| Description |
Low-cost air quality sensors are an accessible and widespread way to monitor atmospheric pollutants. However, these sensors need to be individually calibrated in a process that takes weeks in the field or days in the lab. Temperature and humidity are important covariates when establishing a correlation between sensor measurement and target gas concentration. In this work, flow and temperature controllers were used to precisely manipulate the conditions of gas delivered to the sensors, greatly reducing the manual work of traditional laboratory sensor calibrations. The calibration system consists of two mass flow controllers for gas mixing, a heating system, and three gas hoods to house the sensors. The calibration system's temperature control capabilities were tested on a stream of air by plotting the controller's measurements as it tracks several setpoint changes throughout the system's temperature range. Its calibration capabilities were then tested on low-cost ozone sensors. An ozone generator and monitor were set up at the two ends of the system with three Alphasense OX-A341 sensors located near the monitor. Calibration constants were calculated using multiple linear regression for temperature, sensor measurement, and target gas concentration. These calibrations resulted in an average 𝑅𝑅𝑎𝑎𝑎𝑎𝑎𝑎 2 of 0.987. The temperature control system operated between 75°F and 100°F, tracking five setpoint changes in under 90 minutes and staying within 0.5°F of the setpoint at steady-state. The system has produced high-quality concentration calibrations in a fraction of the time other lab calibration methods take, while providing comparable results. In these experiments, temperature did not end up being a statistically significant covariate for every sensor, as P-values for the three sensors were 0.412, 0.041, and 0.101. However, concentrations measurements at 100°F were all practically zero due to ozone's increased reactivity at higher temperatures. This affected the model's ability to draw a significant relationship between ozone concentration and temperature with the only useful measurements being between 75°F and 85°F. Similar results to the MLR were obtained by performing a linear regression using only sensor measurement as the lone predictor. |
