小样本条件下多应力加速寿命试验预测方法

鉴于导弹中的电子设备价格昂贵、可用于试验的样本量少,在开展加速试验以及寿命预测的实际工作中通常为小样本的背景。文章研究探索小样本条件下多应力加速试验寿命预测方法,分别建立通用对数线性模型、 BAS-BP神经网络模型、灰色–支持向量回归模型,结合多应力加速试验数据在各应力条件下的样本容量分别为 56组、20组、10组、5组的情况下,比较 3种模型的预测效果,分析各模型的适用场合和时机,探索小样本条件下模型的选优问题,为小样本条件下多应力加速试验寿命预测提供有益的借鉴。

Prediction Method of Multi-stress Accelerated Life Test Under Small Sample Condition

In view of the high cost of electronic equipment in missiles and the small number of samples that can be used inexperiments, the actual work of conducting accelerated tests and life prediction is usually in the context of small samples.This article studies and explores the prediction method of multi-stress accelerated life test under small sample conditions,establishes a general logarithmic linear model, BAS-BP neural network model, and gray-support vector regression model.Combine the multi-stress accelerated test data in various stress conditions that the sample capacities are 56 groups, 20groups, 10 groups, and 5 groups respectively, the prediction effects of the three models are compared, the applicable occa?sions and timing of each model are analyzed, and the model selection under small sample conditions is explored. It pro?vides useful reference for prediction of multi-stress accelerated life test under small sample conditions.