A NONLINEAR FINITE ELEMENT MODEL OF A TRUSS USING PINNED JOINTS Joseph D. Dutson and Steven L.

Folkman Mechanical and Aerospace Engineering Department Utah State University Logan, UT 84322-4130

Abstract Researchers at Utah State University have been studying the dynamic behavior of a three bay cantilevered truss. The truss includes eight tang/clevis-type joints with clearance fit pins. The clearance in these joints significantly changes the dynamic behavior of the truss. When strut preloads are minimized the structural damping in the truss increases dramatically. Struts have been subjected to cyclic loading to characterize the dynamic behavior of an individual strut containing a pinned joint. The three bay truss was tested in different gravity environments which alter the dynamic behavior of the truss by changing strut preloads. This paper reports on efforts made to develop a nonlinear finite element model of the truss which simulates the observed truss behavior and provides insights into the damping mechanisms occurring in the joints. The finite element analysis program LS-DYNA3D was used to create a simple model of a single strut which accounts for friction, impacting, and equivalent viscous damping in the joints and gives reasonable agreement with measured data. The strut model was extended to model the entire truss. The predicted results correlate well with measured data, particularly when strut preloads are minimized. Introduction As researchers explore the possibility of constructing space structures, they are confronted with the task of determining a cost effective, safe method for assembling these structures in space. One feasible

approach utilizes a deployable structure assembled with light-weight flexible trusses connected with revolute joints. However, vibration modes are easily induced in these light-weight trusses. Therefore, structural damping is an important consideration in the design of deployable space structures. If the revolute joints in deployable structures are designed such that large preloads are present across the mating surfaces, the behavior of the joint approaches that of a tightly clamped or welded joint. In this configuration, the joints contribute little to the damping of the structure, and the structure could be accurately modeled using linear finite element models. On the other hand, the dynamic characteristics of the structure will be altered dramatically if the joints are designed with a small amount of slop or deadband between the mating surfaces and if joint preloads are small. The deadband and friction characteristics in this type of joint can introduce nonlinearities into the joint behavior. The energy dissipation associated with friction and impacting in the joint results in increased structural damping. Predicting the dynamic behavior of a structure with friction and impacting in the joints is difficult. Large-scale testing of space structures is expensive and often impractical. Testing of a prototype usually occurs far into the design process and may result in expensive redesigns. Also, testing complete structures on the ground may not reveal how the structure will behave dynamically in the microgravity environment of space. However, individual components of the structure are easily tested on the

Associate

Graduate Research Assistant Professor, Senior Member AIAA, Member ASME

Copyright 1996 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

1 American Institute of Aeronautics and Astronautics

ground early in the design process. If the dynamic behavior of individual struts is understood, the overall system might be more accurately modeled. This paper will document the creation of nonlinear finite element models of individual struts of a truss. These strut models allow selection of properties for equivalent viscous damping elements and friction coefficients. The strut models were then extended to model a three bay truss and the ability of the model to predict the dynamic behavior of the truss is reported. Models of Joint Damping The structural damping associated with connections or joints has been studied extensively. Models of joint or interface damping are usually derived from two mechanisms, friction and impacting. Impacting implies that two surfaces, initially separated by some finite distance, come into contact and a momentum exchange occurs. Modeling the energy dissipation associated with impacting is complicated. Crawley, Sigler, and van Schoor1 suggested that a measure of energy dissipation due to impacting is the coefficient of restitution. The coefficient of restitution is a material property, but it is also a function of the geometry of the contacting surfaces. Also, the compliance of the structure attached to the joint can influence the dynamic response. Therefore, the coefficient of restitution by itself is felt to be too simplistic to model damping in a joint. Friction is attributed to either extensional motion or rotation. Models of friction damping generally fall into two categories, microslip and macroslip. In microslip models there is no relative motion between the two mating surfaces; however, small surface imperfections allow localized, microscopic slippage. Microslip damping levels are generally very low and Plunkett2 said that we are still far from being able to predict the damping associated with microslip. Macroslip models assume that no damping occurs until there is relative motion between the two mating surfaces. Relative motion occurs when the forces parallel to the surface interface exceed the Coulomb friction force. The Coulomb friction force is proportional to the normal force between the two interfaces. Den Hartog3 analyzed this classical friction model and determined that for small loads the energy dissipated increases linearly with displacement. Although there are no general models available for predicting joint damping, many

