Rapid Assessment of Incipient Multimodal Faults of Complex Aerospace Systems

Modern aerospace systems integrate a variety of multi-physical components to ensure the high-performance requirements during the operational life. The increasing complexity of those systems determines an exponential growth of multiple and coupled failure modes difficult to predict in advance. This hinders the adoption and integration of new sustainable technologies, for which an accurate and reliable estimate of incipient faults is required to prevent catastrophic events. Model-based Fault Detection and Isolation (FDI) methods allow to infer the health status of aerospace systems using a large quantity of data acquired during the flights, and through evaluations of numerical models of the system. This results in an intensive computational procedure that can be addressed only grounding the aircraft. To address this limitation, we introduce an original methodology to sensitively accelerate fault detection and isolation by reducing the computational demand to identify the health status of the system. Our scheme proposes an original combination of a novel two-step compression strategy to compute offline a synthesized representation of the dynamical response of the system, and uses an inverse Bayesian optimization approach to infer online the level of damage determined by multiple fault modes affecting the system. We demonstrate and validate our FDI algorithm against numerical and physical experiments for the case of an ElectroMechanical Actuator (EMA) employed for the secondary flight control system. Particular attention is dedicated to the case of simultaneous incipient mechanical and electrical faults considering different experimental settings, to investigate the performance and limitations of our algorithm. From numerical experiments, we observe that our methodology allows to accelerate the identification of the health status of the EMA, providing an accurate and reliable identification of incipient fault modes within seconds of computations and leveraging a reduced amount of data. Moreover, the outcomes achieved in the physical experiments validate our FDI strategy, which permits to achieve the exact identification of complex damages within two orders of magnitude faster computational time than state of the art algorithms.