Research On Hybrid Depth Network Based Intelligent Fault Diagnosis And Prediction Method For Rolling Bearings

Rolling bearings are the core parts of mechanical equipment,which play a key role in supporting and reducing friction.They are widely used in many fields such as daily life,industrial production and national defense construction.Bearing performance directly affects the working condition of machinery and equipment.Once bearing failure occurs,it may lead to mechanical equipment paralysis,resulting in heavy losses and even casualties.Therefore,it is of great significance and practical value to carry out intelligent fault diagnosis and prediction for rolling bearings.Based on vibration signal data of rolling bearings and combined with deep learning theory,thesis studies rolling bearings from two aspects: intelligent fault diagnosis and remaining useful life prediction method.The main contents are as follows:(1)The mechanical structure and failure form of rolling bearing are studied,and the generation of vibration signal and vibration characteristics of each structure are analyzed on this basis.Common vibration signal analysis methods and their principles are studied,and common deep learning networks are introduced to provide theoretical basis for fault diagnosis and residual life prediction in the following paper.(2)An intelligent fault diagnosis model of rolling bearing based on multi-task learning was proposed to solve the problems of low fault recognition rate and insufficient use of data correlation in the current rolling bearing fault diagnosis process.In the model,the task of bearing fault diagnosis is divided into two sub-tasks: bearing fault location classification and bearing fault damage degree identification,and hard parameter sharing mechanism is used to make them learn from each other.The bearing data set is used to test the model,and the anti-noise experiment and model comparison are carried out.The experimental results show that the proposed model can correctly identify the bearing fault type and fault damage degree,and has better performance than the corresponding singletask method.(3)Most of the current studies on the prediction of bearing remaining life need to go through complex feature extraction process,and the prediction accuracy is not high.In thesis,a predictive model of residual life of rolling bearings based on Temporal Convolutional Network(TCN)and Attention Gate Recurrent Unit(AGRU)is proposed.The validity of the model is verified by the bearing data set,and the model is compared with the common neural network model.The results show that the model has good prediction effect in the data set.