基于气体传感器阵列的混合气体定量分析

优选CO和H2气体敏感的半导体气体传感器组成阵列，建立实时数据采集系统，结合BP神经网络模式识别技术，实现了混合气体组分的定量分析。讨论了不同响应时间下的阵列输出值、不同的数据预处理算法及不同的神经网络结构等主要影响因素对网络输出结果的影响。结果表明，采用RRD预处理算法对3min响应时间下的阵列输出值进行处理，再输入到有12个隐层神经元数的3层BP神经网络进行训练，预测的效果最好。该处理模式能较准确地完成CO和H2混合气体组分的定量分析。

Gas quantitative analysis based on the gas sensor array

The semiconductor gas sensors sensitive to carbon monoxide and hydrogen were chosen to compose the gas sensor array,and an on-line data acquisition system was constructed.Combining with the pattern recognition techniques of back-propagation(BP) neuron network,the system was used to carry out the quantitative analysis of the partial gas concentration in a mixture.The main effect factors to the outputs of BP neuron network,such as array outputs under different response time,the different pre-processing algorithms and the different structures of the neural network,are discussed in this paper.It is shown that the best prediction results are obtained when the array output under 3 min response time is processed using RRD pre-processing algorithm and used as the actual input of the neural network,then the training and testing of this three-layer BP neuron network with 12 neurons in hidden layer are performed.This processing mode can accomplish the quantitative analysis of the partial gas concentration of the mixture(hydrogen and carbon monoxide) accurately.