An Integrated Surrogate Modeling Approach for Materials and Process Design

In order to describe continuous optimization tasks for the efficient design of materials and production processes from a reasonable data sample size, we propose an integrated surrogate modeling approach. We show the proof of concept by application to a draw bending simulation that describes the relation between the process parameters and the spring-back as the process result. The introduced concept can also be directly applied to experimental data while taking into account the process noise as uncertainty (e.g. for process control). The integrated approach combines three components: Design of Experiments, surrogate process modeling (based on function approximation by regression, e.g. Artificial Neural Networks) and optimization of process or material parameters. The identified parameters enable to rapidly find the optimal operating conditions for real experiments or to constrain them for further detailed simulation studies. Future work involves applications to more complex experiments or simulations to efficiently determine the optimal process or material parameters by sparse and adaptive data samples.