故障模型的可辨识性及其应用

基于系统模型辨识的故障诊断技术有助于定量评估故障的严重程度,对故障诊断技术的发展具有重要的意义,而目前多故障系统的故障辨识理论有待更深层次的研究。从故障模型的可辨识性角度来探讨和丰富故障辨识理论,因为某些并发故障将会带来模型的可辨识性问题。提出了基于Fisher信息矩阵的模型可辨识性考察方法,以及如何改善模型可辨识性的方法,并以转子系统的质量偏心和轴初始弯曲辨识应用为例来说明。

IDENTIFIABILITY OF FAULT MODEL AND ITS APPLICATION

The technique of fault diagnosis based on model identification is beneficial for the quantitative determination of severity of faults. However, the inadequate theory of multiple fault identification needs further investigation. From the viewpoint of model identifiabilty, the relevant theoretical methods are discussed since the simultaneous faults will bring about unidentifiable problem of models. Based on the Fisher information matrix, an investigation method of model identifiability is proposed, which improves the model identifiability. The identification of the mass unbalance and initial bend of an experimental rotor is studied to illustrate the proposed method.