Accelerating optimization and uncertainty quantification of nonlinear SMB chromatography using reduced-order models

Abstract

A parametrized reduced-order model is constructed and employed as a surrogate for the full-order model in optimization and uncertainty quantification of nonlinear simulated moving bed chromatography. The reduced-order model is obtained by the reduced basis method using an efficient error estimation. The complexity of the model is reduced by an empirical interpolation method applied to the nonlinear part of the model. Due to the reduced size and complexity of the surrogate model, the processes of optimization and uncertainty quantification are sped up by a factor of 10.

Propagating input uncertainties into parameter uncertainties and model prediction uncertainties—A review

“…• Crude natural gas flow rate • Chiller heat removal duty Sahraei et al [45] Pilot-scale gasifier • Volatile percentage of the fuel • Solid particle diameters • Angle of multiphase flow jet • Recirculation ratio Zhang et al [46] Moving-bed chromatography • Flow rate Minh et al [8] Crude oil distillation and the corresponding revenue per unit of processes crude oil…”

“…[64] Surrogate models have also been used to cope with the computationally intensive MC simulations involving fundamental models, [8] especially models that are described by partial differential equations (PDEs). [46][47][48][49][50][51]53,54,65] In this approach, an empirical surrogate model is fitted to a set of measurement-error-free fundamental-model responses obtained from a limited number of plausible uncertain inputs (e.g., 10 to 500). [66,67] The fitted surrogated model is then used for model evaluation in step 2 of the algorithm in Table 2, instead of the complex original model.…”

Efficient simulation of chromatographic separation processes

“…Chromatography is an effective process which plays a central role in a great many industrial separation and purification systems. Not surprisingly therefore, the modelling and optimisation of these processes has received a great deal of attention in recent years (von Lieres and Andersson, 2010;Enmark et al, 2011;Close et al, 2014;Zhang et al, 2017;Hahn et al, 2014). A particular case is liquid batch chromatography, which is achieved by the injection of a pulse of solute into the chromatographic column where the differential adsorption results in the separation of the solute between the liquid and solid phases.…”

“…A number of models of increasing complexity have been proposed for modelling chromatography problems (Guiochon et al, 2006), notably the general rate model (GRM) (Guiochon et al, 2006;Püttmann et al, 2016), lumped kinetic model (LKM) (Zhang et al, 2017;Pais et al, 1998) and equilibrium-dispersion model (EDM) (Enmark et al, 2011;Chan et al, 2008). All of these models effectively describe the convection dominated flow of the solute through the bed along with mass transfer with the stationary phase.…”