The Study On The Process Control Of ε-poly-L-lysine Fermentation Based On Neural Network Software Sensor

The process of microorganism fermentation bears the characteristic of nonlining, changability and posteriority, the inner mechanism of which is very complexed. The traditional method cannot measure those important parameters in the process of fermentation on line, which makes it very difficult to control the whole process of fermentation optimizationally.To solve this problem, the thesis, taking the e-poly-lysine as the study subject and the theory of software sensor as the study basis, puts forward the method of neural network software sensor aiming to measure the key biomass in the biochemical process. The key of the method lies in that it takes the structure of the BP neural network as the way to realize the sortware sensor, which will prepare for the later pre-estimation of the parameters. In addition, by improving the standard BP neural network algorithm, the thesis puts forward a kind of BP learning algorithm which is based on the method of golden section and then takes this algorithm to pre-estimate the key variables (the concentration of residual sugar and thallus) that can't be measured directly during the process of the e- poly-L-lysine fermentation. Meanwhile, the net test also testifies the feasiblity and effectiveness of the method, and the good astringency and universality of the network topology. The study provides a new way for the on-line measurement of the key biomass in other process control.On this basis, this thesis designs a digital control system for the biofermentation to realize the real-time control of the main parameters in the process ofe-poly-L-lysine fermentation.