UFDC Home  Search all Groups  UF Institutional Repository  UF Institutional Repository  UF Theses & Dissertations  Vendor Digitized Files   Help 
Material Information
Subjects
Notes
Record Information

Full Text 
ANALYSIS OF A PROPOSED SIX INCH DIAMETER SPLIT HOPKINSON PRESSURE BAR By ELISABETTA LIDIA JEROME A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENT FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1991 ACKNOWLEDGMENTS I wish to thank my committee for their support and advice during this educational process. In particular I thank Dr. C. Allen Ross for his advice, support, and encouragement. Without him this effort would not have been possible. I would also like to thank my husband David for his patience and understanding during the many nights and weekends I spent at my desk or in front of the computer. I would like to acknowledge the financial support of the Air Force Engineering and Services Center at Tyndall Air Force Base. Finally, I wish to acknowledge the help and support of Sverdrup Technology Inc. during my educational endeavors. TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................... ii LIST O F FIG U R ES................................................................................................ v A B ST R A C T .............................................................................................................. viii CHAPTERS 1 INTRODUCTION.................................................................................... 1 O bjective.............................................................................................. 1 Background .......................................... ................ .............................. 1 A approach .............................................................................................. 3 2 MATHEMATICAL BACKGROUND ...................................... .......... 5 SHPB Assumptions ........................................................................... 5 Equations of Motion ............................................. ........................ 8 Solutions of the Frequency Equation............... .......... ........... .. 11 Transient Behavior............................................................................ 19 3 MATHEMATICAL AND NUMERICAL ANALYSIS............................ 21 Scaling Law s.......................................................... ............................. 21 Wave Propagation in an SHPB .......................................... ........... .. 25 Modified PochhammerChree Method....................................... 26 Finite Difference Numerical Simulation..................................... 35 Experimental Method ............................................................. 37 Com prison of R results ........................................................................... 38 Inertia and Friction Effects ............................................................. 70 4 SIXINCH SHPB FEASIBILITY .............................................................. 78 Sum m ary ............................................................................................... 78 Conclusions and Recommendations ................................... ........... 80 page .............................................. 82 REFERENCES APPENDICES A Modified PochhammerChree Program and Sample Outputs......... 86 B Ratio of Fourier Coefficients for Given a/A of 3 Bar Diameters...... 105 C Miscellaneous HULL Plots ................................................................... 107 D HULL Sample Input Files ..................................................................... 114 E Fortran Listing for Inertia Correction............................................... 117 BIOGRAPHICAL SKETCH ....................................................................................... 121 LIST OF FIGURES page 1. Compressive SHPB Schematic ................................................... ............... 1 2. Cylindrical Coordinates ..................................................... ......................... 9 3. PochhammerChree Solutions for First Three Modes................................... 15 4. Tyndall 2inch SHPB. Experimental Trace. Aluminum Specimen Length and Diameter 1.254 inch ...................................................... ............. 30 5. Modified PochhammerChree Results. Incident Stress Normalized to 1.0.................... ........................................................................................ .. 3 1 6. Normalized Longitudinal Stress Given as a Function of Radial Location for Several Fourier T erm s .......................................................................................... 34 7. Experimental Results. 2inch SHPB. Stress Pulses for a 6061T6511 Aluminum Specimen, Diameter = 3.18 cm, Length = 3.18 cm.................. 39 8. Experimental Results. 2inch SHPB. Stress Pulses for a 6061T6511 Aluminum Specimen, Diameter = 2.54 cm, Length = 2.54 cm.................. 40 9. Experimental Results. 2inch SHPB. Stress Pulses for a 6061T6511 Aluminum Specimen, Diameter = 0.56 cm, Length = 0.56 cm.................. 41 10. HULL Output. 2inch SHPB. Stress (TYY) and Strain (EYY) Pulses for an Aluminum Specimen, Diameter = 3.18 cm, Length = 3.18 cm....... 42 11. HULL Output. 2inch SHPB. Stress (TYY) and Strain (EYY) Pulses for an Aluminum Specimen, Diameter = 2.54 cm, Length = 2.54 cm........ 43 12. HULL Output. 2inch SHPB. Stress (TYY) and (EYY) Strain Pulses for an Aluminum Specimen, Diameter = 0.56 cm, Length = 0.56 cm........ 44 13. Experimental Traces at Three Incident Stress Levels (60 MPaTop, 100 MPaMiddle, 167 MPaBottom). Specimen Diameter 1.253 in., Length = .625 in., 2inch SHPB.................................. ............ 46 14. HULL Output. 2inch SHPB. 2inch Diameter and Length, Aluminum Specimen, 257 usec Pulse ......................................................... 48 a 1 15. Modified PochhammerChree Results. A 125 6inch SHPB. A 125 6inch SHPB. 750 sec Pulse.................................................................................................. 49 a 3 16. Modified PochhammerChree Results. A 250 6inch SHPB. 500 pusec Pu lse ..................................................................................................... 50 pWag a 3 17. HULL Output. a = 250 500/sec Pulse, Station 1 (Top) Bar Center, Station 2 (Bottom) Bar Surface ....................................... 51 a 3 18. Modified PochhammerChree Results. A125 6inch SHPB. 250 usec Pulse.. .............................................................................................. 52 a 3 19. HULL Output at Incident Bar Strain Gage. a = 125' 6inch SH PB 250 ysec Pulse.......................................................................................... 53 20. Experimental Traces at Three Incident Bar Locations for Two Striker Bar Lengths. 16inch (Top) and 25inch (Bottom), 2inch SHPB .................... 55 21. HULL Output. 2inch Bar, 250 /sec Pulse, All Stress Components on the Surface at Incident Bar Strain Gage.................................. .......... .. 56 22. HULL Output. 6inch Bar, 500 usec Pulse, All Stress Components on the Surface at Incident Bar Strain Gage.................................. .......... .. 57 23. Modified PochhammerChree Results at Incident Bar Strain G age. 2inch Bar ............................................................................................ 58 24. HULL Output. 6inch Bar, 250 usec Pulse. Surface (Top) and Center (Bottom) Transmitted Stresses ........................... 59 25. HULL Output. 6inch Bar, 500 ,usec Pulse. Surface (Top) and Center (Bottom) Transmitted Stresses ........................... 60 26. HULL Output. 2inch Bar, Longitudinal Stress Near the Impacted End (Top) and at the Strain Gage Location (Bottom) ................. 62 27. HULL Output. 6inch Bar, All Stresses Near Impacted End (Top) and at the Strain Gage Location (Bottom) .................................. ............. 63 28. HULL Output. 2inch Bar, All Stresses Near the Impacted End (Top) and at the Strain Gage Location (Bottom) .................................. ............. 64 29. HULL Output. 2inch Bar, All Stresses at Three Locations at the BarSpecimen Interface ........................................................................ 66 30. HULL Output. 2inch Bar, Stresses in Aluminum Specimen at Two Locations Near the Surface (Top), Near the Center (Bottom). Compare with Figure 29..................................................... .................................... 67 31. HULL Output. 2inch Bar, Transmitted Stresses at Three Radial Locations. Surface (Top), Middle (Middle), Center (Bottom). Compare with Figure 29 ..................................... .......................... .......... 68 32. HULL Output. 2inch Bar, Transmitted Stress with Zero Friction Between Specimens and SHPB ............................................. .............. 74 33. HULL Output. 2inch Bar, Transmitted Stress with Infinite Friction Between Specim en and SHPB......................................................................... 75 34. HULL Output. 6inch Bar, Transmitted Stress with Zero Friction............ 76 page 35. HULL Output. 6inch Bar, Transmitted Stress with Infinite Friction....... 77 36. Modified PochhammerChree Results. Radial Variations of Reflected Stresses. 2inch SHPB (Top), 6inch SHPB (Bottom) ................................ 88 37. Modified PochhammerChree Results. Radial Variations of Transmitted Stresses. 2inch SHPB (Top), 6inch SHPB (Bottom).................................. 89 38. HULL Output. 2inch SHPB. Radial and Longitudinal Velocities at Incident Bar Specimen Interface. (Bar Surface: Top, Middle of the Bar: Bottom, Bar Center: Right)............................................ 108 39. HULL Output. 6inch SHPB. All Transmitted Stresses at Bar Surface.............................................................................................................. 109 40. HULL Output. 6inch SHPB. All Transmitted Stresses at Bar Center. Compare with Figure 37 ...................................................................... 110 41. HULL Output. 2inch SHPB. 1inch Aluminum Specimen. Incident and Reflected Stresses............ ................................................. 111 42. HULL Output. 2inch SHPB. 1inch Aluminum Specimen. All Stresses W within the Specimen ..................................................................... 112 43. HULL Output. 2inch SHPB. 1inch Aluminum Specimen. All stresses in the Transmitter Bar................................................................. 113 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ANALYSIS OF A PROPOSED SIX INCH DIAMETER SPLIT HOPKINSON PRESSURE BAR By Elisabetta Lidia Jerome August 1991 Chairman: C. Allen Ross Major Department: Aerospace Engineering, Mechanics and Engineering Science The objective of this investigation was to determine whether a 6inch diameter Split Hopkinson Pressure Bar (SHPB) is feasible; whether the assumptions made in a typical 2 to 3inch SHPB still apply; and what chan ges, problems, and issues would be associated with an SHPB that size. The effort included a mathematical and a numerical analysis. Experimental data from a 2inch SHPB were used to compare and validate analytical models. A methodology based on the PochhammerChree frequency equation was developed by the author to study the wave propagation and wave inter action in any size SHPB. Additionally, a second order accurate finite dif ference code was used to mathematically understand the wave motion in a typical bar. The effects of friction and inertia were analyzed numerically, and general guidelines and corrections were developed. This investigation concluded that a 6inch SHPB is theoretically feasible and, in fact, identical to any existing system if the input pressure pulse duration is scaled appropriately. The modified PochhammerChree method can be used to assess the validity of any size SHPB and to study the wave interactions. If the specimen length and diameter are kept in the same order of magnitude, one need not worry about friction in the proposed 6inch SHPB. A correction will be needed to account for inertia, especially at strainrates of 50 to 100 sec' and above. The Split Hopkinson Pressure Bar has been and remains a reliable, accurate, and useful tool to study material response to loading, even for a new proposed system six inches in diameter. CHAPTER 1 INTRODUCTION Objective The objective of this investigation is to determine whether a 6inch diameter Split Hopkinson Pressure Bar (SHPB) is feasible; whether the assumptions made in a typical 2 to 3inch SHPB still apply; and what major changes, problems and issues would be associated with an SHPB that size. STRAIN GAGE INCIDENT BAR STRAIN GAGE TRANSMITTER BAR Figure 1. Compressive SHPB Schematic Background The Split Hopkinson Pressure Bar may be used to study materials at high strainrates in tension, shear and compression. A compressive SHPB is shown schematically in Figure 1. An SHPB system consists of a specimen sandwiched between and in contact with two elastic barsincident and trans mitter bar. The incident bar is impacted by a striker bar at a known velocity V STRIKER BAR causing a pressure wave to travel down the bar. This wave is partially trans mitted and partially reflected at the incident bar/specimen interface, and partially transmitted and partially reflected at the specimen/transmitter bar interface. Strain gages attached to the incident and transmitter bars record the strains associated with the reflected, transmitted, and incident pulses. From