Zone of Intrusion Envelopes Under MASH Impact Conditions for Rigid Barrier Attachments (2022)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

45   Simulation Calibration 4.1 Overview MwRSF researchers previously used nonlinear FEA simulation to develop ZOI guidelines for desired scenarios (Reid and Sicking 2010; Stolle, Reid, and Faller 2014). FEA represents a cost-effective means of analyzing multiple impact scenarios. Although full-scale testing is the empirical basis for establishing ZOIs, the associated costs rendered it impractical to evaluate the entire suite of barrier geometry and vehicle combinations desired for this study. Simulations were conducted to fill gaps in existing crash test data to develop final MASH ZOI envelopes. Baseline simulations were developed and validated to reliably predict vehicle-barrier interaction and determine ZOIs. The baseline models were a small subset of crash tests using tuned parameters for vehicle suspension behaviors, vehicle- and tire-barrier friction coefficients, and tolerance adjustments to confirm that models involving similar systems and input param- eters accurately represented all reference crash tests. No new models were developed as part of this effort, and researchers relied on vehicle models mostly developed at the George Mason University Center for Collision Safety and Analysis (CCSA) or through funding provided by the FHWA; in some cases, vehicle models did not match the make and model of the actual crash test vehicles, which may affect model validation. The research team considered these differences throughout simulation and ZOI envelope development. For each baseline model, the research team considered the models to be validated if (1) the predicted ZOI of a stiff vehicle component was within 20% of the reported test value, with a preferred limit of 10%; (2) validation & verification (V&V) comparisons between the full-scale test and simulation were satisfactory; and (3) simulated vehicle damage, occupant risk, and tra- jectory matched corresponding outcomes from the full-scale tests. Multiple simulations were performed using the same vehicle parameters and model with impact conditions consistent with those reported in the full-scale crash test documentation. Optimum values for simula- tion impact parameters were selected to provide a minimum cumulative error for all baseline models. Three baseline models were considered for MASH TL-2 and TL-3 impacts and two for MASH TL-4 and TL-5 impacts. The validation outcomes for the MASH TL-2 and TL-3 2270P vehicles, TL-4 10000S vehicles, and TL-5 36000V vehicles are shown in Appendices D, E, and F, respectively. 4.2 Model Preparation Three simulation models were used: (1) a pickup truck model (RAM) provided by CCSA at George Mason University, (2) an SUT model first funded by FHWA and refined at the National Crash Analysis Center at George Washington University (Mohan and Marzougui 2007), and (3) a tractor-trailer that had been extensively modified by Chuck Plaxico (Plaxico et al. 2010a; C H A P T E R 4

46 Zone of Intrusion Envelopes Under MASH Impact Conditions for Rigid Barrier Attachments Plaxico et al. 2010b). Vehicle models were customized to use MwRSF’s standard unit system (kg, mm, ms base units) with a standard orientation of impact near the model’s Cartesian system origin of (0,0,0) and adjusted such that a planar elevation of zero in the Y-axis was equal to the ground plane. The pickup truck model was validated against test nos. IBBR-1 (Bielenberg et al. 2020), H34BR-2 (Bielenberg 2019), and OSSB-1 (Bielenberg, Faller, and Ronspies 2018). The SUT model was validated against tests nos. 4CBR-1 (Rosenbaugh et al. 2021) and 420020-9b (Sheikh, Bligh, and Menges 2011). The tractor-trailer model was compared against test nos. MAN-1 (Rosenbaugh et al. 2016) and 510605-RYU1 (Buth and Menges 2012). The validations are summarized in Table 13. 4.2.1 Vehicle Model Refinements Vehicle models were adjusted to simplify simulation, improve stability, and identify friction that produced good agreement between simulation and full-scale results. The RAM, F800 SUT, and tractor-trailer model validations and comparisons are described in Appendices D, E, and F, respectively. Initial pickup truck simulations yielded error terminations believed to be attributed to two model issues. The pickup truck model was initially equipped with a simulated dummy, and contacts between the dummy and seatbelts caused snag, which contributed to premature model termina- tion. The dummy was not believed to significantly affect the ZOI prediction, and no objects were located in the ZOI for the simulations evaluated in this study; therefore, the occupant reaction during impact was not necessary, and the simulated occupant was removed. The second model modification related to the bushing element formulation, which displayed hourglassing. The bushing section type formulation was revised, which also improved model stability. Simulations using the initial F800 SUT frequently resulted in error terminations. Initial observa- tions indicated the coarsely meshed box of the truck snagged on the barrier and caused shooting nodes and unrealistic deformations. To resolve this issue, the box was replaced with a 2016 version (NHTSA n.d.). While model stability improved, ZOI estimates did not match test results under any combination of friction or suspension parameters investigated. Analysis of recent full-scale tested trucks consistent with MASH TL-4 impact conditions indicated that truck frames, suspensions, and geometries had evolved since the development of the original TL-4 model, and truck box heights were higher than modeled, particularly on trucks with air ride suspensions. The vehicle box was lifted to match full-scale test vehicles, and modifications were made to include U-bolt failure (Plaxico 2020). The model refinements improved both stability and ZOI predic- tion, but the coarse mesh of the cab, door panels, and fuel tank and inconsistent behaviors of those models often contributed to a lower ZOI estimate. Since most crash-tested vehicles either MASH TL Test No. Test Vehicle Velocity (mph) Angle (deg.) ZOI (in.) Test Model Test Model Test Model 2 IBBR-1 2011 RAM 1500 45.3 45.3 25.6 25.0 21.7 16.0 3 H34BR-2 2018 RAM 1500 64.0 63.9 25.4 25.0 16.7 16.1OSSB-1 2011 RAM 1500 62.8 62.8 24.9 25.0 7.90 10.9 4 4CBR-1 2005 International 4300 57.6 57.6 16.0 16.0 55.4 62.4420020-9b 1991 International 4700 57.2 57.2 16.1 16.1 62.5 67.9 5 MAN-1 2004 International 9200 51.7 51.9 15.2 15.2 22.7 25.2510605-RYU1 1995 GM Tractor 49.1 51.9 14.6 14.6 63.8 57.2 Table 13. Simulation vehicle models and impact conditions (Buth and Menges 2012; Sheikh, Bligh, and Menges 2011; Bielenberg et al. 2020; Bielenberg 2019; Rosenbaugh et al. 2021; Bielenberg, Faller, and Ronspies 2018; Rosenbaugh et al. 2016).

