Liu Shuming, Che Han, Smith Kate, Chang Tian
School of Environment, Tsinghua University, Beijing 100084, China.
School of Environment, Tsinghua University, Beijing 100084, China.
J Environ Manage. 2015 May 1;154:13-21. doi: 10.1016/j.jenvman.2015.02.023. Epub 2015 Feb 18.
Early warning systems are often used to detect deliberate and accidental contamination events in a water source. After contamination detection, it is important to classify the type of contaminant quickly to provide support for implementation of remediation attempts. Conventional methods commonly rely on laboratory-based analysis or qualitative geometry analysis, which require long analysis time or suffer low true positive rate. This paper proposes a real time contaminant classification method, which discriminates contaminants based on quantitative analysis. The proposed method utilizes the Mahalanobis distance of feature vectors to classify the type of contaminant. The performance and robustness of the proposed method were evaluated using data from contaminant injection experiments and through an uncertainty analysis. An advantage of the proposed method is that it can classify the type of contaminant in minutes with no significant compromise on true positive rate. This will facilitate fast remediation response to contamination events in a water system.
早期预警系统常用于检测水源中的故意和意外污染事件。在检测到污染后，快速对污染物类型进行分类对于为实施补救措施提供支持很重要。传统方法通常依赖基于实验室的分析或定性几何分析，这些方法需要较长的分析时间或真阳性率较低。本文提出了一种实时污染物分类方法，该方法基于定量分析来区分污染物。所提出的方法利用特征向量的马氏距离对污染物类型进行分类。使用来自污染物注入实验的数据并通过不确定性分析对所提出方法的性能和鲁棒性进行了评估。所提出方法的一个优点是它可以在几分钟内对污染物类型进行分类，而不会对真阳性率造成显著影响。这将有助于对水系统中的污染事件做出快速的补救响应。
