Design And Implementation Of Rolling Bearing Fault Diagnosis System Based On Deep Learning

When rolling bearings are used for industrial production in mechanical operations,rolling bearings are often faulty components.When rolling bearings fail,due to the unstable vibration signal of the bearing,the signal presents nonlinear characteristics,which makes some rolling bearings at present.The fault diagnosis system or diagnosis method has a weak ability to extract the original signal,and the poor ability to identify the rolling bearing signal leads to a big gap between the final diagnosis result and the expected result.In order to enhance the intelligence and accuracy of the fault diagnosis system,improve the generalization ability of the system so that it can meet the needs of factory big data fault diagnosis.This paper makes the following two improvements for this situation:1.In view of the fact that the traditional neural network rolling bearing fault diagnosis model does not separately identify the frequency domain data information in the fault data,this paper compares the differences between the wavelet transform and the short-time Fourier transform in various aspects in the experiment,and finally chooses the shortest one.Time Fourier transform,a diagnostic model of short-time Fourier transform is proposed in the traditional diagnostic model.The original data is sampled using the sliding window acquisition technology,and then the obtained sample is transformed by the short-time Fourier transform,so that a two-dimensional time-frequency data sample is obtained,and the obtained sample is trained in the neural network model.Experiments are carried out on the bearing dataset of Case Western Reserve University,a classic rolling bearing dataset.The experimental results show that this processing method can still obtain better results than the traditional model in the case of missing or unbalanced data.2.There is a lot of noise in the collection environment when the rolling bearing data is collected,so the collected data samples have great interference to the training of the model.This paper proposes to optimize the network structure and add feature extraction to the fault diagnosis model.Module and long-term memory network,improve the defects of traditional rolling bearing on global feature extraction,enhance the accuracy of diagnosis results,and improve the generalization ability of the model.The experimental results demonstrate that the proposed model has better effect than the traditional model.