these strain data, stresses, strains, and strainrates in the specimen can be computed as a function of time. Although different sizes of SHPBs have been built, the largest one in existence in the USA is 3 inches in diameter. The main reason behind this size limitation is that it is generally assumed that the wave propagation in the bar is onedimensional, implying a high length to diameter ratio. A simpler form of the SHPB has been around since the late 1940s, when techniques and instrumentation were not as sophisticated as they are today. Historically, there may not have been a need to test a material specimen larger than 3 inches in diameter. However, in recent times the need for more economical structures to resist impulsive loadings associated with earthquakes, accidental explosions, and effects of conventional explosives has prompted more research into strain rate effects on the properties of concrete. Considerable data have been generated in 2 to 3inch diameter SHPB systems on concrete specimens containing small or scaled down aggregate. The question which arises from such small tests is, does the strain rate sensitivity, found for the small laboratory specimens, apply to the standard 6inch diameter large aggregate test specimens which are cored or cast in the field? There is a need to use a 6inch diameter SHPB type apparatus to test concrete specimens, which will match the concrete specimens which are cored directly in the field. The SHPB has proven to be a very effective and useful tool to conduct high strainrate material response studies since Kolsky [1] described its use in 1949. Detailed discussions of how a typical system works can be found in numerous papers including Hauser, Simmons and Dorn [2], Lindholm [3], Malvern and Ross [4], Robertson, Chou, and Rainey [5], and many others. Because of the inherent difficulties associated with a system of this kind, an idealized onedimensional wave theory is commonly assumed where stress is uniform over each cross section, plane cross sections remain plane during motion, and stress and strain are uniform throughout the length of the specimen. The above assumptions are adequate if the length of the bar greatly exceeds the diameter, in order to ensure no end effects and the loading pulse is much larger than the transit time of the specimen. However, factors like dispersion, radial and axial inertia effects, and friction between the specimen and the incident and transmitter bars can cause erroneous results if care is not used in either conducting the experiment or analyzing the data. Approximate corrections for the calculated onedimensional stress have been formulated to account for inertial and frictional effects. Kolsky [6] introduced a correction for radial inertia. Davies and Hunter [7] dis cussed both friction and inertia effects. Other papers of interest on this subject are those by Rand [8], Dharan and Hauser [9], Samanta [10], Jahsman [11], Chiu and Neubert [12], and Young and Powell [13]. The original SHPB apparatus was devised to study materials in com pression. More recently it was modified to test specimens in both tension and torsion. Examples of these techniques are discussed by Nicholas [14], Lawson [15], Baker and Yew [16], Jones [17], Okawa [18], Ross, Nash, and Friesenhahn [19], Rajendran and Bless [20], and Ross [21]. Follansbee and Frantz [22] used the PochhammerChree frequency equation to correct the recorded pulses for dispersion as they travel down the bar assuming only the first longitudinal vibrational mode is excited. Approach Before undertaking any research it is imperative to know and under stand what has already been done on the subject. This ensures no duplica tion of work and gives a good working foundation from which to start the research effort. The wave propagation in an SHPB is governed by the equations of motion. Generally, it is assumed the motion is onedimensional, and there fore is uniform in the radial and the circumferential direction. Although the onedimensional wave theory has been shown to be accurate for the existing systems, a mathematical analysis is needed for the proposed 6inch SHPB to assess whether this simple theory still applies. The twodimensional axisym metric cylindrical equations of motion need to be analyzed and solved (approximately), by various methods to understand the wave motion in a bar. There are no known exact solutions to the twodimensional wave propagation problem in a finite cylindrical bar, and therefore one must turn to a numerical scheme, namely a finite difference approximation. Modern computers make it possible to solve large problems in a reasonable amount of time. The plan was to first numerically simulate an existing SHPB in order to compare the results with experimental data and have confidence in the reliability of the numerical approximation. The second step would then be to simulate the proposed 6inch system and gather information on the wave motion. The final step will be a comprehensive look at the different methodologies, approximations, and assumptions to better understand the implications of increasing the size of the SHPB system. From this one can assess whether the data reduction of the current SHPBs still applies to the large bars, and what are the problems associated with a system two or three times the size of the existing ones. CHAPTER 2 MATHEMATICAL BACKGROUND SHPB Assumptions As discussed earlier, a onedimensional wave theory is generally assumed in the analysis of an SHPB. The onedimensional theory assumes longitudinal spatial variations only. Applying Newton's law for a displace ment u in the x direction of the diagram below, gives the equations of motion for an elastic system a  8a  + dx ax I  dx aa (a2u aA + (a + dx)A = pAdx ax atz or upon simplification au a2u ax =p at2 and Hooke's law du = Ee =E, ax where a is the stress, E is Young's modulus, E is strain, t is time, A is the crosssectional area, and p is the density of the material. Equation (la) can be rewritten as follows a2u / 2U Wt (E/p ) ax2 (lb) where E/p is the familiar elastic wave speed squared, or E/p =Co The general solution to (Ib) is of the form u = f(x + Cot) + g(x Cot) corresponding to forward and backward propagating waves. Assuming u = g(x Cot) only, differentiate u and obtain Ou u , Sg' and cog ax at where g' denotes differentiation of the function w.r.t. the argument (xCot). Eliminating g' between these two expressions yields Bu 9u S= Co = CoE at ax Strains are usually measured in an SHPB experiment at some distance from the specimen. To obtain the interface displacements, these recorded strains must be timeshifted to give the strains at the interfaces, and then integrated over time to give S= f Co (ei r) dt and u2 = Co tdt where the subscripts i, r, and t indicate incident, reflected and transmitted. Additionally, ul and u2 are the specimen's displacements at the incident and output bars interfaces respectively. The negative signs give positive dis placements for compressive (negative) strains. Let L be the length of the specimen; then the average strain in the specimen is Es = (U2 Ui)/L. This is sometimes approximated for short specimens and slowly changing stress by assuming equal stresses at the two interfaces, which implies that Et = Ei + Er. The specimen strain then becomes s= (2co/L) f trdt and the specimen strain rate is s = (2co/L) Er. By using Hooke's law, the stresses on either face of the specimen are os2 = E A and Usl = E (Ei + Er) A where is the ratio of the bar and specimen cross section areas. If the As interface stresses are not assumed to be equal, then Et # Ei + Er and the specimen strain and strain rate are given by S=( Co/L) f ( i +Er) dt and s = (co/L) (E i + r) There are some problems associated with such a simplified one dimensional theory, even for long slender rods, and these problems are accentuated for larger diameter and shorter length SHPB systems. First, longitudinal waves are dispersive, (wave speed is a function of frequency), which means the waveform that is recorded at a distance from the specimen is not the same waveform that actually arrives at the specimen. Wave disper sion can be corrected for, but is time consuming and the correction methods are not exact. The onedimensional theory also neglects stress and displace ment variations across the cross section. Since strains are measured only at the surface, this can also lead to errors. Equations of Motion To assess the validity of this simple theory, one can study the equa tions of motion that govern the wave motion in a circular cross section SHPB. The equations of motion are given in tensor form as d2u VT+pb= dt2 (2) where T is the stress tensor, b are the body forces (which are usually safely d2 neglected), and y is the acceleration term. In cylindrical coordinates (Figure 2), equation (2) becomes 02Ur dUrr rr UOO 1 UrO 8Urz p + + at2 8r r r 80 Oz p + r+ +2 (3) at2 8r r aO 8z r Suz a"rz + 1 BOoz + a0z +g at2 Or r 0 aZ r Figure 2. Cylindrical Coordinates Equations (3) represent the threedimensional motion in a cylindrical bar. The modes of vibration associated with an isotropic circular cylinder are in general very difficult to calculate and they depend on the specific boundary conditions as well as on the form of the input loading. If the axis of the cylinder coincides with the zaxis, two general types of vibrational modes exist. One set is symmetric and one is asymmetric. In this analysis, the assumption is made that since the striker bar impacts the incident bar squarely, then the motion should be axisymmetric. This means the solution is independent of theta; then u., a, and oz vanish, and the second of equa tions (3) is satisfied identically. It also means that the only possible modes of vibration will be symmetric. Using Redwood's [23] notation, these modes will be called Mm,n where m is either 1 (symmetric) or 2 (asymmetric) and n ranges from 1 (Young's modulus mode) to infinity. Therefore, in this study of an SHPB, only M1,n modes are considered. The constitutive rela tions for an elastic material can be written as aur (Ur az St +2 U)r (fr aUz ) Or =('t+ ")fi 7+f 9Ur Uz) arz a= ut + u) az ar (4) ouz (r OUr uz = (1 + 2u)z + +r aOrr aUrz aeeoo = r + + rr Or 9z where A and P are Lame's constants. Making use of (4) and neglecting theta dependence, equations (3) can thus be reduced to only two equations 2Ur [2ur 1 Our Ur 2Uz a2ur p = ( + 2) + (2 2+u) + at2 ar r Or r rz az2 (5) 2z 2Uz 1 Uz O2Uz 2Ur 1 OUr p =2 2 r + + 2,u) + ( + u) ++r r  t r r Or 0z2 Lz r 9z (5 The boundary conditions require no stress at the bar surface or arr =0 and Or = 0 at r = a where a is the radius of the bar. Also at a free end, Urz = 0 and Ozz = 0. For a colinear impact of one bar against another, the initial conditions can be in the form of a step change in pressure az = PH(t), arz = 0 where P is a uniform pressure and H(t) is the Heaviside step function, or as a harmonic displacement of uz = sinw t and arz = 0. The condition orz = 0 is correct only if the two bars are of identical diameter and material, so that the interface is a plane of symmetry of the stress state, until a reflected wave reaches the interface. Equations (5) have been solved for an infinitely long bar by Pochhammer and Chree for the propagation of sinusoidal waves and studied by numerous investigators over the years [6, 24, 25, 26, 27, 28, 29]. The solution is called the frequency equation and has the form 2y Jo(ha) (2y2 P) Ji(Ka) a = 0 (6) 2u 2 Jo(ha) Jo(ha) uya JI(Ka) aa A + 2/ aa where h2 = (2r/A)2 [pc /(A + 2u) 1],K2 = (27r/A)2 [(pc/p) 1], A = wave length, y = 2 Jo and J1 are the zero and first order Bessel functions A Cp respectively, w is the angular frequency, and cp is the phase velocity. Solutions of the Frequency Equation There are an infinite number of solutions to Equation (6), each cor responding to a unique mode of vibration. The solutions are exact only for an infinitely long cylinder although, for a cylinder whose length greatly exceeds its diameter, the errors are small. Furthermore, the solutions are not explicit but rather a function relating the phase velocity cp, the wavelength A and Poisson's ratio v. Now (6) can be written in implicit form as 2Uy J1 (Ka) 2y Jo (ha) 2y2 J (a). (7) 2u 2Jo(ha) 2 Jo (ha) =0 [ a2 A + 2/s By making use of the following properties of Bessel functions, a (Jo(ha))= Jl(ha) A (Ji (ha)) = hJo (ha) (J (ha)) /a, equation (7) can be rewritten as 4py (KJo(Ka) Ji(Ka)/a) (KJi(ha)) ((2y2 ) Ji (Ka) L2p. (hJo (ha) + J1 (ha)/a) +2Jo (ha) = 0, A + 2,u which finally becomes y2 JK o(a) 1 1 2+) + (K 2) Jo(ha)= (8) J1 (ca) 2 a 2 hJ1 (ha) Equation (8) has been solved numerically by several investigators [25, 27, 30]. It involves assuming a value of Poisson's ratio and then for varying values of frequencies, ranging from zero to some large number, equation (8) is numerically evaluated to obtain the values of the phase velocity. This turns out to be a very laborious task even with the aid of high speed computers. Results are usually shown in graphical form, namely in curves relating the phase velocity and the frequency, both in nondimensional form for general applicability. Figure 3 [22] shows the solution to equation (8) for a Poisson's ratio of .29. Only the first three vibrational modes are plotted namely M1,1, M ,2 and M ,. Recall that equation (8) only applies to symmetric vibrational modes. Mode M is of most interest since it appears to be the only mode excited in a typical SHPB. As the frequency of this mode approaches zero, or as the wavelength approaches infinity, the phase velocity approaches (v ), which is the elastic, onedimensional