| Evaluation of the fermentation quality in the process of black tea fermentation,rely on tea master to observe the color changes on the leaves,finally,the quality and degree of fermentation is judged by production experience.Vulnerable to the human experience and subjective factors,lack of scientific accuracy and,causing fermentation low or excessive,resulting in flavor is not unified.To accurately grasp the fermenting qualities needed expert review or physical and chemical detection by means of changes in sensory,sensory assessment vulnerability to environmental,psychological and subjective experience and so on,and physical and chemical detection has a long,high costs and a lagging indicator of the shortcomings,in tea processing products valid information is not real-time feedback.In order to overcome the lack of artificial sensory evaluation and physical and chemical detection,improve the quality of black tea ferment the objectivity and accuracy of evaluation,this research was carried out based on nearinfrared spectroscopy,machine vision,smell,Visual evaluation of electrical characteristics of black tea ferment and intelligent technology.According to the limitations of tea fermentation equipment,developed a new type of drum-type oxygen-enriched fermented and matching technology,as black tea ferment experimental platform,discusses Congou integration technology for automated production lines.Main research contents are as follows:(1)Design and performance optimization of drum-type oxygen-enriched fermentation machine.Firstly,based on the principle and technology characteristics of black tea fermentation,a kind of drum type black tea fermentation machine was designed to solve the problem of difficult-to-turn mixing,insufficient oxygen supply,and uneven quality of the traditional fermentation equipment.Fermentation experiments were conducted on this novel drum-type fermentation machine as the platform,the performance parameters of fermentation machine were clarified.Methodologically,with dimensionless comprehensive scores as a measure of fermentation quality,response surface methodology(RSM)and back-propagation adaptive boosting based genetic algorithm(BP-AdaBoost-GA)were used separately to optimize 3 parameters(fermentation temperature x1,fermentation time x2,rotational interval x3)that affect fermentation quality.Also the optimizing effects of RSM and BP-AdaBoost-GA were compared.Results showed the importance degrees of the three parameters ranked as x1 > x3 > x2.With RSM at x1= 25 oC,x2 = 150 min and x3 =20 min.The predicted value and actual value of comprehensive scores were 0.863 and 0.856,respectively,showing relative error 0.8%.With BP-AdaBoost-GA at x1= 27 oC,x2 = 170 min and x3 =25 min,the predicted value and actual value of comprehensive scores were 0.871 and 0.868 respectively,showing relative error 0.3%.When the BP-AdaBoost has 7 nodes in the hidden layer and a prediction error threshold of 0.25,its determination coefficient is greater than of RSM(0.994 vs.0.973),and it has lower root mean square error of prediction(RMSEP)of 0.017 and residual predictive deviation(RPD)equal to 18.456.In this study,both of RSM and BP-Ada Boost-GA were feasible for optimization of fermentation parameters.However,the fitting ability of RSM is limited because it is based on quadratic polynomial regression,while the fitting ability over experimental data is limited.The algorithm combining improved neural network and GA has higher global extremum prediction ability and higher accuracy.Thus,it can be concluded that even though RSM is most widely used method for fermentation parameter optimization,BP-AdaBoost-GA methodology may present a better alternative.In the meantime,the rotation function has both advantages and disadvantages on the fermentation quality of black tea,moderate rotation and mixing material can enhance the quality of black tea and shorten the fermentation time.(2)Study on the quality evaluation of black tea fermentation based on near infrared spectroscopy technology.This chapter explores the method of near infrared spectroscopy for the quality of black tea fermentation,focusing on sensory quality and physical and chemical quality indices(lutein,)in the construction of the predictive model of Theaflavin,Thearubigins,Theabrownine,catechin and phenol ammonia ratio,the paper studies comparing various spectral pretreatment,characteristic wavelength variable optimization and quantitative analysis model,so as to find a more concise evaluation model with higher forecasting accuracy.First of all,to different stages of fermentation for the study of tea products,data preprocessing of near infrared spectroscopy to get,compared to the typical effects of spectral pretreatment on evaluation model,the standard normal variable transformations(SNV)optimal smoothing effect of noise.Furthermore,combined with partial least squares regression(Si-PLS),random frog algorithm(SFLA),competitive Adaptive weighted sampling(CARS)and continuous projection(SPA),filtering out the best quality indexes of characteristic wavelength variable.Optimization quality character of the wavelength are established based on geneticsupport vector machine(GA-SVR)linear prediction models.The model