Parameter estimation for reactive chromatography model by Bayesian inference and parallel sequential Monte Carlo

Abstract

Reactive chromatography overcomes many disadvantages of conventional separation processes by improving yields and reduces costs by combining reaction and separation in a single operation. Model-based optimization is essential to realize industrial applications of this process. Parameters in the model must be estimated accurately, and evaluating the uncertainty of the model parameters is crucial. However, uncertainty quantification of model parameters has not been performed for reactive chromatography processes. In this study, we propose an approach to estimate the parameter uncertainty of reactive chromatography processes using Bayesian inference and parallel sequential Monte Carlo. As an example, the esterification synthesis of acetic acid and methanol catalyzed by a cation exchange resin was considered. Parameter estimation was performed using a reactive chromatography experiment and a non-reactive experiment in which only the products were injected. The results of the analysis showed that using both the reactive and non-reactive chromatography experimental data simultaneously improved the accuracy of the estimation. In addition, correlations between some parameters were revealed.