Design And Implementation Of Fault Diagnosis System Based On Deep Learning

With the progress of science and technology and the development of industrial big data,increase industrial data are collected,and the diagnostic analysis of industrial data has become a trend.However,the traditional way of processing industrial big data requires a lot of manpower and material resources,so the rapidly developing deep learning theory is applied to the diagnosis of industrial data.In actual data collection,most of them are normal data with only a small part of abnormal data samples.the traditional fault diagnosis method is not effective and can not adapt to the industrial development.Therefore,a fault diagnosis framework based on generative adversarial model and convolutional neural network is proposed to solve the above two problems.The framework can balance the unbalanced phenomena in industrial data sets and identify the mechanical failure problems in them.(1)this thesis studies the experimental data for time series data,in order to fully tap the characteristic information of the time series of deeper features in a single time series data,with the improved convolution neural network model and the cycle of combining neural network model to forecast the time sequence,compared with the traditional convolution neural network method,the variable load has a great performance improvement data sets.(2)For industrial imbalance data set,this thesis sampled deep convolution generation adversation network to solve this problem.The deep convolution generation antagonistic network can generate data according to data features,mix the newly generated data with the original data,and achieve better results compared with the direct use of unbalanced data set training.The research framework of data generation and fault diagnosis proposed in this thesis was verified in the CWRU bearing vibration data,and the fault diagnosis method combining the data set enhanced by generation admissive network with the convolutional neural network can improve the accuracy of fault classification.(3)Design and implement a data monitoring system platform,which can display the status of the machine tool online,check the operating efficiency of the machine tool,and query the historical status data.Also can carry on the online warning to the machine tool.