Simulating Rollovers in PC-Crash

aka..."An Analytical Review of Two Decades of Research Related to PC-Crash Simulation Software - Part III"

If you missed Parts I or II of this series, here are links to those: Part I and Part II. In this installment, I review the prior literature related to using PC-Crash software to simulate rollover crashes. In a future installment, I will extend this prior literature and present analysis of a full-scale, steering-induced rollover crash test with PC-Crash.

I have memories from 16 or 17 years ago, of sitting in my office over the weekend, systematically changing input after input into a rollover simulation and realizing a was in over my head. The fairly clear patterns that would usually emerge for planar collisions just weren't there and all I could do was carefully work my way through the ranges for each variable. This exploration so many years ago, led me down many of the paths I explored in later articles I wrote. If you want to understand rollover crashes, I can't think of a better tool than PC-Crash to play with rollover dynamics and to explore the many variables that influence how any particular rollover occurs. After all these years, there's a ton I still haven't delved into. I'd love to dig into the soft soil model in PC-Crash, for example. Maybe someday.

Prior Studies – Rollovers

Studies Describing the Conceptual Models: Steffan [2004] discussed the models within PC-Crash relevant to modeling rollover crashes (tire and suspension models, ground surface modeling, and the vehicle body-to-ground contact model) and then presented a basic validation of PC-Crash for modeling rollovers. This validation consisted of using PC-Crash to model two rollover crash tests, and then, visually comparing the overall vehicle motion between the tests and the simulations. While Steffan obtained favorable visual agreement between the overall vehicle motion in the tests and the simulations, no comparisons were made between the translational and angular velocities and accelerations experienced by the test vehicles and those experienced by the vehicles in the simulations.

Operational Validation Studies: Andrews [2009] published a study at the Enhanced Safety of Vehicles Conference that compared modeling of a staged rollover collision using PC-Crash Version 7.3 to the actual measured and videoed vehicle motion. Andrews concluded that “PC-Crash was successfully utilized to reconstruct the staged rollover collision. The PC-Crash reconstruction showed the speed of the vehicle at the point of roll to be within 2.1 mph of the actual data. The number of rolls and the vehicle path during the yaw phase and the rollover phase were consistent between PC-Crash and the collected data.”

Kiefer [2011] used PC-Crash 8.2 to model 26 instrumented handling tests using 1998 and 1999 Ford Explorers. The handling tests involved rapid steering inputs at vehicle speeds between 30 and 60 mph [48.3 and 96.6 km/h]. The vehicle weights, center of gravity positions, suspension stiffness parameters, tire parameters, and steering angle were measured and used as inputs into the simulations. Kiefer utilized the TM-Easy tire model within PC-Crash. The suspension stiffness coefficients were determined by measuring the change in front and rear ride height of the test vehicles under different loading conditions. PC-Crash does not model the tire compliance separate from the suspension compression, and so, the stiffness coefficients included the compliance of the tires. Kiefer noted that “the PC-Crash linear damping rate was increased to approximately twice the default value…The authors did not have shock curves or other suspension information to use as input for the model. As a result, the suspension damping values were increased from the default until the modeled response approached the test data.” Kiefer concluded that “PC-Crash appeared to be a reasonable tool for modeling gross vehicle response. In addition, PC-Crash correctly predicted whether or not the test vehicle would experience rollover instability in a majority of the cases.”

Combination Validation/Calibration Studies: Rose [2009] reported PC-Crash modeling of the dolly rollover crash test. This modeling utilized PC-Crash 7.3. While more recent versions of PC-Crash have incorporated changes relevant to rollover modeling, certain methodological issues addressed by Rose have not changed. Specifically, Rose found that the dominant variables affecting the agreement between the PC-Crash simulation and the test data were the vehicle-to-ground friction coefficients, the vehicle suspension stiffness and damping, and the car body restitution. Perhaps more significantly, Rose found that obtaining a match between the simulated and actual roll velocity histories led naturally to good agreement with the simulated and actual translational velocity histories. This relationship is mediated through the vehicle to ground friction coefficient. This finding indicates that PC-Crash could be used to reasonably reconstruct the deceleration history for a vehicle during a rollover.

To do this, the reconstructionist would first reconstruct the vehicle’s roll motion spatially based on physical evidence. Then, the vehicle’s speed at the beginning of the rollover could be calculated using a constant deceleration rate. The vehicle’s initial roll rate could be estimated based on models available in the literature or based on simulation. Starting with these initial conditions, the motion of the vehicle could be simulated to match the spatial reconstruction. In generating a simulation that matched the reconstructed roll motion, one would inherently reconstruct the deceleration rate history. The accuracy of such an approach would clearly depend on the accuracy of the underlying spatial reconstruction of the gross vehicle motion.

Heinrichs [2013] described a method of measuring suspension properties for use in PC-Crash and he reported suspension stiffness and damping values for 26 vehicles. When vehicle-specific data is not available for a particular vehicle, PC-Crash calculates default suspension stiffness values that are proportional to the static load at each wheel and a damping coefficient that is related to that stiffness. Heinrichs reported that “the deflection of the suspension under the vehicle's own static weight was lower than PC-Crash defaults. Average pitch deflection was 8 cm and average roll deflection was 6 cm, whereas the PC-Crash ‘normal’ and ‘stiff’ defaults are 15 cm and 10 cm, respectively.” This means that the PC-Crash default suspension stiffness coefficients are generally too low.

In his 2012 study, Heinrichs showed that the reconstructed speeds obtained with PC-Crash for a planar collision were “insensitive to four-fold variations in suspension parameters.” One would expect the suspension parameters to matter more for rollover simulation than for simulating planar impacts and Heinrich’s 2013 study bore this out.

