Multi-objective optimization of a bistable curved shell with controllablethickness based on machine learning  

作　　者：Shiqing Huang Chenjie Zhao Xiaoqian Ning Wenhua Zhang Huifeng Xi Zhiwei Wang Changxian Wang 

机构地区：[1]School of Mechanics and Construction Engineering,Jinan University,Guangzhou 510632,China [2]College of Packaging Engineering,Jinan University,Zhuhai 519070,China [3]MOE key Lab of Disaster Forecast and Control in Engineering,Jinan University,Guangzhou 510632,China [4]Academy of Military Sciences,Beijing 100091,China

出　　处：《Theoretical & Applied Mechanics Letters》2024年第6期467-476,共10页力学快报(英文版)

基　　金：supported by the National Natural Science Foundation of China(Grant Nos.12102143,12172151,and 12172149);National funded postdoctoral researcher program(Grant No.GZC20230962);Guangdong Basic and Applied Basic Research Foundation(Grant No.2024A1515010379);Natural Science Foundation of Guangzhou City(Grant Nos.202201010217,202201020539,and 2024A04J3401);Young Talent Support Project of Guangzhou Association for Science and Technology,the Fellowship of China Postdoctoral Science Foundation(Grant No.2022M711333);the Fundamental Research Funds for the Central Universities(Grant No.21623332);the High Performance Public Computing Service Platform of Jinan University.

摘　　要：Bistable curved shells have become a promising low-cost application in energy absorption fields owing to recentadvances in material and technology.Significant research has been conducted to improve their energy absorptioneffect through forward prediction and single-objective optimization.However,these approaches may not fully explore their functional potential.In this study,we propose a multi-objective optimization framework based on theprinciple of main objective optimization that combines neural networks and genetic algorithms.The energy absorption effect and backward snapping force of the bistable curved shell are improved synchronously.Meanwhile,a reverse design algorithm is developed to generate the preset load-displacement curve,which further expandsthe application of machine learning methods in the field of multi-objective optimization.The combination ofmachine learning and multi-objective optimization is highly effective for building meta-structures with specificperformance requirements and has potential applications in solving complex optimization tasks in various fields.

关 键 词：Bistable curved shell Structure optimization Machine learning Multi-objective optimization Reverse design 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]