AAAR 37th Annual Conference October 14 - October 18, 2019 Oregon Convention Center Portland, Oregon, USA
Abstract View
Six Years of Human and Machine Learning about Electrochemical Sensors
EBEN CROSS, David Hagan, Leah Williams, Douglas Worsnop, Jesse Kroll, John Jayne, Aerodyne Research, Inc.
Abstract Number: 776 Working Group: Air Quality Sensors: Low-cost != Low Complexity
Abstract Calibration methods for low-cost, electrochemical sensors generally fall into one of three categories: (1) Factory calibrations of batches of OEM sensors, (2) Laboratory ‘chamber’ calibrations, in which integrated sensor systems are exposed to known pollutant concentrations under a variety of temperature and humidity conditions, and (3) Ambient field co-location alongside reference monitoring equipment. There are advantages and disadvantages to each, but ultimately, the authenticity of a given calibration approach (or training data set) dictates the extent to which the sensor system can move from place to place and still provide reasonable quantification of the air pollution over time. In this presentation we will highlight some of the lessons that we’ve learned over the past ~ 6 years at Aerodyne and MIT working on integrated sensor systems, focusing on our work with electrochemical sensors. The duration of our ongoing projects (spanning months to years) and variety of environmental domains encountered (from wildfire, to near-road, to urban, to background) presents opportunities to further inform/refine/improve calibration model development moving forward. It’s important to embrace the fact that all sensor systems are imperfect machines attempting to measure tiny perturbations in the concentration of a specific gas molecule in a complex and ever-changing ocean of air. We certainly have a lot more to learn in the years ahead, this presentation aims to focus some of those efforts.