Parallel Enclosed Hole Detection for Introducing Manufacturing Constraints in Topology Optimization

This work presents a distributed approach based on Graph Theory for detecting enclosed holes in density-based topology optimization using the traditional material penalization scheme.The distributed topology optimization framework divides the domain into several subdomains to exploit parallel computing resources.The proposal generates a set of graphs from the empty elements of the meshes of such subdomains, detecting the empty regions and classifying them.Then, it generates a hierarchical distributed graph between the subdomains to obtain a coherent representation of the connectivity design, minimizing inter-node communications in the classification of the empty regions.We use such information to introduce manufacturing constraints preventing enclosed holes in the final designs, which is especially useful in diverse additive manufacturing techniques because such holes can induce failures in the manufacturing process.We validate the proposal using a classical two-dimensional cantilever problem with asymmetric simplifications and parallel computing.