Bearing fault diagnosis method based on prototype network

Abstract

The invention provides a bearing fault diagnosis method based on a prototype network, and belongs to the technical field of bearing fault diagnosis methods. The method comprises the following specific steps: using a deep neural network as a measurement module; convolution kernels from large to small are used as a feature extraction module; a measurement module and a feature extraction module are used to construct a prototype network model of bearing fault diagnosis; training the prototype network model by using the plurality of sample bearing vibration signals; and performing fault diagnosis on the bearing vibration signal by using the trained prototype network model. According to the invention, fault diagnosis can be carried out on the bearing under the condition of small samples, and more accurate fault classification can be carried out on unknown operation state samples.

Description

technical field [0001] The invention belongs to the technical field of bearing fault diagnosis methods, and in particular relates to a bearing fault diagnosis method based on a prototype network. Background technique [0002] Bearing is one of the indispensable parts in mechanical equipment, and it is also one of the most prone to failure parts in mechanical equipment. If the bearing breaks down, it will cause economic losses in the slightest and threaten personal safety in the worst case. Therefore, it is particularly important to carry out fault diagnosis on bearings. When a bearing fails, it can be detected and repaired in time, which is of great help to the normal operation of mechanical equipment and can avoid unnecessary losses. [0003] In the field of fault diagnosis, deep learning has been widely valued due to its powerful nonlinear feature extraction and representation capabilities. However, the premise of achieving high accuracy for fault diagnosis methods based ...

AI Technical Summary

Problems solved by technology

[0003] In the field of fault diagnosis, deep learning has been widely valued due to its powerful nonlinear feature extraction and representation capabilities. However, the premise of achieving high accuracy for fault diagnosis methods based on deep learning requires a large amount of labeled data for training. In practical engineering applications, it is very difficult and expensive to collect a large number of accurate labeled data samples, which leads to the problem of over-fitting of traditional deep learning networks in the case of small samples, resulting in low model accuracy

Images