Development of a Finite Element Model for Simulation of Rollover Crashes
Jingwen Hu,Chunsheng Ma,King H. Yang,Clifford C. Chou,Albert I. King,Robert W. McCoy,Roger J. Chen
DOI: https://doi.org/10.1115/imece2007-44083
2008-01-01
Abstract:Rollover crashes are complex by their very nature, and have stimulated many researches aimed at improved occupant safety. In order to investigate the vehicle crashworthiness during rollovers, several test modes are generally used to replicate different real world rollover scenarios. However, such tests are very expensive, especially during the development stage of a new car line. Computer modeling is a cost-effective way to study rollover crashes. However, a survey of literature showed that only rigid-body dynamics based models have been used for rollover simulations. It is well known that this class of models cannot be used to simulate component deformation and structural collapses. Finite element (FE) method, which has been widely used to simulate frontal and side crashes, was rarely used for simulating rollover crashes, due mainly to the relative long duration of a rollover crash. The objective of this study was to develop an FE model for investigating vehicle crashworthiness during three commonly used rollover tests. An FE model of an SUV was developed in this study. Several sub-models, namely the vehicle structure sub-model, the tire sub-model, the suspension system sub-model, the restraint system sub-model, and the dummy model were generated and integrated together. The structure model was first used to simulate the roof crush test as prescribed in FMVSS 216. The resulting load versus roof crush curve matched well against test results. The integrated model was then used to simulate three laboratory-based rollover test modes, namely the SAE J2114 dolly test, curb-trip test, and corkscrew test. For each test mode, up to 1.5 seconds of simulation time (about 1 full vehicle roll) were computed. The vehicle kinematics, including the angular velocity, lateral acceleration, and vertical acceleration during these three test modes were computed and compared with experimental data. The simulated dummy head accelerations, timing and location of the most severe impact to the dummy’s head were also compared with the experimental results. Results showed very good agreement between the tests and simulations. In order to reduce the computational time, multiple CPUs were used. Approximately ten hours were required to run a 1.5 second rollover simulation on eight CPUs. Thus, simulating rollovers using FE method is quickly becoming a reality.