Research On Fault Detection And Diagnosis Of Rolling Bearings Based On Deep Autoencoder Network

As a key component of rotating machinery,rolling bearings have become one of the main sources of mechanical failure due to the changing working conditions.Therefore,the research on fault detection and diagnosis of rolling bearings is of great significance for ensuring the normal operation of machinery and improving production efficiency.As an emerging algorithm in the field of artificial intelligence,deep learning has achieved excellent results in many fields with its powerful feature extraction and nonlinear function mapping capabilities,but it still needs further development in the field of mechanical equipment health management.This paper focuses on the application of Deep Autoencoder(DAE)in signal bearing extraction,fault detection and fault diagnosis of rolling bearings.The main work and contents are as follows:(1)The algorithm principle of DAE is deeply studied and a feature extraction model based on DAE is constructed.The gearbox vibration signal of different pitting states was extracted by constructing a gearbox test bench.Taking the trace of within-class and between-class scatter matrix and the accuracy of classification models as the feature evaluation factor,the model is compared with Principal Component Analysis,Kernel Principal Component Analysis,traditional time domain and frequency domain feature extraction methods.The results show that DAE has advantages in signal feature extraction.(2)Combining the advantages of DAE in signal feature extraction,DAE is used to construct the characteristic space of the vibration signal under normal bearing condition.The deviation degree of the test signal in the feature space is used as the evaluation indicator of bearing performance degradation.An online fault detection model was constructed by setting early failure thresholds.The experimental results show that compared with sample entropy,dynamic time warping and traditional time domain fault detection methods,this method can realize on-line detection of early faults,which has better sensitivity and practicality.(3)Combining DAE with softmax classification layer,a classification model based on DAE is constructed.Aiming at the selection of DAE network parameters,a fault diagnosis model based on improved Deep Autoencoder network is proposed.The model uses improved Whale Swarm Algorithm with iterative counter to search on the key parameters affecting network output in a global optimal way,and obtain the DAE diagnosis model with optimal network structure.The model is verified by the rolling bearing fault dataset,and the classification models such as Back Propagation neural network,Support Vector Machine and Deep Belief Network are compared.The results show that the proposed diagnostic model has better results.