Nivesh Dommaraju, Mariusz Bujny, Stefan Menzel, Markus Olhofer, Fabian Duddeck, "Cooperative framework for many objective crash optimization", 14th World Congress of Structural and Multidisciplinary Optimization, 2021.

Topology optimization with multiple objectives may yield a large set of Pareto optimal designs. Among other methods used to analyze and summarize the optimal solutions, clustering methods can be used to identify a few representative designs, which can be more easily reviewed by a designer. For example, Dommaraju et al. propose to select diverse designs based on geometric features. To generate the Pareto front, well-studied evolutionary algorithms such as NSGA2 are useful but they are expensive. In contrast, methods such as HCA-SEW which vary the weights, expressing the relative preference, for the objectives are more economical in steering the optimization process towards a Pareto optimal design. A natural extension would be to guide HCA-SEW to find designs on the Pareto front with certain geometric features, e.g., designs similar to a specific reference needs to be selected, or designs within a certain performance range. Since the engineer may not know the possible performance ranges and geometrical diversity, we describe a more useful framework for cooperative optimization by supporting the designer to formulate their preferences.