Research On Bearing Fault Transfer Diagnosis Based On Domain Adapted CNN

Intelligent diagnosis of mechanical faults is an important means to ensure the safe and reliable operation of various engineering projects in an industrial environment,and research on advanced fault diagnosis methods is essential.In the actual operation of the engineering project,the mechanical equipment is affected by many factors,among which the change of the load will directly affect the fluctuation of the vibration signal.Therefore,in order to be able to accurately identify the operating status of different load equipment through intelligent diagnostic methods,this paper uses the CNN model after adding the domain adaptation item to perform variable load bearing fault migration diagnosis.First,the vibration signal data of the bearing is over-sampled to generate samples of different fault categories;then the Fourier transform is performed on the sample data to extract the time-frequency information of the signal data;finally,the time-frequency data of the vibration signals of each state are input into the CNN,Train the model under one load data,and then perform fault diagnosis of other load data to realize the migration fault diagnosis of bearing vibration data.The experimental results show that the fault diagnosis accuracy of the CNN deep migration model basically stabilizes at the 5th and 6th iterations,and the convergence speed is fast.Compared with the fault diagnosis method based on signal decomposition,the operability is stronger,and the fault diagnosis precisionis guaranteed.In mechanical fault diagnosis under variable load,the average fault diagnosis accuracy of the CNN model is 81.49%.However,the above-mentioned bearing fault migration diagnostic model still has room for improvement in the feature expression capability of multiple fault state data,so the above model needs to be improved.Secondly,the previous model training generally assumes that the training and test data have the same distribution.In engineering practice,the rolling bearing vibration signal data is often obtained under different loads,and there are differences in the distribution between the data,which results in low accuracy of the fault diagnosis model.This paper applies the domain-adapted deep migration fault diagnosis model to rolling bearing fault migration diagnosis under variable load,and is committed to solving the mechanical fault diagnosis problems in actual engineering.This method first stitches different two kinds of load vibration signal data,uses the fused load samples to train the CNN model,and uses the model’s predictive label as its pseudo label for the target domain data,which is used to update various parameters in the model training process.The deep migration model can simultaneously obtain the fault characteristics of the source domain and the target domain.By minimizing the maximum mean difference and conditional distribution difference of the source and target domain data,the model’s cross-load adaptive fault diagnosis capability is improved,and the diagnostic accuracy of the deep migration fault diagnosis model is improved.The experimental results show that the same CNN network structure as in Chapter 2 is used to construct a new dual-load data set.The depth generalized features extracted by the domain-adapted depth migration model are more accurate than the features extracted by standard CNN and can effectively overcome them.The interference of different loads on the identification of multi-state faults,the average migration fault diagnosis accuracy of multiple tests reached 90.2%,which has important reference significance for bearing fault diagnosis of large mechanical equipment.