Research On Fault Diagnosis Algorithm Of Rolling Bearing Based On Deep Learning

Rotating equipment is an important part of special equipment,and rolling bearing is one of the important parts of rotating equipment,so the fault diagnosis of rolling bearing is of great significance for the safe and stable operation of special equipment.Most of the traditional fault diagnosis algorithms of rolling bearing are based on signal processing technology.Data analysis of time-domain vibration signals is usually performed,which requires rich signal processing technology and manual experience.Convolutional Neural network(CNN)is one of the classical deep learning models,which has the characteristics of extracting the deep features of data.Convolutional neural network is widely used in face recognition,object detection and other fields.Taking rolling bearing as the research object,this paper explores the application of convolutional neural network in rolling bearing fault diagnosis.This paper proposes a data-driven fault diagnosis model based on deep convolution neural network.The depth of convolutional neural network is designed as 8 layers.The input of the network is the time-domain vibration signal,and the input length is 1024*1.A larger convolution kernel is used in the first layer to obtain a larger receptive field of the vibration signal.After the operation of multi-layer convolution pooling,the deep features of the data can be automatically extracted without manual experience,instead of the artificial feature extraction process,so as to realize fault diagnosis.The recognition rate of the model on the Case Western Reserve University Bearing Database(CWRU)reached over 99%.Experiments are carried out at different sampling speeds and different speeds,indicating that the network has certain generalization ability.Traditional fault diagnosis algorithms only focus on fault types,but few pay attention to the damage degree of fault.The damage degree of fault is of great significance to the prediction of service life of bearings,and the treatment of bearings with different damage degree will be different.The damage degree is a continuous value.Obviously,the classification task is not suitable for damage degree.This paper proposes a multi-task convolutional neural network based on the proposed convolutional neural network.The input signal of the network is also time-domain vibration signal.The output of the network consists of two tasks,one for fault classification,the other for fault damage degree regression,which outputs the numerical value of fault damage degree directly.Using these two indexes to train the network at the same time.The training results show that the network can correctly diagnose the fault type and accurately predict the damage degree of the fault.