Optimization Processes for Automated Design of Industrial Systems

Mechanics was one of the earliest application fields demonstrating the power of rule-based deduction of general formalisms from rather few basic axioms. The usual transformation of design problems into classical optimization problems and their solution by nonlinear programming algorithms follows the same principle, and is thus mainly used by the mechanics and control community. System design on an industrial scale, however, is a much more challenging creative task, which cannot be formalized so easily, which is why it is mostly still human-driven. In order to break up this game stopper, algorithms for multi-criterion optimization, function approximation and statistical sensitivity analysis may be integrated in a common design process. Especially the emerging field of data-driven artificial intelligence (AI) methods may become the pushing game changer. The paper will demonstrate major challenges of industrial design and some solution strategies to bridge the gap between engineering design intuition and formalized problem solution.