The Research On Fault Diagnosis Based On PCA And PLS

with the development of the industrial level and the large-scale application of computer technology,the fault diagnosis method has become the key problem in the industrial production.Were observed and predicted by the production process of the state can before the occurrence of fault timely detection of abnormal conditions,diagnose the fault position,reasonable to eliminate noise interference so as to ensure the normal and safe operation of industrial production process.With the development of computer,a large amount of information can be saved in the process of industrial production.Therefore,how to make use of this information reasonably and how to judge the normal working condition by analyzing the big data has become a hot issue in the research field.Under this background,the data driven fault diagnosis method has been widely used in the metallurgical and pharmaceutical production process.Multivariate statistical analysis is based on principal component analysis(PCA)and partial least squares(PLS).This method does not need to establish a specific mathematical model.By reducing the dimension of high-dimensional data processing,can improve the computational efficiency,so it has important research value and practicality.In this paper,the principal component analysis and partial least squares method are studied,and the simulation is carried out based on the working condition of the wind turbine.The main contents of this paper are as follows:The importance and significance of fault diagnosis technology in the field of automatic control are explained.This paper introduces the development history of fault diagnosis and data driven method in the past half century.The basic expression of principal component analysis(PCA),the basic theory of PCA,and the modeling method based on principal component analysis(PCA)are described.Aiming at the deficiency of the traditional principal component analysis method,this paper studies the reconstruction of the principal component method.In this paper,the basic theory of partial least squares method is introduced,andthe simulation results are compared with the results of principal component analysis and partial least squares method.The research trends of the two methods are analyzed.Summary and prospect.The contents of this paper are summarized and the next research plan of data driven is planned.