Building sensor fault detection and diagnostic system

作者：Devanshu Kumar, Xianzhong Ding, Wan Du, Alberto Cerpa · 年份：2021 · DOI：10.1145/3486611.3491122 · 被引用次数：12 · 研究领域：Anomaly Detection Techniques and Applications、Air Quality Monitoring and Forecasting、Data Stream Mining Techniques

In this paper, we developed a Building Sensor Fault Detection and Diagnostics System. It consists of a smartphone application integrated with airflow and temperature sensors, to enable Facilities' crews to collect flow, supply, and zone temperatures from each zone with a simple walk-through. This data is then uploaded to a cloud for further analysis using machine learning algorithms trained to identify zones' faulty sensors based on comparisons with building data available from the Building Management Systems (BMS). We develop two different data-driven fault classifiers and compare our system with two state-of-the-art sensing fault schemes using the Receiver Operating Characteristic (ROC) curve, showing an improvement in the Area Under the ROC Curve (AUC) of 7.1% and 26.3%, and 12.9% and 55.1% in fault detection performance improvement for a 10% false positive rate.