Towards reliable long-term environmental odour monitoring with electronic noses: compensation of humidity and sensor drift effects.

Implementation of electronic noses at the fenceline of industrial plants is becoming relevant in order to monitor odour emissions from these facilities. Literature studies proved the applicability of these instruments for this purpose; however, their application over long period of time is hindered by the presence of interferences. We propose the application of two pre-processing strategies to compensate the presence of humidity variations and sensor drift, two interferents commonly found in uncontrolled sampling systems. We validated the methodology on two electronic noses installed at the fenceline of a waste treatment plant during 16 months. Different classification models have been developed based on different data pre-treatment and their performance was evaluated through field performance testing. Results show that the implemented methodology effectively reduces drift effects and improves classification accuracy for different odour sources, enhancing the feasibility of these instruments for industrial odour monitoring.