Decentralized Diagnosis of Discrete Event Systems

Fault diagnosis is the operation used to detect a fault, locate its origin and, if necessary, determine its causes. Diagnosis involves comparing instantaneous information from the system with a reference, or model, representing normal and/or faulty operation. When we are only interested in the state of the system at particular instants (events corresponding, for example, to the start and end of operations or the occurrence of faults), discrete-event models are used.

The majority of real systems are large-scale. They are made up of interconnected components that are decentralized in terms of information and geographical location. As a result, centralized diagnostic methods are not well suited to this type of system. These methods associate a single diagnostic module, called a "diagnoser", with a global model. The diagnostician collects the various information from the system in order to make a diagnostic decision. Calculating the diagnosis at a central point increases the complexity of reasoning and generates errors, loss of information and communication delays during transmission to the central point. A decentralized diagnostic approach is therefore more suitable and efficient for diagnosing real systems.

This article contains :

• the general principle of decentralized diagnosis methods for discrete-event systems ;

• the best-known decentralized diagnostic methods.