On the construction of modular observers and diagnosers for discrete-event systems

A modular approach for designing observers and diagnosers for detecting faults in large distributed systems is presented. In particular, the model is based on a discrete-event system framework. The model assumes that the system is composed of distributed components that interact with each other via sets of common resources. Modularity can be achieved by imposing a total ordering on access to the common resources. Further, a component´s own access to a common resource is an observable event. It is then possible to partition component behavior such that an observer of a component´s behavior need not be aware of all the behavior of other components that use the same resources. Procedures for the design of such observers and the subsequent construction of diagnosers are given