mathematical models have been proposed. Ferri4 created a model of a nonlinear sleeve joint and concluded that the three major sources of energy dissipation were damping due to Coulomb friction, damping due to impact, and material damping. He also showed that the overall damping was similar to viscous damping. Lankarani and Nikravesh5 created a model for impacting of two elastic bodies. They concluded that energy dissipation during impact was a function of initial velocity, coefficient of restitution, and material properties. Tzou and Rong6 presented a mathematical model of a three-dimensional spherical joint which included the effects of friction and impacting. Onoda and others7 showed that impacting causes energy to be transferred from lower to higher modes. The excitation of higher modes results in greater structural damping. Although numerous mathematical models of pinned joints have been developed, the literature does not describe attempts to model the dynamic response of pinned joints using commercial FEA programs. Folkman and others8 described a combination of finite elements that could be used to model a pinned joint; however, they did not attempt to construct the model and correlate the results with measured data. Experiment Description Researchers at Utah State University have developed an experiment titled the Joint Damping Experiment (JDX). This project is funded by INSTEP, NASA s In-Space Technology Experiment Program. The objectives of JDX include development of a small-scale pin-jointed truss which was flown as a Get Away Special payload on the Space Shuttle. This truss has allowed researchers to characterize the influence of gravity and joint gaps on the structural damping and dynamic behavior of pin-jointed structures. The objectives also include development of a nonlinear finite element model of the truss to simulate the measured behavior. This model will increase understanding of the structural dynamics occurring in a pin-jointed truss and will facilitate future modeling of such structures. The flight model of the JDX truss is shown in the photograph in Fig. 1. This three-bay truss is mounted to an aluminum base plate to provide a cantilevered boundary condition. A stainless steel tip mass is attached to the other end of the truss to lower the natural frequency of the structure. The truss is

2 American Institute of Aeronautics and Astronautics

excited by electromagnets deflecting and then releasing the tip mass. The resulting free vibrational decay for the twang test is recorded. Only the eight joints at the top of the bottom truss bay are free to act like pinned joints. These eight joints exhibit nonlinear dynamic behavior and are referred to as unlocked joints. Figure 1 illustrates the location of the unlocked joints. The pins in the remaining joints are press fit into place. These joints with press fit pins behave linearly as if they were welded connections and are referred to as locked joints

The truss excitation system preferentially excites three modes in the truss; two bending modes and a torsional mode. The two bending modes are the two lowest frequency modes of the truss and are described as the bend 1 and bend 2 modes. The torsional mode consists of a rotational motion about the long axis of the truss. Figure 3 is a top view of the truss tip mass which illustrates the direction the magnets move to excite the bend 1, bend 2, and torsion modes. Ground based truss twang tests were conducted with the truss in two orientations with respect to the gravity vector. Figure 4 illustrates the 0- and 90-deg truss orientations. In the 0-deg orientation, gravity induced preloads in the struts are minimized, while preloads are maximized in the 90-deg orientation. The influence of gravity on joint damping was observed by comparing the results from the two orientations.

Figure 2. Illustration of the JDX joint design.

Figure 1. Photograph of the JDX flight model truss. The design of an unlocked joint in the flight model truss is shown in Fig. 2. This is a tang/clevistype joint with a clearance fit pin. The two holes in the clevis and the hole in the tang are each press fit with hardened steel sleeves to reduce wear. The pin used is a hardened steel shoulder bolt. Therefore, the pinsleeve interface where impacting occurs is very hard. The diametral gap in the unlocked joints of the truss averages about 0.00063 in.

Figure 3. Illustration of the tip mass and the magnets used to excite truss by deflecting the tip mass.

