AI rescues faulty sensors, opening door to cheaper environmental monitors
Researchers have shown that machine learning can salvage chemical sensors too defective to use normally, achieving 90% accuracy in detecting harmful volatile organic compounds even in humid conditions. The breakthrough could dramatically reduce costs for air quality monitoring in factories, warehouses, and cities by repurposing hardware that manufacturers would otherwise discard.
Originaltitel: Machine Learning for Enhanced Operation of UnderperformingSensors in Humid Conditions
<p>Using a single sensor as a virtual electronic nose, we demonstrate the possibility of obtaininggood results with underperforming sensors that, at first glance, would be discarded. For this aim, wecharacterized chemical gas sensors with low repeatability and random drift towards both dangerousand innocuous volatile organic compounds (VOCs) under different levels of relative humidity. Ourresults show classification accuracies higher than 90% when differentiating harmful from harmlessVOCs and coefficients of determination, R2, higher than 80% when determining their concentrationin the parts per billion to parts per million range.</p>