Multi Points Updated and Distance Filtered Kriging Surrogate Model: Application in EOSS Optimization

We put forward a multi points updated and distance filtered Kriging surrogate model. When building Kriging models, two update approaches are used to select infilling points: optimal points and maximized expected improvement. We use Improved general pattern search (IGPS) algorithm to get these points. In IGPS, search step is substituted by GA and SQP and poll step is retained. To decrease simulation times, we adopt a distance filter to eliminate potentially replicated samples. A satellite orbit parameter optimization problem is formulated, which is solved by the proposed method and STK/Analyzer respectively. The results showed that Kriging models, which use multi update points and distance filter, yield global approximations that are more accurate than Analyzer.