关键词: 自适应采样, 反距离插值, 克里金插值, 雷达截面积

Abstract: When surrogate models are applied to datasets with significant gradient variations,substantial discrepancies often arise between predicted and actual values near extremum points.These inaccuracies severely limit their effectiveness in engineering applications.Although increasing the sample size can mitigate this issue to some extent,the improvement is often marginal,and excessive sampling can significantly reduce computational efficiency.The optimal solution lies in adaptively adding interpolation sample points near the extremum or inflection points of the dataset.A novel sample adaptive construction method based on IDW-Kriging difference analysis was introduced.By leveraging the differences between IDW and Kriging surrogate models,the potential extremum points were identified,and the additional sampling points were adaptively introduced at these critical locations.This approach enabled the construction of a more precise surrogate model dataset.The effectiveness of the proposed method was validated through a jet aircraft case study.Experimental results demonstrated that compared to conventional one-time sampling methods,the proposed adaptive sampling approach improved prediction accuracy by 68% under equivalent data volumes.Moreover,the maximum computational errors at extremum points were reduced by 22.9 dBsm.

Key words: adaptive sampling, inverse distance weighting, Kriging interpolation, radar cross section