基于幂均算子和证据理论的故障诊断方法

将证据理论在处理不确定性信息方面的优势用于故障诊断,可提高诊断的精确度和准确性.从证据理论的角度来看,从每个发动机传感器获得的信息可以看作一条证据,基于多传感器信息的发动机故障诊断即是一个证据融合问题.本文使用证据理论作为描述发动机状态的多传感器信息建模方法.首先,在处理特征值样本数据时,引入幂均算子的方法以提高整个故障诊断系统的准确性;通过量化待测特征值和故障原型之间的距离生成基本概率分配函数;然后引入证据信息量的方法对融合后的结果进行性能评估;最后通过发动机故障案例对该方法进行验证,并与其它方法进行对比,结果充分证明了该方法的真确性与可靠性.

Fault Diagnosis Method Based on Power-average Operator and Evidence Theory

The evidence theory can be used to deal with uncertainty information; thus it is advantageous for fault diagnosis, as it can improve the diagnosis accuracy. From the perspective of evidence theory, the information obtained from each sensor can be regarded as a piece of evidence. Engine fault diagnosis based on multi-sensor information is an evidence fusion problem. In this paper, evidence theory is used as a modeling method for multi-sensor information describing engine state. First, when processing the eigenvalue sample data, the method of power-average operator is introduced to improve the accuracy of the whole fault diagnosis system. The basic probability assignment is generated by measuring the distance between the eigenvalue and the fault prototype. Then the fusion results are evaluated by quantifying the evidence information. Finally, the method is verified using a case of engine failure and compared with other methods, and the comparison results prove the authenticity and reliability of the method.