Diese Website verwendet Cookies. Mit der Nutzung der Website stimmen Sie deren Verwendung zu. Weitere Informationen erhalten Sie in unserer Datenschutzerklärung.

Improving Robustness of Chevrolet Silverado with Exemplary Design Adaptations Based on Identified Scatter Sources

The investigations described here are related to the unstable behavior of crash-simulations due to minor changes in the model. As a consequence the received simulation results become in some way unpredictable, whereby the causes can be various: e.g. modeling failure, contact issues, numerical instabilities, physical instabilities, etc.. To identify and separate these scatter sources the results are analyzed by means of visualizing the standard deviation of scatter itself and computing scatter-modes for selected parts of interest. Latter computations are based on the principle component analysis (PCA), and deliver new virtual crash results representing the most extreme geometrical shapes of the scatter-modes. This improves and speeds up the process of identifying scatter causes. For illustration a realistic application case based on the freely available Chrysler Silverado from the National Crash Analysis Center (NCAC) of The George Washington University is analyzed by means of robust design of the crash model. Therefore 25 simulation runs were performed based on small random part thickness changes (representing production tolerances). The part of interest for the investigations is the variance at the fire-wall. As an outcome major scatter sources in the interaction of power-brake and suspension as well as at the longitudinal rail are found which are strongly correlated to the firewall scatter. Approving the software based prediction exemplary design adaptations lead to a significant reduction of scatter on the firewall. The described mathematical methods are part of the software DIFFCRASH, which was used in this study. Additionally a perspective regarding the integration of these analysis methods into a simulation data management system is given.