集成电路与嵌入式系统2026，Vol.26Issue(5)：57-64,8.DOI:10.20193/j.ices2097-4191.2025.0091
基于SVR与改进NSGA-Ⅱ的MCM散热-成本协同优化设计
Collaborative optimization design of thermal performance and cost for MCM based on SVR and improved NSGA-Ⅱ
摘要
Abstract
As a key technology for improving the performance and integration density of integrated circuits,the layout design of Multi-Chip Module(MCM)packaging has a significant impact on system performance and cost.Traditional design methods relying on simula-tions and experiments are inefficient and difficult to achieve effective trade-offs in complex design spaces.This paper proposes a thermal performance and cost collaborative design strategy integrating automated simulation,surrogate models,and multi-objective optimization algorithms.First,Python secondary development was used to realize the full-process automation of numerical simulations,greatly im-proving simulation efficiency.Then,a high-precision surrogate model between chip spacing and maximum temperature was established,and fast prediction of thermal performance was achieved based on Support Vector Regression(SVR).On this basis,combined with the MCM cost model,the improved NSGA-II multi-objective algorithm was adopted to efficiently search for the Pareto optimal front.The research results show that the obtained Pareto front curve is superior to the original layout scheme,and the automated framework and intelligent algorithms significantly improve the efficiency of MCM collaborative design,providing a basis for the trade-off between ther-mal performance and cost.关键词
多芯片模块/散热仿真/支持向量机回归/改进NSGA-Ⅱ/协同优化设计Key words
multi-chip module/thermal simulation/support vector regression/improved NSGA-LL/collaborative optimization design分类
信息技术与安全科学引用本文复制引用
杨仁贵,顾杰斐,宿磊,李可,明雪飞..基于SVR与改进NSGA-Ⅱ的MCM散热-成本协同优化设计[J].集成电路与嵌入式系统,2026,26(5):57-64,8.基金项目
国家自然科学基金项目(U23B2044) （U23B2044）
国家重点研发计划项目(2023YFB4404203) （2023YFB4404203）
江苏省前沿技术研发计划项目(BF2024010) （BF2024010）
长三角科技创新共同体联合攻关项目(2023CSJGG0204) （2023CSJGG0204）
珠海市产学研合作项目(2320004002629) （2320004002629）
