Design And Application Of Handheld Electronic Nose System For Liquor Recognition

Chinese liquor culture has a long history,and liquors are important consumer goods for Chinese people.However,with the rapid development of liquors market,various problems have emerged.The most serious of these problems are fake and shoddy issues.Therefore,it is necessary to classify liquors and recognize the true and fake liquors rapidly through technical means,so as to promote the healthy development of liquors market and protect the physical and mental health of consumers.Electronic nose(e-nose)is a kind of modern test equipment that simulates biological olfactory systems.Before the advent of e-noses,chromatography,spectroscopy and sensory evaluation methods are usually used to evaluate and identify liquors.These methods are not only time-consuming,but also inconvenient to popularize.E-nose can solve these problems with advantages of fast,portable,high reliability and so on.In this thesis,a handheld e-nose device was designed for the sampling and identification of liquors.Six kinds of Luzhou-flavor liquors were used as sampling materials.The collected liquors data were analyzed and processed,aiming at exploring a fast,reliable program about liquors classification and true-fake recognition.The main research work can be concluded as follows:(1)A handheld e-nose device has been designed and manufactured.The device has a high degree of automation.Users only need to set the sampling parameters after the boot,and the subsequent sampling will be automatically completed by the e-nose device.Through the program to control the air pump,solenoid valve and other actuators,the automatic switching of gas path can be realized,so as to automatically complete the liquor sampling and the cleaning.The overall weight of the system is297 g,the volume is 12*10*5 cm~3 and the average power is less than 5 w,so the design fully meets the requirements of handheld devices.(2)A series of liquors identification solutions have been proposed based on the characteristics of the e-nose device.In order to solve the multi-class classification problem of liquors,six kinds of liquors were classified using support vector machine(SVM)and some other algorithms,and the classification accuracy was above 94%.For the problem of true-fake identification of liquors,a one-class SVM model was used to identify six kinds of liquors respectively.It was found that it was difficult to take into account both the recognition rate of the positive samples and the negative samples.In order to solve this problem,a boosting mechanism was introduced to integrate multiple one-class SVM models,and finally the recognition rate of BNWJ,DHX,LJJ,NLS,QJ,SLX reached 92.9%,93.1%,99.0%,95.7%,87.8%and 100%,respectively.(3)Liquors sampling and identification have been integrated into the handheld e-nose device.Using the advantages of embedded Linux system,the python language was used to realize the data preprocessing,feature extraction and dimensionality reduction,and the K nearest neighbor and SVM algorithms were integrated into the python script,and the on-line recognition of liquor was realized.