基于自适应采样的复杂模型全局近似

针对复杂模型近似处理的问题，提出自适应采样结合曲面曲率的全局近似方法. 采用自适应设计域分割采样方法获取新增采样点，逐步提高源模型的响应面近似模型精度. 引入判定响应面近似模型精度，提出利用几何方法计算曲面曲率，并结合启发式直接搜索算法（DIRECT）搜索响应面模型上的最大曲率点及设计域最佳分割位置. 所提方法可以运用于其他响应面模型，并适合用于大设计域、大数据源模型的近似处理. 函数源模型及复杂电动车模型的近似处理测试结果表明，所提方法具有实用性和有效性.

Global approximation of complex model based on adaptive sampling

A global approximation method combining adaptive sampling and surface curvature was proposed to deal with the problem of approximation processing of complex models. The adaptive design domain segmentation sampling method was used to obtain the new sampling points, and gradually improve the accuracy of the response surface approximation model of the source model. A method to determine the accuracy of the response surface approximation model was introduced, and a geometric method was proposed to calculate the surface curvature, and combined with the heuristic search algorithm (DIRECT) to search the maximum curvature point on the response surface model and the best segmentation position of the design domain. The proposed method can be applied to other response surface models and is suitable for the approximate processing of large design domain and large data source models. The approximate processing test results of function source model and complex pure electric vehicle model show that the proposed method is practical and effective.