基于不同MCMC抽样子集模拟的模型修正方法对比研究  

Comparative Study on Model Updating Methods Based on Subset Simulation with Different MCMC Samples

作　　者：夏志远 陈鹏 唐柏鉴 史慧媛 XIA Zhiyuan;CHEN Peng;TANG Baijian;SHI Huiyuan(School of Civil Engineering,Suzhou University of Science and Technology,Suzhou 215011,Jiangsu,China)

机构地区：[1]苏州科技大学土木工程学院,江苏苏州215011

出　　处：《力学季刊》2023年第3期662-672,共11页Chinese Quarterly of Mechanics

基　　金：国家自然科学基金(52208188,52108236);江苏省自然科学基金(BK20200986);江苏省科协青年科技人才托举工程(2021-077);苏州市建设系统科研项目;江苏省交通运输重点科技项目(2022QD17)。

摘　　要：子集模拟方法作为结构可靠度分析方法,也可应用于工程优化问题,诸如优化设计、模型修正等.为研究基于不同蒙特卡洛马尔可夫链(MonteCarloMarkovChain,MCMC)抽样的子集模拟优化方法(Subset Simulation Optimization,SSO),以有限元模型修正作为优化背景问题,开展其精度和效率的对比研究.介绍标准SSO和子集模拟(Subset Simulation,SS)常见的MCMC抽样方法,并基于上述不同MCMC抽样的SSO开展某局部损伤悬臂梁(10维变量)的有限元模型修正,修正结果与基于遗传算法(Genetic Algorithm,GA)的模型修正方法进行对比;而后将上述不同MCMC抽样的SSO修正方法应用于某四层钢框架有限元模型修正中(11维变量).结果表明,采用随机游走的延迟拒绝修正M-H方法(MMH algorithm with Delayed Rejection,MMHDR)和自适应条件抽样方法(Adaptive Conditional Sampling,ACS)的SSO有限元模型修正具有较好的精度和效率,在工程结构有限元模型修正中更具优势.Subset simulation method,as a technique for structural reliability analysis,can also be applied to engineering optimization problems,such as optimal design,model modification,etc.In order to study the performance of subset simulation optimization(SSO)methods based on different Monte Carlo Markov Chain(MCMC)sampling approach,finite element model updating is used as the optimization background problem to carry out a comparative study on its accuracy and efficiency.The common MCMC sampling methods of standard SSO and subset simulation(SS)are introduced.Based on the above different MCMC sampling SSOs,the finite element model of a cantilever beam with local damage(10 dimensional variables)is performed.The updating results are compared with the model updating method based on genetic algorithm(Genetic Algorithm,GA).Then,the SSO updating methods with different MCMC sampling are applied to the finite element model updating of a four-story steel frame(11 dimensional variables).The results show that the SSO finite element model updating using the random-walking modified Metropolis Hasting algorithm with delayed rejection(MMHDR)and the adaptive conditional sampling(ACS)method has better accuracy and efficiency,and has more advantages applied in the engineering structure finite element model updating problems.

关 键 词：有限元模型修正 子集模拟优化 MCMC抽样 延迟拒绝修正M-H方法 自适应条件抽样 

分 类 号：O224[理学—运筹学与控制论]