Analysis of Sensor Signals for Online Detection of Hydrocarbons in Liquids in the Presence of Interferents

Current applicability of many chemical sensors is limited due to the lack of adequate selectivity to enable real-world applications. Often, the chemically sensitive element of the sensor is only partially selective to any specific target analyte, potentially giving rise to low probability of detection. Other challenges include the need to identify and quantify the target analytes in a mixture, especially in the presence of non-target interferents. In this dissertation, to enhance the selectivity of the sensor, analysis of sensor signals for detection and quantification of mixtures of hydrocarbon compounds in liquids in the presence of interferents using estimation theory and polymer-coated sensor devices is proposed. In particular, signal processing techniques are developed that can be employed for real-time detection and quantification of target analytes (specifically, petroleum hydrocarbons) in the presence of interferents using only a single polymer-coated shear horizontal surface acoustic wave (SH-SAW) sensor device. Estimation theory is used for signal processing because it enables near real-time data processing, minimal computational requirements, and minimal memory requirements for real-world implementations. The proposed techniques are based on bank of Kalman filters (BKFs) and/or exponentially weighted recursive least squares estimation (RLSE). The success of the approach depends on appropriate analytical modeling of the sensor response. A general n-analyte model is formulated that takes into account the responses due to the target analytes and non-target interferents that interact with the polymer coated sensor. The model, which assumes that sorption of one analyte does not prevent sorption of other analytes in the mixture, utilizes two sensor parameters, i.e. response time constant and sensitivity. Non-ideal cases, the non-step-like concentration versus time profile seen by the sensors, as well as concentration-dependent sensitivity are also considered. The proposed techniques are tested using experimental sensor response data collected using polymer coated SH-SAW sensors with actual groundwater samples. The estimated analyte concentrations are compared to the results obtained independently using gas chromatography. Very good agreement (within about 10-15% accuracy) between the estimated and measured concentrations is found, even in the presence of non-target interferents. No complex training data set is required for the proposed technique