Quantifying and improving performance of a low-cost PM sensor used for occupational hygiene applications

Low-cost particulate matter (PM) sensors are useful for measuring exposure to hazardous particles in the workplace but show different levels of structural under- or overestimation (ie bias) and random variation (ie error) for different types of particles. Traditional calibration methods that are based on comparing with reference measurements can improve sensor performance but require substantial quantities of reference measurements, which may be infeasible to collect when sensors are used in many different situations. This study aimed to quantify sensor performance in different occupational situations and compare methods for improving performance without requiring reference measurements. Sensor measurements with paired reference measurements for particle number (n = 253) and mass (n = 33) concentrations were collected for 6 situations, where particles were assumed to consist of primarily cinnamon, flour, quartz, spruce wood, varnish, or welding fumes. Absolute and relative bias and error were calculated for sensor-reported measurements and 4 methods to calculate mass concentrations from particle number counts. Methods included a literature-derived calibration model and 3 physics-based algorithms that included combinations of particle size, density, and shape. For particle number concentrations, the highest relative bias was observed for the smallest size bin (0.3 to 0.5 µm). PM2.5 mass concentrations were underestimated by 72% compared to reference respirable dust measurements. This underestimation was reduced to 4.5% by recalculating mass from particle number concentrations, assuming uniform particle density and shape and attributing size fractions according to the respirable dust convention. Including particle-specific density and shape showed minor additional changes in overall bias (4.4% overestimation) but halved relative bias for 4 situations. Error was comparable for sensor-reported and recalculated mass concentrations (SD ranging from 93% to 106%) but differed between situations. This study addresses challenges in sensor performance in occupational settings. It shows that recalculating mass from particle counts can reduce bias without extensive reference measurements. This practical approach enhances sensor accuracy, aiding in better monitoring and managing hazardous particle exposure, ultimately improving occupational health.