Application of experimental design and radial basis function neural network to the separation and determination of active components in traditional Chinese medicines by capillary electrophoresis

Abstract

Orthogonal design has been used to the optimization of separation and determination of two active components in traditional Chinese medicines by capillary electrophoresis. The concentration of phosphate, applied voltage, organic modifier content and buffer pH were selected as variable parameters. Their different effects on peak resolution were studied by the experimental design method. Optimized separation conditions were obtained and successfully applied to the separation and determination of aconitine and hypaconitine in Aconitum medicinal herbs. Good separation was achieved within 7 min using a buffer system composed of 20 mmol L −1 phosphate and 35% acetonitrile at pH 9.5. The applied voltage was 14 kV and the detection was set at 235 nm. In addition, a radial basis function neural network with a “4-18-1” structure was developed based on the experimental results of orthogonal design and uniform design, and was applied to the prediction of peak resolution of the two active components under the optimum separation conditions given by orthogonal design. The predicted results were in good agreement with the experimental values, indicating that radial basis function neural network is a potential way for the selection of separation conditions in capillary electrophoresis.