Search Results

Finite element dummy models have been more and more widely applied in virtual development of occupant protection systems across the automotive industry due to their predictive capabilities. H350 dyna dummy model [1] is a finite element representation of the Hybrid III male dummy [2], which is designed to represent the average of the United States adult male population. Lower extremity injuries continue to occur in front crash accidents despite increasing improvement of vehicle crashworthiness and occupant restraint system. It is therefore desirable to predict lower tibia injury numbers in front occupant simulations. Though lower tibia loading/index predictions are not studied as much as the FMVSS 208 regulated injury numbers, the tibia indices are injury criteria that need to be assessed during IIHS and Euro NCAP frontal offset occupant simulations. However during front crash simulations, it is very difficult to achieve good correlations or predictions of lower tibia loadings.

Accurate prediction of the responses from the anthropomorphic test devices (ATDs) in vehicle crash tests is critical to achieving better vehicle occupant performances. In recent years, automakers have used finite element (FE) models of the ATDs in computer simulations to obtain early assessments of occupant safety, and to aid in the development of occupant restraint systems. However, vehicle crash test results have variation, sometimes significant. This presents a challenge to assessing the accuracy of the ATD FE models, let alone improving them. To resolve this issue, it is important to understand the test variation and carefully select the target data for model improvement. This paper presents the work carried out by General Motors and Humanetics Innovative Solutions (formerly FTSS) in a joint project, aimed at improving the FE model of the Hybrid III-50 ATD (HIII-50) v5.1.

An integrated stochastic design framework that facilitates practical applications involving time-consuming CAE simulations is described. The probabilistic performance measure that addresses stochastic uncertainties in CAE modeling and simulations is used to support design decision-making. Two enabling metamodeling methods using cross-validated radial basis functions (CVRBF) and a corresponding uniform sampling method are introduced to approximate highly nonlinear CAE model input/output relationships. A vehicle restraint system example is used to demonstrate the effectiveness of the proposed framework and enabling techniques.

Stochastic simulation is used to account for the variation in the manufacturing and assembling processes of the vehicle structure and occupant restraint system. An integrated full vehicle model with belted driver, 50th percentile male Hybrid III dummy, subjected to a 35 mph zero degree impact test, is used to present the scatter in the vehicle crash and occupant restraint performance. Yield stress of a typical mild steel has scatter values between 10 to 20% and a coefficient of variation of 5% is derived for scaling the stress and strain curve. The thickness tolerance has scatter values specified between 5 to 10% and a coefficient of variation of 2% is used in the study. The material properties and thickness of major structural components for absorbing impact energies, such as the motor/occupant compartment rails and upper rails, bumper beam, cradle, and toe pan are the random input variables for the structure.