Fault diagnosing method based on class mean kernel principal component analysis and a BP (Back Propagation) neural network

Abstract

The invention relates to a fault diagnosing method based on class mean kernel principal component analysis and a BP (Back Propagation) neural network. The method comprises the steps of firstly classifying historical data of equipment operation according to fault types to obtain a data set with fault class labels; standardizing the data set, calculating a class mean kernel matrix of the data set and carrying out centralization; calculating eigenvalues and eigenvectors of the centralized class mean kernel matrix and extracting a kernel principal component to obtain a dimension-reduced sample dataset; training the neural network by using the sample dataset; and finally, carrying out fault identification on equipment by using the trained neural network. According to the method disclosed by theinvention, the class mean kernel principal component analysis is combined with the BP neural network; no instruction is given to a detected system; and the fault diagnosis of various production systems can be realized by only using operating data of the detected system. The method has the characteristics of low computational complexity, high diagnosis speed, high accuracy, good generalization andthe like.

Description

technical field [0001] The invention relates to a fault diagnosis method combining mean value kernel principal component analysis and BP neural network, belonging to the technical field of testing. Background technique [0002] With the improvement of industrial automation level, the scale of industrial control system is getting bigger and bigger. A large industrial control system has hundreds of sets of control loops. Therefore, there are more and more factors affecting the normal operation of the system, and the reliable operation of the system is subject to many factors. aspects of the threat. Non-linearity, time-varying, strong coupling, high-dimensionality, etc. have caused the complexity and diversity of the control system, which often affect the whole body. Once a failure occurs, it will not only cause immeasurable losses and harm to social production and the ecological environment, but also seriously threaten the safety of personnel. Therefore, it is of great signif...

AI Technical Summary

Problems solved by technology

[0006] However, the existing fault diagnosis methods are not ideal, and there are disadvantages such as complex calculation, low diagnostic efficiency, low accuracy and generalization, so it is necessary to further develop and improve

Images