Simulation Calibration 47   did not use a fuel tank or the fuel tanks were weakly attached to the vehicle and unlikely to pro- duce significant structural load contributions to failing axles, the model was revised such that the impact side of the vehicle was opposite the modeled location of the fuel tank. No fuel tanks were located on the impact side beneath the driver or passenger side step of the cab for most full-scale crash-tested trucks, and thus this modification was warranted both to improve ZOI estimates and to be consistent with typical crash test conditions. The tractor-trailer was the longest simulation vehicle. Due to mesh geometry, construction uncertainty, lack of test vehicle standardization in the tractor-trailer model or connections, and coarse mesh issues, it was the most difficult to validate. Difficulties reproducing the most extreme ZOI values from full-scale testing were often attributed to unique events or vehicle parameters, including but not limited to bending, fracture, or disengagement of the fifth wheel at the tractor-trailer interface; differences in truck stiffness, construction, and geometry; different lengths of trailers (40- and 53-ft-long models); and unique failures of the trailer body itself. These variations in test results provided a disparity in empirical results for TL-5 vehicles and repre- sented a diverse array of possible outcomes for similar test conditions and barrier geometries. Thus, solutions were non-unique. The tractor-trailer model had additional limitations that led to challenging validation efforts. Modifications included: • Front axle U-bolt connection failure was added to the model to replicate axle release conditions. • Wheel deflation was considered using both timed-failure and release-load conditions. • Fifth-wheel connection was revised to accommodate additional tractor-trailer relative rotation and damage. The model that best correlated with full-scale test ZOIs used model parameters, frictions, con- nections, and geometry similar to the Plaxico-Battelle model, which was previously validated to MwRSF full-scale crash test no. TL5CMB-2 (Rosenbaugh, Sicking, and Faller 2007) and showed better agreement between the simulation and crash test ZOI. 4.2.2 Evaluation Criteria The most important criterion to evaluate the accuracy of the simulation models was the ability to reproduce the ZOI measured in high-speed video analysis and reported in full-scale test reports. Additional criteria were imposed to select between models that demonstrated ZOI ranges within 10% accuracy for the 2270P and 20% accuracy for the 10000S and 36000V models. Accuracy was evaluated both qualitatively and quantitatively. Qualitative assessments included comparing sequential snapshots to verify vehicle kinematic response along with the sequence and timing of key events. Vehicle accelerations, velocities, displacements, and Euler angle plots were quali- tatively compared. Importance was placed on the simulated vehicle’s Euler angles (roll, pitch, and yaw) as these directly affected ZOI. The quantitative aspects were performed according to procedures specified in V&V. These included stability analyses, comparison of simulation and test kinematic data, and application of the Roadside Safety Verification and Validation Program (Ray and Mongiardini 2008). 4.3 ZOI Measurement Procedure For all measurements, the reference point was the top-front corner of the traffic side of the barrier. The front edge of the barrier was used as the reference point for all barrier models, and a virtual intersection was defined when the front and barrier-top faces were connected with a chamfered edge. The downstream views were captured using high-speed cameras typi- cally aligned with the axis of the barrier and equipped with zoom lenses to reduce spherical aberrations. Therefore, downstream views were preferred to collect ZOI data. When available,

