Study On Monitoring Method Of Submerged Fermentation Of Tremella Aurantialba Based On Electronic Nose And Electronic Tongue Techniques

Tremella aurantialba has a long edible and medicinal history. Polysaccharide in Tremella aurantialba has a variety of pharmacological effects such as hypoglycemic action and improving immunity, etc. With the continuous development of submerged fermentation technology, submerged fermentation technology have been widely used in large-scale cultivation of Tremella aurantialba. The submerged fermentation of Tremella aurantialba is very complex. The existing monitoring technology can not meet the need of a comprehensive understanding of bacterial metabolism. With the deepening of its industrial production, the needs of conducting a comprehensive monitoring of fermentation are increasingly urgent. Electronic nose and electronic tongue are intelligent detection technologies which have a rapid development in recent years. The main goal of the research is to monitor submerged fermentation of Tremella aurantialba by electronic nose and electronic tongue. The main contents and conclusions are as follows:(1)The analysis of physical and chemical changes of key indicators in the fermentation broth. Dry cell weight, content of reducing sugar and total sugar has a close relationship with cell proliferation and metabolism. Dry cell weight reflected the changes of the amount of the biomass. After a short adaptation, cell dry weight increases rapidly in the logarithmic growth stage. When entering the stable stage, dry cell weight did not change significantly. As one of the main energy material, the reducing sugar content in the fermentation broth decreased with the increasing of bacterial biomass. On the fourth day of Tremella aurantialba fermentation, the total sugar content increased due to the release of a large number of polysaccharide.(2) Study on monitoring method of Tremella aurantialba submerged fermentation by electronic nose. A set of electronic nose system was developed in this research. Seven TGS gas sensor for volatile components was selected. Tremella aurantialba fermentation process could be divided into three stages, corresponding to the lag phase, exponential and stationary phases. The identification rates in training set of K nearest neighbor(KNN) model combined with principal component analysis(PCA) and independent component analysis(ICA) were 97.14% and 100.00% respectively. when the fermentation process was divided into seven stages depending on the number of days,the identification rates in training set of two different models were 88.57% and 97.14% respectively. Quantitative prediction models based on support vector machine(SVM)and back propagation neural network(BPNN)were constructed,when principle components and independent component as input,key factors of Tremella aurantialba broth as output. The results showed that the SVM model combined with ICA for prediction of total sugar had the best accuracy,in which the correlation coefficient in the prediction set(Rp) and the root mean square errors in the prediction set(RMSEP) were 0.938 and 3.500g/L respectively. The Rp and RMSEP of BPNN model combined with ICA were 0.953 and 2.156g/L respectively,which had the best accuracy for reducing sugar prediction. The performance of the BPNN model combined with ICA for dry cell weight was the best, in which the Rp and RMSEP were 0.958 and 3.044g/L respectively.(3) Study on monitoring method of Tremella aurantialba submerged fermentation by electronic tongue. We applied these three main principal components for three-dimensional figure. The fermentation process was divided into three phases. The identification rates in training set of K nearest neighbor(KNN) model combined with principal component analysis and independent component analysis were all 100.00%. When the fermentation process was divided into seven stages depending on the number of days,the identification rates in training set of two different models were 97.14% and 100.00% respectively. The SVM and BPNN combined with PCA and ICA were used to build models for indicators of fermentation prediction. The results showed that the models established by different methods had the best prediction accuracy of total sugar, reducing sugar and dry cell weight. The SVM models combined with ICA for prediction of total sugar and reducing sugar had the best accuracies, in which the Rps were 0.908 and 0.975 respectively, RMSEPs were 4.350g/L and 4.579g/L respectively. The performance of the BPNN model combined with ICA for dry cell weight was the best, in which the Rp and RMSEP were 0.953 and 3.179g/L respectively.(4) Study on monitoring method of Tremella aurantialba submerged fermentation based on multi-sensor fusion by electronic nose and electronic tongue. The SVM and BPNN combined with PCA fusion and ICA fusion were used to build models for indicators of fermentation prediction. The SVM model combined with ICA for prediction of total sugar, reducing sugar and dry cell weight all had the best accuracies. The Rps of the models for total sugar, reducing sugar and dry cell weight prediction were 0.960, 0.987, 0.970 respectively. The RMSECVs of the models for the three indicators were 2.574g/L, 2.293/L, 2.527g/L respectively. The models based on two different fusion method had better prediction accuracies, than the models based on single technology.The research showed that, monitoring of Tremella aurantialba fermentation based on electronic nose and electronic tongue technologies achieved satisfactory results. The results showed that, electronic nose and electronic tongue had broad application prospects in monitoring of Tremella aurantialba submerged fermentation.