Soft Sensor Modeling Based On Machine Learning Algorithm For Marine Protease Fermentation Process

According to the unique characteristics of the marine protease fermentation process,it is widely used in washing industry,environmental protection,food processing,and medical engineering.However,the fermentation process is a highly nonlinear,time-varying,multivariable,and strongly coupled complex biochemical reaction process.Due to the growth and reproduction of living organisms,the internal mechanism is very complicated.Some key variables(such as cell concentration,matrix concentration,and enzyme activity)that directly reflect the quality of the fermentation process are difficult to measure in real-time by traditional measurement methods.There is a large time lag in off-line assay and analysis,which cannot meet the needs of on-site real-time optimization control.Especially in the actual fermentation process,to increase the production efficiency and product quality of enzyme preparation and reduce the economic cost,the environmental variables of the fermentation process should be controlled in a specific range.However,the process of cell growth,reproduction,and metabolic enzyme production is extremely reflected by the external environment.There is a complex nonlinear dynamic relationship among various state variables of the fermentation process,which is difficult to decouple,and the key variables reflecting the quality of fermentation have serious defects in measurement stability and price,which has become a bottleneck problem that affects the real-time optimization control of the fermentation process.The use of soft sensor modeling is an effective way to solve the above-mentioned measurement problems.Soft sensor modeling can organically combine production process knowledge with automatic control theory,and infer the process variables to be measured through mathematical calculations and estimation methods.The use of soft sensor methods to analyze key process variables not only reduces costs but also speeds up dynamic response and enables real-time estimation of key variables.One powerful tool to achieve this is soft sensors that use online signals from the process to derive new information.New implementation of sensors and measuring techniques could be very useful,but the soft sensor should preferably utilize robust sensors already implemented in the process.Soft sensors are not limited to use a few dedicated hardware sensors but can use different sensors that are best suited for the process to deliver estimates of key variables.This thesis mainly includes the following research contents.Firstly,this thesis introduces the purpose and significance of research on marine protease,and points out that it is difficult to measure the quality variables in the process of fermentation.Secondly,it analyzes various soft sensor modeling methods at home and abroad,and summarizes the corresponding advantages and disadvantages.A comprehensive review of existing data pre-processing approaches,variable selection methods,soft sensor modeling methods and optimization techniques was carried out.A comprehensive analysis of various soft sensor models is presented in tabular form which highlights the important methods used in the field of fermentation.Then the fermentation process of marine protease is analyzed,the specific liquid fermentation process is detailed,the influence of process variables on fermentation and the detection methods of fermentation process variables is described,which is conducive to the selection of auxiliary variables and the effective collection of sample data,so as to prepare for the establishment of the model.Thirdly,the principle of soft sensor technology is introduced and the novel soft sensor model based on a support vector regression(SVR)was proposed in this thesis.To further improve the model’s prediction accuracy,the Grey Wolf Optimization(GWO)algorithm is used to optimize the critical parameters(kernel function width σ,penalty factor c)of SVR model.To study the influence of selecting auxiliary variables on soft sensor modeling,the Successive Projection Algorithm(SPA)is used to determine the characteristic variables and compare them with Grey Relation Analysis(GRA)algorithm.Finally,the Excel spreadsheet data was called by MATLAB programming,and the established SPA-GWO-SVR soft sensor model predicted crucial biological variables.The simulation results show that the SPA-GWOSVR model has higher prediction accuracy and generalization ability than the traditional SPASVR model.The real-time monitoring was processed by MATLAB software for the marine protease fermentation process,which met the requirements of optimal control of the marine protease fermentation process.Fourthly,another novel soft sensor model based on the least square support vector machine(LSSVM)was proposed.The hyper-parameters of LSSVM model were optimized by using the improved version of the particle swarm optimization(IPSO)algorithm.This model uses the exponential decreasing inertia weight(EDIW)strategy to improve the search quality of the standard particle swarm optimization(PSO)algorithm.The primary purpose of this improvement is to experimentally establish the fact that the PSO algorithm under the EDIW strategy is very much efficient if its parameters are appropriately set.The fuzzy c-means clustering(FCM)algorithm is used to cluster the sample data,and the fermentation process is divided into three stages.The IPSO-LSSVM is used to establish the corresponding local submodels for these three stages,and then the sub-models are combined to obtain the final multistage soft sensor model.Through MATLAB simulations,compared with the PSO-LSSVM model,the multi-stage soft sensor model based on FCM and IPSO-LSSVM have high prediction accuracy and stability,which provides a reference method for solving variables that are difficult-to-measure in the same type of industrial production process.Finally,the main research contents and results of this dissertation are summarized,and the deficiencies in the dissertation and the improvement direction of related research in the future are proposed.