Research On Soft-Sensing Methods For Penicillin Fermentation Process

Fermentation process is a non-linear process, involving hundreds of physical processes and chemical reaction. Fermentation process is a non-linear process, involving hundreds of physical processes and chemical reaction. Because of technical or economic reasons, some of the biochemical parameters in fermentation process, such as bacteria concentration, substrate concentration, and production concentration are hard to online measurement, which bring a lot of obstacles to the key quality parameters of real-time monitoring and direct quality control. The soft measurement is the solution of this problem. With the background of penicillin fermentation process, this thesis in-depth study soft sensor modeling and related methods for bacteria. The main researches are the following several aspects:(1) In the deep understanding of fermentation process principle and soft-sensing methods, for the difficulty online measurement of bacteria in penicillin fermentation process, soft-sensing methods based on fuzzy least squares support vector machine is put forward, to on-line estimate bacteria concentrations. Simulation results show the effectiveness of strong antinomies ability.(2) Aiming at to the stresses of penicillin fermentation process, the method based on fuzzy clustering algorithm is put forward, to divide the phases of stable operation for penicillin fermentation process, then establish corresponding model base on each phase of the soft measurement.(3) Aiming at the problems that initial conditions is changed, the values of biochemical parameters (biomass concentration, substrate concentration and product concentration) also change, soft-sensing correction method based on the deviation of initial conditions is presented, let soft-sensing model has strong adaptability to the change of initial conditions for fermentation.Simulation results show that soft-sensing modeling method base on the stage division and fuzzy least squares support vector machine has higher accuracy precision relative to other models.