3 American Institute of Aeronautics and Astronautics

As part of the JDX research, a force-state mapping (FSM) technique was used to characterize individual struts from the truss. The FSM technique involves applying a sinusoidal load to a strut and measuring the resulting axial displacements, accelerations, and applied forces. A strut with both joints locked as well as a strut with one joint locked and the other unlocked were characterized. The FSM tests were conducted on both long (i.e. the diagonal struts in each bay) and short struts. The FSM technique resulted in parameters such as strut stiffness, deadband, and equivalent viscous and friction damping for use in the analytical computer model of the truss. Ferney9 documents the FSM results.
GRAVITY DIRECTION OF EXCITATION

model impacting and friction form a sliding interface. No stiffness is assigned to a sliding interface until contact is made, at which time a very high stiffness is assigned in the direction perpendicular to the surface. After contact, stiffness between the node and surface in the lateral direction is based on the Coulomb friction force. Figure 5 illustrates a strut and the beam elements used to model the strut in LS-DYNA3D. An unlocked joint is located between nodes 2 and 3. The upper half of Fig. 5 shows the elements used to model the unlocked joint. The coordinate system was defined such that the x axis is aligned with the strut and the y and z axes are orthogonal to the x axis. Nodes 1, 2, 3, 4, 8, and 9 all lie on the x axis but are offset in the upper part of Fig. 5 for clarity. Elements 1 and 2 are beam elements used to model the clevis and tang, respectively. Element 10 is the large square in Fig. 5. Element 10 is a rigid element which is actually formed from several solid elements to define contact surfaces in the joint. The width of element 10 is the pin diameter used in the joint. Nodes 3, 8, and 9 were rigidly connected by elements 8 and 9 which penetrate element 10. Under a tensile strut load, node 9 impacts the surface of rigid element 10, while node 3 impacts element 10 when the load is compressive. Nodes 3 and 9 are initially located a distance equal to half the joint deadband away from the surface of element 10. Node 8 is located inside a narrow slot in element 10 and has two functions. First, it is the hinge point for joint rotations. The slot that contains node 8, although only shown in two dimensions, is three dimensional and prevents relative displacement between node 8 and element 10 in the y and z directions. The slot is very narrow (2x10-6 inch) and is a sliding interface for node 8. Second, it provides extensional friction as the joint traverses the deadband. A force FN, applied to both element 10 and node 8, maintains a constant compressive force at the friction interface (assuming lateral shearing forces are not present). Element 7 is a viscous damper which damps oscillations that occur at the friction interface when normal force FN is initially applied. Element 6 provides equivalent viscous damping as the joint traverses the deadband. Nodes 2 and 10 are rigidly connected to element 10 to form a single rigid part. Rotational friction may be present when either node 3 or node 9 is in contact with element 10 and there is relative rotation between element 10 and the rigid line of nodes 3, 8, and 9.

DIRECTION OF EXCITATION 90-DEG TRUSS ORIENTATION

0-DEG TRUSS ORIENTATION

Figure 4. Illustration of the 0- and 90-deg truss orientations for ground tests. The JDX flight model truss was flown on NASA s KC-135 low-G aircraft on October 25-28, 1994 and on the Space Shuttle Endeavor on September 7-18, 1995 in order to test the truss in a micro gravity environment. Comparing these tests with ground tests demonstrates the influence of gravity on the structural damping and dynamic behavior of the truss. These tests also provided experimental data for comparison with finite element model of the truss. Finite Element Model of a Strut It was desired to construct a finite element model of a strut with an unlocked joint which would be reasonably simple while capturing the most important features of the strut s behavior. A model was made that would account for the deadband, impacting, extensional friction, rotational friction, and equivalent viscous damping in the joint. LS-DYNA3D10, a commercial finite element analysis program, was chosen to model the strut. Impacting and friction can be modeled in LS-DYNA3D by a point contacting or sliding along a surface. The point and surface used to

