Sbarciog_benelux2013.dvi

Robust cascaded control of an animal cell culture
Mihaela Sbarciog a, Daniel Coutinho b, Alain Vande Wouwer a a Automatic Control Laboratory, University of Mons, Boulevard Dolez 31, B-7000 Mons, Belgium b Department of Automation and Systems, Federal University of Santa Catarina, 476 Florianopolis, 88040-900, Brazil MihaelaIuliana.Sbarciog, Alain.VandeWouwer@umons.ac.be, Coutinho@das.ufsc.br 1 Introduction
Perfusion operation of cell growing systems provides con- sistent culture conditions, high productivity and low prod- uct residence times. In spite of the increased productivity of the culture, perfusion operation with partial cell retention is hardly used at industrial scale because of the complexity raised by the multivariable nature of the process. Althoughseveral studies exist regarding the necessity and the influ-ence of the bleed stream on the cells growth, not much isknown on how to set this process input or how to use it for control and optimization purposes. Therefore, the ex-isting control implementations on perfusion cultures try to Figure 1: Control structure
achieve the control goal by only manipulating the perfusion rate. We present here a multivariable robust control structure for an animal cell growth process, in which the dilution and the bleed rates are manipulated to control the cell and sub- strate concentrations. The control structure combines a par- tial feedback linearizing control, designed to ensure robust- ness with respect to parametric uncertainty and uncancelled nonlinear dynamics, with predictive control principles.
2 The process model
The animal cell culture used in this study is described by a model, which considers that the cells grow on glucose and glutamine and their death is governed by lactate, ammonia and glutamine concentrations. The model equations are: ξ1 = −bl · Dξ1 + r1(ξ) − r2(ξ) 2) − ar1(ξ ) − r3(ξ ) Figure 2: Closed loop response
goal is achieved by a two-layer control structure as illus- 4 = −Dξ4 + cr1(ξ ) + dr3(ξ ) trated in Figure 1: i) a partial feedback linearizing controller is designed such that the inner loop has approximately a de- coupled linear dynamics; ii) two linear MPC controllers are 1, ξ2, ξ3, ξ4, ξ5 respectively represent the concen- trations of viable cells, glucose, glutamine, lactate and am- used in the outer loops to compute the inputs of the inner glutamine in the influent; D = F/V is the dilution/perfusion The robust performance of the closed loop is illustrated in rate and bl ∈ [0, 1] is the bleed ratio; ri(ξ ), i = 1, 2, 3 are Figure 2, where the controlled outputs follow their respec- reaction rates and a, b, c, d, e > 0 are the stoichiometric co- tive setpoints in spite of a 20% uncertainty on the maximum 3 Control structure and simulation results
References
The control goal is to manipulate the dilution rate D and M. Sbarciog and I. Saraiva and A. Vande Wouwer, Acceler- the bleed ratio bl (or bleed rate defined as Db = bl · D) such ating animal cell growth in perfusion mode by multivariable con- that the biomass concentration ξ1 and the glucose concen- trol: Simulation studies, Bioprocess and Biosystems Engineering, tration ξ2 follow their setpoints defined by ξ re f , ξ re f . This

Source: http://saas.ulb.ac.be/bm2013/papers/139.pdf