RBF网络用于性能退化的可靠性评估  

RBF network based reliability assessment of performance degradation

作　　者：胡锦涛[1] 胡昌华[1] 冯文晖[1] 郑建飞[1] 

机构地区：[1]第二炮兵工程学院,西安710025

出　　处：《电光与控制》2008年第11期67-70,共4页Electronics Optics & Control

基　　金：国家自然科学基金重点课题(60736026);国家教育部新世纪优秀人才支持计划项目

摘　　要：针对无法用数学式准确描述退化量与环境因子间复杂非线性关系的问题,提出了一种基于RBF网络的退化轨迹建模方法。利用RBF网络能够逼近任意非线性函数的特性,进行退化轨迹的建模,并结合失效阈值预测产品寿命。将所有估计寿命作为完全失效数据进行可靠性评估,有效地解决了失效数据少的问题。通过对疲劳裂纹增长数据的RBF网络建模分析,证明RBF网络具有很好的函数逼近能力,寿命预测结果非常理想。经验证,在更少的退化数据下,RBF网络仍能够较好地模拟退化轨迹,寿命预测精度较高。Since it is very difficult to describe the complex non-linear relationship between performance degradation measures and environmental factors by mathematic expressions, we put forward a method for modeling degradation path based on RBF network. The RBF network can approach any non-linear functions, the feature is used for modeling the degradation paths and forecast the lifetime of products based on the critical level of degradation. All of the estimated lifetimes are taken as data of complete failure to implement reliability assessment, thus the problem of lack of time-to-failure data is solved. The fatigue crack growth data are analyzed by using RBF network. The function approaching capability of RBF network is proved, and the result of lifetime forecast is satisfactory. It is validated that the RBF network can model the degradation paths preferably, even under much less degradation data, and the precision of lifetime forecast is high.

关 键 词：RBF网络 性能退化 寿命预测 可靠性评估 

分 类 号：V247.1[航空宇航科学与技术—飞行器设计] TB114.3[理学—概率论与数理统计]