Study On Rotor Vibration Faults Diagnosis Method Based On Information Fusion

In this paper, a classification diagnosis method based on information fusion is proposed.The information fusion method has more reliable understanding and situational assessmentbecause it combined the information in different ways. Studies on single faults in rotor systemat home and abroad is more than coupling faults. And the diagnosis methods for couplingfaults are mainly qualitative analysis rather than quantitative calculation. Therefore, inaddition to research and diagnosis four single faults: rotor rubbing, rotor crack, rotormisalignment, rotor imbalance, this article will also focus on three coupling faults:crack-rubbing coupling faults, imbalance-rubbing coupling faults and crack-imbalancecoupling faults.Firstly, the most common faults and fault diagnosis methods were introduced. Then theinformation theory was used to prove the feasible and availability of information fusion.Three diagnosis methods are proposed in this paper:1. Diagnosis method based on information entropy difference matrix. Calculate thecloseness of information entropy distribution range and regularity between unknown vibrationsignal and the typical vibration faults signal. The diagnosis results showed that this methodcan diagnose single faults and coupling faults and has a high accuracy.2. The fusion information entropy distance diagnosis method is proposed for couplingfaults. Calculating the information entropy distance through fusion power spectrum entropy,singular spectrum entropy, wavelet energy spectrum entropy and wavelet space characteristicspectrum entropy can express the characteristics of vibration faults more comprehensively.Especially on the situation that single faults and coupling faults is more difficult to recognize.The diagnosis process is clear and intuitive. The diagnosis results showed that the entropydistance graph in multiple speeds can successfully separate single faults and coupling faults.3. Using the diagnosis method based on LS-SVM for both single faults and coupling faults. LS-SVM has a faster solution speed and a higher accuracy, so it is more suitable forlarge rotating machine’s faults diagnosis. Collect the vibration signals in multiple speed andmultiple points, fusion their four kinds information entropy and calculate their average valuesas feature vectors. This fusion feature vector can express the characteristics of vibration faultsmore comprehensive, so the LS-SVM classification machine has a higher accuracy. Thediagnosis results showed that the LS-SVM model has a high accuracy and is suitable fordiagnosis of rotor vibration faults.