Mechanical fault intelligent diagnosis method based on migration prototype network under small sample

Abstract

The invention discloses a mechanical fault intelligent diagnosis method based on a migration prototype network under a small sample. The deep convolutional neural network is used to carry out featureextraction and operation state identification on the mechanical signal, sensitive features in the mechanical signal can be effectively extracted, and dependence of a traditional feature extraction process on artificial experience is eliminated. According to the method, the principle of prototype clustering is used, effective features of signals are obtained under the condition that the number of available samples is extremely small, and dependence of a traditional machine learning method on huge data volume is eliminated. according to the method, the transfer learning principle is used, and the generalization ability of the network is further improved by means of related source domain data with different feature distribution. By combining the deep convolutional neural network, the prototype network and the transfer learning thought, fault diagnosis can be effectively carried out on the mechanical equipment under the small sample data, and the fault diagnosis accuracy of the mechanicalequipment under the small sample data is improved.

Description

Technical field [0001] The invention belongs to the field of mechanical equipment fault diagnosis, and specifically relates to a mechanical fault intelligent diagnosis method based on a migration prototype network under a small sample. Background technique [0002] At present, traditional mechanical equipment fault intelligent diagnosis technology still relies on learning sample features from a large number of samples. However, in actual working conditions, due to the safety of the mechanical equipment and the complexity of the working environment of the mechanical equipment, it is difficult to obtain the fault signals of the mechanical equipment, and the number and types of fault signals obtained are small. When the number of available samples is relatively small, it is difficult for the intelligent diagnosis algorithm to fully learn the effective characteristics of the data sample, and the problem of small samples seriously affects the accuracy of the intelligent diagnosis algo...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to provide an intelligent diagnosis method for mechanical faults based on the migration prototype network in a small sample, which does not need to rely on a large amount of labeled data, and overcomes the problem that the intelligent diagnosis algorithm cannot be fully learned due to the lack of fault signals of mechanical equipment. Use mechanical signals less than 1% of the total data volume to train the network, and enable the state determiner to obtain more than 95% state classification accuracy

Images