基于一种新混合群算法(MOHO—SA)的结构主动控制多目标优化研究
2017-06-19潘兆东谭平周福霖
潘兆东+谭平+周福霖



摘 要:基于粒子群(PSO)算法和差分进化(DE)算法提出了一种新的多目标混合群优化算法,对结构主动控制系统的作动器位置、数量与控制器参数进行同步优化.首先,分别采用PSO算法与DE算法进行对应种群的进化,使用庄家法则构造非支配解集,并引入边界点几何中心leader选择机制,同时利用模拟退火算法完成个体进化的二级局部搜索;以随机地震激励下反映结构振动控制效果和控制策略优劣的双指标作为优化目标函数.最后,针对ASCE 9层benchmark模型,采用本文提出的具有二级搜索功能的多目标混合群算法(MOHO-SA)对其主动控制系统进行优化设计,并分别与多目标差分进化算法(MODE)、多目标粒子群算法(MOPSO)、普通多目标混合群算法(MOHA)的优化结果进行对比分析,表明其Pareto解集具有更优的收敛曲线及分布性.
关键词:主动控制;混合群算法;二级搜索;多目标优化;庄家法则;几何中心leader
中图分类号:TU375.3 文献标志码:A
文章编号:1674-2974(2017)05-0020-07
Abstract:This paper proposes a new multi-objective hybrid swarm optimization method for active control system based on particle swarm algorithm and differential evolution algorithm, in which the parameters of controller, and the number of and allocation of actuator are synchronously optimized. The basic idea is as follows: The different algorithms are used to complete the evolution of corresponding population, the non-dominated solution set is achieved based on the dealer principle, and the leader selection based on boundary point geometry center is adopted. Meanwhile, the simulated annealing algorithm is used for the secondary local search, the two indexes reflecting the structural vibration control effect and performance of control strategy are used as the optimization objective function. Finally, a ASCE 9-story benchmark model is used as a numerical example to validate the effectiveness of the proposed method. Compared with the conventional MODE, MOPSO, and MOHA algorithm, the MOHO-SA algorithm has better convergence curve and distribution of the pareto solution sets.
Key words: active control;hybrid swarm algorithm; two level search; multi-objective optimization; dealer principle; geometric center leader selection
在土木工程結构主动控制研究领域中,作动器数量、位置及控制器参数的优化一直是研究热点之一.面对规模宏大、结构复杂、功能多样的超限工程结构,振动控制系统若采用传统的优化方法进行设计必然很难得到最优解.随机类搜索方法如模拟退火法(SA)、遗传算法(GA)及群算法为全局优化算法且可用于离散优化问题,因而被广泛地用于控制装置的位置优化研究[1-10].但以往的研究多是基于特定的外界激励、优化准则及单一优化目……
