【作者】 赵万；

【导师】 卢革宇；

【作者基本信息】 吉林大学 ， 微电子学与固体电子学， 2013， 硕士

【摘要】 本文面向食物品质评价研制了基于金属氧化物半导体型气体传感器阵列、LabVIEW虚拟仪器、PCA和LVQ神经网络的电子鼻系统。通过对五种不同种类的食用酱进行气体成分检测，获得“气体指纹信息”，建立“气味指纹数据库”，在此基础上对未知酱品进行检测，实现了对不同种类食用酱的准确区分。首先，研究电子鼻系统结构，一个典型的电子鼻系统由传感器阵列采样装置即气味信息采集装置、数据采集系统和模式识别算法三部分构成。根据电子鼻的工作结构及工作原理，设计电子鼻的工作流程，设计了面向食物品质评价的电子鼻系统框架。系统由采样系统，控制系统和上位机软件组成。其次，设计电子鼻系统，采用分立式传感器阵列，即选用六种费加罗金属氧化物半导体传感器组成气体传感器阵列。传感器驱动电路与传感器阵列被组装在一块电路板上，其被置于5升的玻璃容器内，测试环境温度由温度传感器来监控，利用水浴加热的方式维持测试环境温度恒定，减小外界环境对传感器特性的影响。数据采集系统通过电缆与传感器阵列驱动电路相连，采用24位高精度模数转换器AD7794作为电子鼻信号采集单元，利用32位微处理器STM32F103ZE设计了电子鼻控制系统，实现对电子鼻各个外设的控制及数据采集。采用微控制器集成的USAR（Tuniversalasynchronous receiver/transmitter）实现RS232通信方式，将数据上传至上位机。采用嵌套Matlab的LabVIEW平台搭建上位机软件，其从控制系统获取传感器阵列的响应数据，进行数据整理，并通过主成分分析（PCA）方法和学习矢量量化（LVQ）神经网络模式识别算法对数据进行分析，得到气味的“指纹信息”，并与数据库中的特征数据进行比对，最终将结果呈现出来。最后，利用设计的电子鼻系统对五种食用酱品进行检测。通过测量五种已知酱品中挥发出的气体建立原始数据库，然后再对未知酱品的进行检测，并将所得到的检测数据与数据库进行分析对照，完成对未知酱品的识别。利用PCA分析方法和LVQ神经网络分析方法分别对六种（五种已知酱品和一种未知酱品）酱品进行分析，可以准确地评价未知酱品的品质情况。实验证明利用传感器阵列对样品的“灵敏度响应”分析样品特征比“电阻响应”好。综上所述，本文研制的面向食物品质评价电子鼻系统可以分别工作在学习模式和检测模式，能够完成对标准样品的学习，建立“气体指纹数据库”，可以完成对食用酱的品质评价。电子鼻未来还需要更多的应用性研究工作，具有很重要的研究意义和广阔的发展前景。更多还原

【Abstract】 An Electronic nose system was developedbased on metal oxide semiconductor gassensor array, the LabVIEW virtual instrument, PCA and LVQ neural network for foodquality evaluation. By detecting five different species of sauce, the "gas"fingerprintinformation were used to establish "odor fingerprint database", then the unknown saucecan be accurately distinguished between different species of sauce.Firstly, a typical electronic nose system consists of sensor arrays or samplingstructure, data acquisition system and pattern recognition algorithms. According to thestructure and working principle of electronic nose, framework of electronic nose wasdesignedfor food quality evaluation. System consists of sampling system, control systemand PC software.Secondly, design of the electronic nose system.By using six figaro metal oxidesemiconductor sensors,a gas sensors array was assembled. Sensors drive circuit andsensors array were integrated on a circuit board, which were placed in glass container(5L).In order to reduce the influence of external environment, a water bath heating system wasused to guarantee the test ambient temperature constant.The temperature was monitoredbyMicro Controller. Data acquisition system and sensors array driver circuit wereconnected by cable. By using24-bit high precision ADC AD7794as electronic nosesignal acquisition unit and32-bit microprocessor STM32F103ZE, control system for theelectronic noses undertakenall the peripherals and data collection. Based on microcontroller integrated USART (universal asynchronous receiver/transmitter) and RS232communication mode, data was uploadedto PC software. Software was built underLabVIEW platform which embedded Matlab scripts.By obtaining the response of thesensors array, sorting dataand through the principal component analysis (PCA) method orlearning vector quantization (LVQ) neural network pattern recognition algorithm toanalyze the data, e-nose got the fingerprint information and compared the characteristicsof the data in the database, then presented eventually results.Thirdly, the e-nose system tested five kinds of sauce. The original database was builtby detecting of five kinds of sauce, and then the unknown sauce was tested. By usingPCA and LVQ neural network analysis method, six kinds of sauce(five kinds of sauce andan unknown sauce)were analyzed. As a result, the unknown sauce can be evaluatedaccurately by e-nose. And the experiment hadproved that sensors array’ssensitivitycharacteristicis more convincible than resistanceresponse.In conclusion, an electronic nose for food quality evaluation was developed, whichcould work in learning mode or detection mode respectively, to be able to complete thelearning of standard sample, to establish "gas fingerprint database", and complete thequality evaluation of food. Electronic nosehas very important significance and prospectand needs more research.更多还原