基于证据理论回归的短波接收机故障预测  

Fault prognostication of shortwave receiver based on evidence regression

作　　者：赵建光[1] 李红波[1] 曾繁景[1] 李铁峰[1] 

机构地区：[1]信息工程大学信息工程学院,郑州450002

出　　处：《电子测量与仪器学报》2012年第7期640-645,共6页Journal of Electronic Measurement and Instrumentation

摘　　要：针对数据驱动故障预测方法中的不确定性表示和处理问题,提出了基于证据理论回归的故障预测方法。将训练样本集中与预测特征相似的样本作为证据源,并根据相似度对各证据源的信任度进行初始化。使用证据理论对证据源进行合成,并通过对不确定信任度的分配,得出剩余可用时间的预测值及其上下限。算法被应用于短波接收机的故障预测中,实验结果表明本算法的精度较高,并具有较强的不确定性表示和处理能力。Uncertainty representation and management is always the key hurdle faced by data-driven prognostication To solve this problem, a remaining useful life （RUL） estimation method based on evidence regression algorithm is pro- posed. The evidence regression method regards the k nearest neighbors as k pieces of evidence, whose beliefs are assigned to be proportional to their similarity to the features under prognostics. Then all the beliefs are pooled using the Demp- ster-Shafer theory. Finally, the estimation of RUL and the corresponding bounds are obtained by assignment of the uncer- tain belief. This method is applied to fault prognostication of shortwave receiver, and the results show that this method is more accurate and is less sensitive to the uncertainty in fault prognostication of shortwave receiver.

关 键 词：故障预测 短波接收机 证据理论 剩余可用时间 

分 类 号：TP806.3[自动化与计算机技术—检测技术与自动化装置]