This book consists of eight chapters. Chapter 2 demonstrates the optimization of a fed-batch culture of hybridoma cells using a GA. The optimal feed rate profiles for single feed stream and multiple feed streams are determined via the real-valued GA. The results are compared with the optimal constant feed rate profile. The effect of different subdivision number of the feed rate on the final product is also investigated. Moreover, a comparison between the GA and DP method is made to provide evidence that the GA is more powerful for solving global optimization problems of complex bioprocesses.
Chapter 3 covers the on-line identification and optimization for a high productivity fed-batch culture of hybridoma cells. A series of GAs are employed to identify the fermentation's parameters for a seventh-order nonlinear model and to optimize the feed rate profile. The on-line procedure is divided into three stages: Firstly, a GA is used for identifying the unknown parameters of the model. Secondly, the best feed rate control profiles of glucose and glu-tamine are found using a GA based on the estimated parameters. Finally, the bioreactor was driven under control of the optimal feed flow rates. The results are compared to those obtained whereby all the parameters are assumed to be known. This chapter shows how GAs can be used to cope with the variation of model parameters from batch to batch.
Chapter 4 develops an on-line neural softsensor for detecting biomass concentration, which is one of the key state variables used in the control and optimization of bioprocesses. This chapter assesses the suitability of using RNNs for on-line biomass estimation in fed-batch fermentation processes. The proposed neural network sensor only requires the DO, feed rate and volume to be measured. Based on a simulated fermentation model, the neural network topology was selected. The prediction ability of the proposed softsensor is further investigated by applying it to a laboratory fermentor. The experimental results are presented, and how the feedback delays affect the prediction accuracy is discussed.
Chapter 5 is devoted to the modelling and optimization of a fed-batch fermentation system using a cascade RNN model and a modified GA. The complex nonlinear relationship between manipulated feed rate and biomass product is described by cascading two softsensors developed in Chapter 4. The feasibility of the proposed neural network model is tested through the optimization procedure using the modified GA, which provides a mechanism to smooth feed rate profiles, whilst the optimal property is still maintained. The optimal feeding trajectories obtained based both on the mechanistic model and the neural network model, and their corresponding yields, are compared to reveal the competence of the proposed neural model.
Chapter 6 details the experimental investigation of the proposed cascade dynamic neural network model by a bench-scale fed-batch fermentation of Saccharomyces cerevisiae. A small database is built by collecting data from nine experiments with different feed rate profiles. For a comparison, two neural models and one kinetic model are presented to capture the dynamics of the fed-batch culture. The neural network models are identified through the training and cross validation, while the kinetic model is identified using a GA. Data processing methods are used to improve the robustness of the dynamic neural network model to achieve a closer representation of the process in the presence of varying feed rates. The experimental procedure is also highlighted in this chapter.
Chapter 7 presents the design and implementation of optimal control of fed-batch fermentation processes using a GA based on cascade dynamic neural models and the kinetic model. To achieve fast convergence as well as a global solution, novel constraint handling and incremental feed rate subdivision techniques are proposed. The results of experiments based on different process models are compared, and an intensive discussion on error, convergence and running time are also given.
The general conclusions and thoughts for future research in the area of intelligent biotechnological process control are presented in Chapter 8.
Was this article helpful?