Abstract

In the virtual product development of a car, numerous design changes are applied and analysed until the final model satisfies given design criteria. We focus on crashworthiness requirements, where this procedure usually results in large development trees with many design changes and corresponding simulation results. Following a path in the obtained development tree, the differences from one simulation to the next consist of one or several design variations, which result in numerous changes in the crash behaviour. These differences are mainly determined by local changes in a subset of the car parts. During the development process each of the design variations has to be analyzed and its influences on the crash results have to be compared and evaluated. To simplify and structure this process, we developed a framework to easily analyze the impact of design variations, including a structured data representation. First, we analyse and store the design variations from one model to the next. Second, in order to detect events, such as anomalies or unusual variations in mesh quantities such as deformations or plastic strain, we analyse and store the changes between the two obtained simulation results. Moreover, we developed a tool to automatically highlight the most relevant parts together with the local areas of high deviations between the two simulation runs. The comparison can be based on arbitrary node- and element data functions, e.g. displacements, plastic strains, or thicknesses. The presented framework is an important step towards automatic design and event detection in an overall simulation data analysis workflow. It allows to systematically analyze a design development tree to detect interesting deviations in the crash behaviour caused by design changes and to store these observations in a structured data representation for further analysis by for example artificial intelligence approaches. We demonstrate the framework on a frontal crash example.