barwave speed, co. As the frequency becomes very large, the phase velocity ap proaches cr, the velocity of Rayleigh surface waves. Although it may not be clear from Figure 3, Cp, the phase velocity, approaches cr from the down side, implying that the phase velocity reaches a minimum which is less than the Rayleigh wave speed at some intermediate frequency. Recall that 2 A+2 2 2 E Cd Ct Co  P P c P h2 = (27t/A)2 (pc/(A + 2,u) 1) and K2 = (23r/A)2 (pc/ 1) and it is easily shown that co < ca and for M,1 cp < co. Then for the M1,1 mode the parameter h is always imaginary since, for all frequencies, cp < ed where Cd is the dilatational wave speed in an infinite medium. a For very low frequencies K is real, but past roughly = 0.65, when c becomes less than the shear wave speed in an infinite medium ct, K is imagi nary. This means that at low frequencies ( < 0.65), the displacement is mainly due to plane transverse waves, since the set of dilatational waves exists only as a surface disturbance. At high frequencies the total displace ment becomes increasingly like a pure surface disturbance [33]. Mode M1,2 and higher may be of interest in a "larger" SHPB and at high frequencies. These modes have what are called cutoff frequencies, i.e., frequencies at which the phase velocity becomes infinite and no real disturbance propagates in that mode. From equation (8) this implies either that y = 0.0 or that J1(Ka) = 0.0. If y = 0.0, equation (8) becomes 1 o 1 1 2 Jo (ha) 2 ct a 2 c2 hJ1 (ha) since ic= 1 = 1~ 1~ , 2 2 2 K2 ) C2 and K 2 = O for Cp = oo, where wco is the cutoff frequency. ct 2 Following the same reasoning h = , for cp = oo and thus one can write Cd" tJO (ti 0a (9) \Jo a c2 Wco J1 ad c Cd Cd If Ji (Ka) = 0.0, it can be said that J, (Ct a = 0.0 (10) [Ct ) Equations (9) and (10) can now be used to determine the cutoff frequencies for any mode M,n . From the solutions of the frequency equation, equation (8), it is possible to calculate the radial and longitudinal displacements for each mode at every frequency, and Poisson's ratio within the cylinder. Bancroft [25] calculated and plotted the amplitude of the longitudinal displacement a and showed that they are (as expected) uniform for =0.0 and the motion a is confined to the outside surface when approaches infinity. This agrees with the observations made earlier about mode M,.1 response in general. Numerous other approximate methods of varying degrees of com plexity have been developed to mathematically treat an SHPB system. They fall somewhere in between the simple onedimensional theory and the two dimensional, coupled differential equations just presented. Unfortunately, those that show good agreement with the PochhammerChree solution have very complicated descriptive equations, while those simple equations agree with the exact theory only over a very limited frequency range. 1.4 1.2 Cd/Co o ^ MODE M1 0 1,3 C 0 1.0 P= = SM Cr = RAYLEIGH VELOCITY 0.8 C2= MODE M P C/Co 1,1 0.6 0.4 0.0 0.5 1.0 1.5 2.0 a/A Figure 3. PochhammerChree Solutions for First Three Modes. When computers were not readily available and usable, a large num ber of approximate theories were developed. Most important to this study are the different methods and approaches used to derive these theories. As already discussed, the simplest analysis is to assume purely axial stress, uniform over each cross section. This leads to the familiar onedimensional wave equation in terms of the axial displacement u,, 2Uz a2 (11) at2 pz which predicts that waves of all frequencies travel at the same constant velocity co = v'O. A better approximation to this theory introduces a cor rection for the radial motion by considering the inertia of the cross section [23]. The approach involves the use of Hamilton's principle which states that the first variation of the integral of the Lagrangian, (T V), with respect to time is zero. Specifically 6 f (T V) dt =0 (12) and (T V) can be expressed as f0 pA +(vR) 2 a EA a dz ] (13) L0 Iii pA ( at ( az atj ^j2 az kinetic energy potential (strain) energy where L is the length of the cylinder, R is the radius of gyration about the zaxis, v is Poisson's ratio, and the velocities in the z and r directions are respectively _uz 8ur au_ and aur at at az But ur can be assumed in this approximation to be of the form v rz so that 8ur aUz S vr t at 0z 0t " The result of substituting equation (13) into (12) and solving via integration by parts is as follows a2Uz 2 a4uz a 2u P a (vR) a2a2t =E (14) P at2 t Z2 ( 14 This equation gives a better approximation of the exact theory than equation (11) at low frequencies. However, for short wavelengths the errors become considerable. A third approximate theory developed by Love [29] is based on the exact characteristic equation (8), where the Bessel functions are expanded in a power series. If a, the radius of the cylinder, is small enough so that ha and Ka are small compared to unity, then powers of ha and Ka higher than the second can be neglected. That is Jo(Ka) = 1 4 (Ka)2 Ji(Ka) = V2 (a) and the phase velocity becomes Cp = 1 v2a2) (15) and may be rewritten as v where c a (16) S=1 v2,2 where Co= V p Co X) a This theory agrees fairly well with the exact theory up to values of = 1 but then rapidly diverges. The assumptions that Kca and ha are small compared to unity imply that the wavelengths of the vibrations are large compared to the radius of the cylinder. Another common theory was developed by Mindlin and Herrmann [26]. It considers shear stresses and strains by assuming first that the radial displacement is of the form Ur = (r/a) u(z,t) with us = 0 and uz = w(z,t). Forces and moments are then calculated from standard engineering mechanics formulae and corrected for shear and iner tia by introducing factors K1 and K2. This leads to equations of the form 2 2 2u 2 22 Au 2 a2U a Kp  8K2( + ) u 4aK2 pa 2 az2 z 3t (17) au 2 a2a) 2 232) 2aa + a2 (A + 2u) =p a az az" at By substituting u = A exp (iyz) exp (iwt) and w = B exp (iyz) exp (iwt) into equation (17), two equations in A and B are obtained, from which A and B can be eliminated to again obtain a characteristic equation which relates phase velocity and frequency for the first two modes. By adjusting K2 and K1, a very good approximation can be obtained for mode M1,i Mode M1,2, on the other hand, shows considerable deviation from the exact theory. Many other theories have been developed and, in general, their com plexity is directly proportional to their accuracy. These analyses are impor tant because of the insight they give into the response of a cylinder subjected to impulsive loading. Knowing how stresses and displacements are dis tributed is crucial to the study of wave propagation and SHPB systems. Transient Behavior Thus far, the discussion has been limited to continuous waves or at least to pulses more than just a few cycles in length. However, in an SHPB experiment, a very short pulse is propagated through the cylinder, giving rise to a transient behavior which cannot be completely determined from the continuous theories; therefore, a different approach must be taken. One way to tackle the transient problem is to use Fourier analysis. By decompos ing the pulse into its continuous component, continuous wave theory can be used on each component at any point down the bar and the new pulse can be obtained by adding the pieces together again. This method is not exact, but it does give a close approximation, especially at some distance from the source. Choosing a mathematical function that describes the input pulse and that can be represented by a reasonably simple Fourier series is very dif ficult. Davies [27] used a trapezoidal pulse while Kolsky [28] assumed an error function. In both cases the PochhammerChree curves are used to obtain the phase velocity for each pulse, and in calculating the new shape of the waveforms, only the fundamental mode M1,1 was assumed to be ex cited. Another approach taken by some investigators in studying transient behavior is based on the concept of "dominant groups" and the method of stationary phase. The idea is that if at the beginning everything is in phase, any time after that all waves are out of phase, and therefore interfere destructively. When the combined effects of all these waves are studied, the 20 main contribution comes from a small "dominant" group whose phase velocities, periods, and wavelengths are almost the same. By focusing on this special group, one can obtain approximate propagation of longitudinal, flexural, and torsional waves in a cylinder. In this study a modified PochhammerChree method is developed using the Fourier analysis approach. It is straightforward and a much simpler numerical simulation than the existing complicated computer codes. This new approach is presented in the next section. CHAPTER 3 MATHEMATICAL AND NUMERICAL ANALYSIS Scaling Laws In order to analyze the propagation of waves in a cylindrical bar, one should ideally solve the threedimensional equations of motion. Unfor tunately, there is no known closed form solution to the problem. Even if axisymmetry is assumed, the twodimensional equations of motion cannot be easily solved. The problem is further compounded when a different material is introduced, such as when a specimen is introduced in an SHPB. The addition of two interfaces between dissimilar materials creates reflections and transmissions which cause waves to superpose, cancel each other, and in general interact with each other in a very complicated way. Pochhammer and Chree independently solved the equations for the propagation of a sinusoidal wave in an infinite bar. The solution, discussed in chapter 2, is the so called frequency equation; it has an infinite number of solutions, one for each mode, and results in a function relating phase velocity, wave length and Poisson's ratio. These solutions are valid and exact for infinitely long bars. If the bar is long enough to eliminate end effects (approximately 10 diameter lengths) and if the interest is in the transient pulse in its first passage, then the infinite assumption is reasonable. The PochhammerChree solutions can be plotted in nondimensional form as shown in Figure 3. It is interesting to note that since A = CpT, then a a ac a S and thus c where T is the pulse duration and is defined as T /c, where L is the striker bar length. Now if a is multiplied by a as T = 2 Ls/co, where L is the striker bar length. Now if a is multiplied by a constant factor and T is also multiplied by that same factor, then a plot of a vs.  will not change. What this means is that mathematically, under 0 the assumptions made, if the diameter is multiplied by a constant and the duration of the input pulse is simultaneously multiplied by the same con a a stant, then the solution for the original a is the same for both values. This applies to a sinusodial, continuous input where the wavelength is con stant, and travels at only one wavespeed relative to the bar radius: a long wave in a large bar will have the same wave speed as a short wave in a small a bar, as long as A stays constant. What happens if the input is a transient pulse? A simple mathemati cal function can be chosen to describe this type of input which can be represented by a Fourier series. A transient pulse is composed of a spectrum of frequencies; the higher frequency components travel more slow ly than the lower frequency components, and thus lag behind and cause the initial sharp pulse to spread. This spreading is called dispersion. The Poch hammerChree solutions give the velocity of each wave depending on its frequency. Thus, one may correct a given pulse for dispersion by repre senting the pulse by a series of frequency components, calculating how far each frequency component has traveled in a certain time, and then reas semble the pulse. This task basically amounts to correcting for phase chan ges within each term of the Fourier series. It is important to be able to correct for dispersion since the SHPB pulses are recorded on strain gages at distances typically 30 to 60 inches from the specimen. Interest is in the response at the specimen itself, but in many cases a strain gage cannot be physically placed there. The dispersion correction technique allows one to predict the shape of the pulse as it travels from the specimen to the strain gage. Further, following the same kind of reasoning as for the continuous wave case, a C a t can be written where n is the number of points A, c nAtco taken to represent the transient pulse and At is the time interval between them. In essence, nAt is the period of the transient pulse, and if a n At stays constant, the case is analogous to the continuous case. This means as long as the bar radius and the input pulse length are increased or decreased by the same amount, the response characteristics of a SHPB system will not change. This fact is important to this study because the main concern is the feasibility of a 6inch SHPB system. What the PochhammerChree equa tions show is that if the incident pulse is of large enough duration relative to the bar radius, then, as far as the longitudinal waves of the M,, mode are concerned, there is mathematically no difference between a 2, 3, 6 or 60inch diameter SHPB. One must then determine what pulse duration would be needed to make a new 6inch system equivalent to the existing 2 and 3inch bars and whether the stress, strain and displacement distribution of that pulse are satisfactory for a usable SHPB. As mentioned earlier, for a typical SHPB system, a uniform stress and strain distribution is assumed, and only strains on the surface are measured. It is then