Research On Gas Detection Node Using Wireless Sensor Networks

Gas detection node using wireless sensor networks is a device that is composed of possessing part selective gas sensor, data processing unit and wireless communication module, can detect the information of gas concentration and transmit detection information by wireless sensor networks. As a new type of gas detection method, it merges sensor technology, electronics, signal processing, computer technology and wireless networks together, can be used to quantitatively analysis of gas composition for single or mixed gas, and it has broad application prospects in food processing and inspection, coal gas detection and medical diagnostics field.In this paper, we mainly have exploratory research in areas of hardware, software and algorithms of the gas detection node of wireless sensor networks. Based on in-depth study on dynamic detection of gas sensor technology, the characteristics of gas sensors, and gas detection pattern recognition, and with the help of the experiment conditions of CETC-49, the experiment system is established. We acquire and train containing different concentrations of hydrogen(H₂)and methane(CH₄)gas mixture, finally achieve quantitative detection of gas.Hardware of gas detection node includes gas sensor and conditioning circuit, DSP processing circuit and wireless communication modules. Centering on DSP processor, it full use of the DSP chip modules, such as ADC, SCI serial communication unit, rich GPIO resources etc. DSP controls the gas sensor heating circuit to achieve dynamic heating of the gas sensor.Software of gas detection node is composed of the pattern recognition algorithm program, gas testing procedures based on DSP and wireless sensor networks data transmission procedures and PC software. Among them, the training process of pattern recognition algorithms written by the MATLAB software is used for the experimental simulation of pattern recognition algorithms; Gas testing procedures based on DSP is used for embedding pattern recognition algorithm, which has been trained well and has better detection result, in the DSP using C language; wireless sensor networks data transmission procedures and the PC software are used to transfer and display information such as the concentration of gas detection separately.This paper researches two pattern recognition algorithms: support vector machine regression and RBF neural network learning algorithm. Based on a large number of experimental data, we respectively do simulation experiment of the two pattern recognition algorithms with MATLAB. We obtain that support vector machine regression testing method is better by establishing of gas concentrations in the quantitative detection model, giving the two algorithms of gas concentrations in the test results and analysis and comparing the two kinds of pattern recognition algorithms in the concentration of the test results. Finally, we choose support vector machine regression algorithm to be embedded in DSP for detecting gas on line.