Neural Network-Based Identification of Nonlinear Adsorption Isotherms

Abstract

The adsorption isotherm is the most important parameter in chromatographic separation process. In this paper, a numerical isotherm estimation method based on neural networks is proposed. As not the absolute values but the slopes of the isotherms are most important in process simulation and optimization, the slopes of the isotherm instead of the absolute values are taken as the outputs of the neural network. A tailored method is used to initialize the neural network. The parameters of the neural network are adjusted to minimize the difference between the simulated and the measured profiles. The issue of the design of the experiments, which decides the amount of the isotherm information contained in the profiles, is discussed. Simulation and experimental studies illustrate the potential of the proposed method