48 Zone of Intrusion Envelopes Under MASH Impact Conditions for Rigid Barrier Attachments measurements were confirmed using overhead and upstream views, although it is typically more difficult to identify accurate ZOI data using these views due to obscurities including dust, debris, shadows, and unclear positions of the barrier with respect to vehicle components at the point of maximum lateral extension. ZOI measurements in each simulation consisted of two-dimensional coordinate measurements, vehicle component type, and corresponding times for each component’s maximum extension. The procedure to collect simulation ZOI data was similar to the procedure to collect ZOI data from full-scale crash tests and was identical to the method used when considering simulations at multiple barrier heights. 4.4 Vehicle Model Calibration 4.4.1 RAM The pickup truck model was used to simulate test no. IBBR-1. The full-scale test comparison is shown in Table 14 (Bielenberg et al. 2020). The steel rail attached to the barrier in test no. IBBR-1 was not included in the barrier model. Maximum Euler angles in the simulation were similar to the test; however, simulated ZOI differed by 26.5%. This was attributed to the absence of the steel rail, which caused additional damage to the vehicle’s front-end components. Simulation and crash test lateral extents are compared in Figure 36. Note the test frame shown in Figure 36 was captured prior to the vehicle significantly interacting with the steel rail as its presence sig- nificantly affected the vehicle’s lateral extent. The full model validation and V&V comparison are shown in Appendix D, Section D-5. Validation Parameters Max. Lateral Extent (in.) Max Euler Angles (deg.) Roll Pitch Yaw Test no. IBBR-1 21.7 -3.0 14.8 30.1 FEA 16.0 -5.5 11.1 31.2 Table 14. Test no. IBBR-1 (Bielenberg et al. 2020) and simulation results. Figure 36. Maximum lateral extent during test no. IBBR-1 (Bielenberg 2019) and calibrated simulation.

Simulation Calibration 49   Figure 37. Maximum lateral extent during test no. H34BR-2 (Bielenberg 2019) and calibrated simulation. Validation Parameters Max. Lateral Extent (in.) Max. Euler Angles (deg.) Roll Pitch Yaw Test no. H34BR-2 16.7 13.7 -2.8 -37.6 FEA 16.1 9.5 -1.6 -34.1 Table 15. Test no. H34BR-2 (Bielenberg 2019) and simulation results. The pickup truck model was used to simulate test no. H34BR-2 (Bielenberg 2019), and the com- parison is shown in Table 15. Maximum Euler angles and lateral extent in the simulation were similar to the crash test. Simulation and crash test lateral extents are compared in Figure 37. The full model validation and V&V comparison are shown in Appendix D-6. The pickup truck model was used to simulate test no. OSSB-1 (Bielenberg, Faller, and Ronspies 2018), and the comparison is shown in Table 16. Maximum Euler angles and lateral extent were similar to the crash test. Simulation and crash test lateral extents are compared in Figure 38. The full model validation and V&V comparison are shown in Appendix D, Section D-7. 4.4.2 F800 SUT The F800 SUT model was used to simulate test no. 4CBR-1 (Rosenbaugh et al. 2021), and the comparison is in Table 17. Maximum Euler angles were similar to the crash test; however, the maximum lateral extent was 12.5% higher in simulation. This difference was deemed acceptable and could be due to a shorter wheelbase on the SUT model decreasing vehicle stability. Simula- tion and crash test lateral extents are compared in Figure 39. The full model validation and V&V comparison are shown in Appendix E, Section E-5. The F800 SUT model was used to simulate test no. 420020-9b (Sheikh, Bligh, and Menges 2011). The gyro sensors failed to record Euler angle data during testing. Therefore, ZOIs and key events are compared in Table 18 and noted to be similar between the simulation and test. Lateral extents are compared in Figure 40; this was the only available view where the maximum lateral extent occurred in camera view. Although obstructed by dust in this view, video analysis revealed the

50 Zone of Intrusion Envelopes Under MASH Impact Conditions for Rigid Barrier Attachments Validation Parameters Max. Lateral Extent (in.) Max. Euler Angles (deg.) Roll Pitch Yaw Test no. OSSB-1 10.8 -12.7 6.6 29.0 FEA 7.9 -11.4 4.9 29.0 Table 16. Test no. OSSB-1 (Bielenberg, Faller, and Ronspies 2018) and simulation results. Figure 38. Maximum lateral extent during test no. OSSB-1 (Bielenberg, Faller, and Ronspies 2018) and calibrated simulation. Validation Parameters Max. Lateral Extent (in.) Max. Euler Angles (deg.) Roll Pitch Yaw Test no. 4CBR-1 55.4 32.8 -6.0 -15.3 FEA 62.4 33.7 -4.7 -15.3 Table 17. Test no. 4CBR-1 (Rosenbaugh et al. 2021) and simulation results. Figure 39. Maximum lateral extent during test no. 4CBR-1 (Rosenbaugh et al. 2021) and calibrated simulation.

