Estimation of fatty acid composition using PLS based model

Abstract

Given sufficiently rich and appropriate pre-treated data, Partial Least Square (PLS) models are able to accurately represent process dynamics. However, when applied to a fatty acid distillation process over broader ranges of operating conditions, the model was found not adequate. To incorporate nonlinear estimation feature, neural networks were incorporated to form Neural Network (NNPLS) and a further modified method known as Nested NNPLS. The results obtained proved that the Nest-NNPLS model provided the best estimation capability and should therefore be developed further as a potential candidate for on-line estimation of chemical product composition.