Automatic parameter classification for dimension reduction as basis for robust parameter identification

This paper presents a fully automatic parameter classification procedure in order to identify the most influencing parameters together with locally interesting parts of the component considered in a certain processing step. The results of this classification approach are used for a parameter space reduction in order to minimize the computational effort for subsequent analysis and optimization tasks based on forecast models. In particular, an outlook on the evaluation of radial basis function metamodels for a robust parameter identification is given. We demonstrate the classification procedure and its benefits by an industrially relevant deep drawing process of a pan with secondary design elements.