Incipient Fault Detection and Diagnosis: A hidden information detection problem

In this paper, we propose an overview of the Fault Detection and Diagnosis concept as the detection and characterisation of a hidden information as usually addressed in Data Hiding approaches. In fact, on one side Fault Detection and Diagnosis is generally studied as the detection and characterization of changes (abrupt or not) in a signal describing the behavior of a system. On the other side, data-hiding depicts the art to embed (hide) an information in a considered host signal. From an attacker point of view, one challenge is to detect and characterize this hidden information. Therefore by drawing a parallel with data hiding, we give a triptych to design and evaluate the performances of Fault Detection and Diagnosis methods. For the particular case of incipient faults nested in noisy corrupted signals (considered as hidden information with statistical transparency performance), we explain the relevance of using data hiding concept for efficient diagnosis purpose. Examples of the preliminary results with such an approach are given.