very important to identify under which conditions the onedimen sional assumption is acceptable. Davies [27] computed the ratio of the longitudinal displacement at the surface to that of the longitudinal displace a ment along the bar axis and showed that for less than 0.1 their dif ference is less than 5 percent. Thus for a continuous pulse, with one value of a A, onedimensionality is assured if A < 0.1. On the other hand, a tran sient wave contains many different frequencies, and in fact the majority of the a s will be greater then 0.1. Thus, if the transient wave f(t)is represented with a Fourier series expansion, we would generally express it in a form such as A o0 Ao + Dncos(nwotqo'.) f(t) 2 Dn cos (n ao t  2 1 where Ao and Dn are Fourier coefficients and p'n is the phase angle which is a function of the wave speed and the wavelength. Specifically, from [31] =c(ncox Co Pn = (P+ (nwx) [( 1i o c [n J where nwo is the angular frequency and c, is the phase velocity of the nth component. Equation (8) and Figure 3 give the relationship between c, and the frequency. The amplitudes in a Fourier series expansion are in decreasing order, meaning that Ao is the largest and Dn+1 is less then Dn. Furthermore, the frequencies of each component are also in order, they start at zero and go to infinity. In summing the Fourier terms, each term will contribute less and less to the overall pulse. This is why truncating the series at some large number, say 55, results in a very small error. Also, at some a point the nth component will have an A 0.1 To ensure onedimen a sionality in a SHPB the terms preceding the one with A = 0.1 must sum up to give almost the complete pulse. In other words, the amplitude D, of the a nth component having =0.1 must be a small fraction of the original amplitude Ao. Notice that it does not matter at which component a X reaches 0.1, but rather the value of the amplitude at that point must be small. If it is at the 10th or the 100th term in the Fourier series it makes no difference. The only question then is how small should this Dn amplitude be with respect to Ao, or what percentage of the whole pulse must be accounted for when a reaches 0.1. One way to answer this question is to look at the values for the existing SHPB systems at the University of Florida and at a Tyndall AFB. A short computer code was written to calculate and Fourier amplitude values for an assumed trapezoidal input based on the PochhammerChree solutions [32]. It was found (see Appendix B) that for each system the amplitudes of the Fourier components had fallen to 59 a percent of the original value by the time A reached 0.1. Therefore, a good minimum amplitude ratio to use for the proposed 6inch system is about 10 percent. This would assure the onedimensionality of the response. In conclusion, because of the form of the equations involved, a pres sure bar system can be mathematically scaled up or down to any degree, without affecting the characteristics of the wave motion. Analyses of the bars and input pulses used both at Tyndall AFB and at the University of Florida, showed that a 6inch system would be equivalent to the existing 2 and 3inch systems, if the input pulse duration was roughly 500 microseconds. The length of the striker bar is related to the pulse duration as 2L T = where co is the elastic wave speed and Ls is the striker bar length. The elastic wave speed co for steel is approximately 5080 m/sec. This means that for a 500 microsecond incident pulse in a steel SHPB, a 1.27 m (50 inch) long striker bar would be required. Wave Propagation in an SHPB Having an understanding of this very important fact, one can now turn to a more indepth look at wave propagation in a cylindrical bar, to broaden the understanding of the overall problem. Three different methodologies were followed to study this problem. The goal was to compare and contrast results, understanding the drawbacks, assumptions and reliability of each one. These three approaches were : A new numerical method developed by the author using the PochhammerChree frequency solution and referred to as a modified PochhammerChree method. A finite difference numerical simulation of the governing equa tions contained in an existing computer code. Experimental results from the SHPB at Tyndall AFB. The following sections are a general description of these three approaches. Modified PochhammerChree Method As previously discussed, Pochhammer and Chree independently solved the twodimensional equations of motion for an infinitely long bar assuming a sinusoidal waveform. Their solution is what is referred to as the frequency equation for longitudinal motion of an elastic rod. It turns out to be a function relating Poisson's ratio, wavelength, and wave speed, and can be solved for each mode of a cylindrical bar for a particular material. This solution is neither simple nor explicit. Beginning with the Pochhammer Chree frequency equation and assuming only the first mode (MI,1) is ex cited, the wave speed of any wave of a given frequency in the bar of interest, since it is dependent only on its wavelength, can be calculated. Using this wavespeed and corresponding wavelength along with given material proper ties, the longitudinal and radial distribution of displacements and stresses may be determined from the assumed displacement functions for the "lon gitudinal" vibrations. One can now turn to the transient pulse being propagated in the bar and describe it in terms of Fourier series; that means representing the waveform by the sum of an infinite number of pulses, each one having a distinct amplitude, a distinct wavelength and traveling at a unique speed. This wave speed can be calculated from the frequency equa tion. Therefore, each piece of the input pulse can be followed as it travels down the bar, and at any time, the complete waveform can be reconstructed numerically by adding up all the pieces. This methodology is currently being used to correct for dispersion in the SHPB at the University of Florida and at Tyndall AFB. To carry the logic a bit further, it appears one can still treat the pulses individually as they encounter the boundary between the specimen and the incident bar, partially reflect and partially transmit through, and eventually encounter the second interface where more reflec tions and transmissions take place. The basic assumption is that as the single Fourier component encounters an interface, its amplitude splits ac cording to the characteristic impedance values of the two media, (which will be derived later), but it will maintain its original frequency and velocity. In other words, each wave component divides each time it encounters an inter face, new components are formed each with a different amplitude, but all with the same frequency. Note also that the sum of these Fourier com ponents produce the total incident, reflected, and transmitted pulses whose wavelengths will also not change. The implications of this assumption are crucial to the conclusion that one can theoretically follow each wave component through multiple reflec tions, transmissions, and any length of travel, and at any time know its amplitude and location. With that information one can then reconstruct the pulse at any time or location and obtain the shape of the response. A computer program has been written for an SHPB following the logic just described and is included in Appendix A. A trapezoidal input pulse was assumed to represent the longitudinal strain at r=a and 2 meters from the specimen. Fiftyfive terms were carried through in the Fourier analysis. At the onset all the terms in the series are assumed to be in phase and their relative phases are calculated as they travel down the bar. After multiple reflections, the pulses are reconstructed at the two strain gage locations on the incident and transmitter bars (both 1 meter from the specimen). At any interface between dissimilar materials, velocities and forces must be con tinuous. Assuming normal incidence waves only, in a perfectly elastic medium, then the relationship between the incident, transmitted, and reflected stresses can be shown to be at = [2 Alp2 c2/(A2p2 c2 + Alpi ci )] oi r = [(A2p2 c2 Al p cl)/(A2p2 c2 + Al pi c) ] i where A is the bar crosssectional area, p is the density, and c is the wavespeed. Subscripts 1 and 2 pertain to the two materials, and the wave is moving from material 1 to material 2. Each term in the Fourier series has a different wavespeed, and since only Mode MI,i is considered in the analysis, these wavespeeds are always less than or equal to the infinite elastic speed VEIp. For the sake of consistency in comparing the various approaches, this effort was based entirely upon a compressive SHPB whose incident and transmitter bars are made of steel, and a work hardened aluminum specimen of varying dimensions was assumed. Thus, the characteristic impedance pc is approximately three times greater in the bars than in the specimen. Conse quently, an incident compressive wave will be reflected as a tension wave at the incident barspecimen interface and the transmitted part will still be in compression. As this compressive pulse reaches the end of the specimen at the transmitter bar, it will be both reflected and transmitted in compression in accordance with the mismatch of characteristic impedances of the bar and the specimen. This means that a compressive wave is now "trapped" in the specimen. It loses amplitude at each reflection, but it remains compressive. It also means that every time the wave reaches the specimenincident bar interface it will also transmit into the incident bar as a compressive pulse. The compressive pulse traveling out of the specimen into the incident bar superposes itself onto the reflecting tensile part and thus reduces its mag nitude. As the compression pulse makes multiple reflections and transmissions within the specimen, the magnitude of the original tensile reflection of the incident pulse is reduced by the magnitude of the trans mitted compressive pulse from the specimen into the incident bar. The length of the compressive incident pulse is of finite time, which means the tensile reflection and the compressive transmission pulses are also of finite length, and in fact are of identical duration as the incident wave. This implies the wavelengths of these transient pulses do not change as they cross interfaces, only their amplitudes are changed. These phenomena are clearly seen in experimental traces (see for example Figure 4), and in the Modified PochhammerChree results (see for example Figure 5) obtained from the program discussed in the previous page; the reflected tensile wave starts out at a certain value, and then in the time required to transit twice the specimen length, it jumps to a lower level due to the arrival of the first portion of the compressive pulse from the specimen. The amplitudes of the reflected and the transmitted pulses add up to the amplitude of the incident pulse. In the experiment of Figure 4 the specimen was plastically deformed, while for Figure 5 the specimen remains elastic. It should be mentioned at this point that in order to be able to directly compare the results, the dimensions of the bars and the locations of the strain gages were the same in all methodologies. Specifically, since the Tyndall SHPB is 2 inches in diameter and the gages are 40 inches on either side of the specimen, both the modified PochhammerChree method and the finite difference method were run with those values. For the 6inch SHPB, the numerical methods assumed strain gages 80 inches on either side of the specimen. C, So F: C, 1: I edW 'SS381S o o o o 0 0 0 0 ln in o I 4 Z Tz I0 II OE I o z rM o to ro NT O d dI I I qtowoooWm6l N iii iii TY S3SS3u.S 03ZnnryVION As mentioned earlier, the results of the PochhammerChree frequency equation also lead to the variation of the stresses and displacements in the radial direction. These variations also depend on the wave speed and the wavelength of the pulse, and there are an infinite number of solutions each associated with a particular mode of vibration. After some tedious deriva tions [33] the amplitudes W(r) and U(r) of the assumed longitudinal and radial displacements uz = W(r) f(t,z) and ur = U(r) f(t,z) are ^ Ji (ha) haJo(hr) (1 f x) /caJo(Kr) W(r)= iyc +I ha J J(ha) (x 1) J (Kc a) J1 (ha) x( ha J (hr) (1 x) 1 /aJ Ji ( r) U 0 ha (x )a J, (ha) (x 1) (2x 1) J1 (K a) where c = a constant determined by the amplitude of vibration, (Co) 12v h=y(fpx 1)V2, K/=y(2x 1)'/ With this information the radial, shear, and longitudinal stresses can be calculated as 2 [^J(ha) Jo (hr) a Ji(hr) =2 ha (1 x) ha1(ha) r J (ha)+ fix 1 Jo (Kr) xfx la J1(K r) 1 x J1 (K a) 1 x r Ji(/ca) ur = 2 ip y ch J1 (ha) (J1(ha) J1 ( a) f (t,z) 2 J(ha) Jo (ha) J1 (K a) z= 2uyc (ha) [(l+xfix) haJl(ha) + (1x)a J( ]f(t,z) Again assuming only Mode M1,1 is excited, the values of the wave speed for each component of the Fourier series under consideration can be calculated. Thus, if the bar is divided radially into a number of increments, it is possible to calculate the stress and displacement distributions in the bar at any point at any time. These calculations can be combined with those of the preceding discussion of the Modified PochhammerChree equation and the result is a complete description of the elastic wave propagation in an SHPB. Care must be taken when solving the equations involving Bessel func tions numerically, since as the argument becomes negative, modified Bessel functions must be used. Specifically Jo (ix) = Io (x) and Ji (ix) = i II (x), or in general , In (x) = (i)n Jn (ix) a