Optimization Of Biochemical Processes Based On Design Of Dynamic Experiments

Nowadays,model-based optimization is the predominant methodology for optimization of the batch biochemical process.The model is at the heart of such approaches.The process mechanism of many actual biochemical processes is unclear or the process is too complex,which makes the establishment and optimization of mechanism model very difficult.And the data-driven model established based on historical data may not include the optimal operating condition or expected operating condition.The purposeful design of experiments(Do E)around preset operating conditions to achieve modeling and optimization of biochemical processes is proposed in this paper.However,traditional Do E cannot deal with the design of time-varying manipulated variables,and there is no effective way to model batch response of the process for optimization.Therefore,based on the design of dynamic experiments(Do DE),combined with the Kriging model and functional principal component analysis(FPCA),targeted sequential modeling optimization algorithms is proposed to model and optimize the batch biochemical processes.The main work is summarized as follows:(1)Aiming at the optimization of feeding process of fed-batch ethanol production,the Do DE based on the Kriging model is given to establish a prediction model between ethanol production and the feeding process.The iterative updating and optimization algorithm of Kriging model is proposed by combining sequential design and Mega-trend-diffusion technique.The algorithm first establishes a global model that meets the accuracy requirements to quickly determine the optimal solution region and then establishes a locally enhanced global model to find the optimal solution.The algorithm is iteratively executed until the optimal solution meets the convergence conditions.The simulation results verify the effectiveness of this method.(2)For batch response modeling of batch biochemical process,a sequential modeling method based on the FPCA is considered.To improve the prediction accuracy,a sequential modeling algorithm of FPCA is proposed.According to the improved convergence conditions,the algorithm adopts the sequential design of the maximum mean square error criterion to update the model iteratively.The effectiveness of the proposed method is verified by modeling the batch concentrations of substances in the two batch biochemical processes of the simulated reaction network and fed-batch penicillin fermentation.The simulation results show that the model built has good data visualization and model interpretation capabilities.(3)In the absence of batch response data,conditional expectation method is used to implement FPCA to model and optimize batch response of biochemical processes.The sequential modeling algorithm is used to obtain the process model meeting the accuracy requirements,and the optimization proposition is constructed based on the model and solved.Through the experimental simulation of a three-reaction biochemical network,the effectiveness of the proposed method for establishing a process batch response model in the absence of data is verified,and the feasibility of the optimization method based on the FPCA model is verified.