Diagnosability analysis and diagnosis aware design

The design of fault monitors, diagnosers and fault protection systems is highly compromised when the requirements are ignored in the design phase of the system. These requirements express in terms of diagnosability and reconfigurability/repairability. These properties can be checked at design phase so that one knows before hand which faults or states the diagnoser will be able to discriminate with the specified instrumentation during operation and which fauls the controller will be able to compensate with the specified actuators. From another perspective, these properties provide the means to identify the set of ­— additional — sensors and actuators that are required to achieve a given degree of diagnosability and reconfigurability/repairability.

The research conducted by the DISCO research team spans over diagnosability for continuous systems, with emphasis on nonlinear uncertain systems, DES and hybrid systems. Importantly, a unified definition of diagnosability, which is model independent, has been established based on the notion of fault signature. Fault diagnosability has also been extended to apply not only to faults but to states and modes (sets of states). Identifiability is also studied as a key property when parameter estimation methods are used for diagnosis. Bridges with diagnosability are analysed.

Special attention is given to distributed systems, such as modern networked systems, for which models are local and no single entity has a global view of the complete system.