Soft Sensing And Decoupling Control For Marine Alkaline Protease MP Fermentation Process

With the development of marine biotechnology,the role of marine microbial fermentation engineering is increasingly important in social life.Marine microorganism has strong cold resistance,pressure resistance and alkali resistance due to its long-term survival in the harsh environment of low temperature,high pressure and high salt.Therefore,the produced enzyme preparation has more prominent characteristics and advantages than the ordinary.In the actual fermentation process,the fermentation process needs to be controlled in order to obtain a higher yield and quality of enzyme preparation to get the best economic benefits.However,the mechanism of marine microorganism fermentation is complex,and each fermentation process parameter has high nonlinearity,time-delay and strong coupling.Particularly,some key parameters in the fermentation process(such as: bacterial concentration,substrate concentration,relative enzyme activity)cannot be directly detected online by hardware instruments,which seriously restricts the development of marine biological enzymes in automation and industrialization.Therefore,researching the on-line detection and decoupling control of marine microbial fermentation process can effectively promote the large-scale production of marine enzyme fermentation industry,which has a good theoretical significance and social application value.For this purpose,under the support of the National Natural Science Foundation of China(41376175)and Jiangsu University Superiority Discipline Construction Project(PAPD),the paper takes a new marine alkaline protease called MP fermentation process as this research object,and studied the soft-sensor and decoupling control methods for the fermentation process.Firstly,on the basis of summarizing various soft-sensor modeling methods,a method for marine alkaline protease MP fermentation process based on improved particle swarm optimization-radial basis neural network(PSO-RBFNN)is proposed,which provides a foundation for further decoupling and optimization control of nonlinear multivariable systems.Secondly,in order to achieve the decoupling and optimal control of multivariable nonlinear systems,a composite control combining decoupled control theory and internal model control is proposed based on the analysis of various nonlinear multivariable decoupling methods,and using the RBFNN to approximate the ?-order integral inverse system of continuous nonlinear systems.Finally,the inverse system is connected in series with the original nonlinear system to achieve the linear decoupling of the original fermentation system,and combines with the internal model control method for optimal control,which improves the robustness and anti-interference ability of the system.The experimental simulation results show that the method has achieved good results.The main research contents of this dissertation are as follows:Firstly,to solve the problem that the key parameters(bacterial concentration,substrate concentration,relative enzyme activity)in the process of marine protease MP fermentation are difficult to detect in real time,a soft parameter modeling method of key parameters based on RBFNN is proposed.Based on the analysis of the modeling principles and parameter learning methods of RBFNN,a design method of optimizing RBFNN parameters using PSO algorithm is proposed.Then,the inertia weight selection method of the PSO algorithm is improved by adopting the Exponential decreasing inertia weight(EDIW),and the improved RBFNN model optimized by PSO algorithm is applied to predict the key parameters of marine alkaline protease MP fermentation process.The experimental results show that the training time of the improved soft-sensor model is shortened by about 40%,and the model prediction accuracy is improved by more than 3%.Secondly,in order to solve the problem that the fermentation system is difficult to control in the actual industry due to its own time-variation,coupling,non-linearity and other characteristics in the process of marine alkaline protease MP fermentation,a decoupling method based on the RBFNN inverse system is proposed.Based on the simplified kinetic model of the marine alkaline protease MP fermentation process,the inverse model is solved according to the reversible analysis of the process model.The RBFNN identification method is used to obtain the inverse process system,and the original system and the inverse system are connected in series to form a complex pseudo-linear system,which realizes the fermentation process decoupled into several single-input single-output(SISO)first-order linear integration subsystems.In order to further introduce the internal model control strategy,the closed-loop control of the composite pseudo-linear subsystem is established,which makes the whole system have strong robustness and good anti-interference ability.Thirdly,On the basis of the multivariate decoupling of RBFNN inverse system,an internal model control method was used to implement closed-loop control of marine alkaline protease MP fermentation process.The simulation diagrams of the decoupled internal model control structure system for the fermentation process are constructed by MATLAB,which proves that the decoupling control method is feasible and has good control effect.