Reliability analysis of high-lift device wear based on indicator error control Kriging

基于指示误差控制Kriging的增升装置磨损可靠性分析

¹ School of Aeronautics, Northwestern Polytechnical University, Xi'an 710072, China
² National Key Laboratory of Strength and Structural Integrity, Aircraft Strength Research Institute of China, Xi'an 710065, China
³ Standardization Department, COMAC Shanghai Aviation Industrial (Group) Co., Ltd, Shanghai 200232, China

Received: 2 August 2025

Abstract

To address the issues that the Kriging surrogate model using the learning function as the convergence criterion cannot reflect the accuracy of predicted failure probability directly and accurately and incur high computational cost in modeling, a reliability analysis method based on Indicator Error Controlled Kriging (IEC Kriging) is proposed in this paper. With Kriging as the basic model, IEC Kriging establishes a sequential transmission relationship among sample information, performance functions, and failure probability. It analyzes the error of predicted failure probability based on the probability of indicator sign misclassification by the Kriging model at each sample point, while considering the corrective effect of candidate sample points on prediction accuracy. A dynamic sample selection strategy is designed with the upper bound of the overall sample indicator error as the criterion, and the sample point with the highest corrective effect is incorporated into the training set to optimize the surrogate model, ensuring the accuracy of prediction results while improving the efficiency of building the surrogate model. The proposed method is applied to the wear reliability analysis of spherical plain bearings in an aircraft high-lift device as a case study. Compared with Kriging methods using the EGO learning function and U learning function, the IEC Kriging demonstrates higher computational efficiency for reliability estimation, which can provide a theoretical method reference for mechanism reliability analysis.

摘要

针对学习函数作为收敛准则的Kriging代理模型存在无法直接准确反映预测失效概率精度与建模计算成本高的问题, 提出一种基于指示误差控制的Kriging(indicator error controlled Kriging, IEC Kriging)可靠性分析方法。以Kriging为基础模型, 通过IEC Kriging构建样本信息、功能函数、失效概率的序列传导关系, 借助各样本点的指示符号判断正误, 分析预测失效概率误差, 并考虑候选样本点对预测精度的修正作用, 设计以总体样本指示误差上界为判据的样本动态筛选策略, 将修正作用最大的样本点纳入训练样本集对代理模型进行优化, 在保证预测结果精度的同时提高建模效率。以某型飞机增升装置关节轴承为例, 运用所提方法开展了机构磨损可靠性分析, 与EGO学习函数和U学习函数Kriging方法相比, 所提方法在可靠性计算效率方面具有显著优势, 可为机构可靠性分析提供理论方法参考。