Fault detection of air quality measurements using artificial intelligence

Abstract

In this work we use Artificial Intelligence (AI) for the detection of faults in air quality measurements. This is crucial in large air quality monitoring networks in particular were fault detection can be a complex and time consuming process. The proposed methodology encompasses several essential steps in anomaly detection. Data preprocessing ensures the quality and relevance of the data by applying techniques like data cleaning, outlier removal, and feature selection. The Isolation Forest model is trained using the pre-processed data, and appropriate hyperparameters are determined through cross-validation. Anomaly detection is performed using the trained model, allowing the identification of abnormal events or instances. The visualization of anomalies provides a clear representation of abnormal patterns, facilitating the interpretation and understanding of air quality data. The proposed methodology can help environmental agencies, researchers, and policymakers in identifying abnormal air quality events, enhancing the accuracy of monitoring systems, and facilitating timely interventions. This methodology can be applied to other industries also, to improve operations and reduce risk.