Modeling And Optimization Of Agrobacterium Fermentation Process

Microbial polysaccharides are bulk fermentation products that have important applications in food field and industrial production.Under the background of increasing demand on reducing energy and material consumption in China,as well as,the increasing internationalization marketing competition of microbial polysaccharide,how to improve the fermentation efficiency of microbial polysaccharides by optimizing fermentation process to achieve energy saving,emission reduction and synergy,and ultimately improve the competitiveness of domestic microbial polysaccharide production industry in the international market has become an urgent issue for experts in the field of computer and control engineering.Because microbial polysaccharides fermentation is a high oxygen consumption process,the Dissolved Oxygen Tension(DOT)is an important controlling parameter for microbial polysaccharides production process.Therefore,how to establish a fermentation process kinetic model based on multi-batch fermentation data,and realize the optimal control of DOT in the fermentation process has become the key to strengthen the production of microbial polysaccharides.Establishing an accurate fermentation kinetic model is an important prerequisite and the basis for fermentation control system design,adaptive and self-learning control,and modeling simulation.Model parameter identification algorithm is one of the important means to obtain accurate fermentation kinetics models.Therefore,establishment of a precise fermentation dynamic model,and selection of a better parameter identification algorithm according to the characteristics of the microbial polysaccharides fermentation model is needed,which is expected to reduce the error between the model prediction and the experimental data.Under the above background,this thesis took fermentation process of curdlan produced by Agrobacterium sp.ATCC31749 as an example.Firstly,an Improved Bird Swarm Algorithm(IBSA)with better performance in convergence speed,optimization precision,and stability was developed.And then,a hybrid model based on the combination of the Least Squares Support Vector Machine(LSSVM)and the mechanism model was established.The developed IBSA in this thesis was used to estimate the parameters of the hybrid model.Then,the optimal dissolved oxygen control strategy was obtained.Finally,a fermentation control system was designed and the optimized dissolved oxygen control model was applied for the fermentation process of Agrobacterium sp.ATCC31749 to verify the developed model.Detail research contents are listed as follows:(1)To overcome the weakness of slow convergence rate,local optimum and poor overall stability of the Bird Swarm Algorithm(BSA)in solving high-dimensional nonlinear complex functions,an Improved Bird Swarm Algorithm(IBSA)was developed.Learning coefficients in the original algorithm were adjusted by a nonlinear function,and the update formula of the bird position was replaced by the Levi flight formula when the birds maintain vigilance and attempt to move to the center of the population.Besides,in the optimization process,the BSA was improved by adding chaotic perturbation to the optimal solution and using the Simulated Annealing Algorithm(SAA)to re-optimize when the optimal solution of the algorithm remains unchanged.The experimental results show that the IBSA is better than the BSA,the Modified Differential Evolution incorporating Cauchy Disturbance(CDMDE)algorithm and the improved Particle Swarm Optimization(PSO-A)in terms of convergence speed,optimization precision,and stability.(2)The IBSA was used to solve the parameters’ values of the fermentation process dynamics model of Agrobacterium sp.ATCC31749,and compared with the group intelligence algorithms such as BSA,CDMDE,and PSO-A.The experimental results show that the IBSA has significant advantages in optimization accuracy and stability,and the IBSA is more suitable for identifying the parameters of the ATCC31749 fermentation process dynamics model.(3)To resolve the weakness of the computational accuracy of the traditional Agrobacterium sp.ATCC31749 fermentation kinetic model,an LSSVM algorithm was developed.This thesis optimized the LSSVM algorithm by adding fuzzy weighting idea and the mixed kernel function method,and then used the optimized LSSVM algorithm to solve the fermentation process dynamics model of Agrobacterium sp.ATCC31749.Firstly,the kinetic model of Agrobacterium.ATCC31749 fermentation process was established,and then the experimental data of DOT of 15%,30%,45%,60%,and 75% were used for training the model,and the IBSA was used to find the best values of parameters under different DOT.And then the relationship between DOT and model parameters was analyzed.Following,a model for predicting the polysaccharide concentration from DOT was established.Finally,the hybrid model and IBSA was applied to optimize simulate the Agrobacterium fermentation for maximizing the concentration of polysaccharide products,and the optimal DOT control curve was acquired.(4)A system which can acquire and control the parameters’ values for the fermentation process was developed and used to verify the effectiveness of the optimized dissolved oxygen control model.The optimal dissolved oxygen control curve was used as control strategy in controlling DOT in the experiment of Agrobacterium sp.ATCC31749 fermentation.The curdlan production arrived at 46g/L,which is higher than 38.7g/L in 45% DOT.The result verifies the effectiveness of the model and the optimal DOT control curve established in this thesis.