Gas Mixture Detection System Based On Neural Network

With the continuous development of protected agriculture, automated monitoring and controling carbon monoxide, carbon dioxide, oxygen and other gas concentrations of the greenhouse have a new development and demand. To solve the detection and environmental control technology of carbon monoxide, carbon dioxide,oxygen and other gas of the greenhouse has become the key technology directions with the development of modern agriculture into high input, high output,and high efficiency. The electronic nose technology and equipment used in protected agriculture to test gas composition of not only meet the urgent requirements to control and detect the carbon monoxide, carbon dioxide, oxygen and other gas in agricultural Greenhouse, but also opens up an important and significant practical value of research.For this purpose, this article developed an electronic nose system to analyze quality and quantity of carbon monoxide, carbon dioxide, oxygen gas.Develop a gas composition detection experimental systembased on discrete sensor arry. Built a data collection system using microcontroller as the core, including software and hardware system. Develop a feature extraction and pattern recognition software based on neural network.For feature extraction recognition system of gas mixture, this article chose the BP neural network and RBF neural network as the pattern recognition method to analyze the quality and quantity of the three gases. Analysis and compare the advantages and disadvantages of both. The final experiment verify the feasibility of the whole system, and verify the correctness of the two neural networks. Although the results there are errors, but errors are in the scope of our project's goals.The electronic nose system achieves our basic requirements.