The Research On Fault Diagnosis Methods Of Motor Bearing Based On Data Driven

Nowadays,industrial processes are more complicated and expanding in scale.Motor bearing is more and more used in industrial production.So,it becomes very important to detect the motor bearing fault effectively and accurately.In order to improve the accuracy of motor bearing fault diagnosis,an industrial process fault diagnosis algorithm based on data called Ensemble Empirical Mode Decomposition(EEMD)-Improved Local Mean Decomposition(ILMD)-Incremental Probabilistic Neural Network(IPNN)-Improved Gravitational Search Algorithm(IGSA)is proposed in this thesis.All of the data used in this thesis are come from Case Western Reserve University bearing data center website.In industrial production,the working environment of motor bearing is noisy,coupled with the vibration interference by other equipment to make the vibration signal contains noise.Therefore,to reduce noise it need to preprocess the data.But traditional methods are not good at handling noise in non-stationary and nonlinear data.This thesis applies EEMD algorithm which has strong ability by calculating the correlation coefficient and setting threshold to decompose non-stationary signal.Because of the end effect of LMD,it proposed an improved LMD method to extract fault features and calculate the sample entropy and energy of PF component as the feature parameters to constitute a fault feature vector as the input of fault classification.IPNN is a type of feed forward neural network based on statistical,and it has strong ability of classification and simple training process that is not need to set the initial weights.In the process of fault diagnosis,the network model parameters have a significant impact on the diagnosis of performance.In order to improve the defect which is slow convergence speed and falling into the most superior easily and improve the accuracy of the classification results eventually,this thesis adopt an optimization algorithm based on time-varying weight and boundary mutation to optimize the threshold of neural network model called IGSA,and improve the accuracy of the classification results.The theoretical research and experimental results show that the proposed method based on EEMD-ILMD-IPNN-IGSA is effective in the diagnosis process of motor bearing fault diagnosis,and the accuracy of it is higher.