(180d) Parameter Estimation from Operation Data for Simulated Moving Bed Chromatography

Chemical products in daily lives, such as sugars, petrochemicals, and chiral compounds, are produced by chromatographic processes. Simulated moving bed (SMB) chromatography is one of the most widely used chromatographic processes. To design and operate an SMB process, a reliable mathematical model should be utilized without incurring excessive experimental costs. To this aim, one potential method is to utilize the data that contain product concentrations under different operating conditions; in many SMB processes in industries, such data are collected and stored in a large amount without effective use. Such data can potentially be used to improve the accuracy of the mathematical model obtained from single-column chromatographic tests.

In this study, based on the model correction method [1], we estimated the model parameters in the SMB model including those in a nonlinear isotherm. We obtained 35 datasets from a pilot plant study including product concentrations under different operating conditions. The SMB process employed the Advanced SMB (ASMB) operation developed by Organo Corporation [2]. As a case study, separation of fructose and glucose was considered. Our parameter estimation scheme with Tikhonov regularization found a parameter set from the data efficiently, and uncertainty of the estimated parameters was quantified as confidence intervals. Moreover, model validation was carried out using the experimental data that were not used in the estimation, which confirmed that our estimation improved the model prediction (Figure 1).

Reference: [1] S. Tie et al., “Experimental evaluation of simulated moving bed reactor for transesterification reaction synthesis of glycol ether ester,” Adsorption, vol. 25, no. 4, pp. 795–807, 2019. [2] Sato et al., SEPARATION BY USE OF NEW ADVANCED-SMB PROCESS, 8th International Conference on Separation Science and Technology, Nagano, Japan 2008