Evaluation of nonlinear modeling based on artificial neural networks for the spectrophotometric determination of Pd(II) with CPA-mK

Abstract

A new method is proposed for the spectrophotometric determination of Pd(II), based on the reaction of Pd(II) with 2-(4-chloro-2-phosphonophenylazo)-7-(3-carboxyphenylazo)-1,8-dihydroxynaphthalene-3,6-disulfonic acid(CPA-mK) in sulfuric acid without heating. Beer’s law is obeyed for 1.0–4.0 μg of Pd (II) in 10 mL of solution. The calibration curve from 1.0 to 42.0 μg in 10 mL of solution is modeled successfully by artificial neural networks (ANNs). The maximum relative error between experimental values and the values predicted by ANNs is 1.5%. In comparison with some mathematical functions, ANNs show better ability for curve fitting, thus greatly extending the applicable range of the calibration curve of this new system. The method has been applied to determine Pd (II) in ore and catalyst samples with a relative error of less than 4% and with a recovery range between 94% and 103%.