Stochastic optimization of simulated moving bed process: Sensitivity analysis for isocratic and gradient operation

Abstract

A random search strategy has been used for designing and optimization of simulated moving bed process (SMB) under isocratic as well as under solvent gradient conditions. The effectiveness of both the process modes has been compared. For predictions of the objective functions, i.e., the minimum of eluent consumption and/or the maximum of the process productivity a mathematical model of the process dynamics has been employed and implemented in the optimization procedure. Four-dimensional space of decision variables corresponding to the flowrates in the SMB zones has been searched in order to find the optimal set of the process parameters. The optimization was constrained to the purity demand in the outlet streams withdrawn in the raffinate and the extract port. The obtained set of random decision variables fulfilling purity constraints was used to construct the operating window of parameters guaranteeing successful separation. For feasible operating points the sensitivity of the purity constraints with respect to the operating parameters has been calculated. The results of calculations indicated that operating conditions, which ensure similar process efficiency, could correspond to different sensitivity of the process constraints. Such an analysis was found to be useful for the selection of process conditions, for which the best trade-off between the process efficiency and its robustness can be achieved. This appears to be particularly important for designing the gradient SMB process, for which robustness of the operating conditions is a factor of a major importance. In order to improve the efficiency of calculations a modification of the original random search procedure based on the Luus–Jaakola algorithm has been proposed.