results show that the variable screening methods such as cars,SFLA and spa can compress variables effectively and improve the accuracy of the model.The predictive RMS error value(RMSEP)of the non-linear model is significantly smaller than the partial least squares(PLS)model,and the correlation coefficient(R)and relative analysis error(RPD)are higher than the PLS model.For the determination of the quality of black tea fermentation,the performance of non-linear model is better than that of linear model.In addition to the evaluation model of lutein(RPD 1.77),the RPD value of the GA-SVR model based on the optimum characteristic wavelength is greater than 2,which indicates that the model has excellent predictive performance and can be used for quantitative analysis of the parameters of each index.Near infrared spectroscopy technology can realize the fast nondestructive testing of black tea,physical and chemical index and sensory score,and adopt the feature variable screening to establish a more concise forecasting model with higher accuracy,and the research results provide the theoretical basis for the practical application of the near infrared spectral rapid detection of the physicochemical qualities of black tea.(3)Study on the fermentation quality evaluation of congou black tea based on machine vision technology.The temporal variation of leaf color in the product during fermentation is an important basis for evaluating the quality of fermentation artificially.The paper took the congou black tea in process of the fermentation as the research object,according to trace the changing regularity of the key pigment materials based on the visual imagery of machine technology,combining linear(PLS)and nonlinear(SVR support vector machine and RF stochastic forest)modeling methods,the quantitative evaluation models of 4 quality indices,such as theaflavins(TFs),thearubigins(TRs),theabrownins(TBs)and sensory quality,were constructed to realize the rapid characterization of key quality indices in fermentation.Firstly,a machine vision acquisition system and GUI analysis software were developed to collect the leaf images of fermented products.Through the image color space transformation,9 color variables(R,G,B,H,S,V,L*,a* and b*)were extracted as the characteristic parameters of the evaluation of fermentation quality,with Zscore standardized pretreatment of the data set,and the change regularity,difference and relativity of image color and quality indexs were analyzed.Then,the quantitative evaluation models of each quality index were established based on the linear and nonlinear methods.Results showed that the color characteristic parameters and the quality index were significantly different in the fermentation stage,and there were significant correlations between them.When predicting 4 quality indices,the RMSEP value of the nonlinear model(RFand SVR)was significantly less than that of the PLS model,and the R and RPD were significantly higher than that of the PLS model,so the nonlinear model performance was better than the linear model,which can characterize the quantitative analytic relationship between image information and quality indices.The performance of non-linear models was comparable,and the RF modeling method was slightly superior to the SVR method.The R value of the RF model of TFs and TRs were 0.89,RPD value were less than 2 and greater than 1.0,which indicated that the performance of the models was general,and can identify the level of physical and chemical indexes and assessment.The RPD value of the RF model of TBs and sensory score were greater than 2.5,which shows that the models have good predictive effect and can be used for quantitative analysis.The results of the study provide a new way of thinking and technical ways to accurately predict the fermentation quality and develop the special instrument in the productions.(4)Study on fermentation quality evaluation of congou black tea base on olfactory visualization Technology.The odor characteristics of the products are the key for judging the quality of fermentation by artificial olfaction.In this study,16 kinds of porphyrin compounds constitute a sensor array are used as gas-sensing materials,Silicagel60 silica gel thin layer chromatography plate is used as the carrier of color-sensitive materials.A set of olfactory visualization imaging system(OVS)is designed for the detection of the organic volatile aroma(VOC)information in the fermentation process of congou black tea.The image of the difference between the aroma substances and the coloring materials before and after the reaction is taken as the characteristic image of the product.Combined with pattern recognition method such as PCA,KNN,ELM and SVR,the average values of R,G and B of 16 groups of coloring materials in the image are extracted as the characteristic value(48 in total)to evaluate the fermentation quality.At the same time,the gas chromatography mass spectrometry(GC-MS)and the expert sensory evaluation are used to determine the sequential changes of the terpene and sensory scores during the fermentation process.Base on the analysis of olfactory visualization data of 108 samples of fermented congou black tea(0?8h),the results show that the