Study On Typical Fault Diagnosis Methods Of Rotor System Based On Current And Vibration Information Fusion

With the development of science and technology,safety and reliability requirements of rotating machinery in industrial production are getting higher and higher and rotor system is the core part of rotating machinery.Once the equipment fails,it will lead to loss of work performance.These failures can cause a chain reaction,resulting in huge economic losses and catastrophic accidents.At present,using single signal diagnosis method is one-sided and fault diagnosis accuracy is relatively low.It is imperative to study fault diagnosis method based on the combination of motor current and vibration signal.In this paper,failure mechanism on three typical faults of rotor system including imbalance,misalignment and parallel misalignment is studied and fault dynamics equation of rotor system is established.It is found that these three faults can cause vibration acceleration change of x direction and y direction and torque change of shaft.They can change magnetic flux,resulting in the change of motor current.Then,fault diagnosis test bed of rotor system is designed and built and test schemes of three faults are worked out.Finally,experimental study is completed.Firstly,current and vibration signals of fault and fault-free state were collected through test bed and preprocessed.The vibration signal was denoised by wavelet soft threshold denoising method.Most of irregular burrs on signal time domain map were basically disappeared,which highlighted fault characteristics of vibration signal.The frequency of current signal was processed by notch filter method.50 Hz frequency components were basically removed,which highlighted fault characteristics of current signal.Then,time domain,frequency domain and wavelet packet energy characteristics of current signal and vibration signal were extracted.It is found that fault sensitivity of frequency domain was higher than that of time domain and fault sensitivity of vibration signal was higher than that of current signal.Considering different signal units measured by different sensors and different proportion of each feature,eigenvector was standardized.Due to high dimension of eigenvector,dimension of eigenvector was reduced by using principal component analysis.Finally,bayesian network for typical fault diagnosis of rotor system was designed,which could carry out fault diagnosis based on vibration signal,motor current signal and fusion information.The results show that the accuracy of fusion information is the highest,vibration signal is the second,and current signal is the worst.It verifies correctness and validity on diagnosis method of bayesian network fusion information.