Chemical separation process monitoring based on nonlinear principal component analysis

Principal component analysis (PCA) is a useful tool to deal with linear relationship among process variables. For many industrial processes with variables containing nonlinear relationship, conventional PEA methods lose their power. Instead, applying neural network,technique, some generalized linear PCA methods are presented. Motivated by the results of [1], this paper discusses monitoring and diagnosis for a chemical separation process. Two neural networks are employed, one of which is used,to model nonlinear loading functions, and another to map principal components onto corrected data set.