论文部分内容阅读
针对昂贵单目标约束优化中真实模型计算费时且现有算法收敛速度慢的问题,提出了动态Kriging优化算法以提高计算效率.该算法首先将所有约束条件转换为一个约束函数,然后采用拉丁超立方体采样(LHS)法进行采样,分别建立真实模型目标函数和约束函数的Kriging代理模型,同时结合真实模型对代理模型估计进行误差矫正,采用非支配个体选择、保留和替换机制不断更新样本库和Kriging代理模型.最后将进化最优种群代入真实模型计算其最优值.通过13个标准函数测试表明该算法具有较高的精确度和稳健性,明显减少了真实模型的评价次数.
To solve the problem of time-consuming real model computation in expensive single-objective constrained optimization and the slow convergence of the existing algorithms, a dynamic Kriging optimization algorithm is proposed to improve the computational efficiency. The algorithm first converts all the constraints into a constraint function and then uses the Latin hypercube (LHS) method, the Kriging agent model of the objective function and the constraint function of the real model is established respectively. At the same time, the error correction of the proxy model estimation is combined with the real model. The non-dominated individual selection, reservation and replacement mechanism are used to continuously update the sample database and Kriging Proxy model.At last, the optimal evolutionary population is substituted into the real model to calculate its optimal value.Through 13 standard function tests show that the algorithm has high accuracy and robustness, significantly reducing the number of real model evaluation.