Research And Application Of Bearing Fault Diagnosis Method Based On Transformer

Fault diagnosis of rotating machinery is an important means to accurately identify the health state of equipment and ensure its safe operation.With the extensive application of artificial intelligence,Internet of Things and other technologies in industrial production,the application of data-driven machine learning theory in rotating machinery fault diagnosis has become a hot research topic.In recent years,Transformer neural network represented by attention mechanism has shown strong ability to extract and identify key features.The rise of Transformer provides a new idea for rotating machinery fault diagnosis technology.This thesis takes rolling bearing as the research object,takes Transformer as the theoretical basis,and integrates self-supervised learning theory and transfer learning theory to study the fault diagnosis method of rotating machinery.The main research work of this thesis is as follows:(1)Existing research has only used attention mechanism(AM)in combination with deep networks,or to replace certain components of these deep networks.This reliance on deep networks severely limits the feature extraction capability of AM.To solve the above problems,in this thesis,a bearing fault diagnosis method is proposed based on a signal Transformer(Si T)with pure AM.First,the raw one-dimensional vibration signal is segmented and a new segmented learning strategy is introduced.Second,linear encoding and position encoding are performed on the segmented subsequences.Finally,the encoded subsequence is fed to the Transformer for feature extraction to achieve fault identification.The validity of the proposed method is verified using the Case Western Reserve University dataset and the self-priming centrifugal pump bearing dataset.Compared with existing methods,the proposed method still achieves the highest average diagnostic accuracy without any data preprocessing.The results demonstrate that the proposed Si T based on pure AM can extract features and identify faults from the raw vibration signal,and has superior diagnostic performance.(2)In view of the fact that the construction of a Transformer diagnosis model requires a large amount of labeled data,and the actual industrial process can obtain less labeled data,which leads to Transformer’s limitations such as over-fitting and low diagnostic accuracy on small data sets.A self-supervised learning-based Transformer(SSFormer)is proposed for rotating machinery bearing fault diagnosis under few samples.In mask self-supervised learning,unlabeled samples can be used to mine robust representations of fault signals and potential relationships between sub-sequences to obtain a pre-trained model with well-generalized parameters.Then,a few labeled samples are fine-tuned in a supervised learning manner to enable SSFormer the discrimination ability to identify different bearing fault types.The effectiveness of the proposed method is fully validated on the multistage centrifugal fan dataset and the Case Western Reserve University motor bearing dataset.The experimental results demonstrate that SSFormer is effective in reducing the number of labeled training samples,and compared to existing methods,SSFormer has superior diagnosis performance under the condition of limited labeled samples.(3)On the basis of the above research,a bearing fault diagnosis application system based on Transformer is designed.The system is based on the EAIDK610 embedded device as the hardware foundation,and combines the two Transformer bearing fault diagnosis algorithms studied above to design the fault diagnosis software.The system has many functions such as data loading,fault diagnosis,data clouding,diagnosis result visualization,and supports human-computer interaction in the touch screen.The system test results show that the system has fast response and high fault diagnosis accuracy,which verifies the feasibility of applying Transformer fault diagnosis models in embedded devices and has certain practical application value.