系统工程与电子技术

• 可靠性 • 上一篇    下一篇

基于空间相似性和波动阈值的退化模型一致性检验方法

许丹, 陈志军, 王前程, 陈云霞, 康锐   

  1. 北京航空航天大学可靠性与系统工程学院可靠性与环境工程技术重点实验室, 北京 100191
  • 出版日期:2015-01-28 发布日期:2010-01-03

Method for consistency check of degradation model based on spatial similarity and fluctuation threshold

XU Dan, CHEN Zhi-jun, WANG Qian-cheng, CHEN Yun-xia, KANG Rui   

  1. Science & Technology on Reliability & Environmental Engineering Laboratory, School of
    Reliability and Systems Engineering, Beihang University, Beijing 100191, China
  • Online:2015-01-28 Published:2010-01-03

摘要:

针对退化模型中存在模型一致性检验不充分问题,借鉴数理统计知识和灰色系统理论,提出一种基于空间相似性和波动阈值的退化模型一致性检验方法。首先,考虑基于正态总体的波动阈值一致性检验,确定退化模型和验证试验结论具有动态一致性,即退化模型结论在容许的波动阈值内;然后,考虑基于灰色关联的空间形状相似性一致性检验,确定退化模型曲线之间的相似度;最后,根据退化模型同时满足基于正态总体的波动阈值一致性检验和基于灰色关联的空间形状相似性一致性检验,认为退化模型通过一致性检验;反之,认为不通过一致性检验。通过实例建模和试验数据对比,验证了所提方法的有效性。

Abstract:

To solve the problem of insufficiency of consistency check in the degradation model, we propose a method of consistency check of the degradation model based on spatial similarity and fluctuation threshold, using the knowledge of mathematical statistics and grey system theory for reference. Firstly, we take into consideration the consistency check of the fluctuation threshold based on the normal population to identify the dynamic consistency between the degradation model and validation test results, namely, the results of the degradation model are within the allowable fluctuation threshold. Secondly, we consider the consistency check of space shape similarity based on the grey correlation to determine the similarity among degradation model curves. At last, if the degradation model satisfies the consistency check of the fluctuation threshold based on the normal population and the consistency check of space shape similarity based on the grey correlation, we can conclude that the degradation model meets the consistency check. Otherwise, we conclude that the degradation model does not meet the consistency check. By comparison of modeling and experiment data, the validity of the proposed method is verified.