Simultaneous ultraviolet spectrophotometric determination of sodium benzoate and potassium sorbate by BP-neural network algorithm and partial least squares

Abstract In this study, we proposed a reliable method to quickly and simultaneously determine the content of sodium benzoate and potassium sorbate simultaneously by ultraviolet (UV) spectrophotometry with BP-neural network algorithm(BP-ANN) and partial least squares regression(PLS). The pure reagents were used to prepare a series of sodium benzoate and potassium sorbate solutions and used deionized water as a reference solution. The calibration models were constructed by using 36 reference samples according to the orthogonal design and the prediction model set consisted of 9 sets of randomly configured mixed solutions. The results used BP-ANN and PLS showed that the root mean square error prediction (RMSEP) of sodium benzoate and potassium sorbate were 0.129, 0.09 and 0.155, 0.089, and the correlation coefficient (R2) were 0.9997, 0.9994 and 0.9998, 0.9995, respectively. The recovery of the actual samples by both methods was over 97%.