基于信息熵贴近度和证据理论的旋转机械故障诊断方法

从信息融合的思路出发,建立反映振动能量的旋转机械故障状态的信息熵特征,如奇异谱熵、功率谱熵、小波空间状态特征谱熵和小波能谱熵。通过试验,建立了旋转机械典型故障下的信息熵期望值,即获得基于信息熵的故障诊断标准特征向量。由于传感器的不确定性和故障的多样性,提出采用D-S证据理论来对4种信息熵进行信息融合。根据越相似模式间的距离越短的思路,提出采用信息熵贴近度来建立证据理论的基本可信度分配,以基于基本可信数的决策方法来作为故障模式识别方法。通过实例计算,证明基于信息熵贴近度和证据理论的旋转机械故障诊断方法是故障模式定量识别的一种可行的新方法。

A Rotational Machinery Fault Diagnosis Method Based on Close Degree of Information Entropy and Evidence Theory

Starting from information fusion, the paper sets up information entropy features which show the vibration energy of the fault of rotational machinery such as singular spectrum entropy, power spectrum entropy, wavelet space state feature spectrum entropy and wavelet power spectrum entropy. Through experiments, the expectation values of information entropy with typical faults of rotational machinery, namely standard feature vectors for fauh diagnosis based on the information entropy, were established. Because of the uncertainty of sensors and the diversity of faults, the four information entropies are fused by the DS evidence theory. According to the thought that the shorter the distance is, the more similar the models are, the basic probability assignment is set up by the close degree of information entropy for the evidence theory, and a decision-making method based on the basic probability number is used as the fault model recognition method. An instance proves that the rotational machinery fauh diagnosis method based on close degree of information entropy and evidence theory is valid and feasible for recognizing fault models in quantification.