4 American Institute of Aeronautics and Astronautics

Element 10 is actually a composite of several elements formed by combining six solid elements into a single part. These six solid elements are combined in such a way that a slot is left in the center of the part. Figure 6 is a three-dimensional cutaway view of the unlocked joint model which shows five of the blocks used to construct element 10. Care must be taken when constructing element 10 so that node 8 cannot slip out of the slot. Note that node 8 can penetrate slightly into the sliding interface of the slot. If the solid blocks used to construct element 10 do not overlap each other, this penetration of node 8 could allow the node to slip out of the slot at the interface between two blocks. Thus in Fig. 6, block B is shaded to show how it overlaps A, C, D, and E.
FN 10 y 1 x z 9 10 UNLOCKED JOINT ASSEMBLY 8 FN 1 2 6 8 9 7 3 2 4

equivalent viscous damping, deadband, and impacting are all included in the LS-DYNA3D joint model. Figure 7 shows the expected quasi static forcedisplacement relationship for the unlocked joint. KC represents the strut stiffness when the gap is closed in compression while KT represents the stiffness in tension. DB is the width of the deadband. For a
z ELEMENT 10 y x NODE 10

C
NODE 8

E
PART A B, C, D, E DESCRIPTION

END BLOCK OVERLAPPING BLOCKS USED TO FORM FRICTION SLOT

ELEMENT NUMBERS 1 2 3 4 5

Figure 6. Three-dimensional view of the unlocked joint model.

(2 & 3)

6 7 NODE NUMBERS

ELEMENT NUMBER DESCRIPTION 1, 5 2, 4 3 6, 7 8, 9 10 HUB/CLEVIS BEAM ELEMENTS TANG BEAM ELEMENTS TUBE BEAM ELEMENT VISCOUS DAMPING ELEMENTS RIGID BAR ELEMENTS RIGID SOLID ELEMENT

Figure 5. Finite element model of a strut with an unlocked joint. Although the model shown in Fig. 5 is simplistic, it captures many of the desired features of an unlocked joint. Extensional and rotational friction, 5 American Institute of Aeronautics and Astronautics

APPLIED FORCE

KT W KC DISPLACEMENT

By plotting the measured force-displacement results from a quasi-static pull test of a strut, the stiffness of the strut can be obtained from the slope. The force-displacement curve for a strut with both joints locked is nearly linear. Therefore, the struts with both joints locked can be modeled using simple beam elements. The force-displacement curves for struts with one joint locked and one joint unlocked (locked-unlocked) are nonlinear and demonstrate hysteresis similar to that shown in Fig. 7 when the loading is quasi static. Figure 8 is an example of the forcedisplacement curve for a short locked-unlocked strut subjected to a quasi static (0.1 Hz) applied load. The figure demonstrates how parameters can be obtained for use in a finite element model. Again, KC and KT represent the strut stiffness in compression and tension, respectively. DB is the width of the deadband. Due to strut misalignment the observed deadband is less than the expected value. The FSM tests showed that the deadband predicted from the force-displacement curve was generally 0.0004 to 0.0007 inches less than the expected deadband. The joint deadband in the model was set equal to the measured deadband rather than the deadband predicted by measuring the hole and pin diameters. W, the width of the quasi static hysteresis loop, represents two times the extensional friction force as the joint moves through the deadband. Stiffness values were chosen for the five beam elements shown in Fig. 5 such that the overall strut stiffness would be equal to the average of KC and KT. The stiffness of the tubing was easily calculated because it has a constant, known cross section. The tang and clevis stiffness values could not be estimated by hand calculations. Therefore, stiffness values were selected such that the overall model strut stiffness would be approximately the same as the measured strut stiffness values for long and short struts as well as for struts with both joints locked and struts with one joint unlocked.

DB SYMBOL KT KC DB W DEFINITION STRUT STIFFNESS IN TENSION STRUT STIFFNESS IN COMPRESSION LENGTH OF THE DEADBAND WIDTH OF THE DEADBAND (1/2 FRICTION FORCE)

Figure 7. Expected quasi static force-displacement curve for unlocked joint. perfectly aligned strut with identical hole diameters, the deadband width is equal to twice the difference of the hole diameter and the pin diameter. Finally, W represents the width of the hysteresis loop and is twice the friction force for the quasi static case. At higher velocities the hysteresis loop is wider than the quasi static loop shown in Fig. 7 due to the viscous damping losses. Model Parameters from Force-State Mapping Force-state mapping (FSM) tests reported in reference 9 were used to characterize individual struts from the JDX truss. During FSM tests a sinusoidal load was applied to the right end (node 7 in Fig. 5) of the strut while the left end (node 1 in Fig. 5) was constrained. The applied force as well as the resulting displacements and accelerations (of node 7) were measured. These data were used to produce a map of the force-displacement-velocity domain. The FSM tests provided parameters for the finite element model of a single strut.

