一种快速收敛的动态松弛协同优化方法  

Dynamic Relaxation Cooperative Optimization Method with Fast Convergence

作　　者：陈静[1] 吕玉超[2] 王利敏 

机构地区：[1]桂林理工大学机械与控制工程学院,广西桂林541000 [2]桂林理工大学信息科学与工程学院,广西桂林541000

出　　处：《系统仿真学报》2018年第1期96-104,共9页Journal of System Simulation

基　　金：国家自然科学基金(51365010)

摘　　要：针对CO优化过程中计算成本较大,远离最优解的初始点收敛速度较慢的问题,提出一种快速收敛的动态松弛协同优化方法。该方法的计算过程分为两个阶段:加速收敛阶段对松弛因子的计算方法进行改进,采用各学科优化解与优化解均值之间的不一致信息构造松弛因子;优化求解阶段以加速收敛阶段的最优解作为初始点,选取符合一致性精度要求的松弛因子进行协同优化,求得全局最优解。通过典型数值算例和减速器多学科设计优化问题对该方法进行验证,结果表明,该方法能够有效降低计算成本,加快远离最优解初始点的收敛速度。To solve problem of high computational cost and low convergence speed of initial points away from the optimal solution in collaborative optimization, a new dynamic relaxation cooperative optimization method with fast convergence is presented. Two-phase optimization is adopted in this method. In the accelerating convergence phase, the calculation method of relaxation factor is improved, and the inconsistent information between the optimal value of disciplines and its mean value is used to construct the relaxation factor. The optimization solution of thefirst phase is adopted as the initial points in the optimization solution phase. The relaxation factor satisfying the consistent precision requirement is selected for cooperative optimization, and the global optimal solution is obtained. A typical numerical example and the reducer MDO problem are adopted to test this optimization method. Experimental results show that the proposed method can greatly reduce the computational cost and accelerate the convergence speed of the initial points away from the optimal solution.

关 键 词：多学科优化 协同优化 松弛因子 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]