Integrated real-time optimization and control of multiunit batch processes
Francesco Rossi,a,b,* Daniel Casas Orozco,c Gintaras Reklaitis,a Flavio Manentib
a Purdue University, School of Chemical Engineering, Forney Hall of Chemical Engineering, 480 Stadium Mall Drive, West Lafayette, IN 47907-2100, United States
b Politecnico di Milano, Dipartimento Chimica, Materiali e Ingegneria Chimica “Giulio Natta”, Piazza Leonardo da Vinci 32, Milano 20123, Italy
c Universidad de Antioquia UdeA, Engineering Faculty, Chemical Engineering Department, Environmental Catalysis Research Group, Calle 70 No. 52-21, Medellín, Colombia
Despite the historical tendency for conversion of batch operations to their continuous processing equivalents, batch manufacturing remains in widespread use in a number of CPI sectors, such as pharmaceuticals and fine chemicals, fermentations and bio-chemical products, and special polymers. Indeed batch processing remains important on the industrial landscape and is likely to remain significant for the foreseeable future. Many processes operating in the batch mode produce high added-value products, are characterized by intrinsic non-ideal thermodynamic behavior and exhibit strongly non-linear dynamics, features which provide challenges for design and in operation. To address these challenges, there continues to be an ongoing body of research seeking safe and economical ways to design, optimize and control batch operations.
Several papers in the literature describe how to (efficiently) apply model-based online optimization and control strategies to batch units, but typically the focus is on a single process unit: a single reactor (Smets et al., 2004), distillation column (Greaves et al., 2003), crystallizer (Acevedo et al., 2015) or more specialized, unconventional equipment. There is limited literature on approaches to simultaneously consider multiple units or even an entire batch plant. For this reason, the authors of this contribution propose the development of integrated online optimization and optimal control methods for treating an entire integrated batch production line.
Specifically, we propose to: (I) model the entire batch process as a whole, also including design choices as independent variables (activation/deactivation of entire units, unit feeds & draws, unit to unit internal material/thermal streams, etc.); (II) specify a suitable performance function for the entire plant (either economic or performance based); and (III) apply an online optimization & control algorithm to the framework described in (I) and (II).
Depending on the perspective taken, the methodology proposed can be considered either as an online batch process design method or as an online model-based optimal control/optimization strategy. From a design perspective, the approach dynamically and optimally allows: (I) deciding when to stop/activate a unit and how long to keep it inactive; (II) choosing how to manage interconnections among units; (III) selecting optimal values for product draws and feeds (for fed-batch units only); (IV) determining the batch cycles in which the multiunit batch process will operate. From an operational perspective the online optimization/optimal control approach should promote: (I) the effective compensation of disturbances at the entire plant scale; (II) the selection of the optimal production levels based on the market demand; etc. All of these aspects need to be optimized together, accounting for their interactions. Therefore, the proposed integrated real-time optimal design and optimization/optimal control strategy seeks to guarantee improved performance compared to the conventional sequential approach, which combines offline plant optimal design with individual units online control/management developed separately.
The methodology discussed is validated and tested on a novel batch process for the production of β-pinene (Bauer et al., 1988), which includes a (catalytic) batch reaction followed by filtration and distillation steps. It is applied to all the batch plant units simultaneously and its performance is compared to the conventional sequential (offline optimal design & online optimization/optimal control) approach. Based on the cases presented, lessons will be drawn regarding the situations in which the simultaneous and sequential approaches have advantages.
Acevedo, D., T. Yanssen, and Z.K. Nagy. "Multiobjective Optimization of an Unseeded Batch Cooling Crystallizer for Shape and Size Manipulation." Industrial & Engineering Chemistry Research 54 (2015): 2156-2166.
Greaves, M.A., I.M. Mujtaba, M. Barolo, A. Trotta, and M.A. Hussain. "Neural-network approach to dynamic optimization of batch distillation-Application to a middle-vessel column." Chemical Engineering Research and Design 81 (2003): 393-401.
Smets, I.Y., J.E. Claes, E.J. November, G.P. Bastin, and J.F. Van Impe. "Optimal adaptive control of (bio) chemical reactors: past, present and future." Journal of process control 14 (2004): 795-805.
Bauer, K., D. Garbe, and H. Surburg. "Ullmann’s encyclopedia of industrial chemistry." Ullmann's Encyclopedia of Industrial Chemistry 11 (1988).