6 American Institute of Aeronautics and Astronautics

100. 80. 60. 40. FORCE (LBS) 20. 0. -20. -40. -60. -80. -100. -.003 DB = 0.001 IN W KT = 71,700 LB/IN KC = 72,700 LB/IN

-.002

0.0 .001 .002 -.001 DISPLACEMENT (INCH)

.003

Figure 8. Measured quasi static force-displacement plot illustrating determination of parameters. The width of the force-displacement hysteresis loop is related to the energy dissipated per cycle in the strut. This width can be modeled by either friction or viscous damping. The equivalent viscous damping in the unlocked joints was chosen such that the width of the model hysteresis curve was approximately equal to the width of the measured hysteresis curve from a FSM test with dynamic loading. Figure 9 illustrates force-displacement curves which compare results from a single strut finite element model with measured data. The comparison in Fig. 9 is for a 35 Hz sinusoidal load applied to a short locked-unlocked strut. The force shown is the force applied to node 7 (see Fig. 5) while the displacement is the axial displacement of node 7. Although there are differences between the two curves, the areas (and the energy dissipated per cycle of the strut) are nearly the same. Figure 10 shows a comparison between the measured and predicted displacement of node 7 as a function of time. The lowest frequency oscillations are a result of the 35 Hz applied force. The natural frequency of the strut causes the higher frequency oscillations. This higher frequency strut mode causes the irregular hysteresis loops in Fig. 9. Figure 10 shows that the model predicts higher amplitude high frequency oscillations. We were unable to achieve better agreement between the measured data and the model by changing the properties of the model. It is suspected that a more detailed model of the joint may be needed to get better agreement.

Figure 9. Measured and predicted hysteresis curves.

Figure 10. Measured and predicted time-displacement curves. Finite Element Model of the Truss The single strut finite element models were extended to model the entire JDX truss. Other than the sliding interfaces at the unlocked joints, the truss was modeled using beam and plate elements. The expected deadband in a joint can be computed as twice the average clevis and tang hole diameters minus the pin diameters. The actual deadband at each unlocked joint is influenced by strut misalignment. It was not possible to measure the effective deadband of each unlocked joint after the truss was assembled. An estimate of effective deadband for each joint was obtained by using two times the smallest of the two clevis hole diameters and the tang hole diameter minus two times the diameter of the pin.

7 American Institute of Aeronautics and Astronautics

A parameter called global damping is used in LS-DYNA3D to provide a small amount of damping for each node in a deformable structure. Global damping was used to represent low level material damping. In essence, global damping defines a viscous damper between each node of the structure and ground. The equivalent viscous damping for each node is proportional to the mass assigned to the node. In order to find the initial deflected position of the truss in either the bend 1 or bend 2 directions, a 40 lb ramped force was applied to the tip mass for 0.2 seconds, then the force was held constant for 0.3 seconds to allow the structure to come to rest. A large value of global damping was used while the truss was being deflected so that all truss vibrations would damp out quickly. At 0.5 seconds the global damping was decreased and the force was removed from the tip mass to allow the truss to vibrate freely. The displacements, velocities, and accelerations for each node were stored at 3000 samples per second which was the same sampling rate used in measured data. A truss model with all of the joints locked was used to determine an appropriate value for global damping. The global damping was adjusted in the truss model until the results matched the measured data for the truss with all joints locked. When the truss was excited in the bend 1 direction, a global damping parameter of 3.0 lb-sec/in modeled the measured data well. The global damping was set to 20.0 lb-sec/in during the 0.5 second truss deflection period. Figure 11 illustrates the deflection of the center of the tip mass in the bend 1 direction. Figure 12 shows the acceleration of the center of the tip mass for both the measured data and the LS-DYNA3D model. The release of the tip mass for the model was shifted to 0 seconds in Fig. 12 to match the measured data. It can also be seen from Fig. 12 that the locked truss natural frequency predicted by the model matches the measured results for the bend 1 direction. The truss model was modified to include eight joints unlocked. Figure 13 compares the model and measured results for a test in the bend 1 direction in a micro gravity environment. When power to the magnets in the flight model truss is turned off, the magnetic force decays in an exponential fashion. The time constant for the decay is not known, but it is approximately 0.01 seconds. This decay occurs during the first, short peak in the acceleration data. The analytical model uses an instantaneous release of the

