Modeling of Commuter Category Aircraft Seats under Crash Loading

Transcription

1 Modeling of Commuter Category Aircraft Seats under Crash Loading Copyright 2003 SAE International David Bowen and Hampton C. Gabler Rowan University ABSTRACT This paper describes the development of a non-linear finite element model of a commuter category aircraft seat designed to explore the issue of energy absorption in severe, but survivable, crashes. Using a reference seat, the paper presents a description of the model and the results of finite element modeling of the seat at increasingly severe impact velocities. The paper presents the results of a parallel experimental program, conducted to validate the model, in which instrumented crash dummies were drop tested in the reference seat at the same impact velocities as the simulation. Experimental results are reported for passenger lower lumbar loading, peak pelvic acceleration, and seat structural loading. INTRODUCTION The Federal Aviation Administration (FAA) Technical Center has undertaken an extensive research program to explore and improve the crashworthiness of aircraft in severe, but survivable, crashes. Using the FAA Dynamic Vertical Drop Test Facility, FAA researchers have conducted full-scale tests in which aircraft cabins containing instrumented crash dummies were subjected to vertical impact velocities up to 30 ft/sec. Previous tests have included the Beechcraft 1900C Airliner shown in Figure 1 [McGuire and Vu, 1998], a Shorts 330 Airliner [Abramowitz et al, 1999], a Metro III Aircraft [McGuire et al, 1993] and several Boeing 737 fuselage sections [Abramowitz et al, 2002]. These tests have demonstrated a significant probability of severe lower lumbar injury at these velocities. More recently, Rowan University in collaboration with FAA, has conducted a series of aircraft seat drop tests to compare and contrast the passenger crash protection provided by three different commuter category aircraft seat designs. Using the Rowan University Drop Tower, these component tests have shown significant differences in the ability of individual seat designs to protect their occupants from serious lower lumbar injury. Figure 1. FAA Drop Test of a Beechcraft 1900C Airliner Fuselage [McGuire and Vu, 1998] What is not clearly understood, however, are the nature of the engineering design features which lead to seat failure for some designs, and crash survivability for other seat designs. Clearly, if these design differences could be more fully understood, the more successful design features could be incorporated into future seat designs as a means of reducing crash related passenger injuries. Although experimental tests are extremely valuable as a means of determining the relative crashworthiness of different aircraft designs, these tests do not in general provide the level of engineering insight necessary for seat redesign. These insights can be better obtained through the development and execution of detailed structural models of the aircraft seat under impact loading. Development of a structural model would allow prediction of structural impact response at impact conditions other than those directly tested including simulations of crash response from different vertical impact velocities, alternate impact orientations, and multiple passenger sizes. Such a model would also serve as one component of a whole aircraft model for simulation of more complex crash events.

2 OBJECTIVE The objectives of the research program presented in this paper were to develop and validate a comprehensive finite element structural model of a production commuter category aircraft seat under impact loading. The resulting model will serve as a computational tool for the future evaluation and development of energy absorbing seats for commuter category aircraft. DEVELOPMENT OF THE FINITE ELEMENT MODEL The Beechcraft 1900C passenger seat, shown in Figure 2, was selected for modeling based upon its superior performance in experimental tests. In both full systems tests and component drop tests of three commuter category aircraft seats, the Beechcraft seat was observed to result in the lowest peak axial lumbar loads and peak pelvic acceleration. With our long-term goal of investigating energy-absorbing seats, we believed that this seat design would provide a good starting point. GEOMETRIC MODELING The geometry of the model was developed by a complete teardown and measurement of the reference seat. The seat frame is composed of high-strength steel. The seat cushion is composed of foam covered by a loosely woven fabric. Using the seat geometry developed during the seat teardown, a threedimensional solid model of the reference aircraft seat was developed using the SolidsWork 3-D modeling package. The geometric model of the seat, with the seat foam and seat cover removed, is shown in Figure 3. Figure 2. Beechcraft 1900C Passenger Seat MESH GENERATION Using the SolidWorks representation of the seat, a finite element mesh of the seat was created using the Hypermesh code. HyperMesh is an industry-standard finite element pre-processor and post-processor that works seamlessly with LS-DYNA and other finite element codes. Since most of the geometry was thin and of constant thickness, the finite element mesh was generated using shell elements. The seat frame is constructed of thinwalled tubing which is well suited for modeling with shell elements. The only exceptions were the tips of the fins around the bolt holes, where the structure is somewhat thicker than the tubing walls. Fortunately, that section of the frame does not appear to be particularly susceptible to bending in a vertical drop test because of its vertical orientation. Figure 4 shows the final Hypermesh model of the reference seat. Figure 3. Geometric Model of the Reference Seat

