工程结构优化的进化策略-高斯过程协同优化方法  

Engineering structural optimization using a cooperative optimization method based on evolution strategy and gaussian process

作　　者：苏国韶[1] 武振兴[1,2] 燕柳斌[1] 

机构地区：[1]广西大学土木建筑工程学院工程防灾与结构安全教育部重点实验室,南宁530004 [2]中建钢构有限公司,深圳518040

出　　处：《计算力学学报》2013年第5期610-615,共6页Chinese Journal of Computational Mechanics

基　　金：国家自然科学基金(51069001);广西理工科学实验中心重点(LGZX201001)资助项目

摘　　要：针对采用仿生全局优化方法进行复杂工程结构优化时数值计算量浩大导致的计算代价过高的公开问题,将自适应协方差矩阵进化策略(CMAES)全局优化算法、高斯过程(GP)机器学习技术与有限元方法相结合,提出了基于自适应协方差矩阵进化策略-高斯过程协同优化算法(CMAES-GP)的结构优化方法。该方法利用全局寻优性好且寻优效率高的CMAES算法进行全局最优搜索,当搜索进入局部寻优阶段时,采用回归性能优秀的GP模型对适应度函数进行动态拟合,进而利用GP模型替代有限元分析进行个体适应度评价,以减小局部寻优阶段的有限元重分析次数,从而实现有效降低工程结构优化计算代价的目的。算例研究表明,与传统结构优化方法相比较,本文方法具有全局性好、计算效率高的优点。The optimization of complex structures with large number of degrees of freedom using bionic optimization will have the huge computational volume and will be very time consuming. For reducing the computational burden of bionic optimization, a novel cooperative optimization method （CMAES-GPFEM） based on Covariance Matrix Adaptation Evolution Strategy （CMAES） algorithm, Gaussian process （GP） machine learning and Finite Element Method （FEM） is proposed. In the proposed method, CMAES algorithm is applied to search the globe optimization solution. GP model is applied to approximate the fitness function when the CMAES algorithm going into the statue of searching local minimum. Thus, the computational cost induced by reducing the numbers of finite element analysis during the local searching process. Compared with the traditional optimization and bionic optimization method for structural optimization,the proposed method is much more economical to achieve the global solution.

关 键 词：结构优化设计 进化策略 高斯过程 近似模型 

分 类 号：TU323.4[建筑科学—结构工程] O224[理学—运筹学与控制论]