Davies [27] showed that if the ratio is kept below 0.1 then the radial variation is less than 5%. In an SHPB there are an infinite number of pulses a whose W values span from zero all the way to infinity. As discussed earlier, it is sufficient to make sure that the amplitude of the Fourier components have fallen to a small fraction of the initial value, when the value of  reaches and goes beyond 0.1. That is, most of the contributing components a of the Fourier series must fall below = 0.1. A subroutine was added to the Modified PochhammerChree program to calculate the radial variations for the longitudinal and the radial stresses. The results of the longitudinal stress for a 2inch and a 6inch bar are plotted in Figure 6. As expected, the lower ( lower values of n), Fourier series terms S1 C' ii /H I C0 t o C I / / /  I I J I I of cu Ci Ocn 0e I UU LL IP i0 0o SLI I I I I Hn 1) 011 I II I I I I I I II I I I II I o 000000000 000000000 SS3I.S IVNIanlIIONO1i 3:ZInVVON (the ones whose values are low), show a nearly uniform distribution of a stress, while as the wavelengths become smaller ( larger ), the radial variations become progressively greater to a point where they change sign. If the amplitudes of these higher terms have fallen to a very small value, then their contribution to the overall pulse will be negligible. Plots of the radial variations of reflected and transmitted stresses are shown in Appendix A for a 2 and a 6inch SHPB (Figures 36 and 37). It is also apparent that as the diameter of the bar is increased and the input wavelength is kept the same, the stress will show uniformity for only a few terms. This fact is additional proof that the onedimensionality of an SHPB is dependent on the ratio of the radius to the wavelength, and if kept the same as for the existing SHPBs, then the wave propagations should be mathematically identical. Although the higher Fourier terms are highly nonuniform, their con tributions are very small. Furthermore, if one normalizes the amplitude values of the radial and the shear stresses to those of the longitudinal stres ses at the center, the results show the radial and shear stresses to be of many orders of magnitude smaller than the longitudinal values. This can also be observed in the finite difference calculations, and it is further proof of the onedimensional nature of the wave propagation in the bar. Also, as expected, when the diameter of the bar is increased to 6 inches, the radial and shear stresses also increase, but remain in the same order of magnitude, and can thus still be safely neglected. Finite Difference Numerical Simulation The twodimensional equations of motion in cylindrical coordinates can be solved using an existing hydrocode such as HULL [34]. HULL has a twodimensional Lagrangian module that solves conservation of mass, momentum, and energy simultaneously with the definitions of the deforma tion tensor, constitutive relations, and the stress deviators. A centered finite difference approximation is employed. The topic of constitutive rela tions is the most crucial as it includes the relationship between the elastic and plastic deviatoric strains and stresses, and the relationship between pressure, density, and internal energy. The accuracy of any hydrocode is highly dependent on the implementation of the constitutive relations. In this effort only metals were considered, namely steel and aluminum. Both are modeled in HULL as elastic/plastic materials, although the stress levels were kept low enough to keep the steel bars in the elastic region. Both materials are also considered isotropic. Stresses are decom posed into a spherical and a deviatoric component, and the generalized Hooke's law is used. Yielding is treated following the von Mises criterion, which assumes that yielding starts when the combinedstress distortion ener gy equals the yield value of distorsion energy in a tension or compression test. Specifically, it relates the second deviatoric stress invariant to a specified yield function. Once yield is reached, the deviatoric stresses are corrected and reset to be normal to the yield surface in stress space. The total stresses can then be calculated. Using HULL, a trapezoidal pressure pulse was input on the incident bar and the specimen was specified as an elastic/strainhardening material. The HULL output stations included the locations of the strain gages on the bars, plus one within the specimen itself. Sliding interfaces with no friction were used between the two materials for most calculations. To study the effects of friction, locked interfaces that allow no sliding to occur were used. The bars are made of steel and remain elastic at all times. The first runs were made to model as closely as possible the experimental data available for aluminum from the Tyndall AFB SHPB. Appendix D contains a HULL sample input file. HULL was also run to investigate the complicated wave interaction phenomena at the specimenbar interfaces, especially in cases when the specimen is smaller in diameter than the bars. The next step was to study the effects of a 6inch diameter bar, with two input pulses. One input pulse with a wavelength long enough to give a onedimensional response was used. Additionally, one with a much shorter wavelength was used to show how the stresses and displacements vary in the radial direction, and disperse after a short travel, making the conventional SHPB analysis invalid. Experimental Method How does one know how good or even reliable the HULL output is? In order to answer this question, a direct comparison was sought between existing experimental data obtained at Tyndall AFB, and the calculated results from the HULL code. Specifically, numerous compression tests were run on aluminum 6061T6511 specimens between steel incident and trans mitter bars. The input pressure was varied to assess strain rate sensitivity, and stresses and strains were recorded at strain gages located in the incident and transmitter bars. It was found that (as already known) this particular aluminum is strain rate independent; that is, the stress strain curve is not affected by the rate at which the load is applied. This is very important because the HULL code does not take strain rate into account. Therefore, a comparison will be meaningful between the experimental data and the com puter results for an aluminum specimen. The available data was obtained from a 2inch diameter SHPB and varying specimen sizes. HULL was then run with a new material added to the material library to match as closely as possible the 6061T6511 aluminum properties. In particular, aluminum is modeled as an elastic/plastic material, and one would expect to observe that stress strain relationship in the output. One should be able to recreate the relationship between stress and strain since that is what the SHPB is designed and used for. The rationale was to prove the validity of the numerical tool by comparing existing data with the results of the code, and then move on to the proposed 6inch bar and have the confidence to trust and draw conclusions from those results. Comparison of Results A number of compressive tests were conducted at Tyndall AFB using a 2 inch SHPB on aluminum specimens. The diameter and length of the specimen, as well as the input pressure were varied to assess their influence. The results for three cases are shown in Figures 7, 8, and 9. To compare experimental and numerical results, HULL was run matching as closely as possible the conditions in the Tyndall tests. One needs to realize however, that no matter how good the numerical model is, there are some parameters one cannot predict or reproduce. For example, the input pulse is not in reality a perfect square pulse, nor is it perfectly uniform along the cross section as numerically modeled. Furthermore, the constitutive equations for steel and especially for aluminum (which goes beyond yield in the course of the test), are theoretical models and by no means exact. Finally, the effects of friction are not included in the numerical scheme. Having said that, one is then looking for a qualitative match in the data, more so than exact values of stresses or strains. What this means is that certain trends that have been observed experimentally should show up numerically. For example, as the specimen gets smaller compared to the incident bar, experimental results show that the transmitted stress also be comes smaller. Consequently, the reflected stress increases and its shape becomes more uniform as the initial high peak starts to blend in with the rest of the reflected pulse. The HULL calculations also show these phenomena. Although the actual amplitude values do not match, the relative change between tests does. In general, the experimental data gives higher trans mitted stresses, but it does so consistently. Experimental results are shown in Figures 7, 8, and 9 and can be compared to the numerical results from HULL shown in Figures 10, 11, and 12. The results of the HULL calculations show very good agreement with the experiments, which point out the validity of the HULL code. Ei 00 I *G '0 rij 4O ! ce C 0 4"S *4 I3 o o o 0 o 4 1 edW 'SS3UIS In j o o i o o o a o o o0 0 0 0 0 0 I I edW 'SS3EIS S E I. C/3 ct2 6 r ~II UI 13 j eE ..c~ P, o0 Uo *a <* t a 03 o ia a Uj  I 1 I S: o . < In SI  a E z a  cu a iu Cu RdW 'SS3u.LS 2.01E+00 1.6gE+00 1.21E+00 8.01E01 4.01E01 0.0 4.01E01 8.01E01 1.21E+00 1.61E+00 2.01E+00 0.00 1.00E03 8.00E04 6.00E04 4.00E04 2.00E04 00 2.00E04 4.00E04 6.0OE04 8.00E04 1.00E03  0.00 0.11 0.22 0.33 0.44 0.55 0.66 0.77 0.88 0.99 1.1 T (MSEC) SPUT HOPKINSON BAR 0.4 CM ZONING Problem 4.0000 SSTA= 1 XO= 2.33 YO= 99.80 MAX= 5.372E04 MIN=8.220E04 0.11 0.22 0.33 0.44 0.55 0.66 0.77 0,88 0.99 1.1 T (MSEC) Figure 10. HULL Output. 2inch SHPB. Stress (TYY) and Strain (EYY) Pulses for an Aluminum Specimen, Diameter = 3.18 cm, Length = 3.18 cm STA= 1 XO= 2.33 YO= 99.80 MAX= 1.287E+00 MIN=1.969E+00 2.01E+00 1.61E+00 1.21E+00 8.01E01 4.01E01 0.0 4.01E01 8.01E01 1.21E+00 1.61E+00 2.01E+00 0.00 1.00E03 8.00E04 6.00E04 4 00E04 2.00E04 0.0 2.00E04 4.00E04 6.00E04 8.00E04 1.00E03 0.00 0.11 0.22 0.33 0.44 0.55 0.66 0.77 0.88 0.99 1.1 T (MSEC) SPUT HOPKINSON BAR 0.4 CM ZONING Problem 3.0000 STA= 1 XO= 2.33 YO= 99.80 I MAX= 6.267E04 MIN8.215E04 0.11 0.22 0.33 0.44 0.55 0.66 0.77 0.88 0.99 T (MSEC) Figure 11. HULL Output. 2inch SHPB. Stress (TYY) and Strain (EYY) Pulses for an Aluminum Specimen, Diameter = 2.54 cm, Length = 2.54 cm STA= 1 XO= 2.33 YO= 99.80 MAX= 1.500E+00 MIN=1.968E+00 2.01E+00 1.61E+00 1.21E+00 8.01E01 4.01E01 0.0 4.01E01 8.01E01  1.21E+00 1.61E+00 2.01E+00 0.00 1 00E03 8 OOE04 * 8.00E04 4 00E04 4.00E04 2.00E04 0.0 2.00E04 4 00E04 6.00E04 8.00E04 1.00E03  0.00 STA= 1 X0= 2.33 YO= 99.80 MAX= 1.666E+00 MIN=1.973E+00 0.11 0.22 0.33 0.44 0.55 0.66 0.77 0.88 0.99 1.' T (MSEC) SPLIT HOPKINSON BAR 0.4 CM ZONING Problem 6.0000 STA= 1 XO= 2.33 YO= 99.80 MAX= 6.938E04 MIN8.238E04 0.11 0.22 0.33 0.44 0.55 0.66 0.77 0.88 T (MSEC) 0.99 1.1 Figure 12. HULL Output. 2inch SHPB. Stress (TYY) and Strain (EYY) Pulses for an Aluminum Specimen, Diameter = 0.56 cm, Length = 0.56 cm It makes sense physically that as the specimen gets smaller, more and more of the incident pulse is reflected back, to the limit case when there is no specimen and the reflected trace is identical and reverse of the incident. It also seems reasonable that when the specimen diameter is less then half of the SHPB diameter, little of the incident pulse gets transmitted into the specimen, which means that even less of this transmitted pulse will get back to the reflected pulse and therefore its basic shape will not be changed. Hence the appearance of a distinct peak in the reflected trace is reduced when the specimen is small. It has already been shown in Figure 5 that the Modified Pochhammer Chree method also shows the characteristic "peak" in the reflected trace even though the solution is elastic. The only difference between an elastic and an elasticplastic specimen is in the magnitude of the stress that may be applied. Therefore, the only difference in the shape of the traces in the modified PochhammerChree method will be a higher compressive reflec tion out of the specimen, resulting in a lower reflective plateau after the initial "peak", than for the corresponding nonelastic case. To show this, three experimental tests were run with the same specimen configuration at different levels of stress; one level was kept in the aluminum elastic region, the other two exceeded the yield value for aluminum. Figure 13 shows how after the initial peak in the elastic case (top figure), the reflected pulse almost goes down to zero, while as the input stress is increased (middle and bottom figures), the plateau after the peak increases also. Another issue to be explored experimentally and numerically is the effect of decreasing the pulse duration with respect to the diameter of the bar. As discussed, the equations governing an SHPB show that increasing the ratio will affect two important things. The first thing is that the pulse will disperse more; that is, the contribution from the slower traveling com ponents will be greater, and the overall waveform will distort and spread out more as it travels. The second thing is that the wave propagation will be less one dimensional in nature, which means that the longitudinal stress will have <REFLECTED 0.0 o1 <INCIDENT TRANSMITTED 50 25.0 i In 50.0  0.0 250.Ous 500.Ous 750.Ous TIME <REFLECTED 0.0 i 50.0  /<TRANSMITTED <INCIDENT 100.0  I I I I I I I I I I I I 0.09 250.Ous 500.0us 750.0us TIME I T r n I m I I I I I I I I I _ 100.0  .REFE CTED U) TIME Figure 13. Experimental Traces at Three Incident Stress Levels (60 MPaTop, S aiddle, 167 MPaBottom). Specimen Diameter 1.253 Length = .625 in, 2inch S B 0.05 250.0us S00,us 750.0us Figure 13. Experimental Traces at Three Incident Stress Levels (60 MPaTop, 100 MPaMiddle, 167 MPaBottom). Specimen Diameter 1.253 in., Length = .625 in., 2inch SHPB larger variations along the radius and the radial and shear stresses will be a less negligible as increases. Numerically one can observe the effects of decreasing the input wavelength by using HULL and the Modified Pochham merChree method. The following runs shown in Table 1, were made to test the methodologies against what it is analytically believed to be true. Table 1. HULL and Modified PochhammerChree Runs DIAMETER WAVELENGTH a FIGURES (inches) microsecondss) A 2 250 1/125 3, 14 6 750 1/125 15 6 500 3/250 16, 17 6 250 3/125 18,19 Figures 5, 15, 16, and 18 show results from the Modified Pochhammer Chree program, while Figures 14, 17, and 19 are HULL output traces. Notice that HULL appears to have two different coordinate systems from one figure to the next. The reason for this is that the runs were made on different computer systems (VAX and CRAY) and each system has a slightly different version of HULL. In one z is the longitudinal coordinate, and in the other, y is the longitudinal coordinate. This discrepancy should not a cause confusion. It can be seen that the pulses disperse more when s is greater. a The results also confirm that if X is the same, regardless of the bar 2.41E+00 STA= 1 XO= 2.33 YO= 99.80 1.81E+00 MAX= 1.743E+00 MIN=3.323E+00 1.21E+00 6.01E01 S0.0 6.01E01 S1.21E+00 1.81E+00 2.41E+00 3.01E+00 3.61E+00 0.00 0.11 0.22 0.33 0.44 0.55 0.66 0.77 0.88 0.99 1.1 T (MSEC) SPUT HOPKINSON BAR 0.4 CM ZONING Problem 5.0000 1.50E03 STA= 1 XO= 2.33 YO= 99.80 1 20E03 MAX= 7.268E04 MIN1.386E03 9.00E04 3.00E04 0.0 3.00E04 6.00E04 9.00E04 1 20E03 1 50E03 0.00 0.11 0.22 0.33 0.44 0.55 0.66 0.77 0.8B 0.99 1.1 T (MSEC) Figure 14. HULL Output. 