CHEMICAL-COMPOSITION ANALYSIS OF CARRAGEENANS BY INFRARED-SPECTROSCOPY USING PARTIAL LEAST-SQUARES AND NEURAL NETWORKS

The use of partial least squares and neural networks is described for the multivariate analysis of the chemical composition of carrageenans by infrared spectroscopy. The content of K-, l- and lambda-carrageenan is determined in standard mixture samples and in production batches of carrageenans. Various models of PLS and neural networks are examined using a designed ternary mixture of different carrageenan forms and the use of up to 36 independent variables selected from the infrared spectrum of the carrageenans. Both models can be applied for an accurate and rapid chemical composition analysis, although the neural networks model performs slightly better and appears to be more consistent.