Reaction–diffusion equation driven topology optimization of high-resolution and feature-rich structures using unstructured meshes. (June 2023)

• Record Type: Journal Article
• Authors: Li, Hao; Yu, Minghao; Jolivet, Pierre; Alexandersen, Joe; Kondoh, Tsuguo; Hu, Tiannan; Furuta, Kozo; Izui, Kazuhiro; Nishiwaki, Shinji
• Is Part Of: Advances in engineering software. Volume 180(2023)
• Publication Date: 2023-06
• Subjects: Topology optimization -- Reaction–diffusion equation -- Maximum length-scale constraint -- PDE filter -- Lattice designing -- Unstructured mesh -- FreeFEM-PETSc-ParMmg; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2023.103457
• Languages: English
• ISSNs: 0965-9978

Abstract

Abstract: Motivated by the need for porous structures in the design of biodegradable implants, as well as for the diverse and competitive designs in architecture, this paper builds upon the recent advances in the single-scale (macroscopic) topology optimization (TO) approach and proposes a level-set method (LSM) for the design of lattice structures. The key idea is to introduce the maximum length-scale constraint, realized by a PDE filter, into the reaction–diffusion equation (RDE)-based LSM which can end up with feature-rich shape, starting from scratch. Then, the proposed TO technique is coupled with distributed computing algorithms based on PETSc, interfaced in the open-source software FreeFEM using unstructured meshes, to deliver both two- and three-dimensional designs with complex-shape geometry. The parallel efficiency is validated by solving large-scale 3D benchmarks with 20 million tetrahedral elements on a cloud-based cluster. Furthermore, we integrate local mesh refinement ( h -adaptation scheme) into this workflow, allowing one to reach high-resolution and feature-rich 3D designs on a desktop, paving the way to future investigations, aimed at including the maximum length-scale constraint in large-scale multiphysics topology optimization at a reasonable cost. Graphical abstract: Highlights: A reaction–diffusion equation-driven TO method is built for feature-rich designs. The maximum length-scale constraint is realized using a variational method. A fully-distributed TO framework is built using FreeFEM-PETSc-ParMmg. Unstructured and/or adaptive meshes can be used in this framework. A large-scale high-resolution 3D problem can be solved on a desktop.