Joint Calibration Modeling Method For Multiple Components Of Fermentation Processes Based On Near-infrared Spectroscopy

The near-infrared spectroscopy technology takes the advantages of fast analysis speed,high measurement accuracy,and multiple measurement modes,which has been increasingly used in engineering fields of China and oversea countries.As a new energy source,bioethanol becomes more and more important for our world development nowadays.Biological fermentation is an important way to produce ethanol.However,the existing ethanol fermentation process mainly uses the offline measurement manner,which has the disadvantages of inaccuracy and long time consumption.Therefore,it is of great significance to realize online detection.This dissertation first introduces the main steps of applying the near-infrared spectroscopy to measure the component parameters of a biological fermentation process and the existing spectral pretreatment methods,including the classical modeling methods such as partial least squares(PLS)and partial robust M regression(PRM)methods,along with the model evaluation indices of correlation coefficient(R~2),root mean square error for prediction(RMSEP),ratio of performance to standard deviation(RPD).For online detection of substrate concentration,product concentration and biomass in the process of ethanol fermentation,a joint calibration modeling method is proposed for using the near-infrared spectroscopy.By analyzing the correlation between substrate concentration,product concentration and biomass during ethanol fermentation,a joint modeling method is given based on the coupling between components.In addition,the PRM method is also used to model each component of the ethanol fermentation process,respectively.Model evaluation indices are used to compare these two modeling methods.Experimental results show that,compared with the single-component modeling method,the joint modeling method can not only improve the modeling efficiency,but also obtain better prediction effect.For the glutamic acid fermentation process in the production of monosodium glutamate,an offline measurement scheme based on a near infrared spectroscopy and an online detection experiment platform for the production workshop are designed.Considering the coupling between the substrate glucose,product glutamic acid and yeast during the glutamic acid fermentation process,a joint calibration modeling method is proposed for spectral modeling,and a particle swarm optimization(PSO)algorithm is proposed for establishing the model parameter optimization method,which can significantly improve the model prediction accuracy.Experimental results verify the feasibility and advantages of the proposed method...