基于自适应协方差矩阵进化策略的结构可靠度计算

摘要

Covariance matrix adaption evolution strategy (CMA-ES) is a guidable random optimization method with excellent ability of searching the global best solution in great depth and width of the solution space.Aiming to the shortcoming of genetic algorithm (GA)and particle swarm optimization algorithm ( PSO ) in the high calculation cost, CMA-ES algorithm is applied to calculate structure reliability based on the geometry implication and the checking points method.Study results show that this method is feasible and has the merits of excellent global optimization performance and fast convergence.Comparing with GA and PSO algorithm, CMA-ES algorithm could increase the computational efficiency remarkably with less function evaluations.It provides a new way for the structural reliability calculation.%自适应协方差矩阵进化策略(CMA-ES)算法是一种引导式随机优化算法,兼顾了深度搜索最优解和广度搜索解空间的能力.针对采用遗传算法(GA)、粒子群优化算法(PSO)等仿生优化算法求解复杂结构可靠度时往往遇到计算代价过高的问题,基于结构可靠度指标的几何涵义并结合验算点法,提出了结构可靠度计算的自适应协方差矩阵进化策略方法.研究结果表明,该方法是可行的,具有全局性好、收敛速度快的优点,与遗传算法、粒子群优化算法相比较,可大幅度地提高计算效率,为结构可靠度计算提供了一条新的途径.