电子鼻的深度神经网络算法实现可燃气体检测的研究OA北大核心CSTPCD

A gas identification algorithm for electronic noses that combines the Graham angle field with a deep residual convolutional neural network is proposed,which can improve the identification accuracy of common mixed dangerous components in gas.Graham angle field(GAF)transformation is performed on the gas sample data of the electronic nose,so that the two-dimensional sensor response data received by the host computer can be input to the convolutional neural network(CNN)after dimension enhancement in the form of three-dimensional data,thus giving full play to the advantages of CNN's strong feature extraction ability,fast model convergence and high rec-ognition accuracy.The experimental results show that in the presence of interfering gases,the detection accuracy of the algorithm for CO and CH4 reaches 93.04%and 92.43%,respectively.Compared with the conventional principal component analysis,linear difference analysis and support vector machine identification methods,the proposed algorithm has the advantages of high anti-interference and high detection accuracy,and provides an intelligent identification algorithm with good application prospects for the high reliable and specific detection of combustible gases in the actual environment when interfering gases exist.