Adaptive Machine learning for Backup Air Quality Multisensor Systems continuous calibration

Abstract

Air Quality Multisensor Systems (AQMS) have shown to be able to provide, if properly calibrated, high quality data in terms of indicative measurements. In this work, we show how this capability can be used for using them as a backup tool for reference analyzers providing a way to obtain meaningful data in the case reference data become unavailable due to failures. Field calibration robustness is still debated but the stream-like availability of reference data make possible to continuously update the AQMS calibration function to sensors and concept drifts so to provide optimal performances whenever needed.