A bottleneck-aware two-stage evolutionary algorithm for heat pipe-constrained component layout optimization

Electronic industry aspires to design portable and high-performance products to meet the market demand, where intensive heating components appear to contradict small-sized boards with limited capacity of heat-conducting. Driven by the real-life requirements and challenges, this paper proposes a bottleneck-aware two-stage evolutionary algorithm, multi-start variable neighborhood descent with genetic algorithm (MSVND-GA), to address the heat pipe-constrained component layout optimization (HCLO) for minimizing the maximum heat pipe power. The HCLO problem is decomposed into heat pipe assignment (HA) and component location (CL) sub-problems to reduce the complexity. Firstly, the proposed algorithm applies the MSVND with a problem-specific integer encoding scheme to the HA, in which five bottleneck-aware neighborhood structures are adopted to find high-quality heat pipe assignments. Afterwards, the GA serves to identify coordinates of components for obtained assignments, thus ensuring generated solutions meet practical engineering constraints in the CL. The proposed MSVND-GA is tested on both the CEC 2022 benchmark and extended instances, and is compared against state-of-the-art methods. Extensive experimental results indicate the superiority of the MSVND-GA in the optimality, stability, and feasible rates. Furthermore, the proposed algorithm won the first prize in the HCLO competition at 2022 World Congress on Computational Intelligence.