(657a) Model Predictive Control of 4-Zone Simulated Moving Bed Chromatography for the Separation of Bicalutamide Enantiomers: Experimental Validation

Simulated moving bed (SMB) technology
has been applied in petrochemical, pharmaceutical, and fine chemical industries
to isolate products in high purity and high yield since it was introduced in
1960s [1]. In recent decades, different control concepts were applied to SMB
processes to optimize the operating conditions dynamically [2-4]. Since the SMB
process is operated periodically, i.e. the process outputs do not reach a
steady-state, it is difficult to observe the process states in detail.
Therefore, the control of SMB process is still challenging.

To estimate the SMB process states,
a simple cell model, a flux limiter, and competitive Langmuir/Bi-Langmuir
isotherms were used. The cell model is equivalent to the explicit finite
difference method, and the nonlinear isotherm models can be solved Newton’s
method. The computation time can be remarkably reduced by parallel computation
algorithms. From the estimated concentration profiles in the SMB process, the
actual and estimated process outputs, e.g. the product stream concentrations, can
be compared to decide the model parameters, i.e. isotherm parameters and
dispersion coefficients. A model predictive control concept was applied to search
the optimized future operating conditions to obtain the desired target purity
and yield.

Bicalutamide is a racemic drug for
prostate cancer (CASODEX, AstraZeneca). Since (R)-bicalutamide is more
efficacious than (S)-bicalutamide, a four-zone SMB process with chiral stationary
phase (Chiralpak IA, 20µm) was used to purify the eutomer, (R)-bicalutamide
[5]. In this work, an analytical HPLC unit was used to analyze the observed
stream concentrations. To design the four-zone SMB process, basic information
obtained from the batch preliminary experiments was provided. The four-zone SMB
process performed in a pilot plant scale (4 columns, 2.5 cm I.D., 15 cm length)
was successfully controlled to obtain the target (extract) and waste
(raffinate) components up to 99% purity (Figure 1).

Figure 1. Controlled operation histories (feed concentration = 14 g/L, Pu_Extr:
desired purity of extract product, Pu_Raff: desired purity of
raffinate product, R_S1: raffinate stream impurity concentration ratio
in recycle stream, R_S4: extract stream impurity concentration ratio
in recycle stream).

[1]   Broughton, D.B.,
Neuzil, R. W., Pharis, J. M., Brearley, C. S. (1970). The parex process for
recovering paraxylene. Chem. Eng. Prog., 66, 70.

[2]   Natarajan, S.,
Lee, J.H. (2000). Repetitive model predictive control applied to a simulated
moving bed chromatography system. Comput. Chem. Eng., 24, 1127.

[3]   Schramm, H.,
Grüner, S., Kienle, A. (2003). Optimal operation of simulated moving bed
chromatographic processes by means of simple feedback control. J. Chromatogr.
A, 1006, 3.

[4]   Grossmann, C.,
Langel, C, Mazzotti, M., Morari, M., Morbidelli, M. (2010). Experimental
implementation of automatic ‘cycle to cycle’ control to a nonlinear chiral
simulated moving bed separation. J. Chromatogr. A., 1217, 2013.

[5]   Kaemmerer, H.,
Horvath, Z., Lee, J. W., Kaspereit, M., Arnell, R., Hedberg, M., Herschend, B.,
Jones, M. J., Larson, K., Lorenz, H., Seidel-Morgensten, A. (2012) Separation
of Racemic Bicalutamide by an Optimized Combination of Continuous
Chromatography and Selective Crystallization. Org. Process Res. Dev., 16, 331.