Recombinant Bacteria For Production Of Human Collagen Fermentation Process Modeling And Control

The process of microbial fermentation is of high nonlinear and time variation, whose rules are very complex. Some important process variables, such as biomass, substrate concentration, product concentration, etc., cannot be measured on line, which makes it more sophisticated to model and control the fermentation process and a satisfactory control performance cannot be achieved with the traditional control methods. With the development of the intelligence control, its application in fermentation process is increasingly attracting people's attention. The present paper studies the process modeling and controlling of recombinant Escherichia coli producing human-like collagen in fed-batch culture based on the theory of neural control system of intelligence control.Firstly, the paper, with fermentation process of recombinant E. coli producing human-like collagen as its subject of investigation, gives an comprehensive overview of the two neural network models, BP and RBF. After a detailed derivation of their working rules, the computer programming methods are given. To counter the performance deficiency of these models discovered during the study, three effective improved algorithms are given: fixed network parameters of BP, self-suitable network parameters of BP, and the ameliorated RBP. The paper then introduces these three new improved algorithms and gives the computer programming methods in detail. By means of MATLAB 7.0, comparison was made, between the performances of these three new improved algorithms, such as iterations, convergence time, etc. and those of other traditional methods. The results showed that these three new algorithms were effective.Secondly, after training the neural network using the experimental data of recombinant E. coli in fed-batch culture, the process models based on these three improved algorithms were established. These process models can be used to estimate and predict the state variables during fermentation process. They can also be used to calculate the change tendency of substrate concentration, product concentration andcell density with the initial conditions and control points, which is especially significant in guiding the practical work.Finally, based on the structure characters of the neural network, the nonlinear internal model controller was designed and its efficiency was testified. With the nonlinear internal model controller, fermentation process of recombinant E. coli producing human-like collagen is studied. When the practical conditions acted as the practical variables to control cell growth, the quality indexes of the correspondent control system were evaluated through experiments. The results showed that the internal model controller based on the neural network can be used in the process of recombinant E. coli in fed-batch culture to produce human-like collagen, and its proved to be fine in its robustness.