基于径向基代理模型的序列优化中自适应再采样策略

针对径向基插值代理模型样本点预测误差为零时无法获得误差函数进行序列再采样优化的问题，将样本点分布约束引入序列再采样过程，利用潜在最优解加速收敛性，提出一种适用于径向基插值代理模型序列优化的再采样策略，该策略兼顾仿真模型的输出响应特性与样本点的空间分布特性。仿真结果表明，使用该在采样策略后，算法寻优效率和精度均优于传统基于代理模型的优化方法，在对最优解进行有效预测的同时，能显著减少原始模型计算次数。

daptive infill strategy of RBF metamodel based sequential optimization

Taken into account that radial basis function interpolation surrogate model predicts accurately through all the sampling points and it is difficult to obtain the error function to make sequential optimization, constraint of sampling point distribution was applied in the process of infilling strategy. Taking advantage of the convergence performance of potential optimal, a novel infill strategy was proposed which can used for the RBF interpolation surrogate model based optimization. In proposed method, the input response and the spatial distribution properties were considered. Numerical experiment indicated that the proposed strategy lead to higher optimization efficiency and precision than the traditional surrogate model based optimization. The optimum point can be well predicted and the number of calling times can be reduced by the proposed method.