A Method of Fault Classification and Diagnosis Based on Nonlinear Similarity Index

Abstract

The invention discloses a fault classification and diagnosis method based on a nonlinear similarity index, which aims to solve three key problems faced when implementing the fault classification and diagnosis method in an actual industrial process: First, refer to available training samples of fault types Second, the available fault training data is in the initial stage of the operation state transition, and the data nonlinearity is strong; third, the sampling data of different faults will overlap in the spatial distribution. The inventive method firstly selects the characteristic variable of each reference fault type, and selects the characteristic variable that can best distinguish the fault from normal data. Then, only the feature variables are used to match the similarity between the online fault data window and the reference fault data window. The similarity calculation method adopted is based on the nonlinear similarity index of window data, and the fault type detected online can be determined according to the minimum nonlinear similarity value. Compared with the traditional classification and diagnosis method, the method of the present invention significantly reduces the variable dimension through the selection of characteristic variables, which not only greatly reduces the constraints of insufficient training data, but also eliminates the impact of non-characteristic variables on the calculation of nonlinear similarity indicators. negative impact. In addition, the method implements fault similarity matching through nonlinear similarity matching of window data in spatial distribution, which can avoid misclassification of overlapping data to the greatest extent.

Description

technical field [0001] The invention relates to a fault diagnosis method, in particular to a fault classification diagnosis method based on a nonlinear similarity index. Background technique [0002] The increasingly fierce market competition has put forward higher and higher requirements for the continuous and normal operation of the production process. Timely detection of faults in the production process and correct identification of fault types occupy an important position in the entire production automation system. For modern large-scale and complex industrial production systems, accurate mechanism models are difficult to obtain. Instead, due to the wide application of DCS, the production process can measure and store massive industrial data. These sampling data hide information on process operation status and product quality, which provides a solid foundation for data-driven process monitoring methods. In recent years, data-driven process monitoring methods have receiv...

AI Technical Summary

Problems solved by technology

In this way, the constraint of insufficient sample size on the establishment of models is greatly reduced.

Images