Modeling Comparison between Novel and Traditional Feed Modes of Simulated Moving Bed

Abstract

A new feed mode for simulated moving bed (SMB) is proposed. The solution out of zone II flows into a tank to dissolve the solid raw materials and is then fed into zone III. The feasibility of the new feed mode was examined with a model SMB operation. The results showed that this mode could effectively separate mixtures. As the raw materials are dissolved with the solution from zone II, the solvent consumption is reduced significantly in comparsion with the traditional SMB operations; whereas the productivities and purities change insignificantly. When the flow rates of raffinate and extract stream are similar, the concentration plateaus of two solutes in column have similar and suitable heights, otherwise an overly high concentration could cause the solutes to be crystalized and plug the column. The volume of solution tank showed an insignificant effect on the separation. Increasing the tank volume can make a smooth change of concentrations of solutes in the tank, which is beneficial to the operation.

Novel Simulated Moving-Bed Cascades with a Total of Five Zones for Ternary Separations

“…The separations by the SMB cascades are investigated using a rigorous SMB model with periodic switches of the four ports. − The change of component i in column j during each switch interval is described by the transport–dispersive equations as follows: ∂ C i , j ∂ t + 1 − normalε normalε ∂ q i , j ∂ t = prefix− u j ∂ C i , j ∂ Z + D L , i ∂ 2 C i , j ∂ Z 2 ∂ q i , j ∂ t = k f , i ( q i , j * − q i , j ) where q i * is the solid concentration in equilibrium with the liquid concentration C i , and j is the serial number of the column starting from the first column in zone I. The adsorption equilibrium follows Henry’s law as follows: q i * = H i C …”

“…The initial and other boundary conditions are identical to those in conventional SMBs. , The convective term ∂C / ∂Z is approximated using the five-point biased upwind finite difference scheme, and the dispersive item ∂ 2 C / ∂Z 2 is approximated using the five-point centered scheme . Thus, the model is discrete to ordinary differential equations, which are solved using the ODE15s solver in Matlab.…”