(308e) A Centralized/Decentralized Control Approach for the Multicolumn Countercurrent Solvent Gradient Purification (MCSGP) Process
AIChE Annual Meeting
2015
2015 AIChE Annual Meeting Proceedings
Computing and Systems Technology Division
Networked, Decentralized and Distributed Control
Tuesday, November 10, 2015 - 9:50am to 10:10am
A
centralized/decentralized control approach for the Multicolumn Countercurrent
Solvent Gradient Purification (MCSGP) process
Maria M.
Papathanasioua,d, Muxin Suna, Fabian Steinebachb,
Thomas Müller-Späthc, Massimo Morbidellib, Athanasios
Mantalarisa, Efstratios N. Pistikopoulosd*
aDept. of Chemical Engineering, Centre for Process
Systems Engineering (CPSE), Imperial College London SW7 2AZ, Lodnon, U.K
b Institute for Chemical and Bioengineering, ETH
Zurich,
Wolfgang-Pauli-Str. 10/HCI F 129,
CH-8093 Zurich, Switzerland
AG, Technoparkstr. 1, CH-8005 Zurich, Switzerland dArtie McFerrin Department of
Chemical Engineering, Texas A&M University, College Station TX 77843 *stratos@tamu.edu
Keywords: periodic processes,
multi-parametric, control, stability
Industrial applications, such as
separation processes and pressure swing adsorption systems (PSA) are
characterized by cyclic operation profiles and they often involve tight
constraints on product purity. In order to maintain the process under optimal operation,
advanced computational tools need to be developed. The latter include
optimization strategies and optimal control policies that are designed to meet
the process constraints. During the past few years, several research groups
have been focusing on the development of such tools, aiming to tackle issues
related to system periodicity and nonlinearity [1-6].
In this work we present a
centralized/decentralized control approach based on the principles of the
Multicolumn Countercurrent Solvent Gradient Purification (MCSGP) process. MCSGP
is an industrial, semi-continuous chromatographic separation process used for
the purification of various biomolecules [7]. The system is described by a
highly nonlinear Partial Differential and Algebraic Equation (PDAE) model that
operates under a cyclic profile. Following our previously presented work [8, 9] we demonstrate a seamless
procedure for the development of advanced decentralized multi-parametric
control strategies for the system at hand.
The examined system shifts
periodically between two distinct configurations (Figure 1). Figure 1a
illustrates the continuous operation mode, where the two columns operate
connected to each other. This is usually followed by batch operation as shown
in Figure 1b. For further details regarding the complete system operation the
reader is referred to Krättli,
et al. [10].
In this study we develop independent control strategies for each column that
are then linked during the interconnected operation, tracking the integral of
the outlet concentrations. In particular, the outlet of column A is treated as
a disturbance by the controller operating on column B during continuous
operation (Figure 1a).
Figure
1
Chromatographic system based on the principles of MCSGP for (a) continuous and
(b) batch operation mode, considering the modifier concentration as input, the
integral of the three outlet concentrations as outputs and the feed composition
as measured disturbance.
The periodic operation profile and
the disturbances occurring from the feed stream render the preservation of the optimal
operating conditions a challenging task. Therefore, to ensure optimal operation
throughout the process cycle, the controller stability needs to be guaranteed. Following
the framework presented by Pistikopoulos, et al. [8]
the controller development is based on the derivation of a linear state space
model that describes the system dynamics. The optimal periodic steady state and
input strategy for the tracked process intervals are calculated off-line,
minimizing the cost function of the optimization problem [11]. The online control scheme can
then be applied by driving the system to the pre-computed optimal periodic
steady state by using multi-parametric Model Predictive Control (mp-MPC)
(Pistikopoulos, 2009).
To prove the stability of the
cyclic process we introduce a transformed system as suggested by to Huang,
et al. [12].
The designed controllers are then
tested in-silico, in a ?closed-loop' fashion, using the original process model [8].
Due to the system dynamics a repetitive output time delay is observed that
corresponds to the time that the components require to reach the column exit [13]. This can be either pre-computed
or related to the output and/or control horizon during the design of the
control policies. Here we examine both cases and we demonstrate the evolution
of the time delay with respect to both the output and the control horizon. It
can be observed that both strategies are equally effective, however the second
case allows greater freedom in the controller design as the setpoints are tracked
independently from the time delay. The suggested strategy can efficiently track
the desired outputs and retain the process under optimal operating conditions.
The designed controllers guarantee stable, cyclic operation and the system time
delay is successfully eliminated.
Acknowledgements
Financial support
from the European Commission (OPTICO/G.A. No.280813) is gratefully acknowledged.
References