多组分气体定量识别的电子鼻系统设计  

Quantitative identification of multi-component gases based on e-nose system

作　　者：马宏莉 王涛 李泊龙 曾敏 杨建华 杨志 MA Hongli;WANG Tao;LI Bolong;ZENG Min;YANG Jianhua;YANG Zhi(Key Laboratory of Thin Film and Microfabrication,Ministry of Education,Department of Micro/Nano Electronics,School of Electronic Information and Electrical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]薄膜与微细技术教育部重点实验室上海交通大学电子信息与电气工程学院微纳电子学系,上海200240

出　　处：《传感器与微系统》2022年第5期84-86,90,共4页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(61971284,61671299,21703267);上海交通大学“深蓝计划”基金资助项目(SL2020ZD203,SL2020MS031);自然资源部第二海洋研究所基本科研业务费专项资金资助项目(SL2003)。

摘　　要：为同时实现多种单一气体和混合气体的定量识别,设计了一种新型便携式电子鼻系统。系统选用4个商用气体传感器作为传感器阵列,在多组分气体识别过程中,针对训练样本不平衡的特点,提出了一种新的基于反向传播神经网络的层次分类器(BPNN-HC)以提高模式识别的准确率。在每一种模式下,利用偏最小二乘回归(PLSR)分别建立多元回归模型以实现对应组分的体积分数估计。实验结果表明:该电子鼻系统可同时实现6种单一气体及3种二元混合气体的定量识别。In order to realize the quantitative identification of multiple single gases and mixed gases at the same time,a new portable electronic nose(e-nose)system is designed.In the designed e-nose system,the sensor array is composed of four commercial gas sensors.Considering the imbalance of training samples in the process of gases identification,a novel back-propagation neural network-based hierarchical classifier(BPNN-HC)is proposed to improve the accuracy of pattern recognition.In each pattern,by utilizing partial least-squares regression(PLSR)algorithm,the multiple regression model is established to realize the volume fraction estimation of the corresponding components.The experimental results show that the quantitative identification of six kinds of single gases and three kinds of binary mixed gases can be realized simultaneously by employing the designed e-nose system.

关 键 词：电子鼻系统 气体识别 体积分数估计 反向传播神经网络 偏最小二乘回归 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]