Detection of contaminants using a MEMS FAIMS sensor

Abstract

Detecting the presence of contaminants in water is a critical mission, but thorough testing often requires extensive time at a remote facility. A MEMS implementation of a FAIMS (High-Field Asymmetric-Waveform Ion Mobility Spectrometry) sensor has recently been developed, and is capable of promptly analyzing water on-site. In this thesis, we apply two well-established statistical target detector algorithms to the detection of chlorite in water. The matched filter and the adaptive cosine estimator (ACE) are subspace detectors that possess complimentary geometric properties. We address several significant challenges in implementing these detectors, including the estimation of the covariance given the limited amount of data available and the design of a target signature subspace in response to the fact that the signature does not scale linearly with the contaminant concentration. In addition, we consider the need for dimension reduction through the use of wavelets. We evaluate each of the detectors on FAIMS data of pure and chlorite-contaminated water.