atlanta, Ga. Thomas, H. R. (2000). “Schedule accelerati on, work flow, and labor productivity.” J. Constr. Engrg. and Mgmt. , 126(4), 261–267. Thomas, H. R. and Zavrski I. (1999). “Cons truction baseline produc tivity: theory and practice.” J. Constr. Engrg. and Mgmt. , 125(5), 295-303. Thomas, H. R., Riley, D. R., and Sanvido, V. E. (1999). “Loss of labor productivity due to delivery methods and weather.” J. Constr. Engrg. and Mgmt. , 125(1), 39–46. Thomas, H. R., and Sanvido, V. E. (2000). “Rol e of the fabricator in labor productivity.” J. Constr. Engrg. and Mgmt. , 126(5), 358–365. Thomas, H. R., Horman, M. J., de Souza, U. E. L., and Zavrski, I. (2002). “Reducing variability to improve performance as a lean construction principle.” J. Constr. Engrg. and Mgmt. , 128(2), 261-267 Thomas, H. R., Minchin, R. E., and Chen, D. (2003a). “Role of workforce management in bridge superstructu re labor productivity.” J. Mgmt. In Engrg. , 19(1), 9-16. Thomas, H. R., Horman, M. J., Minchin, R. E., and Chen, D. (2003b). “Improving labor flow reliability for better productivi ty as lean construction principle.” J. Constr. Engrg. and Mgmt. , 129(3), 251-261. Vorster, M. C., and De La Garza, J. M. (1990). “Consequential equipment costs associated with lack of availability and downtime.” J. Constr. Engrg. and Mgmt. , 116(4), 656-669. Vorster, M. C., Beliveau, Y. J., and Ba fna, T. (1992). “Linear scheduling and visualization.” Transp. Res. Rec. 1351, Transportation Research Board, Washington, D.C., 32-39. Wakefield, R. R., and Sears, G. A. (1997). “Petri nets for simulation and modeling of construction systems.” J. Constr. Engrg. and Mgmt. , 123(2), 105-112. Zink, D. A. (1986). “The measured mile: Proving construction inefficiency costs.” Cost Eng. , April, 19–21.

PAGE 218

200 BIOGRAPHICAL SKETCH Jaehyun Choi was born on April 18, 1973 a nd raised in Seoul, Korea. He graduated from Chungbuk National University in Korea in February 1999 with a Bachelors of Engineering degree from Architect ural Engineering. After graduation, he proceeded to the University of Florida in Ga inesville, where he has been working towards his masterÂ’s and doctoral degree in the C onstruction Management program of Civil Engineering since January 2001. During his graduate study, he has been a fforded the opportunities to get involved in two research projects granted by the Na tional Science Foundation and the Florida Department of Transportation. Other than his doctorate dissertation, he has published three refereed proceedings, and, at the time of graduation, two of his journal papers have been accepted for publication by prestigious journa ls in the area of Civil Engineering. He has also been appointed as a teachersÂ’ assist ant for various classes at the undergraduate and graduate level. In recognition of his academic achievements and contributions to the university, he received an Outstanding Intern ational Student Award from the University of Florida in April 2006. Upon receiving his doctoral degree, he will begin employment as a project controls specialist for the J acobs Engineering Group, Inc. in Greenville, South Carolina.