应用径向基代理模型实现序列自适应再采样优化策略

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,使用该再采样策略后,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。

Adaptive resampling strategy of sequential optimization based on radial basis function surrogate model

Taking account of that it is difficult to obtain the error function to make sequential re-sampling optimization when the predicted error of sampling points in radial basis function(RBF) interpolation surrogate model is zero, the constraint of sampling point distribution was applied in the process of sequential re-sampling. Taking advantage of the convergence performance of potential optimal solution, a re-sampling strategy which is suitable for the sequential optimization of RBF interpolation surrogate model was proposed. The strategy matches the input response property of emulation model with the spatial distribution property of sampling points. Simulation results indicate that the optimization efficiency and precision of the proposed strategy is higher than that of the traditional optimization method based on surrogate model. The optimum point can be well predicted and the number of computational times in primitive model can be reduced obviously by the proposed strategy.