A CPU-parallel thermal topology optimization framework for arbitrary structures using flood fill algorithm and implicit identification model

Jiawei Wu

doi:10.1016/j.cma.2025.118165

Return

A CPU-parallel thermal topology optimization framework for arbitrary structures using flood fill algorithm and implicit identification model

2025-06-24

PRE

Jiawei Wu

Gao Liang (Liang Gao)

王林 (Lin Wang)

周小强 (Xiaoqiang Zhou)

高杰 (Jie Gao)

DOI：10.1016/j.cma.2025.118165

Original

Original request for help

Save

Abstract

En 中文

Recently, advancements in industries such as semiconductor manufacturing have brought the essential issue of heat transfer to the forefront of engineering challenges. As technological demands escalate, existing design methodologies are proving increasingly inadequate to meet performance requirements. In the current work, the primary intention is to develop a promising and efficient multi-core CPU-parallel heat transfer topology optimization design framework. Firstly, an implicit identification model is developed for automatically recognizing model shapes using Boolean operations for the signed distance field, which intends to simplify the design process by skipping the complex manual meshing stage as much as possible. Secondly, the Flood Fill algorithm, inspired by the simulation of water spreading, is introduced into thermal topology optimization to eliminate isolated structures that are generated near the tips of the design domain. Meanwhile, the Multi-stage Volume Relaxation Strategy is considered here to ensure that the topology optimization rapidly converges toward the global optimum solution for the thermal problem. Thirdly, the multi-core CPU-parallel computation framework for thermal topology optimization, which can considerably accelerate the optimization efficiency, is developed with the derivation of sensitivity analysis. Finally, the effectiveness in approaching global optimum and computational efficiency of the CPU-parallel algorithm are demonstrated through several sink structures to showcase excellent engineering applicability of this framework.