Study On On-line Monitoring Technology Of Yogurt Fermentation Quality Parameters Based On Dielectric Properties

Yogurt is a food with a long history,good nutrition and health care function.With the increase of social demand,the output of yogurt is increasing year by year,and the production of yogurt tends to be automated.In the automated production process,in order to achieve high consistency and reproducibility of the product,it is necessary to online monitor the quality parameters of the yogurt fermentation process to achieve online monitoring of the yogurt fermentation process.However,the current online monitoring technology is limited to the judgment of fermentation end point,and there is little online monitoring of yogurt quality.Therefore,it is of great significance to study the online monitoring technology that can simultaneously realize yoghurt quality detection and fermentation endpoint judgment.In this paper,yoghurt was the research object.Firstly,the dielectric properties of yoghurt fermentation between 1 MHz and 3 GHz were studied.Then,the dielectric spectrum of yoghurt fermentation process was determined by stoichiometry: The sample weres divided by joint X-Y distance(SPXY),Savitzky-Golay convolution smoothing(SG smoothing)method and standard normal transformation(SNV)method were used to preprocess the spectrum.After the feature variables of full spectrum(FS)were extracted by principal component analysis(PCA)method a nd successive projection algorithm(SPA)method,the partial least squares(PLS),support vector machine(SVM)and extreme learning machine(ELM)models were established to mobitor yogurt p H,major component content and texture characteristics.Then,the performance of the model was analyzed with different modeling processes and the optimal model was selected.The research results are as follows:(1)With the increased of frequency,the dielectric constant ε′ and dielectric loss factorε″ of yoghurt were decreasing;with the prolongation of fermentation time,the dielectric constant ε′ of yoghurt had no obvious regularity,but dielectric loss factor ε″ was gradually increasing.(2)The SPA-SVM model had the best monitoring effect on the p H of yogurt,with Rc and Rp were 0.9835 and 0.9939,respectively,and RMSEC,RMSEP were 0.1305 and0.0756,respectively.(3)The FS-SVM model had the best monitoring effect on fat,protein and lactic acid content in yogurt.The model for monitoring fat content had Rc and Rp were 0.9386 and0.8445,respectively,and RMSEC,RMSEP were 0.0730 and 0.0610,respectively.The model for monitoring protein content had Rc and Rp were 0.9396 and 0.8702,respectively,and RMSEC,RMSEP were 0.0578 and 0.0446,respectively.The Rc and Rp of the model for monitoring lactic acid content were 0.9899 and 0.9861,respectively,and the RMSEC and RMSEP were 0.0124 and 0.0110,respectively.None of the models selected in this paper can be used to monitor non-fat milk solids and total solids content in yogurt.(4)The PCA-SVM model was more suitable for monitoring the firmness of yogurt,with Rc and Rp of 0.9287 and 0.9637,respectively,and RMSEC and RMSEP of 23.7503 and 17.8438,respectively.The SPA-SVM model was more suitable for monitoring the consistency,cohesiveness and viscosity of yogurt.The Rc and Rp of the model for monitoring the consistency of yogurt were 0.9709 and 0.9850,respectively,and the RMSEC and RMSEP were 150.8626 and 121.3996,respectively.The Rc and Rp of the model for monitoring the cohesiveness of yogurt were 0.9115 and 0.9708,respectively,and RMSEC and RMSEP were 18.0369 and 16.4670,respectively.The Rc and Rp of the model for monitoring yogurt viscosity were 0.8901 and 0.9757,respectively,and RMSEC,RMSEP were 18.6235 and 10.1780,respectively.