A topology-preserving level set method for shape optimization (Q1777048)

The level set method has been shown to be very effective in solving shape optimization problems. While the level set method has its strong points -- one being that it gives a representation that is topology-independent -- it is not obvious how to extend it to problems where there are constraints. The constraints such a bound of the size of a connected component of the shape, or requirement that it has a fixed number of connected components are not easy to work with and the aim of the paper is to attack these problems. The authors' approach starts with the concept of subdomain neighborhood. The neighborhood of a subdomain will detect the nearness of other subdomains, and will thus allow to take action to prevent geometrical or topological changes. This strategy can be formulated as a penalty functional. Namely, the authors propose a logarithmic barrier penalty for this purpose. They then deduce a descent direction which enable to find minimizer for the obtained penalized optimization problem. Some numerical examples are presented to illustrate the method.