tip mass. The release of the truss in the LS-DYNA3D model was shifted to occur at about 0.1 seconds so as to coincide with the second acceleration peak in the measured data. The model predicts well the natural frequency in the bend 1 direction (which is a function of amplitude), the high frequency hash in the acceleration data, and the structural damping of the truss. The results for the bend 2 direction are similar to the bend 1 results and thus are not included in this paper.

Figure 11. Bend 1 displacement of tip mass in locked truss model.

Figure 12. Bend 1 acceleration for a locked truss in 0deg orientation. Figure 14 shows the measured and predicted results for the bend 1 direction in the 0-deg orientation (see Fig. 4) in a 1 G environment. The measured data shows a significant decrease in damping while the predicted decay is very similar to the micro gravity

8 American Institute of Aeronautics and Astronautics

environment. Figure 15 compares the bend 1 results in the 90-deg orientation when gravity induced strut preloads are maximized. In this case the model does a good job of simulating most of the effects of high gravity preloads. However, structural damping in the model is too low. The cause of the discrepancies in the 1 G environment tests is currently unknown.

Figure 15. Bend 1 acceleration for an unlocked truss in 90-deg orientation. Onoda and others7 showed that impacting in joints can excite higher frequency modes in a structure. Accelerations predicted for the truss model were obtained at 25,000 samples per second in the bend 1, micro gravity test to examine higher frequency mode content. Figure 16 shows a plot of the Fourier transform of a 0.01 second segment of the predicted decay. The truss model predicts that higher frequency modes are being excited in the truss with unlocked joints. An attempt is currently being made to use high frequency accelerometers to confirm the presence of higher frequency modes in the JDX truss.

Figure 13. Bend 1 acceleration for an unlocked truss in micro gravity.

Figure 14. Bend 1 acceleration for an unlocked truss in 0-deg orientation. Figure 16. Bend 1 frequencies for an unlocked truss in micro gravity. The truss model torsion mode was excited and compared with the measured data. Two 20 lb forces were applied to the arms of the tip mass in order to twist the truss. Figures 17 and 18 compare the 9 American Institute of Aeronautics and Astronautics

measured and predicted torsion tests for a locked truss and an unlocked truss, respectively. The results were shifted in time in order to have a similar amplitude peak in the predicted output occur at the same time as a measured peak. It is seen from Fig. 17 that the global damping chosen for the bend 1 mode is too large for the torsion mode. An accurate model of the torsion mode requires a new global damping parameter.

Figure 17. Torsion test acceleration for a locked truss.

Figure 18. Torsion test acceleration for an unlocked truss in micro gravity.

10 American Institute of Aeronautics and Astronautics

It is informative to look at the frequencies being excited in the truss during the torsion tests. Figures 19 and 20 show the locked truss frequencies for measured and predicted torsion tests, respectively. Both figures were generated with about 0.2 seconds of data after the release of the tip mass. The torsion mode is seen at approximately 110 Hz. Although not at the same frequencies, both figures show that higher frequency modes are being excited in the locked truss torsion test. A variety of modes in the tip mass and torsion arms could produce the observed response. Figures 21 and 22 illustrate the frequencies excited in the measured and predicted torsion tests for a truss with eight unlocked joints and a truss orientation of 0-deg.. In both cases the 110 Hz torsion mode disappears and only the higher frequency modes can be seen. It is significant that a mode that would be predicted by a linear model of the truss can disappear when a few clearance fit pinned joints are included in the structure. The cause of this response is unknown. It is, however, interesting that the model and measured data agree in the disappearance of the torsion mode when the truss uses a few unlocked joints. Figure 20. Predicted torsion test frequencies for locked truss in 0-deg orientation.