3 of the seat fabric. The constitutive properties of the steel and fabric materials were obtained from standard tabulations of material properties. Constitutive properties are listed in Table 1. MODELING THE DUMMY Figure 4. Hypermesh Model of the Reference Seat To determine the crash loads on a passenger located in the reference seat, a finite element representation of a crash test dummy was added to the seat model as shown in Figure 5. The 7300-element crash test dummy model was adapted from a finite element model of a Hybrid-III dummy developed by Altair Engineering. Modifications included the inclusion of a lower lumbar load cell model for later comparison with drop test results. FINITE-ELEMENT MODELING. The Hypermesh finite element mesh along with boundary conditions, loads and constitutive properties was assembled into a finite element model for input to the LS-DYNA3D code. LS-DYNA3D is a generalpurpose, explicit finite element program used to analyze the nonlinear dynamic response of three-dimensional structures. The completed finite element model was comprised of approximately 13,500 elements 6,200 elements for the seat structure and 7,300 elements for the crash test dummy. Table 1. Constitutive Properties Material Young s Modulus (10 9 Pa) Yield Stress (10 6 Pa) Ultimate Stress (10 6 Pa) Material Behavior Steel Elasto- Plastic with Strain-rate Effects and Failure Fabric 100 none none Elasticfabric with Rigidity The model used computationally efficient Belytschko- Tsay type 2 shell elements wherever possible. The Belytschko-Tsay shell elements are computationally efficient, but are susceptible to hourglassing errors. The fully integrated shell element formulation (LS-DYNA3D Shell Type 16) elements were used in regions where necessary to reduce hourglassing problems. The seat frame is composed of high-strength steel. The seat cushion is composed of foam covered by a loosely woven fabric. However, as the foam was very soft, it was assumed to produce only negligible forces on the dummy and was not included in the model. For this study, the seat pan and seat back cushions were modeled as a membrane with the constitutive properties Figure 5. Beechcraft Seat Model with the Dummy FE Model INITIAL AND BOUNDARY CONDITIONS The nodes corresponding to the seat attachment points to the fuselage side wall and seat floor were allowed to rotate but not to translate. As described below, the seat / dummy system was given an initial velocity corresponding to the impact velocity in each drop test. The concrete impact surface was modeled as a rigid surface. DROP TOWER EXPERIMENTS This section presents the results of a parallel experimental program conducted to validate the model. In the experimental component of the research program, instrumented crash dummies were drop tested in the

4 (a) 1-meter drop test (b) 1.5-meter drop test (c) 2-meter drop test Figure 6. Seat / Dummy Assembly at Drop Heights just Prior to Drop Testing reference seat at increasingly severe impact velocities. As described later in this paper, the results of these laboratory tests were compared with the results of numerical simulations conducted under the same crash loadings. As shown in Figure 6 and Table 2, drop tests of the reference seat were conducted at heights of 1 meter, 1.5 meters, and 2 meters at the Rowan University Drop Tower. The target impact velocities in Table 2 were computed from kinematics. The seat was replaced between each test. Test Number Table 2. Drop Test Matrix Target Drop Height (meters) Target Impact Velocity (m/s) m m TEST APPARATUS In each test, the reference seat was installed on a rigid platform. In an aircraft, one side of the Beechcraft 1900C seat mounts to the floor while the opposite side of the seat mounts to the side wall of the fuselage. For our test series, the seat platform was constructed with seat attachment track on a rigid side wall to accommodate the seat side wall attachment, and a second strip of seat track on the floor to accommodate the seat floor attachment. An instrumented Hybrid-II crash test dummy was placed in the reference seat. The lap safety belt was securely buckled around the test dummy. In each test, the seat platform with seat and dummy was raised to the target height as shown in Figure 6. Upon reaching the target height, the dummy-seat-platform assembly was dropped in free fall onto a rigid concrete impact pad. A unique quick release mechanism, designed to minimize any lateral motion of the seat platform upon release, was used in each test. INSTRUMENTATION The Hybrid II dummy was extensively instrumented to determine the potential for passenger injury resulting from the crash loading. The head and chest were each instrumented with three Endevco 7231C-750 single axis accelerometers aligned along the x-, y-, and z-axes. The pelvis was instrumented with an Endevco 7231C- 750 single axis accelerometer aligned with the z-axis. Only a single accelerometer could be installed in the pelvis, because of lack of space. The lower lumbar spine of the dummy was fitted with an R. Denton Model 1708 six-channel load cell. The load cell measured axial load, shear forces along two axes, and moments about the x-, y-, and z-axes. The seat platform was instrumented with a single axis accelerometer aligned with the z-axis. DROP TOWER TEST RESULTS Table 3 presents a summary of the impact test results for each channel. Maximum values shown in this table are absolute values. The tests were quite severe. At a drop height of only 1-meter, the peak acceleration on the platform was over 170 G. The platform accelerometer, rated at a maximum of 750 G, failed during the 1.5 and 2-meter tests.

