Najeh, HoudaHoudaNajehSingh, MahendraMahendraSinghPloix, StephaneStephanePloix2022-03-062022-03-062021https://publica.fraunhofer.de/handle/publica/27113810.1007/978-3-030-76477-7_14Building systems are equipped with building automation system, supervisory controllers, and a lot of sensors which make them complex systems. In this context, technical malfunctions can have a huge impact on building operation and occupant's comfort. To make a resilient building management system, it is important to identify the severity, cause, and type of each fault using fault diagnosis techniques. The first step for a building diagnostic framework is the designing of tests. To make a test, data are required from different sensors. Due to a battery problem, sensors yield gaps. The delay between two data sent depends on the measured value and the type of sensor and the question that arises is from which delay, a sensor becomes faulty? This is a challenge. In a building system, there is no universal test, but there are contextual tests with limited validity. These local contexts are measured with potentially faulty sensors, and the problem is how to conclude about a test that can be valid or not knowing that validity can only be tested with possibly faulty sensors? This is also a complex problem to solve. A test is characterized by thresholds i.e. the behavioral constraints which are either satisfied or unsatisfied. Uncertainty is related to the validity constraints. Indeed, it is difficult to set a threshold for the level of validity from which we can conclude if a test is valid or not. In this work, the proposed methodology of diagnosis comprises the diagnosis from the first principle because it allows us to determine the minimum diagnoses with explanation at component level. Moreover, the diagnostic results are calculated from a set of tests, each one is defined by its level of validity and the problem is how to conclude in terms of diagnosis and how to take into account the level of validity in the diagnosis? The objective of this work is to highlight these challenges as well as to provide a strategy about how to solve them. A real application has been studied for validation: a platform at the University of Southern Denmark.enbuilding systemdiagnosissensorsfaultvaliditydata groupscompleteness levelconfidence level303600book article