The numerical simulation is an inherent process of the development of the passive vehicle safety. Robust and predictable computational models are the base of the successful application of numerical simulations. This study is focused on the assessment of the quality of dummy models used in occupant simulations. The progress of those models was remarkable over the past years. By increasing the quality, the potential of further improvements declines. Hence, the assessment of model improvements and their impact on the quality of simulations is getting more and more complicated. Major improvements of sub-parts do not necessarily improve the overall performance of a model. Therefore, a standardised objective evaluation of models would ease the definition of priorities of model updates.
Objective rating tools could help to solve this problem. These tools are calculating the level of correlation between two signals, usually coming from test and simulation. All signal ratings can be merged to a global rating of a loading case. However, the analysis of only one loading case is not sufficient to calculate a reliable and a robust quality score of a dummy model. A more comprehensive approach is required to provide a valid rating for all relevant loading conditions. Furthermore, it must distinguish between good and poor models and should correlate with user experiences.
This paper provides guidelines of defining boundary conditions of an overall quality rating of dummy models. The LS-Dyna ES-2 dummy model was used as a demonstrator of the new approach.
The study analyses the possibilities of an objective rating tool. Various tests with dummy parts, sled tests as well as dummy certification tests were analysed to define a set of characteristic loading conditions of the ES-2. Furthermore, the extraction of the most relevant dummy responses was an essential part of the evaluation, too. Finally, all defined scenarios were applied to different releases of the same dummy model. The calculated quality scores were verified with the experiences of users of the model.
The findings of this feasibility study are limited to the LS-Dyna ES-2 model. However, they can easily be transferred to other ES-2 models. If another side impact dummy or a dummy for a different crash scenario (e.g. frontal impact) is used, then the selection of loading cases and signals must be revised.