Abstract: For the complexity of rotor vibration process and the randomization of vibration fault generated, and difficulty for getting vibration fault samples, a fault diagnosis method based on power spectrum entropy and Support Vector Machine(SVM) with the SVM advantages of small sample, generalization and overall under information entropy was put forward. Four typical failures of rotor vibration was simulated based on rotor experiment and vibration fault data under more points and multi-speed is collected. The power spectrum entropy has been calculated through analyzing and processing the datum as fault vector, and the SVM model has been gained. And the validity of this method for distinguishing fault types, fault severity and fault location is proved to be effective by calculation and analysis of rotor vibration fault signals. 

Key words: Information entropy; Power spectrum entropy; Support vector machine (SVM); Information fusion; Rotor vibration; Fault diagnosis

摘要： 针对旋转机械振动过程的复杂性和振动故障产生的随机性以及振动故障样本获取难的问题,在信息熵理论的基础上,融合了支持向量机(SVM)小样本、全局性和泛化性好的优点,提出了过程功率谱信息熵(功率谱熵)SVM的故障诊断方法。结合转子实验台,得到了4种典型振动故障在多测点多转速下的数据,通过计算提取了其功率谱熵特征值作为故障样本,即故障向量,并建立SVM诊断模型,对转子振动故障的类别、严重程度和部位识别诊断,验证了该方法在转子振动故障诊断方面效果良好。 

关键词: 信息熵;功率谱熵;支持向量机;信息融合;转子振动;故障诊断