基于传感器阵列解耦合的气体种类识别及浓度检测方法  

Gas classes identification and concentration detection method based on decoupling of sensor array

作　　者：董典典 黄正兴 李中洲[2] 余隽[2] 冯仕玮 DONG Diandian;HUANG Zhengxing;LI Zhongzhou;YU Jun;FENG Shiwei(School of Control Science and Engineering,Dalian University of Technology,Dalian 116024,China;Faculty of Medicine,Dalian University of Technology,Dalian 116024,China)

机构地区：[1]大连理工大学控制科学与工程学院,辽宁大连116024 [2]大连理工大学医学部,辽宁大连116024

出　　处：《大连理工大学学报》2024年第3期323-330,共8页Journal of Dalian University of Technology

基　　金：国家自然科学基金资助项目(61874018).

摘　　要：采用传感器阵列进行气体识别时,某个传感器出现故障将导致整个系统不能使用.受集成学习Bagging方法的启发,提出了传感器阵列解耦合方法.当硫化氢、氨气和丙酮这3种气体在某路传感器数据无法被正常采集的情况下,使用逻辑回归方法作为基分类器仍然能达到100%的气体分类正确率.鉴于良好的分类效果,提出了基于类别先验知识的浓度检测方法.将决策树回归作为基回归模型的传感器阵列解耦合浓度检测方法平均绝对百分比误差为2.28%,结果验证了传感器阵列解耦合方法的可行性.When the sensor array is used for gas identification,fault with a sensor will cause the whole system to be unusable.Inspired by the ensemble learning Bagging method,the decoupling method of sensor array is proposed.In the case that three gases,hydrogen sulfide,ammonia and acetone data cannot be correctly collected from a certain sensor,100%of gas classification accuracy can still be achieved using logistic regression method as the base classifier.In view of the good classification effect,a concentration detection method based on prior knowledge of classes is proposed.The concentration detection method based on the decoupling of sensor array uses the decision tree regressor as the base regression model and yields 2.28%of mean absolute percentage error,and the results verify the feasibility of the decoupling method of sensor array.

关 键 词：解耦合 气体识别 集成学习 浓度检测 先验知识 

分 类 号：TP212.9[自动化与计算机技术—检测技术与自动化装置]