Modeling of Fermentation Systems and Its Applications
| Autor: | Kewei Zuo, 左克偉 |
|---|---|
| Rok vydání: | 2001 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 89 The main purposes of this study are to establish a hybrid modeling procedure of fermentation systems and to investigate its applications. It is desired to integrate scattered experimental data into useful information with less efforts and lower costs. An HNN model possesses advantages of both analytical and black-box models. The modeling procedure is standardized and simple, and shows a good capacity to describe complex systems. In addition, an HNN model can be utilized to explain the behavior of a fermentation system. Therefore, it is an excellent modeling tool. Besides, the analytical part of an HNN model can be modified according to the actual conditions. The modeling method is flexible. However, no investigation has been reported to modify an existing HNN model to describe systems with different operating types, environmental conditions or reactors. In this study, we proposed the application of HNN models to simulate fermentation systems with different operating types. In the beginning, an ANN is established using the batch experimental data. After that, the system equations for a fed-batch or continuous system are utilized as the analytical part and combined with the ANN to give an HNN model. The HNN model can be employed to describe a fed-batch or continuous fermentation system according to the system equations. Phase plane plots are frequently adopted to implement stability analyses of fermentation systems. However, quite a few continuous runs are required to make a phase plane plot and the costs are very high. In this study, batch runs are carried out to make an HNN model and the HNN model can be modified to simulate continuous runs. Several simulated runs can be obtained easily to make an accurate phase plane plot. By using the proposed method, the costs of the stability analyses are reduced substantially. On the other hand, the effects of imperfect mixing in a bioreactor on the environmental conditions are also considered. The imperfect mixing often leads to pH / temperature / DO deviation. It will be useful to predict the state variation due to the environmental change. Two HNN models in company with the model fusion method predict the state variation very well when a pH pulse occurs. It is possible to improve the productivity of a fed-batch fermentation system by adjusting the feeding strategy. The conventional way is to carry out an off-line optimization and then follow the optimum feeding profile during the cultivation. Nevertheless, this method will be less effective when model errors or disturbances occur. Hence, we propose a semi-realtime optimization procedure to solve the problem. The semi-realtime optimization carries out a re-optimization procedure every one hour to determine new set-points. The new set-points are then employed to modify the feeding rate. The results are outstanding. A fermentation system in an imperfect mixing reactor can be divided into the biological and hydrodynamic parts. An ANN is obtained from the experimental data of a cultivation in a stirred tank reactor. A tanks-in-series model is employed to describe the mixing properties of the fermentation system. The parameters are determined by a tracer response method. A modification according to the KLa is also employed. By combining these parts, a hybrid model is obtained to describe the fermentation system in a modified airlift reactor for bacterial cellulose production. Due to the open structure and standardized modeling procedure of the HNN model, it is suitable for the modeling purposes on industrial production. The applications illustrated in this study form a template which is simple and flexible. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|