Process to Improve Optimization with Combined Robustness Analysis Results

Today’s engineers are facing a challenge of creating both robust and optimized solutions to fulfil a variety of requirements. While being able to create designs that can handle parameter variability and still produce predictable results, on the other hand the models also need to be optimized with respect to lower production and development costs. Thus following the growing demand to create both robust and optimized solutions, new processes and techniques are needed allowing the engineer to improve his CAE model. Not only to speed up the development process but also to improve the quality of the model. To manage the above mentioned challenge a new process is introduced creating a direct link from a Principal Component Analysis (PCA) driven robustness analysis towards an LS-Opt driven metamodeling based optimization. It is shown that the results of the robustness analysis can be directly fed into the optimization task to improve both the quality of the metamodel as well as the therewith calculated optimization results. For illustration a publically available application case is analyzed by means of robust design and optimization of the crash model. Therefore 77 simulation runs were performed based on random thickness variations and analyzed with the above mentioned methods. The described PCA based methods are part of the software DIFFCRASH.