Modeling And Optimization Of Nosiheptide Fermentation Process

Fermentation is an industrial process that produces useful products by microbial growth and metalism. It plays an important role in national economy, such as chemical industry, agriculture, food industry and pharmaceutical industry. With the rapid development of fermentation industry, the demand to improve benefits of fermentation production and overall level of fermentation industry is becoming increasely urgent, while optimal control is an important way to realize the demand.Nosiheptide is a sulfur peptide antibiotic that can be produced by fermentation. It can be used as feed additives and has a broad market prospect, but the production and substrate conversion ratio is low currently. Taking nosiheptide fermentation process as the background, comprehensive and systematic research is made for modeling and optimization in nosiheptide fermentation process. Furthermore, fermentation modeling and optimal control system is designed in this dissertation. The main work is summarized as follows:Based on analysis of process characteristics and main factors in nosiheptide fermentation, mechanism model of nosiheptide batch fermentation is first established according to fermentation kinetics and material balance principles. The unknown parameters in the model are identified using differential evolution algorithm. Mechanism model of nosiheptide fed-batch fermentation, which lays a solid foundation for the further development of nosiheptide fermentation hybrid model, is then developed based on mechanism model of nosiheptide batch fermentation.Due to complexity of fermentation process, the mechanism model is developed after simplification to the process, so it is not precise enough. To overcome this disadvantage, a hybrid model for depicting nosiheptide fermentation process is developed, in which mechanistic modeling method and black modeling method are combined. The mechanism model is used to describe the basic knowledge, and neural network ensemble is used as black model to compensate for the error between the mechanism model and the practical process. Thus the disadvantages of the two kinds of modeling methods are complementary, and accuracy and generalization ability of the hybrid model are improved effectively. Elman neural network, a typical dynamic network, is used as individual network in the neural network ensemble, in order to reflect the dynamic characteristics in fermentation process better. The weight of each Elman network is decided by partial least squares regression. Furthermore, confidence bounds are used to analyze the uncertainty of the hybrid model.Fermentation optimization often has several objectives, so multi-objective methods are studied. Based on mulit-objective differential evolution algorithm, adaptive chaotic mulit-objective differential evolution algorithm(AC-DEMO) is proposed, combining with adaptive and chaotic principles. In the improved algorithm, chaotic initialization, adaptive mutation and chaotic migration operators are introduced. Numerical simulation results of test functions show that the algorithm has a good approximation and uniformity index and is suitable to solve complex multi-objective optimization problems.In nosiheptide process, the main problems are low production and low conversion ratio. So objectives of nosiheptide fermentation optimization are selected as maximum product quantity and minimum substrate consumption for high conversion ratio. Then decision variables and constraint conditions are decided. In the constraint conditions, model uncertainty restriction is specially considered. Using the developed hybrid model to describe nosiheptide fed-batch fermentation, the optimization problem is solved by AC-DEMO.On the basis of fermentation modeling and optimization study, modeling and optimal control system of nosiheptide fermentation process is designed. Hardware structure and data flow of the system is given and main function modules are analyzed emphatically. The system is realized by mixedly programming, combining merits of C#in interface design, strong data access ability of Oracle and powerful computation and simulation function of Matlab.Finally, future directions for research on modeling and optimization of fermentation process are discussed after summarizing the whole work in this thesis.