Application Of Generalized Grey System And Nondestructive Analysis Technology In Tobacco Production And Processing

Tobacco leaf production and processing stage is regarded as the most important process for quality generation in the production,and it refers to the whole process which is from harvesting to be as the raw material for cigarette,including harvesting,flue-curing,grading,threshing and redrying,and aging.However,at present,there are still many problems in the production and processing of tobacco leaves.Modulation dependent on experiencehighly and the lack of objective and quantitative modulation standards and methods are particularly obvious.Therefore,it is very important to find some accurate,fast and objective methods for distinguishing,analyzing,monitoring the variation of tobacco leaves quality during production and processing.Tobacco leaf as a kind of biological sample,has extremely complex chemical compositions.The chemical composition,thecontents of components,the interaction among different components and theirdynamic change in production and process are largely unknown,and there is no solid physical and/or chemical lawto use.Although the traditional chemical methods cannot meet the requirements,the method of complex system,named as generalized grey system theory,can be used to analysis.Generalized grey system is a method to describe the complexity of environmental analysis and can solve the complex analytical problems processed from measurement space of variables,which is developed from the white,grey and black analytical systems.Nondestructive analysis technology as a rapid and non-destructive technology,it can obtain a large number of variables continuously compared with the traditional methods,which is especially suitable for tobacco production process withlarge sample size,multi factor analysis and strong continuity.Therefore,based on the idea that the high quality tobacco leaves should have reasonable and harmoniousinternal chemical compositions,this paper had discussed the availablity of the study on the generalized grey system theory combined with the nondestructive analysis to harvest,bulk curing,grading and purchasing,threshing and redrying and aging processes.The aims of this study are to make full use of the effect of tobacco curing and processing,shape coordinating quality characteristics of tobacco leaves and provide a theoretical reference for the further study of tobacco modulation quantitative system.The results of this study are as follows:(1)Timely harvest is one of the key links and key technologies for the production of high quality tobacco.It is important to accurately determine the maturity of tobacco leaves in the field for timely harvest and scientific modulation.The methods of near infrared spectroscopy(NIRS)combined with model population analysis random forest(MPA-RF)and image recognition technology combined with the extreme learning machine(ELM)were used to identify the maturity of tobacco leaves rapidly and the lowest recognizedaccuracy were 90% and 82%,respectively.In order to improve the accurate rate of recognition,the NIR data and image data of tobacco maturity were fused and constructed the ELM model to analyze.The accurate rate of recognition of tobacco leaves maturity can reach more than 96% which is benefit to the quantitative and scientific discrimination of tobacco leaves maturity.(2)It is very necessary to monitor the change of tobacco leaves compositions real-timely and accurately regulate the baking conditions.Firstly,the method of NIRS combination with adaptive evolution-extreme learning machine(SaE-ELM)was used to monitor dynamic changes of the three key indicators(moisture content,starch and chlorophyll)in tobacco flue-curing process.Secondly,a multivariate calibration method,variable permutation population analysis-partial least squares(VPPA-PLS),was proposed to establishquantitative analysis model of the main chemical components of tobacco leaves.Finally,the fitted change rules of key indicators and the main chemical components can provide a reference to precise control in tobacco flue-curing.(3)Flue-cured tobacco as an important economic crop,the quality inspection and classification must be strictly regulated.However,the current grading and purchasing mainly depend on the human senses and cause grading inaccurateby subjective factors which seriously restricts the development of tobacco grading technology.Firstly,we had proposed a multi-classification method,named as Monte Carlouninformative variable eliminationrandom forest(MC-UVE-RF),combined with NIRS that can realize the automatic grouping and grading of tobacco leaves.The lowest accuracy was 88.46%.Due to the use of a single technique to identify easily-confused grade tobacco leaves obtaining poor results,the fused ELM model constructed by NIRS data and image data can acquire greatly improved classification accuracy and reach 96.67%.(4)Since there is no valid method for evaluation the internal quality of tobacco strips in the process of threshing and redrying tobacco,calibration transfer between different types of near infrared spectroscopy instruments and different tobacco physical state is proved to be feasible theoretically firstly.Then on this basis,a calibration transfer method,canonical correlation analysis coupled with wavelet transform(WTCCA),had been presented to transferring the on-line spectra of tobacco strips to the laboratory model of tobacco powder.The mean predictive errors of tobacco strips’ nicotine,total sugar,reducing sugar and total nitrogen in single grade selection line and mixed grades selection line were decreased 42.03%,29.43%,15.63%,30.10% and 8.16%,11.38%,6.66%,19.44%,respectively.It will be favor for improving the online prediction accuracy and the quality control in threshing and redrying process.(5)Tobacco aging is a slow fermentation process under natural conditions after tobacco flue-curing and redrying.However,the research on the rule of tobacco aging is lagged behind.Therefore,we had made the full use of the accumulated temperature and humidity and physicochemical index data to min the implicitrules and associations in them by chemometrics methods and established the simulation time transfer model of accumulated temperature and humidity in warehouse and in the smoke box.This time transfer model combined with sensory quality indicators can realize the classification of aging warehouses initially.Aging time more than 18 months in the warehouse is defined as a slow warehouse.On the contrary,it is called a fast warehouse which can assist researching the tobacco aging technology to a certain extent.