Loading...

Dynamic Modeling of a Continuous Bioreactor Using Hybrid Neural Network

Rousta, Hamed | 2021

265 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54226 (06)
  4. University: Sharif University of Technology
  5. Department: Chemical and Petroleum Engineering
  6. Advisor(s): Pishvaie, Mahmoud Reza
  7. Abstract:
  8. Dynamic modeling and control of bioprocesses, especially bioreactor and fermenter equipment, has always faced many challenges due to their complexity and high uncertainty. One scientific approach to modelling of such complex systems is the use of neural networks. Recently, an approach called hybrid networks has become popular, whose structure incorporates some of the basic rules, such as the mass balance of reactive species. Practically, it can also be used as a ‘software sensor’ or non-measurable state estimator. In this project, a bread yeast production bioreactor is used as a reference process and a hybrid neural network is used to estimate unknown parameters of biochemical reactions. In the current work, it was attempted to design different estimators for estimation of inaccessible states. In this regard, three types of observers were designed, Luenberger-Like, Kalman filter and neural network. Considering the importance of ethanol concentration and oxygen concentration in the above-mentioned process, a model predictive control was adopted to control them. The designed model used is formulated as a state space. Also, according to the constraints and limitations of the process, the issue has been considered and studied in both restricted and unrestricted forms. Also, the objective function is formulated as a quadratic objective function to speed up the controller calculations, and quadratic programming is used for the constrained model. Finally, the performance response of this controller was compared with the proportional integral (PI) controller and it was observed that in all states, the MPC controller has much better performance. The MPC controller was also used with various forms of observer and state feedback. The results show that not only is the response of this controller to the hybrid neural network observer better than other observers, but it is also very close to answering with state feedback
  9. Keywords:
  10. Fermenters ; Model Predictive Control ; Luenberger-Like Estimator ; Bioreactor Simulation ; Hybrid Neural Network ; Leunberger-Like Observer

 Digital Object List

 Bookmark

No TOC