In that study, Heinrichs used PC-Crash Version 9.1 to explore the sensitivity of PC-Crash rollover simulations to variations in the suspension and damping. Specifically, Heinrichs ran simulations for each of the 26 vehicles he tested using both default and measured suspension parameters. He stated that the “simulations consisted of a steering maneuver to the right followed by a hard steering maneuver to the left…the lowest initial vehicle speed was found that resulted in a rollover. After the critical rollover speed was determined, new simulations were run using the measured suspension parameters. The initial speed for these simulations was a randomized value above the critical rollover speed. The results of these simulations were used as trajectory data; it was treated as field data for a reconstruction using the program default suspension values. For each vehicle, an initial vehicle speed was found that gave an operator-assessed best fit to the trajectory data. Vehicle speeds, steering, braking, and friction coefficients could be altered, but geometry, inertial and suspension parameters were fixed.” Heinrichs found that “the critical speed required for a vehicle to roll in a PC-Crash simulation was about proportional to the suspension stiffness.” He further found that the sensitivity of PC-Crash to variations in suspension stiffness for rollover reconstruction was minimized if evidence from all three phases of the rollover (loss of control, trip, and roll) was incorporated in the simulation.

In these findings, Heinrichs inherently distinguishes between two purposes for which PC-Crash could be used. For the purpose of predicting rollover, Heinrichs found that the results were sensitive to the suspension parameters. For the purpose of reconstructing a rollover – where the reconstructionist can incorporate evidence from all three phases of the rollover – the sensitivity to suspension parameters was minimized.

Studies that Assume the Validity of PC-Crash: Berg [2003] used PC-Crash to model various rollover test procedures. He stated that “PC-Crash is a sufficient instrument to simulate these rollover crash tests.”

Viano and Parenteau [2004] reported that “a multi-disciplinary effort was conducted to define the most relevant tests that reflect real-world rollover crashes and injuries. That work prompted a series of rollover tests, including new procedures in which vehicle and occupant kinematics were studied. Also, each test was simulated in mathematical models to study other parameters and scenarios in rollover crashes.” This team utilized and validated PC-Crash software for its mathematical modeling of the vehicle motion for the rollover tests. They used MADYMO to simulate the occupant motion. The article by Viano and Parenteau reports that PC-Crash and MADYMO simulations were used “to assure robust testing, sensing and algorithms. The mathematical models were applied to each specific test condition, validated and used for evaluation of parameters influencing rollover sensing requirements. The simulations were found to be robust representations of a vehicle rollover…Excellent comparability was demonstrated between the tests and the simulation.”

Schubert [2006] reported reconstructions of 3 rollover crashes. He used PC-Crash to simulate the events and to derive crash sensor signals for the events. He stated that “insights derived from these events can aid in development of robust rollover detection algorithms and calibrations.”

Ootani and Pal of Nissan Motor Company published a pair of studies in 2007 in which they used PC-Crash to simulate real-world, soil-tripped rollovers from the NASS-CDS database. In these studies, they noted that “through a process of iteration the overall kinematics of the vehicle movement before and after the crash was captured. The output of this PC-Crash simulation was then used as the initial input conditions (i.e., speed, deceleration, etc.) of a detailed finite element analysis.”


Andrews, Stanley B., et al., “A Comparison of Computer Modeling to Actual Data and Video of a Staged Rollover Collision,” Enhanced Safety of Vehicles (ESV) Conference, 2009, Paper Number 09-0346.

Berg, Alexander, et al, “Rollover Crashes – Real World Studies, Tests and Safety Systems,” Enhanced Safety of Vehicles (ESV) Conference, 2003.

Heinrichs, B., Mac Giolla Ri, B., and Hunter, R., "Sensitivity of Collision Simulation Results to Initial Assumptions," SAE Int. J. Passeng. Cars - Mech. Syst. 5(2):807-832, 2012, doi:10.4271/2012-01-0604.

Heinrichs, B., Goulet, J., Fix, R., and King, M., "Measuring and Modeling Suspensions of Passenger Vehicles," SAE Technical Paper 2013-01-0774, 2013, doi:10.4271/2013-01-0774.

Kiefer, Aaron, Bilek, David, Moser, Andreas, Webb, Andrew, “A Comparison Study between PC-Crash Simulation and Instrumented Handling Maneuvers,” SAE Technical Paper Number 2011-01-1121, 2011, doi:10.4271/2011-01-1121.

Ootani, R. and Pal, C., “Effective Numerical Simulation Tool for Real-World Rollover Accidents by Combining PC-Crash and FEA,” SAE Technical Paper 2007-01-1773, 2007, doi:10.4271/2007-01-1773.

Ootani, R. and Pal, C., “Soil Trip Rollover Simulation and Occupant Kinematics in Real World Accident,” SAE Technical Paper 2007-01-3680, 2007, doi:10.4271/2007-01-3680.

Rose, Nathan A., Beauchamp, Gray, “Analysis of a Dolly Rollover with PC-Crash,” SAE Technical Paper 2009-01-0822, 2009, doi:10.4271/2009-01-0822.

Schubert, P., “Real-World Rollovers Reconstructed from Interviews and Measurements,” SAE Technical Paper 2006-01-0060, 2006, doi:10.4271/2006-01-0060.

Steffan, H. and Moser, A., "How to Use PC-CRASH to Simulate Rollover Crashes," SAE Technical Paper 2004-01-0341, 2004, doi:10.4271/2004-01-0341.

Viano, David C., Parenteau, Chantal S., “Rollover Crash Sensing and Safety Overview,” SAE Technical Paper Number 2004-01-0342, 2004, doi:10.4271/2004-01-0342.

#accidentreconstruction #simulation #forensics #forensicengineering #crashreconstruction #expert