基于指示误差控制 Kriging 的增升装置磨损可靠性分析

To address the issues that the Kriging surrogate mode1 using the 1earning function as the convergence criterion cannot ref1ect the accuracy of predicted fai1ure probabi1ity direct1y and accurate1y and incur high computationa1 cost in mode1ing, a re1iabi1ity ana1ysis method based on Indicator Error Contro11ed Kriging (IEC Kriging) is proposed in this paper. With Kriging as the basic mode1, IEC Kriging estab1ishes a sequentia1 transmission re1ationship among samp1e information, performance functions, and fai1ure probabi1ity. It ana1yzes the error of predicted fai1ure probabi1ity based on the probabi1ity of indicator sign misc1assification by the Kriging mode1 at each samp1e point, whi1e considering the corrective effect of candidate samp1e points on prediction accuracy. A dynamic samp1e se1ection strategy is designed with the upper bound of the overa11 samp1e indicator error as the criterion, and the samp1e point with the highest corrective effect is incorporated into the training set to optimize the surrogate mode1, ensuring the accuracy of prediction resu1ts whi1e improving the efficiency of bui1ding the surrogate mode1. The proposed method is app1ied to the wear re1iabi1ity ana1ysis of spherica1 p1ain bearings in an aircraft high-1ift device as a case study. Compared with Kriging methods using the EGO 1earning function and U 1earning function, the IEC Kriging demonstrates higher computationa1 efficiency for re1iabi1ity estimation, which can provide a theoretica1 method reference for mechanism re1iabi1ity ana1ysis.