基于RBF神经网络的混合气体智能检测系统研究

基于传感器阵列和神经网络构造智能系统用于检测混合气体的低质量分数.传感器阵列获取质量分数为1×10-6～5×10-6范围的H2,C2H4,C2H2混合气体响应和质量分数为5×10-5～3×10-4范围的CO响应.通过RBF神经网络学习改善低质量分数混合气体检测的灵敏度.把传感器响应作为神经网络输入,神经网络输出为H2,C2H4,C2H2和CO的质量分数.实例分析表明,系统能较好地克服低质量分数混合气体检测过程中普遍存在的交叉灵敏度,得到满意的检测结果.

Study on the Intelligent Measurement System of Mix Gases Based on RBFNN

An Intelligent Measurement System was constructed based on gas sensor array and RBF neural network to measure the quality fraction of mix gases. The former was used to acquire the signals responding to H2, C2H4 and C2H2 with a concentration of 1×10-6～5×10-6 and CO with a concentration of 5×10-5～3×10-4, and the latter was introduced to improve the selectivity of the gas sensors. With a whole pattern of different sensors as input nodes of the network, the quality fraction of H2, C2H4 , C2H2 and CO could be correspondingly predicted as the output of the network. It can overcome the cross sensitivity among heterogeneous gas sensors and lead to more satisfied measurement results.