Simulation Calibration 51   box top-front corner achieved maximum lateral intrusion. The full model validation and V&V comparison are shown in Appendix E, Section E-6. 4.4.3 Tractor-Trailer The tractor-trailer model was used to simulate test no. MAN-1 (Rosenbaugh et al. 2016), and the comparison is shown in Table 19. The initial roll was much higher in the full-scale test, while pitch and yaw were more comparable. It was believed the model’s front axle stiffness may have pre- vented vehicle roll at the outset of simulation. Simulation and crash test lateral extents are compared in Figure 41. The full model validation and V&V comparison are shown in Appendix F, Section F-5. The tractor-trailer model was used to simulate test no. 510605-RYU1 (Buth and Menges 2012). The gyro sensors failed to record Euler angle data during testing. Therefore, ZOIs and key events are compared in Table 20 and noted to be similar between the simulation and test. The test dis- played increased trailer roll, which generated a ZOI 10.3% larger than the simulation. Simula- tion and crash test lateral extents are compared in Figure 42. The full model validation and V&V comparison are shown in Appendix F, Section F-6. Validation Parameters Max. Lateral Extent (in.) ZOI Time (ms) Parallel Time (ms) Test no. 420020-9b 62.5 840 264 FEA 67.9 820 270 Table 18. Key events comparison, test no. 420020-9b (Sheikh, Bligh, and Menges 2011) and simulation. Figure 40. Maximum lateral extent during test no. 420020-9b (Sheikh, Bligh, and Menges 2011) and calibrated simulation. Validation Parameters Max. Lateral Extent (in.) Max. Euler Angles (deg.) Roll Pitch Yaw Test no. MAN-1 22.7 16.3 5.2 -15.2 FEA 25.2 11.7 4.2 -13.6 Table 19. Test no. MAN-1 (Rosenbaugh et al. 2016) and simulation results.

52 Zone of Intrusion Envelopes Under MASH Impact Conditions for Rigid Barrier Attachments Figure 41. Maximum lateral extent during test no. MAN-1 (Rosenbaugh et al. 2016) and calibrated simulation. Validation Parameters Max. LateralExtent (in.) ZOI Time (ms) Parallel Time (ms) Test no. 510605-RYU1 63.8 1,453 840 FEA 57.2 920 700 Table 20. Test No. 510605-RYU1 (Buth and Menges 2012) and simulation results. Figure 42. Maximum lateral extent during test no. 510605-RYU1 (Buth and Menges 2012) and calibrated simulation.

Simulation Calibration 53   4.5 Summary of Calibrated Simulation Parameters Researchers reviewed previous simulations to determine which vehicle and barrier model parameters were likely to produce the most accurate replication of full-scale testing. The pickup truck model was modified to remove the simulated occupant, which was causing snag and subsequent error terminations. Additionally, the bushing element formulation was modified to reduce hourglassing; both changes improved model stability. The SUT model ini- tially used a cargo box with a coarse mesh, but this was replaced with a newer model (NHTSA n.d.) with a more finely meshed box. The box was lifted to match more recent vehicle geometries and made to include U-bolt failure. Other coarse mesh elements, including the fuel tank, still caused simulation inaccuracies, and thus the vehicle was reoriented to impact on the non-fuel tank side. The tractor-trailer was the largest vehicle model and involved more contacts and connections than the other models, which produced highly unique results. Front axle U-bolts, wheel deflation, and the fifth-wheel connection were modified to produce more accurate results. Parameters were gleaned from the simulation review detailed in Section 2.4. Friction values obtained from simulation reports ranged from 0.05 to 0.4 for vehicle-barrier, 0.1 to 0.45 for tire-barrier, 0.2 to 1.4 for barrier-ground, and 0.9 for tire-ground. In one set of validated tractor- trailer simulations, the vehicle-barrier, tire-barrier, and tire-ground friction coefficients were 0.2, 0.45, and 0.7, respectively. In general, suspension failure was not considered critical to simulated system performance.