Modeling And Optimization For Fermentation Processes Based On Dynamic Metabolic Flux Analysis

The fermentation process is an important foundation for the industrialization of biotechnology.The modern fermentation engineering technology integrates genetic engineering and metabolic engineering,and has made great achievements in the fields of chemical industry and medicine.An accurate and effective process model is the key for control,optimization and monitoring.Nowadays,kinetic models or intelligent models can only roughly capture the macroscopic characteristics of fermentation process.Without analysis and utilization of the metabolism mechanisms,most existing models can not regulate the microelements by macro operation variables.This dissertation conducted a thorough research on fermentation metabolism network and proposed a fermentation process modeling and optimization strategy based on Dynamic Metabolic Flux Analysis(DMFA).The specific research contents are as follows:(1)Calculate flux distribution in metabolic networks.Metabolic fluxes determine the physiological features of cells.The DMFA method is employed to quantify the metabolic fluxes.Firstly,the flux equilibrium equation is established based on the unsteady characteristics.At the same time,the metabolic flux is written as a linear combination of a set of free fluxes.Then the DMFA moment is introduced to linearize the free fluxes in each time interval,and the metabolic flux function is obtained based on time series solved by least-squares equation.Finally,the DMFA method is validated by penicillin simulation and lysine laboratory scale experiment.(2)Develop a mapping model between macro operation variables and metabolic fluxes.Due to the nonlinear and multi-stage characteristics,a multi-model modeling strategy with double-level information is employed.First,the fermentation process is clustered into stages by the Gaussian mixture model(GMM)algorithm.Within each sub-model,a weighted partial least squares(WPLS)method is employed to update the model.Maximum likelihood method achieved by Expectation-Maximization(EM)is used to estimate the parameters of each sub-model,and the global output is obtained by fusing all the sub-models.Eventually,the penicillin simulation and lysine laboratory scale experiment are employed to demonstrate the effectiveness and universality of the proposed method.(3)Optimize the operation based on the macro-micro operation model.The models of fermentation processes used for optimization aim at the description of macroscopic mechanisms rather than metabolic mechanisms.Based on the mapping model developed above,this dissertation proposes a novelty optimization strategy by combining macro information with micro information.According to the metabolic requirements of energy,bacteria,products,byproducts and other substances in cell growth stage and product synthesis stage,the optimization objective is designed specifically,and the multi-objective particle swarm optimization(MOPSO)algorithm is introduced to solve the optimization problem.The application of the optimization strategy in penicillin fermentation illustrates its advantages.