A Multi-Class Imbalanced Fault Classification Method Based on Reinforcement Learning and Knowledge Distillation

Abstract

The invention discloses a multi-category unbalanced fault classification method based on reinforcement learning and knowledge distillation. The method combines hierarchical clustering, knowledge distillation and reinforcement learning algorithms to solve the problem of multi-category unbalanced fault classification. For multi-category fault classification problems, firstly, hierarchical clustering is used to cluster multi-categories into several clusters according to the characteristics of similarity between samples of homogeneous categories and large differences between samples of heterogeneous categories in imbalanced problems, According to different clusters, the student network is established for fine-grained classification, and then the knowledge distillation method is used to take into account the global information. Finally, the sample weight is iteratively learned in combination with reinforcement learning, so as to improve the classification effect of unbalanced faults. In this process, it is necessary to design a reasonable reward function to cooperate with the fine-grained knowledge distillation classifier to optimize the sample weight. Compared with other comparative methods, the method of the present invention has good effect and applicability.

Description

technical field [0001] The invention belongs to the field of industrial process monitoring, and in particular relates to a multi-class unbalanced fault classification method based on reinforcement learning and knowledge distillation. Background technique [0002] In machine learning or deep learning classification, the imbalance of the number of class samples is a very common problem, which widely exists in various fields, such as bioinformatics, smart grid, medical imaging, fault diagnosis. Most existing classification methods are based on the assumption that the underlying distribution of observed data is relatively balanced. However, real industrial datasets often violate this assumption and exhibit skewed distributions or even extremely unbalanced distributions of class sample sizes. For example, data-driven fault classification is an important part of industrial process monitoring, which exhibits unbalanced skewed distributions due to the different frequencies of fault...

AI Technical Summary

Problems solved by technology

However, real industrial datasets often violate this assumption and exhibit skewed distributions or even extremely unbalanced distributions of class sample sizes

Images