Experimental Research on Distinguishing Volatile Gases by Electronic Nose

Experimental Research on the Identification of Volatile Gases by the Electronic Nose of the Journal of Jiangsu University of Science and Technology (Natural Science Edition) Zou Xiaobo, Wu Shouyi, Fang Ruming (School of Biological and Environmental Engineering, Jiangsu University of Science and Technology, Zhenjiang 212013, Jiangsu) Using an electronic nose system composed of metal oxide semiconductor gas sensor arrays, the gas sensor array and data processing analyzer of the electronic nose were studied in depth, and the electronic nose was used for 5 different concentrations of ethanol solutions (0.5%, 1%, 1.5% , 2%, 5%) analysis, elaborated the experimental process in detail, at the same time using I3P neural network to identify and analyze the sample, the neural network's correct rate of judgment is 90%, the test correct rate is 80%. Electronic nose; BP nerve Network, gas sensor array At present, the quality of food such as alcohol, cosmetics, tobacco, tea, etc. is mainly judged by human senses, and sensory evaluation mainly depends on human physical and psychological conditions, which is an ingenious technology in itself.

In a sense, due to the influence of subjective factors such as experience and emotion, the evaluation results of sensory evaluation methods vary considerably with the identification of individuals, even if the same person also changes with their physical state and emotions, etc. Different results. Therefore, it has been an expectation for many years to require an objective and accurate method for identifying the smell and smell instead of artificial products to smell the smell and volatile substances. When we have a clearer understanding of the human olfactory process, we can develop an electronic nose, an instrument that can objectively measure and discriminate the smell and other volatile substances. The electronic nose is different from ordinary chemical analysis instruments (such as chromatograph, spectrometer, capillary electrophoresis instrument), etc. It is not the qualitative and quantitative results of one or several components in the tested sample, but the overall information of the volatile components in the sample , Also known as "fingerprint" data.

It simulates the nose of humans and animals and smells "the overall breath of the target. It can not only detect different signals according to different odors, but also compare these signals with the signals in the database established by learning. Identify and judge.Therefore, it has a nose-like function, which can be used to identify the authenticity of products and control the process from raw materials to the entire production process, so as to ensure product quality.

The main application of the electronic nose developed in this laboratory is to identify and classify the volatile odors of alcohol, tea, fruits and other foods, and the purpose is to classify the quality and freshness. Since the electronic nose device in this study is in the development stage, it is now mainly used to discriminate some quantitative solutions.

1 The principle and composition of the electronic nose The working principle of the electronic nose is based on simulating the formation of the human sense of smell. It mainly consists of a gas sensor array and a data processing analyzer (for example). The gas sensor array is functionally equivalent The human olfactory sensory cells that overlap each other produce olfactory signals; the data processing analyzer is functionally equivalent to the human brain and has the functions of analysis, judgment, and intelligent interpretation. Odor molecules are absorbed by the gas sensor array in the electronic nose, which generates a signal. The generated signal is processed, processed and transmitted by various methods, and the processed signal is judged by the pattern recognition system.

According to the requirements for sensors used in electronic noses, the sensors used in electronic noses developed in this research are SnQz metal oxide semiconductor sensors (M0S). After the reaction, the resistance of the metal oxide semiconductor is usually changed, and an electrical signal is generated. At present, it is the world's largest production and most widely used gas sensor. Its advantages are low selectivity, good stability, low energy consumption, long life and strong corrosion resistance. The electrical signal generated by the sensor is amplified by the electronic circuit and converted into a digital signal and input to the computer. The intensity of the sense of smell measured can be used to fund the project of Jiangsu Yan Application Yue Foundation Fund by the absolute voltage, resistance or conductance output by each sensor ( BJ97061) Zou Xiaobo (1975-), *, Hunan Luo 1 Luo, Ting. Master of Science and Technology University.

It can also be used to indicate relative signal values ​​such as normalized resistance or conductance values, that is, to compare the nature of smell with their rate of change. The signal output by the sensor array is collected, processed, and processed by special software, and finally obtained The quantitative quality fi factor determines the quality of the tested samples, such as authenticity, pros and cons, and whether they are qualified or not. Therefore, the key point of this division is the mathematical method used by the data processing software. In the electronic nose, the commonly used data processing methods include multivariate statistical data processing methods and artificial neural network data processing methods.

2 Test device and data collection gives a block diagram of the electronic Yi-style test device under development in this laboratory. The entire system includes a test box and a gas flow package: control valve, vacuum pump, and two sets of gas sampling devices. Test circuit (A / D conversion board, Ao machine interface), fan, heater. The software used includes data acquisition software, experimental observation control software and MATLAB for processing and pattern recognition of the collected signals.

2.1 Test chamber The test chamber is made of stainless steel, and the design is designed to make those materials (such as wires, etc.) that emit odor and easily react with gas outside the test chamber. There is a set of sensor arrays on the bottom plate of the test chamber, and eight thick film tin oxide sensors produced by Figaro Corporation are used to form the sensor arrays. After the gas sample enters the test box, the gas sample is fully mixed with the gas in the test box through the stirrer to achieve the same gas concentration around each sensor. The electronic nose system can detect two kinds of gas samples, one is a fixed M sample. If you want to measure the gas at a specified concentration, according to the calculation, use a micro sampler to determine 1 to extract this sample; the other The situation is when measuring the volatile gas of a certain sample (such as the volatile fruit aroma of strawberries and bananas, the odor of fish and meat, etc.). The volatile gas of the test sample can be obtained by the flow headspace method.

