Robustness Evaluation Crashworthiness Simulation Results

The numerical robustness of simulation results from explicit time integration is an important topic. We know for real world applications of passive safety and crashworthiness that we have some numerical noise, but the interesting question is if that really does influence significantly our simulation results. Furthermore, today the robustness of the designs against naturally given input scatter, in loading conditions, geometry or material become part of the virtual product development process. Then, the prognosis of the variation of important simulation results using stochastic analysis procedure is necessary. Again the question arises how much of the calculated variation is coming from numerical noise. The paper will present a procedure of numerical robustness evaluation using stochastic analysis to quantify the scatter of simulation results. Using coefficients of determination, a procedure of deselecting variation defined by correlation to physical input scatter and "undefined" variation is introduced. The breakthrough in practical application and the acceptance of stochastic analysis for robustness evaluations was achieved by using linear and quadratic correlations and the corresponding measures of determination as well as by projection of statistical measures on the finite element structure.