证据理论在电机故障诊断中的应用

Application of evidence theory in fault diagnosis for electric machine

D-S证据理论作为一种非精确推理算法具有独特的优势,非常适用于存在大量不确定性因素的电机故障诊断工作。提取故障电机的状态特征量,并将其按时域、频域、奇异值分解为多个子参数空间。在此基础上,采用并行BP神经网络及模糊聚类系统对电机故障进行局部诊断。将每个局部诊断结果作为独立的证据体,构造相应的信度分配函数。结合电机故障的信息融合诊断模型,将基于D-S证据理论的决策融合的方法应用于电机故障诊断。通过对案例进行分析,实现了利用多证据体的融合信息对电机故障状态进行诊断,其诊断结果验证了D-S证据理论在提高电机故障诊断的准确性和灵敏性方面的作用。

D-S evidence theory has its own unique advantages as an inaccurate reasoning means, and it＇s in point of fault diagnosis for electric machine having large inexact factors. Firstly, the state features of fault electric machine are extracted and decomposed into a number of sub-spaces according to time domain, frequency domain and singular value. On the basis, partial diagnosis is realized based on BP neural networks and fuzzy cluster systems. The independent evidences can be obtained using the results of partial diagnosis, and the belief assignment function of corresponding evidence is constructed. To diagnose the States of electric machine by fusing some evidences＇ information, this paper combines information fusion model of fault electric machine and makes decision based on D-S evidence theory. The testing results show that the application of D-S evidences can improve the accuracy and delicacy of fault diagnosis.