Diagnosis of discrete-event systems by separation of concerns, knowledge compilation, and reuse

Model-based diagnosis of discrete-event systems (DESs) requires the reconstruction of the behavior of the system to be diagnosed, which is computationally expensive and, therefore, time-consuming. Accordingly, most approaches propose a trade-off between off-line and on-line computation: suitable knowledge, derived off-line from the model of the system, can be exploited on-line based on the actual observation. This way, a large amount of model-based reasoning is anticipated off-line, thereby making the on-line task considerably lighter. The essential novelty of this paper, which aims to support the diagnosis of asynchronous DESs, lies in the ability to exploit not only the general-purpose diagnostic knowledge compiled off-line but also the special-purpose knowledge generated on-line for the solution of previous problems, thereby pursuing processing reuse. To this end, compatibility checking is required: the solution of a new diagnostic problem can exploit the solution of another problem provided the latter subsumes the former.