Identification of composite sensor faults in structural health monitoring systems using long short-term memory networks

Fault identification (FI) is an integral part of sensor fault diagnosis in structural health monitoring (SHM) systems. However, current FI approaches often overlook composite sensor faults, i.e. different sensor fault types occurring simultaneously within an individual sensor. As a result, actual fault occurrences in real-world SHM systems may be underestimated. This paper introduces an FI approach utilizing long short-term memory networks, addressing composite faults. The FI approach is validated using sensor data recorded by a real-world SHM system. The results demonstrate the capability of the FI approach to identify composite sensor faults, thus enhancing the reliability and accuracy of fault diagnosis.

DDC Class

690: Building, Construction

Funding(s)

Explainable fault diagnosis for smart cities

Verbundprojekt: Automatische Bewertung von Monitoringdaten von Infrastrukturbauwerken mithilfe von KI und IoT - IDA-KI -, Teilvorhaben: Technische Universität Hamburg

Options

Identification of composite sensor faults in structural health monitoring systems using long short-term memory networks