Study On Fault Diagnosis Of Beer Brewing Process Based On KPCA

With the progress of science and technology and the improving of degree of automation,the structure of system is increasingly complex,so how to ensure the reliability of the system and security has become a problem of public concern.In this paper,based on beer fermentation of the batch fermentation process and the method of industrial process fault detection is the multivariate statistical analysis.The main research contents and work are as follows:1.This paper analyzes the present situation of the research on fault monitoring at home and abroad,and introduces the basic idea and geometric meaning of PCA.The process of fault detection based on PCA is also studied.2.The PCA and the locality preserving projection(LPP)method are introduced.This method takes into account the main change direction of the global data of PCA and the advantages of LPP in expressing the local structure of data.In the process of converting data from high dimension to low dimension,it takes into account the advantages of LPP,and does not affect the relationship between samples.Similarly,by projecting the data into the direction of the largest variance,the global structure is taken into account.Finally,the application of orthogonal constraints,which is subject to the singularity of the interference,the smaller the amount of calculation.3.The kernel function is introduced to solve the problem of missing report in KPCA-LPP algorithm.Using kernel function mapping nonlinear data into a high dimensional space,and by using the local structure of high dimensional samples improved distance measure samples,fault monitoring model of the KPCA-LPP method and established.4.Finally using for simulation the fermentation process of beer.The validity of the method is proved.