5 Table 3. Drop Test Peak Measurements - Absolute Value Channel Sensor Attachment Orientation Filter 1-m 1.5 m 2-m Units Number (Hz) drop drop drop 1 Head x-axis G s 2 Head y-axis G s 3 Head z-axis G s 4 Chest x-axis G s 5 Chest y-axis G s 6 Chest z-axis G s 7 Pelvis z-axis G s 8 Platform z-axis G s 9 Lower Lumbar x-axis pounds 10 Lower Lumbar y-axis pounds 11 Lower Lumbar z-axis pounds 12 Lower Lumbar x-axis inch-pound 13 Lower Lumbar y-axis inch-pound 14 Lower Lumbar z-axis inch-pound Peak pelvic acceleration ranged from 20 G in the 1- meter test to 80 G in the 2-meter test. The peak axial lower lumbar load in the 1-meter test was 1157 pounds. By comparison, an axial lower lumbar load of 1500 lbs corresponds to a 9% probability of detectable spinal injury [Chandler, 1985]. The lower lumbar axial and shear load cell channels failed during both the 1.5-meter and 2-meter tests. RESULTS OF IMPACT SIMULATIONS To validate the finite element model, a simulated drop test was conducted from the same drop heights as the three physical experiments. The results of the computer simulations were compared with measured dummy responses from the physical experiments. Ideally, the computer simulation should agree with the physical experiments. COMPARISON OF SIMULATED DROP TESTS WITH EXPERIMENTAL RESULTS The results of each simulation and corresponding experiment are compared in the plots which follow. For this study, the experimentally measured and computationally simulated vertical (z-axis) acceleration of the pelvis, chest, and head were compared. In addition, the experimentally measured and numerically simulated axial lumbar loads were compared. Both experimental and numerical simulation results were filtered with a 60 Hz low pass cutoff frequency before being plotted. Figure 7, Figure 11, and Figure 15 present the pelvic acceleration along the vertical axis for each of the three test conditions. There is good agreement between simulation and experiment in terms of both the magnitude and time of the peak. Both the numerical simulation and experiments show that, as would be expected, peak pelvic acceleration increases as drop height is increased. Pulse width for both simulated and measured pelvic acceleration were relatively close. Figure 9, Figure 13, and Figure 17 present the vertical head acceleration for each of the three impact conditions. Both the simulations and the experiments predict that peak head acceleration increases with increasing drop height. The peak magnitude is similar for the 1-meter and 1.5-meter cases. However, the simulation under predicts the peak magnitude in the 2- meter drop test. A comparison of the video from the 2- meter drop test with an animation of the simulation shows that, in the simulation, the dummy head rotates backward after impact a motion not observed in the test. This is likely due to the fact that the model uses a Hybrid-III dummy while the experiment uses a Hybrid-II dummy. The Hybrid-III dummy neck is substantially more flexible than the Hybrid-II neck which may allow for the head rotation observed in the simulation. This may result in higher loadings along head axes other than the local z-axis compared here. Figure 8, Figure 12, and Figure 16 present the vertical chest acceleration for each of the three impact conditions. As with the head comparison, this is believed to be due to rotation observed in the simulation that was not observed in the experiment. These differences in rotational rates are believed to be a function of the dummy joint properties. The dummy joint properties will be revisited in future work on the model.

