Study On Main Components And Sensory Quality Management Of Agricultural Tobacco Smoke

Studying the chemical components of tobacco smoke is the key work to judge the sensory quality of tobacco.Clarifying the chemical components of tobacco is conducive to further understanding the influence of sensory regulation of tobacco on the formation of aroma,so as to provide solutions for the sensory quality regulation of tobacco and provide a basis for the sensory quality management of tobacco manufacturers.However,the evaluation of tobacco sensory quality needs to rely on professional evaluation experts,due to environmental,personal physiological and other factors will lead to differences in the evaluation results.Therefore,the application of deep learning methods such as neural network model in tobacco sensory quality assessment is of great significance in tobacco product research and sensory quality management to correlate the chemical components of tobacco smoke with tobacco sensory quality,so as to improve the limitation of manual evaluation.In view of the above considerations,the main tobacco growing areas in Yunnan Province are as follows: The tobacco leaves of Baoshan City,Qujing City and Dali City were used as raw materials.Pyrolysis experiments at different temperatures were carried out by using high-pressure micro fixed bed and GC-MS,and combustion experiments were carried out by using self-made smoking machine and GC-MS to study the chemical components in the mainstream tobacco smoke.After classifying the chemical components,the sensory correlation analysis of tobacco was conducted and the neural network model was established.Meanwhile,the model was verified by sensory quality evaluation of manual cigarette smoking.The main research contents and results are as follows:(1)The composition analysis of tobacco smoke of different types and origin shows that the types and amounts of chemical components in mainstream smoke are also different with different analysis methods,and the conclusion is drawn: The quantity and type of chemical components of flue gas obtained by pyrolysis experiment are smaller than those obtained by combustion experiment.The chemical components of tobacco from different producing areas are also different to some extent,resulting in different aroma and taste.It is concluded that H01 has more aroma and taste.(2)Threshold exploration of co-pyrolysis products.In this study,147 sensory related compounds were detected by literature retrieval.By calculating the relative contribution of the measured compounds combined with the retrieved threshold values,the characteristic compounds with the highest relative contribution of each sample and the top 10 among all compounds were obtained.(3)The neural network model of correlation between chemical components of tobacco smoke and tobacco sensory quality.In the process,147 sensory compounds related to chemical composition in mainstream flue gas and the relative value of sensory contribution of corresponding functional groups are taken as the input layer of the model,and the sensory evaluation index corresponding to measured chemical components is taken as the output layer,with 2 hidden layers and 10 hidden nodes.Topological structure is used to directly identify each value of sensory evaluation and absorption.It is concluded that the total value of R-square regression reaches 0.9915,indicating that the BP neural network model has high accuracy and can be used for conventional cigarette mainstream smoke sensory prediction and evaluation.The results of artificial evaluation experiment were compared with those of BP neural network model,and the model prediction calculation was basically consistent with the results of evaluation experiment.In this paper,literature review and research were conducted on the correlation between cigarette smoke chemical composition and cigarette sensory quality,and experiments and characterization analysis were carried out on the components of mainstream cigarette smoke,so as to clarify the concepts and theories related to tobacco sensory quality.By analyzing the correlation between cigarette smoke chemical composition and cigarette sensory quality,BP neural network model was established to calculate tobacco sensory quality.At the same time,the calculated model was verified by artificial evaluation experiment.Based on the above experimental research conclusions and the current industry status,the sensory evaluation value of tobacco can be quickly quantified through neural network calculation in the process of tobacco processing,which is of great significance in the work of tobacco processing and quality improvement.