Modeling and monitoring of batch processes using principal component analysis (PCA) assisted generalized regression neural networks (GRNN)

Abstract

Multivariate statistical methods namely, principal component analysis (PCA) and partial least squares (PLS), which perform dimensionality reduction and regression, respectively, are commonly used in batch process modeling and monitoring. While PCA is used to monitor whether process input variables are behaving normally, the PLS is used for predicting values of the process output variables. A significant drawback of the PLS is that it is a linear regression formalism and thus makes poor predictions when relationships between process inputs and outputs are nonlinear. For overcoming this drawback, a formalism integrating PCA and generalized regression neural networks (GRNNs) is introduced in this paper for conducting batch process modeling and monitoring. The advantages of this PCA-GRNN hybrid methodology are (i) process outputs can be predicted accurately even when input-output relationships are nonlinear, and (ii) unlike other commonly used artificial neural network (ANNs) paradigms (such as the multi-layer perceptron), training of a GRNN is a one-step procedure, which helps in faster development of nonlinear input-output models. A two-module software package has been developed for implementing the PCA-GRNN formalism and the effectiveness of the proposed modeling and monitoring formalism has been successfully demonstrated by conducting two case studies involving penicillin production and protein synthesis.