Fritz, HenriekeHenriekeFritzPeralta Abadia, JoseJosePeralta AbadiaLegatiuk, DmitriiDmitriiLegatiukSteiner, MariaMariaSteinerDragos, KosmasKosmasDragosSmarsly, KayKaySmarsly2021-11-082021-11-082022Structural Integrity 21: 143-164 (2022)http://hdl.handle.net/11420/10804Smart structures leverage intelligent structural health monitoring (SHM) systems, which comprise sensors and processing units deployed to transform sensor data into decisions. Faulty sensors may compromise the reliability of SHM systems, causing data corruption, data loss, and erroneous judgment of structural conditions. Fault diagnosis (FD) of SHM systems encompasses the detection, isolation, identification, and accommodation of sensor faults, aiming to ensure the reliability of SHM systems. Typically, FD is based on “analytical redundancy,” utilizing correlated sensor data inherent to the SHM system. However, most analytical redundancy FD approaches neglect the fault identification step and are tailored to specific types of sensor data. In this chapter, an analytical redundancy FD approach for SHM systems is presented, coupling methods for processing any type of sensor data and two machine learning (ML) techniques, (i) an ML regression algorithm used for fault detection, fault isolation, and fault accommodation, and (ii) an ML classification algorithm used for fault identification. The FD approach is validated using an artificial neural network as ML regression algorithm and a convolutional neural network as ML classification algorithm. Validation is performed through a real-world SHM system in operation at a railway bridge. The results demonstrate the suitability of the FD approach for ensuring reliable SHM systems.en2522-560X2022143164Artificial neural network (ANN)Convolutional neural network (CNN)Fault diagnosis (FD)Machine learning (ML)Signal processingStructural health monitoring (SHM)Wavelet transformMLE@TUHHIngenieurwissenschaftenJournal Article10.1007/978-3-030-81716-9_7Journal Article