Abstract:
|
This presentation focuses on the needs of industry for engineering optimization tools. We clarify that industrial demands go far beyond what typical academic research in evolutionary computation is focusing on - and some demands can not even be met today. We also clarify that, in order to meet industrial requirements, the focus needs to be broader than just agorithmic issues. Even more important are usability, stability and fail-safeness, and the compatibility of the optimizer with common finite element or CFD simulation tools. As a matter of fact, we will also show extremely demanding examples where the evolution strategy (after Schwefel and Rechenber, further extended by NuTech Solutions) performs much better than competing algorithms both in terms of final solution quality and speed of optimization (i.e., number of function evaluations). The algorithm is today being used by major automotive OEMs in Germany, including BMW, Volkswagen, and others.
|