2inch SHPB. 2inch Diameter and Length, Aluminum Specimen, 257psec Pulse o d000o dd000o d0  I I I I I I l I I I I S3SS3 IS C3ZrrnVW ON I C   C3 3f CO 0 0 YC ' r. IS LL) U a w Oi I > Q 0I 0 I I I I I I I I 1 I I 0 0 0 0 0^ 8 i s r "^ ^ ^ ' n^nro Sc~c o SSMuLS iaMGIONI O3ZrWNON T (MSEC) Figure 17. HULL Output. = 500 psec Pulse, Station 1 (Top) Bar Center, Station 2 (Bottom) Bar Surface I O2 c 0 0 E I 0. Z< CD W 00 z z N E I I I I I I I I I I I I I I I I 0 LI u4) N7 P 4n I l I I I TI I S3SS3ylS a3ZnVMviON I ~ : II 1+ 0 o Bo T ai 00 x =0 N LV N 00 t 0 I Cb 0 W 4 X 0 0 0 0 0 0 e'4 0 0 oi N N ID I I (B)) ssais diameter, the waveforms are mathematically the same; this is shown in Figures 5 and 15. Figures 18 and 19 indicate that an input pulse of 250 microseconds for a 6inch SHPB is not recommended. It appears dispersion becomes a problem, even after a very short time, and also the onedimen sionality of the motion is questionable. Experimentally, one can achieve different wavelengths by changing the length of the striker bar. The result ing waveforms for a 16inch and a 25inch striker bar are shown in Figure 20 ( all three bar locations are on the incident bar ). Notice the more pronounced dispersion for the shorter wavelength. As discussed, along with more wave dispersion, the one dimen sionality of the motion in a bar is also affected by the change of A values. a Again, one can turn to the two numerical schemes and observe how, as  increases, the radial and shear stresses increase, and the longitudinal stress shows more variations along the radial direction. Figures 21, and 22 show all stress components on the surface of the incident bar at the strain gage location for a 2inch bar and a 250 microsecond pulse, and a 6inch bar with a 500 microsecond pulse. Incident and reflected stresses are shown. Al though the radial and shear components of the stresses are higher for higher values of , they nevertheless remain negligible when compared to the longitudinal stress. Figures 23, 24, and 25 show the pulses at the center and at the surface of the transmitter bar at the strain gage location. Figure 23 is the result of the modified PochhammerChree method for a 2inch bar and 250 microseconds. Figures 24 and 25 are HULL output for 250 and 500 microsecond input pulses respectively. Figure 23 shows a difference of about 1.5% between the two curves. The variation in stress along the cross section is roughly 5% for Figure 24 and 2.5% for Figure 25. Therefore, the a uniformity of the longitudinal stress is degraded as A is increased, although for the proposed 6inch bar and a 500 microseconds pulse, the wave motion should be very close to onedimensional. distances from striker bar end 0.0001 0.0003 0.0005 0.0007 Time (sec) 0.0001 0.0003 0.0005 0.0007 Time (sec) Figure 20. Experimental Traces at Three Incident Bar Locations for Two Striker Bar Lengths. 16inch (Top) and 25inch (Bottom), 2inch SHPB 1 I 0.0001 1 I 0.0001 0.0009 0.0009 2.41E+00 1.81E+00 1.21E+00 6.01E01 0.0 6.01E01 1.21E+00 1.81E+00 II 2.41E+00 3.01E+00 3.61E+00  0.00 REFLECTED STA 1 XO= 2.33 YO= 99.80 7 7 YY DENT * 0.11 0.22 0.33 0.44 0.55 T (MSEC) 0.66 0.77 0.8B 0.99 1.1 Figure 21. HULL Output 2inch Bar, 250 usec Pulse, All Stress Components on the Surface at Incident Bar Strain Gage 0.75 T (MSEC) Figure 22. HULL Output. 6inch Bar, 500 ,sec Pulse, All Stress Components on the Surface at Incident Bar Strain Gage Wm I uJm LU N o S I I I I I I I I I I I I I 0 LLJ I o' d^ 1 c o Io 66I I I I I  I ill iiI I r 2  \ i ^ * ^~~~~~ ~~ rM LC MK ^ O ior ac i iK o d d d d l ^ ^ S3SS3I1S 1N3aIONI a3ZInVINON 3.00x 10' STA= 4 XO = 7.410E+00 MAX = 2173E01 YO 4.998F'.; ,: ",:"', o.oo " 03.0010 6.00x10" 9.DO00xD' 1.20 1.50 1.80 Z 2.10 2.40 2.70 0.00 0.15 0.30 0.45 0.60 0.75 0.90 1.05 '.20 1.35 15 TIME (MSEC) 3.00x 0' O.03.00 x 10' YO4 STA= 5 XO = 1.900E01 MAX 2 69 l1 YO 4.99BE+02 MI 2. 1 0 0.00 Z 6.00x I ' 9.00x1 0' 9.00 ID" 1.20 Z LA 2 1.50 2.10 1.80 L HP = "hoop" stress I 2.40 2  i  i  i   2,70 0.00 0.15 0.30 0.45 0.60 0.75 0..90 1.05 1.20 1.35 1. TIME (MSEC) Figure 24. HULL Output 6inch Bar, 250 ,sec Pulse. Surface (Top) and Center (Bottom) Transmitted Stresses 1.21E+00 8.01E01 4.01E01 0.0 4.01E01 8.01E01 1.21E+00 1.61E+00 2.01E+00 2.41E+00 2.81E+00 0.00 1.21E+00 8.01E01 4.01E01 0.0 4.01E01 8.01E01 1.21E+00 1.61E+00 2.01E+00 2.41E+00 2.81E+00 0.00 0.15 0.30 0.45 0.60 0.75 T (MSEC) 0.15 0.30 0.45 .60 0.75 0.90 1.05 1.20 1.35 1.5 T (MSEC) Figure 25. HULL Output 6inch Bar, 500 ,sec Pulse. Surface (Top) and Center (Bottom) Transmitted Stresses STA= 3 XO= 7.41 YO= 399.80 MAX= 6.370E01 MIN=2.442E+00 STA 4 X0= MAX 5.477E01 0.90 1.05 1.20 1.35 1. A tool like HULL offers the advantage of being able to "look" at the wave response, stresses, accelerations, etc., anywhere in the bar. In particular, two questions regarding the operation of an SHPB have been of interest for some time. The first one has to do with the initial oscillations that are observed experimentally and whether they are due to something inherent to the equations of motion, or whether they are the result of some experimental characteristic like friction, or striker misalignment. The second question has to do with what really goes on at the interfaces between the specimen and the bars, especially when the specimen is smaller in diameter. To answer the first question one can look at the HULL stress output at points very close to the impacted ends. Figures 26 through 28 show curves for various diameter bars and various input pulses at two locations on the inci dent bar, one near the impacted end, and another one meter from it. The a following observations can be made; in the cases where A is larger, the oscillations in the input pulse are more pronounced and they take longer to damp out. Further, if one compares these results with the experimental traces (Figure 20), the similarity in the response is striking. Since the finite difference calculation does not take into account friction, misalignments, etc., it can be deduced that the so called "end effects" are due to inherent characteristics of the wave motion not to the way the experiment is carried out. That means higher frequency modes are excited at the impacted end of a the bar. If one refers back to Figure 3, the cases with larger show more oscillations because the response is further to the right on the xaxis, and thus more modes of vibrations are possible. Nevertheless, as predicted by Davies [27] and others, these high frequency modes damp out quickly, so that after about ten diameter lengths the only mode of vibration present is mode M1,1. The second question presents more of a problem; the wave interac tion at the barspecimen interfaces is extremely complicated, especially if the specimen yields at some point during the process. Looking at the stress a co a, I c=._e C.4 o o i s i3J oo o o o o o o o o a x X X X X xxx xx x X X X 0 0 0 0 0 x O O O x 0 0 0 0 0 0o 0o 0 0 0 0 6 o o 0 0 0 0 'o (33/3NNO) ZZi (33/3cNa) ZZL 1 .0 STA= 1 XO = 7 410E+00 MAX = 9.756E0 YO 5.000E+00 MIN 3 328E+0( 1.00 5.00' 10' 0.00 5 .00 0 ' 100 2 z z 1.50 2.50 z 2.00 z z HP "hoop" stress a 0.00 0.i5 .30 045 0.60 0.75 0D 1.05 1 20 135 TIME (MSEC) 1.50 STA= 2 XO = 1.900E01 MAX 9. YO 3.002E+02 MIN 3. 1.00 5.00x 10  0.00 5.00x 10'  1 .00 1.50 Z 2.00 2.50 3.50 0.00 0.15 0.30 0.45 0.60 0.75 0.90 1.05 1.20 1.3 TIME (MSEC) Figure 27. HULL Output. 6inch Bar, All Stresses Near Impacted End (Top) and at the Strain Gage Location (Bottom) 64 1 50 Sl X0 2 32tO MAx 1 217E00 YO 5000E+00 MIN 3 25E 0N 5.00 10 p 1Z S00 5 00O '1  150 2. 00 r z 2 50 J.oo ;z HP = "hoop" stress = o 3 s0 000 0a15 030 0 45 060 0.75 090 1,0 1 20 135 I 5 TIME (MSEC) SPLIT HOPKINSON BAR LOCKED NODES Problem 5 1000 3 50 3.50 I 2 xO 2 .32E+00 MAX 2 474E+00 O 1 002E402 MIN 3 323E 00 2.80 1 4( z 7 0C. 10' 0 00 0 15 0 0 0 45 060 0.75 090 1.05 1 20 1 5 I TIME (MSEC) Figure 28. HULL Output. 2inch Bar, All Stresses Near the Impacted End (Top) and at the Strain Gage Location (Bottom) components along one such interface from a HULL calculation, it is very difficult to make any definite statement. To help in the understanding of the complicated stress wave interaction, Figure 29 shows the important features and magnitudes of all stress components along the incident barspecimen interface. The longitudinal stress (uniform and in compression, of mag nitude 167 MPa sometime prior to this interface), at the interface varies from almost zero at the bar surface to 100 to 80 MPa at the center. Shear and radial stresses become of the same order of magnitude as the lon gitudinal stress. What is worth noticing is that at the center aeo or the "hoop" stress is very large, in fact larger than the input stress. Also at this point the displacements in the longitudinal and radial direction are small. It appears then, that shear and circumferential stresses play an important role at the edge where the specimen ends against the bar face. In a sense, the incoming stress has to "squeeze" itself through a smaller cylinder therefore it is not surprising to find other than longitudinal stresses at the interface. What is remarkable is how quickly it recovers to an almost uniform state of stress within the specimen only one centimeter from the interface (Figure 30), and in the transmitter bar at the strain gage location (Figure 31), where the longitudinal stress is the only non zero component. Figures 30 and 31 cor respond to the same SHPB as in Figure 29. Notice that a00 in Figure 30 is only about 50 MPa while it peaked at 220 MPa at the interface; that amounts to a reduction of 77% in one centimeter. Appendix C contains more HULL output traces which can further explain the above discussion. Finally, Bertholf and Karnes [23] pointed out that the first interface between the bar and the specimen effectively filters out most of the oscilla tions in the incident pulse, and thus in the cases when these oscillations are large for one reason or another, it is wise to use only the transmitted stress to calculate the response in the specimen. The analytical tools also show this filtering effect of the specimen, and it appears the use of the transmitted pulse only, may be a worthwhile alternative for the proposed 6inch SHPB. 0 6 soo 2l01E0. 001E StcEor S 0EOf 0I0'o O0  OI401 40 Eon wow sorol 2 076of I 00 0 t o6 01 O 0:2 O. 0 0o** 0 s o00 o 0.8 000 T (inco Figure 29. HULL Output 2inch Bar, All Stresses at Three Locations at the BarSpecimen Interface STATION 1 STATION LOCATIONS 3 SPECIMEN t STATION 3 I H.0 \ I ot 0oi oU 0* o aO or o 0 0 i T acm) STATION 2 ooo~~~~~~~ ~ 05621aJ11 ss ooa o HULL Output 2inch Bar, Stresses in Aluminum Specimen at Two Locations Near the Surface (Top), Near the Center (Bottom). Compare with Figure 29 9 00x10 ' 2.00 1.50 5.00x10 1.00 1.50 2.00 250 3.00 0. Figure 30. 3,0010* 5A 3 X 2.328a*00 WX 2 774E+08 Y 3.000E+02 MIN 4 50E+SO .80. 10 s50.to' 540o000 0o1 030 0o45 0.60 0.75 0.90 .0 1.20 1,35 TnE (.EC) 3 0010, 2 70., ' 2 TO"ID* 3600' 4 lD' E r  00oo 0' I 01.10' 2 70l0' 360 l00 " 454010 ' Figure 31. HULL Output. 