Simultaneous spectrophotometric determination of maltol, ethyl maltol, vanillin and ethyl vanillin in foods by multivariate calibration and artificial neural networks

Abstract

Maltol (MAL), ethyl maltol (EMA), vanillin (VAN) and ethyl vanillin (EVA) are food additives, and they have well defined UV spectra. However, these overlapped seriously, and it is difficult to determine them individually from their mixtures without a pre-separation. In this paper, chemometric approaches were applied to resolve the overlapping spectra and to determine these compounds simultaneously. The analysis of these four compounds was facilitated by the use of an orthogonal array data set consisting of absorption spectra in the 200–350 nm ranges obtained from a calibration set of mixtures containing these compounds. With this dataset, seven different chemometric models were built, such as classical least squares (CLS), principal components regression (PCR), partial least squares (PLS), and artificial neural networks (ANN). These chemometric models were then tested by the use of a validation dataset constructed from synthetic solutions of these four compounds. The analytical performance of these chemometric methods was characterized by relative prediction errors (RPE) and recoveries. The proposed methods were successfully applied to the analysis of commercial food samples. It was found that the radial basis function artificial neural networks (RBF-ANN) gave better results than other chemometric methods. PLS, PCR, DPLS, and DPCR also give satisfactory results, while CLS and DCLS perform poorer. It was also found that there was no advantage to pre-treat spectra by taking derivatives. The four compounds, when taken individually, behaved linearly in the 1.0–20.0 mg l −1 concentration range, and the limits of detection (LOD) for MAL, EMA, VAN and EVA were 0.39, 0.56, 0.49 and 0.38 mg l −1, respectively.