摘要
污泥组分复杂、来源广泛,异味强度高,其异味特征随异味组分变化明显,因此尝试利用电子鼻实现不同污泥异味特征的快速识别。基于主成分分析(PCA)与线性判别分析(LDA)手段,电子鼻可以有效识别来自农产品加工厂的大豆污泥、化工染料厂的印染污泥和城镇居民区市政污泥的异味特征;加入厌氧消化市政污泥、脱水处理市政污泥的异味信息后,电子鼻训练集数据从240条增加到400条,LDA响应数据识别的正确率降低4.2百分点。将电子鼻分别结合多种机器学习算法判别5种污泥的异味特征,发现多元逻辑回归、K近邻与支持向量分类(SVC)等机器学习模型适用于大样本空间的数据判别,提高了电子鼻对污泥异味特征识别的准确率,解决了电子鼻训练样本数据量受限的问题。其中结合LDA降维的线性核函数SVC模型(LDA-SVC_(linear))可视化分类程度高,实际可应用范围更广。研究结果可为污泥异味特征快速判别提供有效手段。
Sludge is characterized by its complex composition,wide-ranging sources and strong and distinct odors,which varies significantly with different odor components.Therefore,a fast identification method for sludge odors using electronic nose was developed,and it was found that the electronic nose was effective in distinguishing the odor characteristics of soybean sludge from agricultural processing plants,dyeing sludge from chemical factories and municipal sludge from urban residential areas,with the aid of principal component analysis(PCA)and linear discriminant analysis(LDA)combined.Upon incorporating odor information from anaerobically digested municipal sludge and dehydrated municipal sludge,the training dataset for the electronic nose increased from 240 to 400 samples,while it resulted in a 4.2 percent point reduction in the accuracy of LDA response data recognition.Combining the electronic nose with various machine learning algorithms for the identification of odor characteristics in five different types of sludge,it was found that the models of multinomial logistic regression(MLR),K-nearest neighbors(KNN)and support vector classification(SVC)were suitable for discriminating data in large sample spaces.This significantly improved the accuracy of odor characteristic identification by the electronic nose,addressing the issue of limited training sample data.Notably,the linear kernel SVC model combined with dimensionality reduction through LDA(LDA-SVC_(linear))exhibited high visualization of classification and a broader practical application range.These findings could provide a fast detection method for discriminating odor characteristics in sludge.
作者
张珊珊
楼紫阳
王川
张瑞娜
宋佳
王罗春
ZHANG Shanshan;LOU Ziyang;WANG Chuan;ZHANG Ruina;SONG Jia;WANG Luochun(School of Environmental and Chemical Engineering,Shanghai University of Electric Power,Shanghai 200090;School of Environmental Science and Engineering,Shanghai Jiao Tong University,Shanghai Engineering Research Center of Solid Waste Treatment and Resource,Shanghai 200240;Shanghai Environmental Sanitation Engineering Design Institute Co.,Ltd.,Shanghai Institute of Environmental Engineering and Design Science Co.,Ltd.,Shanghai 200232)
出处
《环境污染与防治》
CAS
CSCD
北大核心
2023年第9期1234-1239,1247,共7页
Environmental Pollution & Control
基金
国家重点研发计划项目(No.2018YFC1900704)
“科技兴蒙”上海交通大学行动计划专项(No.SA1600213)。
关键词
污泥异味
电子鼻
快速检测
机器学习
支持向量分类模型
sludge odor
electronic nose
rapid detection
machine learning
support vector classification model