2inch Bar, Transmitted Stresses at Three Radial Locations. Surface (Top), Middle (Middle), Center (Bottom). Compare with Figure 29 The two analytical methodologies used have advantages and disadvantages, and the use of one over another depends on the particular need and applica tion. The finite difference solution is the most accurate if the constitutive relationships of the materials involved are adequate. It is also the most time consuming and expensive. The Modified PochhammerChree method developed in this study gives a good qualitative look at the wave propagation in a typical SHPB in a very short time. It can be reliably used to assess the onedimensionality of the elastic wave motion, as well as dispersion effects and cross sectional variations and values of all the stress components. It is a tool that can also be used to investigate if a particular experimental setup is affected by friction or inertia by predicting what would happen in the ab sence of friction and inertia. Furthermore, a typical HULL calculation took anywhere between three and six central processing unit (CPU) hours on a CRAY YMP computer at an average cost of $800.00 per job! The modified PochhammerChree method takes seconds to run on most machines at an average cost of less than $2.00. Although the Modified PochhammerChree method assumes a purely elastic response, it does not affect its usefulness since one would not use any of these analytical tools to investigate true material response, but rather to study the reliability of a certain SHPB configuration. These numerical codes are not intended to replace the ex perimental method, but rather to study the validity of the data reduction, the implications of a geometric change, the effects of a different input pulse, and so on. After all, if the true constitutive relations of the specimen material were known one would not need to study its response with an SHPB. The Modified PochhammerChree method can give the results it was designed to give with any elastic specimen, as long as the characteristic impedance of the material and the bars are known. Inertia and Friction Effects The analytical methods discussed are useful tools to study the wave propagation of any SHPB, and within their limitations and assumptions can assess the one dimensionality and therefore the usefulness of different bar sizes and configurations. However, the reliability of an SHPB can be influenced by radial and axial inertia and by friction between the specimen and the elastic bars. Neither the modified PochhammerChree method nor the finite difference code HULL take those into account. Approximate cor rections have been derived to account for the effects of radial and axial inertia [7] and friction [8]. The objective of this investigation is to under stand when and how inertia and friction become important factors and whether the SHPB diameter is a key parameter on inertia and/or friction contributions. In other words, one would like to know whether increasing the diameter of a SHPB will increase the effects of inertia or the effects of friction. Inertia effects become more pronounced as either the initial stress in the input bar becomes high, or the strain rate is increased [24]. The former causes very large stress and strain gradients which from the equations of motion, make the inertia term become important [35]. High strainrates do not appear to cause inertia effects per se, rather the way the higher strain rates are achieved can cause inertia effects to become important. Specifical ly, if one increases the input stress to obtain a higher strainrate, and this input stress is in the form of a step function, then apparently the combina tion of high input stress, and the oscillations caused by radial waves resulting from the short rise time of the input, causes inertia effects to become quite important. Bertholf and Karnes [24] eliminated one of these two causes by, numerically, using a ramp input function with a much longer rise time. Their results show that doing this effectively eliminates the inertia effects, even though the same levels of strain rates were achieved. This means that the magnitude of the input stress is not a major cause of inertia effects, rather the short rise time associated with the high input seems to significantly contribute to inertia effects. Therefore, since in an SHPB one is limited to a step input pressure, there appears to be a maximum achievable strain rate beyond which the validity of the experimental data becomes questionable. Bertholf and Karnes calculated a maximum allowable strain rate for their SHPB. They claim that since their numerical calculations are based on constitutive equations which are rate independent, their results can be ap plied directly to different diameter systems. For a 1inch diameter bar, a length L over diameter d (L/d) = 0.3 in the specimen, and a step input stress, Bertholf and Karnes calculated a maximum strain rate of 400 1/sec. For a ramp input, this value is much higher for the reasons discussed earlier. If the diameter is increased by a factor n, then the new maximum strain rate is 400/n 1/sec. This scaling cannot be applied directly to the proposed 6inch bar because the L/d of the specimen will not necessarily be 0.3. In fact it will be about one for concrete specimens to ensure the same aggregate density in all directions. Therefore one needs to go back to the equations that Davies and Hunter [7] derived to calculate the contribution of inertia based on the kinetic energy due to both axial and radial motion. Their result is (L2 2 d2 d'e (19) Us = Ub+Ps(Vs (1 where Ub is the measured stress, subscript s pertains to the specimen and L2 d2 Sand vare the axial and radial inertia contributions respectively. 6 8 Equation (19) can be rewritten in nondimensional form as S ab d2 E/d t2 ps d2 L L (20) Ub Ub 6 Ld J The left hand side is the error in the measured stress. If one assumes a nominal Poisson's ratio of 0.3, the contribution of the term L/d can be calculated relative to Bertholf and Karnes L/d of 0.3. This leads to the conclusion that for a nominal L/d = 1.0, the error in the measured stress will increase by a factor of 40. This means that the maximum strainrate they calculated should be reduced even further. This would lead to unacceptably low strainrates. Therefore, the thing to do is accept the errors due to inertia effects, and correct them using equation (19) in the data reduction algorithm. Furthermore, if one looks at typical strain rate plots from actual SHPB experiments, it appears that in most cases the strain rate very quickly levels off to some constant value. What that means is that the derivative of the strain rate will eventually approach zero, and it may be that the inertia effects will not be as important after the initial 100 or so microseconds. Friction presents more of a problem in the sense that its effects are not easily quantifiable or easy to predict. It is shown from numerical cal culations that the measured stress will be larger than the true stress, result ing in an artificially high strength in the material being analyzed. But how much? Early work and studies done by Davies and Hunter [7] and by Saman ta [10] show that in general, if the diameter and the length of the specimen are of the same order of magnitude, then friction contributes to only a few percent error. It appears that the important parameter is the ratio of the area constrained by friction over the volume of material in the specimen. In other words, a thin specimen will be more affected by friction than a thick one. Since the area constrained by friction is d2/2 (two faces) and the volume is n d2 L/4 then the key parameter in determining friction effects is 2/L. The larger this value the greater the error due to friction. This means that increasing the diameter of the SHPB bars and specimen will not worsen friction effects. Although an exact value for friction effects is not obtainable, one can assess the order of magnitude of the error by running HULL with "locked" interfaces between the bars and the specimen, and then compare the results with the other runs where the interfaces were friction less. (There is no option in HULL for modeling friction). The results then bound the friction problem as the "locked" case represents a worst case scenario. Figures 32 and 33 show the transmitted stress for a 2inch bar with zero and infinite friction, respectively. Figures 34 and 35 show the same thing for a 6inch bar. As expected, the cases with infinite friction show a higher transmitted stress, implying an artificially higher strength in the specimen. It is interesting to notice that the error due to friction in the 2inch bar is about 16.4% and in the 6inch bar is only 5.6%. This agrees with the criterion derived above which states that the error is proportional to 2/L. It appears the larger SHPB set up will be less susceptible to friction in light of the fact that the specimen length will be increased along with the diameter. However, one should not imply from this that friction introduces errors in the existing systems of the order of 1020%. That figure is an upper bound for a case where the specimen is attached to the bars. In reality, the friction coefficient will be somewhere less then 1.0, especially if care is taken in lubricating the ends. Therefore, the errors due to friction are negligible in typical SHPBs and will be so also for a proposed 6inch system. If, however, one was to test thin specimens, then the data reduction should take friction effects into account. 2.51E+00 2.01E+00 1.51E+00 1.01E+00 5.01E01 0.0 5.01E01 1.01E+00 1.51E+00 2.01E+00 2.51E+00 0.00 0.11 0.22 0.33 0.44 0.55 T (MSEC) 0.66 0.77 0.88 0.99 1.1 Figure 32. HULL Output. 2inch Bar, Transmitted Stress with Zero Friction Between Specimens and SHPB STA= 3 XO= 2.33 YO= 300.08 N SI o _ >jw o a in F i 0 w N o0 0 0 ox 0 0 0 0 *N 0a ( ) ss (GN) SS3H.S 0.75 T (MSEC) Figure 34. HULL Output. 6inch Bar, Transmitted Stress with Zero Friction T b b 0  X 0 0 0 o 0 0 o 0 6 0 I I I 77 0 "   ^ o IN N N 0 o 00 I *I (e) 0  i   ] 6  .0 CHAPTER 4 SIXINCH SHPB FEASIBILITY Summary The objective of this effort was to assess the feasibility of a 6inch SHPB. There is an abundance of experimental data for existing SHPB sys tems, and it was used in this study to test the accuracy and reliability of the numerical finite difference code HULL. The first part of this work researched the multitude of efforts con ducted on the subject in the past years. In particular, it was established that it was important to look at the twodimensional equations of motion as the simplified onedimensional case may not be adequate for a larger system. The PochhammerChree solution for an infinitely long bar was used as the baseline for the mathematical analysis. Early papers on inertia and friction effects were studied, and the methodology extended to the 6inch bar sys tem, primarily to have a feel for the magnitude of the errors introduced by those factors. Taking a closer look at the governing equations and at the Pochham merChree frequency equation, it became apparent that the problem of wave propagation in a cylindrical bar is independent of scale if the ratio of the diameter over the wavelength of the traveling pulse is kept constant. This is not a new idea, but the implications in this study are crucial. It means that in theory at least, an SHPB can be scaled up or down as long as the input pulse is also scaled up or down. It does not mean the SHPB response is one dimensional, rather it implies that since the simple onedimensional analysis has been proven reliable for the existing systems, it should apply just as well to the proposed 6inch SHPB. The next part of this effort looked at three methodologies in order to compare and contrast the results. The goal was to have a firm understanding of the implications of increasing the size of an SHPB. The first two included a finite difference numerical technique, and experimental traces obtained at Tyndall AFB. The third method, called the Modified PochhammerChree method for elastic systems, consisted of computationally following each Fourier component of the input pulse as it reflects and transmits at the barspecimen interfaces. Each pulse has a unique velocity and wavelength, which enables one to calculate its position in time. The complete pulse can then be reconstructed at any time. Furthermore, the distribution of dis placements and stresses along the radial direction can also be calculated if the wavespeeds of the Fourier components are known. It was shown how this methodology gives surprisingly good results, considering the many or ders of magnitude difference in computing time compared to the finite difference technique. All three methodologies reaffirmed the initial statement of scale in variance of any SHPB. It was shown how if the input pressure pulse has a wavelength of 500 microseconds, a 6inch SHPB will behave as a onedimen sional wave propagation system, and thus all the current data reduction schemes and the current assumptions will be valid. Finally, the effects of friction and inertia were analyzed. It turns out friction causes only a few percent error, even for a large 6inch bar. Inertia however, will affect the results of an SHPB, and must be accounted for in the data analysis of a new 6inch system. Inertia effects increase as strainrate increases. Conclusions and Recommendations The following conclusions can be drawn from this analytical study. 1. A 6inch SHPB is theoretically feasible, and in fact identical to any existing system if the input pressure pulse duration is scaled appropriately. Specifically, a 500 microsecond wavelength is recommended. 2. The modified PochhammerChree method can be used to assess the validity and study the wave propagation in any size SHPB. The results give the incident, transmitted and reflected pulses for different diameters, wavelengths and materials. This method can be applied to any input pres sure if it can be described by a Fourier series. 3. If the specimen length and diameter are kept in the same order of magnitude, one need not worry about friction in the proposed 6inch SHPB. It will cause only a small error in the analysis. 