Experimental dataset for an AHU air-to-air heat exchanger with normal and simulated fault operations

Abstract

Fault Detection and Diagnosis (FDD) is an important tool in building commissioning. Providing a consolidated dataset for FDD benchmarking is necessary to accurately evaluate the FDD prediction accuracy and detect anomalies. In this study, we provide an experimental dataset for an air handling unit containing two ducts linked by an air-to-air heat exchanger. The dataset is composed of nominal and faulty operations of the system, including the ground truth in order to investigate various faults in 52 cases. The dataset was obtained by measuring a representative system with real climate variations like the ones obtained by Building Automation Systems. The transition between nominal and fault sequences was continuous, as in real operating conditions. An uncertainty evaluation was carried out to provide confidence bounds in the experimental dataset.