Research On Mechanical Equipment Fault Diagnosis Method Based On Deep Learning And Transfer Learning

As one of the important problems in the manufacturing industry,fault diagnosis is currently developing in the direction of intelligent diagnosis.Due to the improvement of data availability,and recent artificial intelligence algorithms have played an important role in various fields,fault diagnosis research based on artificial intelligence algorithms has become more and more active.However,in engineering practice,it is difficult to obtain high-quality labeled data;at the same time,due to the continuous change of machine working conditions,the data distribution in different periods is often different.All of these pose great challenges to fault diagnosis.Therefore,according to the characteristics of the fault diagnosis data of mechanical equipment,this paper combines deep learning and transfer learning to improve and apply it to the fault diagnosis of mechanical equipment to obtain a high-precision diagnosis algorithm.First,a fault diagnosis method for mechanical equipment based on deep convolutional neural network is proposed.In the fault diagnosis method,the feature extraction of fault data has a great influence on the effect of fault diagnosis,and the process of feature extraction often requires rich experience.Therefore,the signal-to-image method is used,and the convolutional neural network is applied to fault diagnosis,so that features can be automatically extracted,and the accuracy of fault diagnosis is also improved.Secondly,a semi-supervised fault diagnosis method for mechanical equipment based on deep transfer learning is proposed.Since the working conditions of mechanical equipment often change,under the new working conditions,the amount of label data that can provide the training model is small.For this situation,a semi-supervised fault diagnosis method for mechanical equipment based on deep transfer learning is proposed.In this method,the fault signal is converted into an image.Then train a deep network with CORAL to extract image features.Finally,the transfer component analysis(TCA)method maps the generated features to the kernel Hilbert space,and the k-Nearest Neighbor(KNN)method classifies the features in this space.In this paper,different transfer learning methods are combined and used in the field of mechanical equipment fault diagnosis to solve the problem of fault diagnosis with only a few labels.Then,an unsupervised fault diagnosis method for mechanical equipment based on deep transfer learning is proposed.Due to changes in the working conditions of mechanical equipment,under the new working conditions,the amount of label data that can provide the training model is few.Therefore,a transfer learning fault diagnosis method based on adversarial process is proposed.In this method,the fault signal is converted into an image,and then a feature generator and two classifiers are trained.The generator and the classifier are trained against each other to obtain a domain adaptation model.An unsupervised fault diagnosis model is established to solve the problem of no labels in fault diagnosis.Finally,the full text is summarized.On the problem of mechanical equipment fault diagnosis,deep learning and transfer learning are combined according to the characteristics of fault data,and supervised,semi-supervised,and unsupervised mechanical equipment fault diagnosis methods are proposed in turn.And the direction of further research on the problem of fault diagnosis is pointed out.