Artificial neural networks and multivariate calibration for spectrophotometric differential kinetic determinations of food antioxidants

Mixtures of food antioxidants, butylated hydroxyanisole (BHA), butylated hy droxytoluene (BHT) and propyl gallate (PG), were simultaneously analyzed wi th spectrophotometry, based on their different kinetic properties. These an tioxidants react differentially with Fe(III), and the reduced product of wh ich, Fe(II), will be complexed by chromogenic reagent 2,2'-dipyridyl. The d ifferential kinetic spectra were monitored and recorded at 510 nm, and the data obtained from the experiments were processed by chemometric approaches , such as artificial neural network (ANN), classical least squares (CLS), p rincipal component regression (PCR) and partial least squares (PLS). A set of synthetic mixtures of antioxidants was evaluated and the results obtaine d by the applications of these chemometric approaches were discussed and co mpared. It was found that the ANN method afforded better precision relative ly than those of CLS, PCR and PLS. The proposed method was also applied sat isfactorily to the determination of antioxidants in several commercial food products. (C)1999 Elsevier Science B.V. All rights reserved.