基于逆高斯过程的竞争失效模型  

Competing Failure Model Based on Inverse Gaussian Process

作　　者：丁力 蔡风景 DING Li;CAI Fengjing(College of Mathematics and Physics,Wenzhou University,Wenzhou,China 325035)

机构地区：[1]温州大学数理学院,浙江温州325035

出　　处：《温州大学学报（自然科学版）》2021年第3期38-48,共11页Journal of Wenzhou University(Natural Science Edition)

基　　金：国家社会科学基金项目(15BTJ030)。

摘　　要：将软失效建模为逆高斯过程,将硬失效建模为威布尔分布,得到了包含退化失效和突发失效的竞争失效模型.针对复杂系统高性能、成本高、只有小样本的情况,采用退化数据以及截尾试验数据建立模型并对其进行极大似然估计,通过渐进正态性构造模型参数的区间估计.数值模拟结果说明本文所建立模型的可行性,将该结果与只采用精确失效数据建立模型的结果进行对比,说明本文模型的优越性.经真实数据分析,验证本文基于逆高斯过程的竞争失效模型拟合状况良好.In this paper,by modeling the soft failure as the inverse Gaussian process,and the hard failure as a Weibull distribution,we obtain the competing failure model involving both catastrophic and degradation failures.In view of the high performance,high cost and only a small sample of the complex system,degradation and truncated test data are adopted to establish the model and conduct the maximum likelihood estimation,and asymptotic normality is used to calculate the interval estimation of model parameters.The results of the numerical simulation prove the feasibility of the established model,and its comparison with the model based on accurate failure data proves the superiority of the established model.Finally,the real data analysis verifies that the competing failure model based on inverse Gaussian process fits well.

关 键 词：竞争失效模型 截尾试验数据 极大似然估计 区间估计 

分 类 号：O213.2[理学—概率论与数理统计]