Intelligent Fault Diagnosis And Prediction Of Performance Degradation Of Motor Bearing

With the rapid development of electric automation technology,electrical equipment in the power system increasingly has the characteristics of complication,large-scale,high-speed,and automation,which puts higher requirements on a higher level of evaluation of equipment safety and performance degradation.As the most basic components in the motor,bearings with the highest probability of failure in the components will directly affect the performance of the entire rotating equipment.If accurately and effectively determine the type of initial failure and predict the health status or damage of the motor bearing tardily,it will affect the formulation of the maintenance strategy of the entire motor system,and easily cause catastrophic accidents.In summary,accurately predicting the trend of fault diagnosis and performance degradation is beneficial to make reasonable judgments on the operation of the bearings,which is significant for controlling the damage within the scope of operation.Furthermore,it has important practical significance to make plans for the overhaul and maintenance of the electrical equipment.Recently,motor bearing industry in China has developed a variety of techniques of fault diagnosis and performance degradation assessment.In practical engineering applications,what is difficult is that accurately extract feature information and fault diagnosis of motor bearing.Especially in the process of performance degradation assessment of bearings,the performance degradation index that is the basis for performance degradation assessment and prediction is still imperfect.It has always been a hot topic for researchers to study new performance degradation assessment indicators.Due to the points mentioned above,this paper mainly focuses on analyzing the fault vibration signals of the motor rolling bearings,which including the characteristic extraction,fault identification,performance degradation and other issues of the rolling bearings.The main work is described as follows:Firstly,aiming at severe noise in the working environment of motor bearings,the concept of compressive sensing is used for reference.Orthogonal matching pursuit algorithm is introduced to preprocess the original vibration signal.Under a certain degree of sparsity,the compressed information can completely reconstruct the signal,which can effectively remove the noise interference.It is all known that the compressed signal can completely carry the characteristic information of the original signal with a certain degree of sparseness,which effectively removes the noise interference.So,it is that the decomposed noise signal of bearing after noise reduction using wavelet packet number to solve frequency band energy spectrum,which is feature vector for pattern recognition.The comparison of the eigenvalues obtained before and after noise reduction shows that the compressive sensing algorithm has a favorable effectiveness in feature extraction.Secondly,mainly address the problem of failing to directly derive the type of fault in the fault diagnosis of motor bearings with eigenvalues.On the one hand,with the help of the nuclear limit learning machine model,an optimized wavelet-core limit learning machine rolling bearing fault diagnosis method is proposed with the wavelet kernel function is introduced into the limit learning machine.On the other hand,considering the WKELM penalty factor and the parameters of the kernel function have a great influence on the classification accuracy and performance,the whale algorithm based on von Neumann topology is employed to optimize the WKELM's penalty factor and kernel function's parameters and construct a classifier model.It can be seen from the comparison of experiments that this method can effectively extract fault feature information of rolling bearings and has high diagnostic accuracy.Finally,for the problem of insufficient capacity that the traditional indicators reflect the evolution of the performance degradation of motor bearings,this paper regards the construction and statistics of nuclear T~2 and SPE as the evaluation index of bearing performance degradation according to the signal data of motor-bearing life cycle,which is applied to the field of equipment performance evaluation,then the performance degradation indicators in entire service life cycle of the bearing is analyzed in detail.On the basis of statistics,as for accurately and effectively predicting the bearing performance degradation index and fluctuation range,a support vector based on fusion kernel principal component analysis and fuzzy information granularity is proposed.The results show that this above method can predict the trend of bearing performance degradation and the fluctuation range of change earlier and more accurately.