6 Figure 7. Pelvic Acceleration, z-axis, 1-meter drop, Figure 11. Pelvic Acceleration, z-axis, 1.5-meter drop, Figure 8. Chest Acceleration, z-axis, 1-meter drop, filtered using Class 60 filter Figure 12. Chest Acceleration, z-axis, 1.5-meter drop, Figure 9. Head Acceleration, z-axis, 1-meter drop, filtered using Class 60 filter Figure 13. Head Acceleration, z-axis, 1.5-meter drop, Figure 10. Lower Lumbar Axial Load, 1-meter drop, Figure 14. Lower Lumbar Axial Load, 1.5-meter drop,

7 Figure 10, Figure 14, and Figure 18 illustrate the relationship between drop height and lumbar axial load. As shown in Figure 10, the FE model captures the time, but over predicts the magnitude of the peak axial load. Because the lumbar force transducer failed in the 1.5- meter and the 2-meter drop test Figure 14 and Figure 18 present computational results only. Figure 15. Pelvic Acceleration, z-axis, 2-meter drop, Figure 16. Chest Acceleration, z-axis, 2-meter drop, Figure 19. Rear View of Side Wall Attachment Deformation after 2-meter drop test Figure 17. Head Acceleration, z-axis, 2-meter drop, Figure 20. Rear View of Side Wall Attachment Deformation after simulated 2-meter drop test Figure 18. Lower Lumbar Axial Load, 2-meter drop, COMPARISON OF VISIBLE DEFORMATION Figure 19 and Figure 20 show the actual and simulated visual post-crash deformation of the seat after a two-

8 meter drop test. Even at a drop height of two meters, visible damage to the seat was surprisingly minimal. As shown in Figure 19, damage in the 2-meter drop, the most severe test was primarily restricted to bending of the seat sidewall attachment. The model appears to capture the same damage to the seat side wall attachment which is apparent in the post-crash photograph. FUTURE WORK Further efforts will explore the following refinements to the model: The FE model will be modified to either explicitly use a Hybrid-II dummy or to improve the existing crash dummy joint properties to better represent a Hybrid- II dummy. In particular, the neck of the dummy model needs to be modified to correctly capture the kinematics of the inflexible Hybrid-II neck. Future enhancements to the model will also include an explicit model of the side wall attachment points. In the current, the side wall is assumed to be rigid. However, our drop tests have shown this is not a strong assumption: the 2-meter test resulted in visible deformation of the seat side wall attachment track. Future drop tests will include measurement of lap belt loads for comparison with the model. In addition, future tests will measure platform acceleration in the x- and y-axes to verify our assumption that the platform exhibits no lateral motion during free-fall. CONCLUSIONS This research program has successfully developed a non-linear finite element model of a Commuter Category Aircraft seat subjected to severe, but survivable, crashes. The research program used a Beechcraft 1900C passenger seat as a reference commuter category aircraft seat for model development and testing. The finite element model has been used to simulate impact tests of the reference seat with an instrumented dummy at increasingly severe impact velocities. The research effort has also completed a parallel experimental program, conducted to validate the model, in which instrumented crash dummies were drop tested in the reference seat at the same impact velocities as the simulation. serve a promising computational tool for the future evaluation and development of energy absorbing seats for commuter category aircraft. ACKNOWLEDGEMENTS The authors wish to acknowledge Allan Abramowitz, Gary Frings, Tong Vu, and John Zvanya of the FAA Technical Center for their support of this research effort. The authors are also grateful to Christopher Molnar, the Rowan University graduate research assistant, who conducted the drop tests. Our thanks as well to the following undergraduate engineering research assistants for their contributions to the project: Alan Courtright, Matthew Hammill, Jeremy Lamb, Vern Schwanger, and Mark Seidman. REFERENCES 1. Abramowitz, A., Ingraham, P. A., and McGuire, R., Vertical Drop Test of a Shorts 3-30 Airplane, U.S. Department of Transportation, FAA, DOT/FAA/AR- 99/87 (November 1999) 2. Abramowitz, A., Smith, T.G., Vu, T., Zvanya, J.R., Vertical Drop Test of a Narrow-Body Transport Fuselage Section with Overhead Stowage Bins, SAE Paper (November 2002) 3. Chandler, R.F., Human Injury Criteria Relative to Civil Aircraft Seat and Restraint Systems, SAE Paper (1985) 4. McGuire, R.J., Nissley, W.J., and Newcomb, J.E., Vertical Drop Test of a Metro III Aircraft, U.S. Department of Transportation, FAA, Report No. DOT/FAA/CT-93/1 (June 1993) 5. McGuire, R. and Vu, T., Vertical Drop Test of a Beechcraft 1900C Airliner, U.S. Department of Transportation, FAA Report No. DOT/FAA/AR- 96/119 (May 1998) The model was shown to be an good predictor of pelvic acceleration. Reasonable, although not as close, agreement was observed between numerical simulation and experiment for the head, chest, and lumbar body regions. The numerical model appears to capture the trend of passenger response as a function increasing impact velocity. With further refinement, the model will