Data Processing Approaches to Manage Humidity and Drift When Using an E-Nose for Monitoring Odours from Waste Treatment Plants

The use of electronic noses (e-noses) for monitoring odour emissions from industrial plants, also known as Instrumental Odour Monitoring Systems (IOMS), has recently gained increasing relevance. However, the application of these instruments in this field is still limited due to the complexity of the operating environment, which is characterized by the presence of numerous interferents, such as frequent variations in humidity levels and, particularly in long-term applications, sensor drift. This study proposes the implementation of specific data pre-processing techniques to account for sensor variations related to humidity interference and drift, considering an IOMS installed at the fenceline of a Waste Treatment Plant (WTP) to monitor odour emissions generated by different sources. The results demonstrate the effectiveness of the proposed data preprocessing approaches, which allow the classification models to achieve good accuracy even in the presence of sensor drift, with Matthews Correlation Coefficient (MCC) above 0.8. Overall, the findings highlight the need and importance of applying dedicated data processing frameworks that explicitly account for sensor interferences and variability over time.