Figure 21. Measured torsion test frequencies for unlocked truss in micro gravity.

Figure 19. Measured torsion test frequencies for locked truss in 0-deg orientation.

Figure 22. Predicted torsion test frequencies for unlocked truss in micro gravity.

11 American Institute of Aeronautics and Astronautics

Conclusions A truss structure was described which has been used to characterize the influence of gravity and joint gaps on structural damping and dynamic behavior of pin-jointed structures. A finite element model of a single strut with a pinned joint was constructed in LSDYNA3D. The model included extensional and rotational friction, equivalent viscous damping, and impacting in the joint. The results of force-state mapping tests were used to determine appropriate parameters for the finite element model. The single strut model was extended to model the pin-jointed truss structure. The truss model results correlated well with measured data from tests conducted in a micro gravity environment; however, the model did not predict as well the truss behavior when gravity caused strut preloads. The finite element model predicts that impacting in the pinned-joints excites higher frequency modes in the truss, thereby increasing structural damping. Much work remains to be done to determine the effect of each joint parameter on the overall structural damping of pin-jointed structures. Nevertheless, the ability to predict many of the observed behaviors has been demonstrated. Additionally, a procedure for estimating model parameters such as joint deadband, friction, and equivalent viscous damping from tests characterizing individual joints has been demonstrated. Acknowledgments This research was performed under the NASA INSTEP program, funded through NASA Langley Research Center (LaRC) under Contract NAS1-19418. The support of Mark Lake at LaRC as technical monitor is gratefully acknowledged. References Crawley, E. F., Sigler, J. L., and van Schoor, M. C., Prediction and Measurement of Damping in Hybrid Scaled Space Structure Models, Space Systems Laboratory, Dept. of Aeronautics and Astronautics, MIT, Report SSL 7-88, Cambridge, July 1988. Plunkett, R., Friction Damping, Damping Applications for Vibration Control, American Society of Mechanical Engineers, New York, 1980, pp. 65-74. Den Hartog, J. P., Mechanical Vibrations, 4th ed., McGraw-Hill, New York, 1956. 12 American Institute of Aeronautics and Astronautics
3 2 1

Ferri, A. A., Modeling and Analysis of Nonlinear Sleeve Joints of Large Space Structures, AIAA Journal of Spacecraft and Rockets, Vol. 25, No. 5, 1988, pp. 354-365. Lankarani, H. M., and Nikravesh P. E., A Contact Force Model With Hysteresis Damping for Impact Analysis of Multibody Systems, Journal of Mechanical Design, Vol. 112, Sept. 1990, pp. 369-376. Tzou, H. S., and Rong, Y., Contact Dynamics of a Spherical Joint and a Jointed Truss-Cell System, AIAA Journal, Vol. 29, No. 1, Jan. 1991, pp. 81-88. Onoda, J., Sano, T., and Minesugi, K., Passive Vibration Suppression of Truss by Using Backlash, Proceedings of the 34th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference (LaJolla, CA); also AIAA Paper 93-1549. Folkman, S. L., Rowsell, E. A., and Ferney, G. D., Influence of Pinned Joints on Damping and Dynamic Behavior of a Truss, Journal of Guidance, Control, and Dynamics, Vol. 18, No. 6, Nov-Dec., 1995, pp. 1398-1403. Ferney, B. D., and Folkman, S. L., Results of Force-state Mapping Tests to Characterize Struts Using Pinned Joints, Proceedings of the 36th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference (New Orleans, LA); also AIAA Paper 95-1150. LS-DYNA3D User s Manual, Livermore Software Technology Corporation, Livermore, California, 1995.
10 9 8 7 6 5