The continuous-addition-of-reagent technique as an effective tool for enhancing kinetic-based multicomponent determinations using computational neural networks

Abstract

The batch and continuous-addition-of-reagent (CAR) techniques were tested to determine whether the approach used to mix the sample and reagents influences accuracy in kinetic multicomponent determinations based on computational neural networks (CNNs). Both techniques were used to obtain kinetic profiles for ternary mixtures of related sulphur-containing amino acids ( l-cysteine, N-acetyl- l-cysteine and dl-homocysteine) by reaction with the copper(II)–neocuproine complex, from which CNN inputs were acquired at a fixed sampling frequency. Once the influence of chemical and computational variables was established, trained networks were used to estimate the amino acid concentrations in mixtures with relative standard errors of prediction in the range 15–45% and 1–4% for the batch and CAR technique, respectively. The accuracy of the CAR results was found to depend largely on its peculiar response curve, which is a result of the special way in which the sample and reagents are mixed.