Prediction of the competitive adsorption isotherms of 2-phenylethanol and 3-phenylpropanol by artificial neural networks

• ANN a “soft” model to determinate the adsorption isotherm of a competitive system.
• The adjusted error signal was applied to train the network rapidly and efficiently.
• The ANN was more accurate than the “hard” model (the empirical mathematical method).

Artificial neural networks (ANNs) were regarded as data-mapping networks with strong nonlinear fitting abilities. A 2-6-2 network was used to determine the competitive adsorption isotherm of 2-phenylethanol (PE) and 3-phenylpropanol (PP). The ANN results were forms of data mapping rather than theoretical mathematical model. The ANN architecture was established after training with a set of experimental data. The established ANN was applied to predict the adsorption isotherms of PE and PP. The selection of parameters for the ANN was discussed. The results indicate that ANN has excellent potential for use in non-linear chromatography for the prediction of adsorption isotherms.