Adaptive sensing with a microsphere array-based electronic nose

This thesis describes the development and evaluation of a semi-automated fluorescence-based microsensor system designed for continuous adaptive vapor sensing. This system has been developed as an upgrade to the system originally designed in the Walt laboratory over a couple decades ago following mammalian olfaction principles. As in olfactory systems, multiple sensor types respond to numerous vapors, thereby producing diverse patterns that are processed using pattern recognition. The microsensor array contains thousands of fluorescent microbeads that are packed at high density on an optical fiber bundle platform. Indicators attached on the microbeads, including solvatochromic and pH-sensitive dyes, porphyrins, and vapochromic compounds, distinctly change their emission properties and intensity upon exposure to different vapors. Furthermore, a fluorescent microbead probe designed to detect reactive organophosphate chemical warfare agents alerts low ppm levels within few seconds after exposure. The multi-sensor arrays are inert to humidity and many other backgrounds.; The current vapor sensing device is an intelligent system trained to adapt in many different situations as it selects optimal conditions needed to solve each classification problem posed. Fluorescence intensity from different sensors is acquired over time and multiple wavelengths are monitored using various pulse durations and profiles to build training databases that are used for vapor detection and discrimination. The scope of adaptive sensing is to acquire the minimal amount of data necessary to solve a given vapor classification problem. This approach has been developed to reduce sensor exposure and data processing times. Generally, difficult vapor classification problems are solved using data acquired using multiple pulses and multiple sensors and wavelengths, while a simple presence/absence classification problem can be solved with little data. In addition, the optical nose system can acquire reproducible sensor responses for weeks or months using an approach called adaptive exposure. During adaptive exposure, light intensity is attenuated to minimize sensor photobleaching and multiple subsections are inspected instead of the entire array to maximize array usage time.