2.2 In order to obtain the whole process of sensor and gas reaction, the author wrote a dynamic acquisition program and dynamic acquisition dialog box with VB, such as the dynamic acquisition dialog box. When the set software is opened, the dynamic response value of each sensor is displayed in the test window. After pressing the "Start Collection" button, enter the file name, and the sensor response value will be interrogated at every specified collection time (such as 1000ms) is automatically written into the file to form a data file for later data processing, until the "knot and pick" button is pressed.

Circle 3 Dynamic acquisition dialog 3 Test the ethanol solution with the same concentration. Each time when the temperature of the test box (the device used to place the gas sensor array and the measured gas) is stable, use a 100 micro sampler to take 20PL from the prepared solution and evaporate it from the evaporator in the test box. At the same time, turn on the electric fan to make the boil-off gas fully mixed evenly in the test chamber (that is, the gas concentration, temperature, humidity, etc. around each sensor are the same). During this process, the software has been collecting data on the reaction of the sensor with the gas. In the experiment, it should be noted that after each M is completed, the sensor needs to be cleaned and reduced for the next measurement of S, because the sensor generally has an oxidation-reduction reaction with the measured gas during the measurement of fi. If it is not washed and The reduction process will affect the next measurement result. Test 10 times for each solution, that is 5X10 = 50 times.

3.1 Data processing and feature extraction The gas sensor is a sensitive device whose surrounding environment is a sensitive factor. The oxygen partial pressure, temperature, and humidity in the ambient atmosphere can directly affect the performance characteristics of the gas sensor. Four types of gas sensors are provided for gas sensors. The signal processing method of the sensor (take the absolute value, relative value, differential value, and integral value of the sensor and gas reaction), such as extracting the maximum value (3WJ maximum differential value UnvJ, the average and The average value of the values ​​obtained in the last 50s is taken as the stable value.

3.2 The processing of characteristic data by artificial neural network Because the relationship between the response value of the sensor and the measured gas concentration is very complex, it is difficult to express it with a clear mathematical relationship, and the artificial neural network can not only imitate human logical thinking, but also Imitating people's image thinking is a typical non-parametric data processing method. It can automatically master and mine "gray boxes" or "black boxes" hidden inside things through self-learning (training) that cannot be expressed by clear mathematical formulas. Therefore, the neural network technology is used to establish the mapping relationship between the sensor array response signal and the measurement solution. The P network can be used to establish the mapping relationship between the sensor array characteristic value and the test solution concentration. The sensor characteristic value is used as the network input (20 input nodes), and the solution concentration is the network output (5 output nodes). Among the 5 solutions (50 samples in total), 8 samples were used to form the training set (40 training mode pairs in total), and the remaining 2 samples were used in the test set (10 mode pairs in total). Train and test the BP neural network. The corresponding training condition is that the number of hidden layers of the network is 18, the learning factor is 0.035, the dynamic factor is 0.0015, and the maximum number of trainings is 5000. The final result is that after the network training, the correct rate of the judgment of the training set is 90%, and the test set is tested, the test correct rate is 80%. 4 conclusion set, and successfully extracted the characteristic value of the solution gas.

BP neural network is used to establish the mapping relationship between solution gas and sensor response value.

However, it can be seen from the test results of the test set that the accuracy rate is only 80%. This is due to poor control of the test conditions (such as ambient temperature, humidity, and sealing of the experimental device), resulting in large data dispersion; and BP network The parameters and structure need to be further optimized.

Gao Daqi, Wu Shouyi. Research progress of artificial olfaction and its application prospect in food aroma evaluation. Journal of Agricultural Machinery,〗 998 (4): Feng Wei, Gao Daqi, Hu Shangxu. New progress in artificial olfaction sensor array technology and pattern classification methods. Analytical Chemistry, 1996 (7): Gao Daqi. Method and device for analyzing inherent quality of cigarettes based on artificial smell. Zhenjiang: Jiangsu University of Science and Technology, 1998. Gao Daqi. Pattern classification method based on neural network. Hangzhou: Zhejiang University, 1996. Kang Changhe, Tang Shengwu. Gas and moisture sensitive devices and their applications. Beijing: Science Press, 1985. (Responsible Editor Wang Liwei) 1 By analyzing the composition and principle of the electronic nose, a set of electronic nose system testing devices was established.

2 Successfully realized the automatic acquisition of electronic nose data by computer

Universal Backup Camera can fit virtually any vehicle,such as SUV,PICKUP,RV,SEDEN.  High definition image, Resolution: 420 TV lines; Effective pixels: 628 X 586 pixels. low power consumption design . The waterproof car rearview backup camera with IP68 rating,means when mounted,it is fit to operate in almost all weather conditions. 170 degree image angle , make you parking and reversing more convenient.

Universal Backup Camera

Universal Backup Camera,Wired Mental Backup Camera,Car Backup Camera,Oem Backup Camera

Shenzhen Sunveytech Co.,LTD , https://www.sunveytech.com