Research On High Precision Fault Diagnosis And Fault Size Evaluation Of Rolling Bearing

With the progress of modern industry,machinery and equipment are developing towards high speed,intelligence,and precision.As a critical component in these devices,rolling bearings are all the more important for their smooth and safe operation,so it is essential to have an accurate and reliable rolling bearing fault diagnosis system.This is important to ensure the excellent operation of the equipment and reduce the occurrence of accidents.In this paper,a residual network model for the diagnosis of rolling bearing faults is established based on the vibration signal of the bearing,using rolling bearings as a research object.The author has studied the method of accurate assessment of rolling bearing fault size.The main work of this paper is as follows.This paper first illustrated the advantages of extracting signal features from residual networks by analyzing the residual network features.Combining the characteristics of the vibration signal of the faulty bearing,the author proposed a network model of rolling bearing fault diagnosis based on a residual network.The model takes the bearing vibration time-domain signal as data input,uses the residual network for feature extraction,and classifies the different data features to achieve a high-precision diagnosis of bearing failure.This paper goes on to analyze the internal structure parameters of the network and to determine the laws of influence of network structure parameters on diagnostic accuracy.Finally,the network model was validated using real-world data,which showed that the model could accurately and effectively diagnose rolling bearing faults.In response to the problem that the bearing fault characteristics are not extracted comprehensively enough,and the underlying network data are not adequately expressed,the author proposed an improved rolling bearing fault diagnosis residual network model.The model performs a time-frequency conversion of the input data,converting the time-domain signal to a time-frequency signal to express the characteristics of the bearing failure time domain and frequency domain.The direct network connection structure is then improved to a full network segment connection,enabling it to pass the underlying network information directly to the output layer.Finally,the author improved the activation function to improve the noise immunity of the rolling bearing fault diagnosis residual network model.On this basis,the model’s fault diagnosis performance was analyzed by various evaluation indicators,and it was found that the model’s fault diagnosis performance was significantly improved and superior to other fault diagnosis models.In response to the problem that the network data is not sufficiently intuitive,a visual analysis of the network feature layer and diagnostic process is performed using visualization techniques to demonstrate the process by which fault features change in the network.Finally,the validity of the network model was experimentally verified.In order to accurately assess the bearing fault size,this paper proposes a method to assess the rolling bearing spalling fault size accurately.This method evaluates the bearing fault size based on the bearing fault signal and combined with the bearing fault geometry,which is simple in theory and pragmatic to use.