Development of Temperature Dependent Retention Models in Ion Chromatography by the Cascade Forward and Back Propagation Artificial Neural Networks

摘要

The most important part of the complex ion chromatography method development process is retention modeling. It tries to integrate the demands for high quality ion chromatography with the demands for low consumption of chemicals, fast analysis, and the short time of method development. This work compares the properties of the cascade forward and back propagation artificial neural network in the development of temperature dependent retention models. The retention times of bromate, bromide, nitrite, iodide, and perchlorate were modeled in relation with temperature of separation process, concentration of hydroxide eluent competing ion, and eluent flow rate. Artificial neural networks were optimized in term of selecting the optimal training algorithm, optimal number of hidden layer neurons, activation function, and number of experiments needed for modeling procedure. The retention model based on cascade forward methodology exhibited superior predictive ability and, therefore, should be the method of first choice for the temperature dependent optimization in ion chromatography.