Orthogonal projection approach and continuous wavelet transform-feed forward neural networks for simultaneous spectrophotometric determination of some heavy metals in diet samples

Abstract

Simultaneous spectrophotometric determination of a mixture of overlapped complexes of Fe3+, Mn2+, Cu2+, and Zn2+ ions with 2-(3-hydroxy-1-phenyl-but-2-enylideneamino) pyridine-3-ol(HPEP) by orthogonal projection approach-feed forward neural network (OPA-FFNN) and continuous wavelet transform-feed forward neural network (CWT-FFNN) is discussed. Ionic complexes HPEP were formulated with varying reagent concentration, pH and time of color formation for completion of complexation reactions. It was found that, at 5.0×10−4molL−1 of HPEP, pH 9.5 and 10min after mixing the complexation reactions were completed. The spectral data were analyzed using partial response plots, and identified non-linearity modeled using FFNN. Reducing the number of OPA-FFNN and CWT-FFNN inputs were simplified using dissimilarity pure spectra of OPA and selected wavelet coefficients. Once the pure dissimilarity plots ad optimal wavelet coefficients are selected, different ANN models were employed for the calculation of the final calibration models. The performance of these two approaches were tested with regard to root mean square errors of prediction (RMSE %) values, using synthetic solutions. Under the working conditions, the proposed methods were successfully applied to the simultaneous determination of metal ions in different vegetable and foodstuff samples. The results show that, OPA-FFNN and CWT-FFNN were effective in simultaneously determining Fe3+, Mn2+, Cu2+, and Zn2+ concentration. Also, concentrations of metal ions in the samples were determined by flame atomic absorption spectrometry (FAAS). The amounts of metal ions obtained by the proposed methods were in good agreement with those obtained by FAAS.