Diagnosing systems modeled with piecewise linear constraints

Abstract

Most approaches to model-based diagnosis use incomplete constraint satisfaction techniques and are therefore incomplete. It is shown here how diagnostics may be considered as a general constraint satisfaction problem. Provided the system to be repaired is modeled with piecewise linear constraints on real variables, it is possible to give a clear characterization of the conflict set. Furthermore, a complete algorithm to find every conflict is described. This algorithm is an original and elegant solution to the diagnostic problem. Moreover, it can be used in order to build a compiled form of the initial model. Once this compiled model has been obtained, all the possible faults of a given system can be identified quite easily.<>