Abstract

During the last years simulation data management has been adopted by more and more automotive OEMs and other industries. However only in few cases the gathered data has been used in order to automatically acquire knowledge beyond individual simulations. Running DOEs, parameter studies and optimizations has always been in the domain of specialized optimization and stochastic tools. These tools are rather complicated in operation for the average user and their integration into the processes within a simulation data management system is usually not an easy task. Furthermore, the application of the methods provided by such software on the analysis of unstructured data, which is generated in large quantities by simulation data management systems, has been approached very rarely. In this paper we shall outlay the possibility given by a simulation data management system to handle such tasks. It will also be demonstrated how the analysis methodologies known from standard optimization tools can be applied to the simulation results directly integrated in the workflows of a simulation data management environment. On the side of model preparation , where all the input data is Administrated within the SDM system it will be shown how to manage also parameters which are directly linked to the input data. This gives great opportunities to integrate with standard optimization tools for setting up DOEs, parameter studies and so on. Furthermore, this can be extended by integrating some of the functionality of these tools directly with the SDM system such that a DOE can be completely set up from within the SDM System. This simplifies the process significantly and helps users who are not so familiar with expert tools. On the other hand a SDM system is also used to manage large amounts of result and even physical test data. It will be shown how the integration of a data analysis framework with additional visualizations options can aid the engineer to discover trends and relationships in their simulation data. For instance, relationships between input and output variables of a DOE can be visualized with correlation plots and response surfaces. In addition, different outlier detection algorithms of scalar values are implemented, so that larger data sets can be examined. In cooperation with partners, investigations in outlier analysis at simulation level has been performed. In this paper the conceptual idea and the results, with focus on the integration in a SDM system, will be presented.