nonlinear ELM and SVM models are better than the linear KNN model,the prediction performance of the SVM model is equivalent to that of the ELM.The correct rate of the prediction samples is 85.18%,and that of the over-fermentation samples is 100%.A quantitative analysis model of terpene and sensory score based on olfactory visualization of color-sensitive signals are established by two kinds of classical quantitative analysis modeling methods,PLS and SVR.The RPD value of SVR model is greater than 2,which indicats that the model have a good prediction effect,and can be used for the quantitative analysis of the terpene aroma substances and sensory quality during the fermentation process.The olfactory visualization technology combined with chemometric methods(mode identification and quantitative analysis)can be used to evaluate the fermentation quality of congou black tea.(5)Study on fermentation quality evaluation based on electrical characteristics of Congou black tea.This paper first proposed a new ideas for the evaluation of fermentation quality of black tea by using the electrical characteristics as a new nondestructive testing technology.First,the testing system and analysis software of electrical characteristics based on LCR were designed.The variation regularity of the parameters of electrical characteristics in the fermentation process were clarified by detecting the electrical characteristics of the fermentation process samples,it was found that the substances that hinder charge transfer during fermentation were gradually increasing.Second,the internal relevance between the parameters of electrical characteristics and quality index(sensory score,theaflavins,thearubigins,theabrownins)were clarified through correlation analysis,and based on the correlation coefficients,a method of principal component synthesis association index were established,and then screened out that electrical characteristics of characteristic frequency was 0.2KHz during the fermentation of Congou black tea.The characteristic frequency in low frequencies(?1KHz)was confirmed and characteristic electrical parameters were loss factor(D)and reactance(X)on account of Monte Carlo information Variable elimination-Competitive adaptive weighting(MCUVE-CARS)joint optimization method.A PLS quantitative analysis model for predicting each fermentation quality index was established based on the characteristic electrical parameters selected by MCUVE-CARS,and the RPD value of prediction model of sensory score,theaflavins,thearubigins,theabrownins was 2.593?1.517?1.851 and 2.920 respectively,indicating that the model has good performance and can realize the quantitative evaluation of the fermentation quality index.Finally,discriminant model of fermentation moderation about Congou black tea was modeled using different pattern discriminant methods(KNN?ELM?SVM),results showed that ELM model was optimal and model calibration sets and projections set classification rate both reached 100% when the main fraction PCs was 3 as well as hidden layer node N was 20,so the prediction of the moderate level of fermentation quality can be realized by the combination of electrical characteristic parameters and ELM discriminant model.Above all,based on the electric characteristic parameters and metrology method,the purpose of the fermentation quality detection,identification and quantification evaluation can be achieved.(6)Preliminary study on automatic production line and integration technology of Congou black tea.Preliminary research on automation and informatization production line were carried out by adopted functional and modular structure design,integrated drum-type oxygen-enriched fermentation equipment and fermentation quality evaluation technology,and the first domestic tea automation production line of Congou black tea was developed also.Through experimental and demonstration of some tea-producing areas,a production line of assembling components and matching technology were worked out,the line can improve the quality of Congou processing and equipment utilization.According to the results of evaluation technology of fermentation quality and applying machine vision integrated with production line,image characteristic information in the process of fermentation products could be acquired,so that preliminarily realized the prediction and evaluation of sensory quality scores during processing.The aim of this study was to improve the scientificity and timeliness of the fermentation quality evaluation during the processing of Congou black tea.Results provided the basis for instruments intelligent sensory characterization of fermentation quality and special instruments development of fermentation quality testing,and provided an effective approach and new ideas to expand the tea sensory evaluation methods.Meanwhile,provided a theoretical basis and data support of research and development of expert decision-making systems and standardized processing technology that in the future of large scale,intelligent processing.It is of great significance to promote the implementation of the national strategy of "Thirteen-Five" about tea industry technology equipment upgrades and "machine substitution". |
PDF Full Text Request