Adaptive Water Sensor Signal Processing: Experimental Results and Implications for Online Contaminant Warning Systems

Real-time contaminant detection and classification in a drinking water pipe using conventional water quality sensors: Techniques and experimental results

“…Methods are sought to identify the contaminant signal in form of water quality changes and separate it from background variations ½ b mFðtÞ and instrument noises ½nðI t n ; sÞ. Yang et al (2007) described an adaptive monitoring process for signal separation using sensor output normalized against those in an adaptive time window (T). When ½mFðt þ 1Þ À mFðTÞ ( I T i and is set at a constant, the relative change (Dc) in sensor output between at time t and the adaptive time window is obtained.…”

“…This generates the inversed asymmetric U-shape pattern as observed in the experiments of aldicarb, glyphosate, potassium ferricyanide, colchicine, and biological contaminants. When the dispersion effect is sufficiently large, the asymmetric pattern at the upstream sensor stations is transformed into the dispersive inversed V-shape along the flowing distance (Yang et al, 2007).…”

“…Previous studies Mehrsheikh et al, 2006;Ray et al, 1981;Arnstein et al, 1949;Mason et al, 1990;and Miles, 1991) described chlorination pathways and molecular structure changes that are consistent with the sensor measurements and kinetic analysis results. In chlorinated water, aldicarb is oxidized to yield aldicarb-sulphoxide and aldicarb-sulphone with a half-life t 1/2 * < 1 min (Miles, 1991;Mason et al, 1990) leading to the rapidly established steady-state chlorine loss (Yang et al, 2007). On the other hand, glyphosate follows more complex oxidation pathways resulting in differential responses between total and free chlorines.…”

“…In the case of conventional water quality sensors, better detection accuracy and contaminant selectivity are achieved by using advanced sensor data processing techniques or through developing hardware such as the new generations of optical sensors for bacterial and chemical contaminations (e.g., U.S. EPA, 2005;NRC, 1995). The first approach relies on data mining and other similar statistical techniques such as multi-variant cluster analysis (Klise and McKenna, 2006), adaptive data treatment (Yang et al, 2007), and control chart-based robust statistics (Kroll and King, 2006). Szabo et al (2007) reported detectable sensor responses to the presence of various contaminants in pilot-scale pipe flow experiments.…”

“…Haught et al (2005) reported from bench-scale testing the minimum concentration threshold of contaminants above which their presence can be registered in water quality sensors. Recently, Yang et al (2007) proposed a real-time event adaptive detection, identification and warning (READiw) method using an iterative detection procedure (Fig. 1).…”