Sensor-based estimation of BTEX concentrations in water samples using recursive least squares and Kalman filter techniques

This work investigates sensor signal processing approaches that can be used with a sensor system for direct on-site monitoring of groundwater, enabling detection and quantification of BTEX (benzene, toluene, ethylbenzene and xylene) compounds at μg/L (ppb) concentrations in the presence of interferents commonly found in groundwater. A model for the sensor response to water samples containing multiple analytes was first formulated based on experimental results. The first signal processing approach utilizes only RLSE (recursive least squares estimation) whereas the second, a two-step processing technique, utilizes both RLSE and bank of Kalman filters for the estimation process. The estimation techniques were tested using actual sensor responses to contaminated groundwater samples. Results indicate that relatively accurate concentration estimates (within ±15–23% for benzene) can be obtained in near-real time using these techniques. The two-step processing technique gave more accurate results. This approach allows the use of a single sensor, even for multiple analyte detection and quantification.