Design And Realization Of A Universal Electronic Nose System For Detecting The Tea Aroma

Tea is not only a popular drink,but also a carrier of a unique tea culture.In recent years,with the prosperous development of tea market,various serious problems have appeared,such as shoddy products,random labeling,and adulteration.These damage consumers'benefits and confidence,and disrupt the order of the tea market.What's the worse,it restricted the further development of the tea industry.However,the existing tea detection methods have some shortcomings and cannot be quickly and accurately detected on a large scale,which brings difficulties to market supervision.Therefore,it is necessary to develop a device for rapid and accurate detection of tea.Electronic nose has attracted a lot of attention as a powerful detection tool,and has been extensively applied in a lot of fields,such as food,environment,and health in recent years.In this study,a universal electronic nose system containing 18 metal oxide semiconductor(MOS)gas sensor arrays is developed for tea detection.Since MOS sensors are prone to sensor drift during use,an algorithm based on transfer learning is proposed to improve the accuracy of electronic nose classification results.Firstly,select a series of gas sensors which is suitable for tea detection.According to the selected sensors,carry out the hardware design of the electronic nose.Transmit the sensors signal to the single-chip microcomputer chip,and then store the data in the memory card through RS485 communication to achieve data collection and transmission.Then,the 18 sensors in the developed electronic nose are screened,and 10 sensors array suitable for tea detection are obtained.The rest of 8 redundant sensors is shielded.Based on the electronic nose after screening the sensors,two experiments were conducted,collecting 6 different kinds of tea data and 6 different producing areas and harvest times of West Lake Longjing green tea data.The classification of 6 different kinds of teas to verify the universality of the developed electronic noses,and the differentiation of 6 different types of West Lake Longjing teas to verify the accuracy of the developed electronic noses.Finally,the proposed TrLightGBM algorithm which combined transfer component analysis and LightGBM is used to process the two sets of experimental data.The experiment results show that the developed electronic nose has good performance.It can not only be used to distinguish different kinds of tea,but also can be used to distinguish different producing areas and harvest times of the same kind of tea.The TrLightGBM method can improve the performance of electronic nose classification results and has good generalization.Therefore,the developed electronic nose combined with the proposed algorithm can identify the quality of tea quickly and accurately,and has great potential in tea detection and maintaining market order.