Research On Rolling Bearing Fault Diagnosis Method Based On Convolutional Neural Network

Rolling bearings are widely used in high-speed rail,aerospace,wind power generation and other fields.Once a rolling bearing fails,it will directly cause mechanical equipment to fail to work normally,causing huge economic losses.Therefore,the condition monitoring and fault diagnosis of rolling bearings has always been a hot topic at home and abroad.This paper focuses on the fault diagnosis of rolling bearings,and carries out the following research work: modeling and analysis of rolling bearing vibration mechanism,using vibration signal analysis methods to extract rolling bearing fault features,building a convolutional neural network model for rolling bearing fault diagnosis,and rolling bearing fault recognition based on transfer learning.The basic theory of rolling bearing failure and the feature extraction method of rolling bearing failure are studied.Based on the mechanical geometry of the rolling bearing,the load distribution and the number of rolling elements,the vibration mechanism model of the rolling bearing was established;the vibration signal analysis method was used to analyze the vibration signal of the faulty bearing from three aspects: time domain,frequency domain and time-frequency domain.Perform vibration signal analysis.The convolutional neural network combined with time-frequency analysis is used to diagnose the rolling bearing faults.The short-time Fourier transform analysis method is used to process the rolling bearing vibration signals to obtain the rolling bearing fault time-frequency data set.This rolling bearing data set is used to train the convolutional neural network model,and the convolutional neural network model Parameters;The validity of the convolutional neural network model is verified by the measured faulted rolling bearing dataset.The transfer learning algorithm is used to optimize the convolutional neural network model.Construct various types of rolling bearing data sets into source domain data sets and target domain data sets;use the source domain data sets to pre-train convolutional neural network models to obtain convolutional neural network models with shared bearing fault characteristic information;use target domain rolling bearing data The set is used to fine-tune the convolutional neural network model to realize the rolling bearing fault diagnosis,and the effectiveness of the method is verified by comparison with other algorithms.