4. The new proposed 6inch SHPB will need a correction in the data analysis to account for inertia. This is true especially if the input pulse is a step function and if strain rates higher then 1020 1/sec are desired. Davies and Hunter [7] derived an equation to correct for inertia which has been shown to be fairly accurate. It is recommended this equation be incor porated in the existing data reduction algorithm. A Fortran subroutine has been written by the author, is included in Appendix E, and can be directly integrated in the existing data reduction programs in use at the University of Florida and at Tyndall AFB. By following the guidelines in steps 1 through 4, the test engineer has in effect removed or corrected for any rate, friction, or twodimensional effects due to the testing apparatus itself, and thus the results will permit the 81 observation of the true material behavior. The intrinsic characteristics of the material will be separated from any other effects inherent to the way the specimen is being tested. Therefore, the Split Hopkinson Pressure Bar has been and remains a reliable, accurate and useful tool to study material response to loading, even for a new proposed system 6 inches in diameter. REFERENCES 1. Kolsky, H., "An Investigation of the Mechanical Properties of Materials at Very High Rates of Loading," Proceedings of the Physics Society, Vol. 62, pp. 676700, 1949. 2. Hauser, F.E., Simmons, J.A., and Dorn, J.E., "Response of Metals to High Velocity Deformation," Proceedings of the Metallurgical Society Conference, Vol. 9, pp. 93101, 1960. 3. Lindholm, U.S., "Some Experiments with the Split Hopkinson Pressure Bar," Journal of the Mechanics and Physics of Solids, Vol. 12, pp. 317335, 1964. 4. Malvern, L.E., and Ross C.A., "Dynamic Response of Concrete and Concrete Structures," Final Technical Report, AFOSR Contract F4962083K007, University of Florida, Gainesville, FL, May 1986. 5. Robertson, K.D., Chou, S. and Rainey, J.H., "Design and Operating Characteristics of a Split Hopkinson Pressure Bar Apparatus," Techni cal Report AMMRC TR 7149, Army Materials and Mechanics Re search Center, Watertown, MA, November 1971. 6. Kolsky, H., Stress Wave in Solids, Dover Publications, New York, 1963. 7. Davies, E.D.H., and Hunter, S.C., "The Dynamic Compression Testing of Solids by the Method of the Split Hopkinson Pressure Bar," Journal of the Mechanics and Physics of Solids, Vol. 11, pp. 155179, 1963. 8. Rand, J.L., "An Analysis of the SplitHopkinson Pressure Bar," U.S. Naval Ordnance Laboratory, White Oak, MD, NOLTR67156, 1967. 9. Dharan, C.K.H. and Hauser, F.E., "Determination of StressStrain Char acteristics at Very High Strain Rates," Experimental Mechanics, Vol. 10, pp. 370382, 1970. 10. Samanta, S.K., "Dynamic Deformation of Aluminum and Copper at Elevated Temperatures," Journal of the Mechanics and Physics of Solids. Vol. 19, pp. 117124, 1971. 11. Jahsman, W.E., "Reexamination of the Kolsky Technique for Measuring Dynamic Material Behavior," Journal of Applied Mechanics, Vol. 38, pp. 7582, 1971. 12. Chiu, S.S. and Neubert, V.H., "Difference Method for Wave Analysis of the Split Hopkinson Pressure Bar with a Viscoelastic Specimen", Jour nal of the Mechanics and Physics of Solids. Vol. 15, pp. 177193, 1967. 13. Young, C. and Powell, C.N., "Lateral Inertia Effects on Rock Failure in Split HopkinsonBar Experiments," 20th U.S. Symposium on Rock Mechanics, Austin, TX, pp. 299303, 1979. 14. Nicholas, T., "An Analysis of the Split Hopkinson Bar Technique for StrainRateDependent Material Behavior", Journal of Applied Mechanics. Vol. 1, pp. 277282, March 1973. 15. Lawson, J.E., "An Investigation of the Mechanical Behavior of Metals at High Strain Rates in Torsion," Ph.D. Dissertation, Air Force Institute of Technology, WrightPatterson AFB, OH, June 1971. 16. Baker, W.E. and Yew, C.H., "StrainRate Effects in the Propagation of Torsional Plastic Waves", Journal of Applied Mechanics, Vol. 33, pp. 917923, 1966. 17. Jones, R.P.N., "The Generation of Torsional Stress Waves in a Circular Cylinder", Quarterly Journal of Mechanics and Applied Mathematics, Vol. 12, pp. 208211, 1959. 18. Okawa, K., "Mechanical Behavior of Metals Under TensionCompres sion Loading at High Strain Rate," International Journal of Plasticity, Vol. 1, pp. 347358, 1985. 19. Ross, C.A., Nash, P.T., Friesenhahn, G.H., "Pressure Waves in Soils Using a SplitHopkinson Pressure Bar," Final Technical Report, ESL TR8129, AFESC, Tyndall AFB, FL, July 1986. 20. Rajendran, A.M., and Bless, S.J., "High Strain Rate Material Behavior," Technical Report, AFWALTR854009, AFWAL, Kirtland AFB, NM, December 1985. 21. Ross, C.A., "Split Hopkinson Pressure Bar Tests," Final Technical Report, ESLTR8882, AFESC, Tyndall AFB, FL, March 1989. 22. Follansbee, P.S. and Frantz, C., "Wave Propagation in the Split Hopkin son Pressure Bar", ASME Journal of Engineering Materials and Tech nology, Vol. 105, pp. 6166, 1983. 23. Redwood, M., Mechanical Waveguides, Pergamon Press, New York, 1960. 24. Bertholf, L.D. and Karnes, C.H., "Twodimensional Analysis of the Split Hopkinson Pressure Bar System," Journal of the Mechanics and Physics of Solids, Vol. 23,pp. 119, 1975. 25. Bancroft D., "The Velocity of Longitudinal Waves in Cylindrical Bars," Physical Review, Vol. 59, pp. 588593, April 1941. 26. Mindlin, R.D. and Herrmann G., A OneDimensional Theory of Com pressional Waves in an Elastic Rod, Pro. 1st U.S. Nat. Cong. App. Mech., Chicago, pp. 187191, 1951. 27. Davies, R.M., "A Critical Study of the Hopkinson Pressure Bar," Philosophical Transactions Royal Society, Vol. A240, pp. 375457, 1948. 28. Kolsky, H., "The Propagation of Stress Pulses in Viscoelastic Solids," Phil. Mag. Vol. 1, pp. 693710, 1956. 29. Love, A. E. H., 1927, The Mathematical Theory of Elasticity, Dover Publications, New York. 30. Karal, F.C., "Propagation of Elastic Waves in a Semiinfinite Cylindrical Rod Using Finite Difference Methods," Journal of Sound Vibration, Vol. 13, pp. 115145, 1970. 31. Gong, J.C., Malvern, L.E., and Jenkins, D.A., "Dispersion Investigation in the Split Hopkinson Pressure Bar," Journal of Engineering Materials and Technol ogy, Vol. 112, pp. 309314, July 1990. 32. Jerome, E.L., "Feasibility of a 6inch Split Hopkinson Pressure Bar (SHPB)," Final Technical Report, ESLTR9044, AFESC, Tyndall AFB, FL, August 1990. 33. Hopkinson, B., "A Method of Measuring the Pressure Produced in the Detonation of High Explosives or by the Impact of Bullets," Philosophi cal Transactions of the Royal Society of London, Vol. 213, pp. 437456, 1914. 34. Osborne, J.J., and Matuska, D.A., "HULL, Finite Difference Code," Orlando Technology, Inc., Shalimar, Florida, 1987. 35. Rinehart, J.S., Stress Transients in Solids, HyperDynamics, Santa Fe, New Mexico, May, 1975. 36. Chree, C., "The Equations of an Isotropic Elastic Solid in Polar and Cylindrical Coordinates, Their Solutions and Applications," Cambridge Philosophical Society, Transactions, Vol. 14, 1889, pp.250369. 37. Pochhammer, L., "On the Propagation Velocities of Small Oscillations in an Unlimited Isotropic Circular Cylinder," Journal fur die Reine und Angewandte Mathematik, Vol. 81, 1876, pp. 324326. APPENDIX A MODIFIED POCHHAMMERCHREE PROGRAM AND SAMPLE OUTPUTS This program uses the PochhammerChree frequency equation solu tion to the twodimensional equations of motion in cylindrical coordinates to analyze the wave propagation of an arbitrary trapezoidal input pulse in a bar of varying dimensions and material properties. The wave is carried through a typical compressive SHPB. Multiple reflections and transmissions at the interfaces between the specimen and bars are calculated and the waves reconstructed in the incident and in the transmitter bar. The program logic is organized as follows: The user is asked to enter the areas, densities, and wave speeds for the bars and the specimen. This information is used to calculate the charac teristic impedances of the materials. The user is asked to enter the pulse duration, number of time steps (each 0.5e6 seconds long), the diameter of the bar, and the dimensions of the specimen. The input pulse is represented via a Fourier series. 55 terms are retained. Amplitude of the incident pulse is normalized to 1.0. Each com ponent has a unique phase angle, frequency, and amplitude and they are all in phase at the onset. The PochhammerChree solution for mode M1,i is used to calculate the wave speed of each Fourier component in the bar material as well as in the specimen material. Each pulse is then started at a distance of 2 meters from the specimen and is followed to the incident barspecimen interface through the multiple reflections and transmissions that occur at that point and is fol lowed into the specimen to the specimentransmitter bar interface where more reflections and transmissions take place. Since the speed of each 86 component of the pulse is known, and since the characteristic impedances are known, the program calculates the amount that gets reflected and trans mitted each time, as well as how long it takes for each component to travel any distance. All 55 terms are treated in this way, and then the waveform is reconstructed both in the incident and in the transmitter bar at the strain gage locations both 1 meter from the specimen. The output is in numerical form. The incident, transmitted, and reflected waves are represented by values onehalf microsecond apart. These data can be easily plotted with any available software. LOTUS 123 was used by the author. Additionally, the program calculates the cross sectional variations of the stresses derived from the PochhammerChree frequency solution. Be ssel functions are evaluated numerically. For each radial value the 55 terms making up the traveling wave are added for each time increment. The por tion of each stress dependent on r, discussed in chapter 3, are essentially multiplied to each Fourier component at each time step. All the terms, at their particular frequency are added up in time to form the complete waveform at each radial location. Longitudinal, radial, and shear stresses are calculated, and their amplitudes are all normalized to the longitudinal stress at the center. Figures 36 and 37 are examples of cross sectional variations of the longitudinal stress for a 2inch and a 6inch SHPB. Reflected and trans mitted pulses are shown. 88 It) 0 0 v) oo N 0 0 0 N tla 0 0 u < 8 o 7)0 a a 7 o N 5  S 0c / C) Mo. r V) m JJ 1 I T o Lu I (n ^ I P S S c 0 l C3 0 d a 0 0 0 0i0 Sc3WUS O3iKSNVil Si0 ? ? UU) CC F F1 S 5 S O21Sn5 T 5 L LL) J ss 03 ,. a 0al CG'lu 0 S 0 "OI 3 ^^ s 1 Z i \ =s 1 . .^ i ^ \ 1 i na\ C THIS PROGRAM USES POCHHAMMER AND CHREE C FREQUENCY SOLUTION OF THE TWO DIMENSIONAL C EQUATIONS OF MOTION IN CYLINDRICAL COORDINATES C TO ANALYZE THE WAVE PROPAGATION OF AN ARBITRARY C TRAPEZOIDAL INPUT PULSE IN A BAR OF VARYING C DIMENSIONS AND MATERIAL PROPERTIES. C C FURTHERMORE, IT CALCULATES THE RESPONSE DUE C TO THE MULTIPLE REFLECTIONS AND TRANSMISSIONS C AS THE WAVE ENCOUNTERS A SPECIMEN OF DIFFERENT C MATERIAL, SAME DIAMETER, AND SPECIFIED LENGTH. C C THE OUTPUT CONSISTS OF THE INCIDENT, REFLECTED C AND TRANSMITTED PULSES AT STATIONS ONE METER C FROM THE SPECIMEN BOTH ON THE INCIDENT AND ON THE C TRANSMITTER BAR. SOLUTIONS ARE GIVEN AT THE CENTER C OF THE BAR (FILE FOR001.DAT), AND AT THE SURFACE C (FILE FOR010.DAT) TO CHECK FOR RADIAL UNIFORMITY C OF THE MOTION. ALL AMPLITUDES ARE NORMALIZED. C dimension d(3000,40),f(3000),phl(200),ph2(200) dimension time(3000,40),u(3000),fr(3000),c(200),dl(200) dimension frr(3000),ut(3000),frrr(3000),utt(3000) dimension sig(100,200) s0.014 pi3.14159 C WAVE SPEEDS, AREAS, AND DENSITIES ARE ENTERED TO C CALCULATE CHARACTERISTIC IMPEDANCES write(*,*)' input speeds, densities, and areas' write(*,*)' col,co2,(m/sec),rol,ro2,A1,A2' write(*,*)' bar is 1, specimen is 2' Read(*,* )col,co2,rol, ro2,al,a2 a2.0/(pi**2*s) tr2*al*ro2*co2/(ro2*co2*a2+rol*col*al) ref(a2*ro2*co2al*rol*col)/( ro2*co2*a2+rol*col*al) tr2=2*a2*rol*col/( ro2*co2*a2+rol*col*al) ref2ref write(*,*)tr,ref,tr2,ref2 n=1 write(*,*)' input nn,wo (1/sec),diameter (mm),spec. length(m)' read(*,*)nn,wo,diam,zl2 diadiam/1000.0 C STRAIN GAGE LOCATION IS ASSUMED AT ONE METER C FROM IMPACTED END, AND ONE METER FROM EITHER C SIDES OF THE SPECIMEN zl1.0 2131.0 z241.0 C 55 TERMS ARE USED IN THE FOURIER ANALYSIS imm55 do 5 jl,mm n=n+2 ne(n1)/2 C CALCULATE THE AMPLITUDES OF EACH PULSE C AS THEY CROSS THE INTERFACES BACK AND FORTH C C AT STATION ON THE INCIDENT BAR: d(j,1)=(a*(1)**ne/n**2)*sin(n*pi*s) d(j,2)=ref*d(j,1) c THIS NEXT PIECE IS THE FIRST REFLECTION BACK C WHICH WILL ADD TO THE REFLECTED STRESS d(j,3)=tr*ref2*tr2*d(j,l) d(j,6)=d(j,3)*ref2**2 C AT STATION ON THE TRANSMITTER BAR: d(j,4)=tr*tr2*d(j,1) c PLUS WHAT BOUNCES BACK AND FORTH IN THE SPECIMEN C AND ADDS TO THE TRANSMITTED STRESS d(j,5)=tr*ref2*ref2*tr2*d(j,1) d(j,7)=d(j,5)*ref2**2 5 continue C SUBROUTINE ANGLE CALCULATES PHASE CHANGES, VELOCITIES C AND WAVELENGTHS FOR EACH TERM IN THE FOURIER SERIES call angle(dia,col,co2,wo,nim,phl,ph2,c,dl) C SUBROUTINE CROSS CALCULATES THE CROSS SECTIONAL VARIATION C OF THE DISPLACEMENTS AND STRESSES. ONLY LONGITUDINAL STRESSES C ARE BROUGHT BACK INTO THE MAIN PROGRAM C SIGMA RR AND SIGMA ZZ ARE OUTPUTED TO FILE FOR002.DAT call cross(dia,c,dl,mm,sig) c c CALCULATE THE TIMES EACH PULSE TAKES TO C REACH THE STRAIN GAGES ni do 10 jl,mm nn+2 time(n,l)=zl*((1.0/col)+phl(n)) time(n,2)=time(n,l)+2.0*z13*((1.0/col)+phl(n)) time(n,3)=time(n,2)+z12*( (1.0/co2)+ph2(n))*2.0 time(n,4)=time(n,l)+z24*2.0*( (.0/col)+phl(n)) 1 +zl2*((1.0/co2)+ph2(n)) time(n,5)=time(n,4)+2*z12*((1.0/co2)+ph2(n)) time(n,6)=2*z12*ph2(n)/(0.5*1.e6) 10 continue 