基于BP和SOM神经网络的电子鼻识别方法研究

Identify Method for Electronic Nose Based on BP and SOM Neural Network

电子鼻是一种模仿生物嗅觉的综合仿真系统,它可以用来辨别许多复杂的样本,其中用于辨别和分析气体化学成分的仿真系统应用较为广泛,而对复杂混合气体的分析判断和定性识别是电子鼻技术应用的重要方面。本文在分析研究电子鼻原理和基本构成的基础上,重点运用误差回传神经网络(BP)和自组织特征映射网络(SOM)神经网络进行电子鼻系统的定性识别,对三种气体传感器(一氧化碳CO、二氧化硫SO2、二氧化氮NO2)输出的数据进行了仿真、分析和识别,仿真结果表明这两种方法的识别准确率都能达到100%。并且自组织特征映射网络(SOM)算法的识别能力在整体上要优于误差回传神经网络(BP)算法。

Electronic nose,a novel system,is used to measure the chemical composition of gas, which is designed like the biological olfactory system.To identify the complicated odor is the important aspect of the application electronic nose. Based on the study of the theory and constituent of the electronic nose system, Back-Propagation Neural Network(BP) and Self-Organizing Feature Map(SOM),the two kinds of neural network models' application to the qualitative analysis in an electronic nose system are utilized in the paper. And the dates output from three gas sensors(CO,SO2,NO2) are emulated, analyzed and identified. The result shows that preciseness rate of the two recognitions reaches 100%. Through this emulation, the identify capacity of SOM is better than BP in entirety.