Abstract

The modeling of adsorption processes appears quite frequently in the chemical industry, petrochemical plants and refineries, for example for separation and purification of mixtures in Simulated Moving Bed (SMB) units. In the mathematical formulation, the accurate determination of the model parameters is an important step for the design of chromatographic conditions for continuous separation in SMB processes. This work is aimed at the estimation of the model parameters in adsorption processes, using a chromatographic column for the separation of glucose and fructose. The Bayesian framework for inverse problems is investigated through the implementation of Markov Chain Monte Carlo methods (MCMC) and a critical comparison against the classical Maximum Likelihood approach, with the minimization of the objective function via two different stochastic techniques, namely the Particle Collision Algorithm (PCA), and the Particle Swarm Optimization (PSO) is performed. Different cases are presented in order to investigate the statistical significance of the estimates obtained, and perform comparisons between the solution via Bayesian inference and via the minimization of the objective function with the stochastic methods. The results demonstrate that the Bayesian approach employs less computational effort to achieve estimates with comparable statistical information.