(528g) Closed-Loop Optimal Control Operation of an Industry-Scale Bioreactor and Experimental Validation | AIChE

(528g) Closed-Loop Optimal Control Operation of an Industry-Scale Bioreactor and Experimental Validation

Authors 

Shah, P. - Presenter, Texas A&M University
Sheriff, M. Z., Purdue University
Bangi, M. S. F., Texas A&M University
Kwon, J., Texas A&M University
Kravaris, C., Texas A&M University
Botre, C., Texas A&M University
Hirota, J., Kaneka North America LLC
Chemical processes worldwide operate in a way that maximizes the profitability while minimizing the material costs. Currently, most industrial plants achieve their targets by manipulating the inputs heuristically. But in recent times, there has been a transition to industry 4.0 where tools like artificial intelligence, machine learning, and real-time optimal control are utilized to achieve efficient and improved performance [1]. As a case study, we have developed an online controller that determines the optimal operating conditions for an industrial-scale bioreactor while considering process constraints. Controlling a fermenter is challenging due to the high sensitivity of the micro-organisms to slight changes in inputs and the culture media [2]. Hence, a sophisticated model predictive control (MPC) scheme is needed that optimizes the operation of the fermenter, reaching the product and cost targets while also ensuring the safety of the process by considering plant constraints [3].

Optimal control of any process is highly dependent on its accurate modeling. For this case study, we utilize a recently developed hybrid model for a biochemical fermenter [4]. This model reasonably predicts the process states like biomass and substrates in both batch and fed-batch operations. A multi-objective control problem is solved, which maximizes the product amount, minimizes the operating cost, and maintains the substrate at a fixed concentration. The optimization problem is set up in GAMS, and a framework is developed that links it to the graphical user interface (GUI), where the operator can feed real-time values [5]. This shrinking horizon MPC is re-initialized with the actual plant values of the internal states and inputs (during abnormal operation) through this framework. The developed algorithm effectively determined the optimal input trajectories of temperature and flow rates while ensuring that the bounds for the inputs and their rate of change are maintained. This algorithm uses the hybrid model for the state predictions for the corresponding manipulated inputs and was shown to be within the normal operating range. This approach was then validated using multiple batches of experimental data from an industry-scale bioreactor, and an overall 8% increase in productivity was achieved. The closed-loop operation of this framework helps in re-initializing the targets at every instant, and the MPC runs after every sampling time, providing the operator with the next set of inputs to be applied to the bioreactor.

The developed MPC algorithm is computationally inexpensive, accounts for batch-to-batch variability, and can improve the profitability of the process by determining the optimal operating conditions. A framework is also developed that links the operator to the machine using a GUI, helping in real-time monitoring of the process. The control algorithm is implemented on an industrial-scale bioreactor, and the performance of the proposed framework is demonstrated.

References:

  1. Sansana, J., Joswiak, M. N., Castillo, I., Wang, Z., Rendall, R., Chiang, L. H., & Reis, M. S. (2021). Recent trends on hybrid modeling for Industry 4.0. Computers & Chemical Engineering, 151, 107365.
  2. Petersen, L. N., & Jørgensen, J. B. (2014, June). Real-time economic optimization for a fermentation process using model predictive control. In 2014 European Control Conference (ECC) (pp. 1831-1836). IEEE.
  3. Nagy, Z. K., & Braatz, R. D. (2003). Robust nonlinear model predictive control of batch processes. AIChE Journal, 49(7), 1776-1786.
  4. Shah, P., Sheriff, M. Z., Bangi, M. S. F., Kravaris, C., Kwon, J. S. I., Botre, C., & Hirota, J. (2022). Deep neural network-based hybrid modeling and experimental validation for an industry-scale fermentation process: Identification of time-varying dependencies among parameters. Chemical Engineering Journal, 135643.
  5. Ferris, M. C., Jain, R., & Dirkse, S. (2011). Gdxmrw: Interfacing gams and matlab. Online: http://www. gams. com/dd/docs/tools/gdxmrw. pdf.