Fault Diagnosis of Hybrid Dynamic Systems

In view of the growing need for availability, safety and performance of automated production systems, it is necessary to develop a diagnostics module that can detect faults affecting the operation of these systems, and locate their origin or source. The aim of the diagnostic function is to enable corrective action to be taken so that these systems return to nominal operation. This is why a diagnostics module is necessary to improve system performance and productivity, and to limit the consequences of failures, which can be catastrophic for property and human life.

Automated production systems are less and less represented by purely discrete or purely continuous behavior, but rather by a mixture of the two. These dynamic systems are known as Hybrid Dynamic Systems (HDS). These are either continuous systems to which discrete commutations, i.e. commands, are associated, or discrete-event systems to which certain continuous evolutions are associated.

SDH fault diagnosis therefore requires consideration of the interactions between continuous and discrete dynamics, in order to detect and isolate faults that affect continuous behavior (parametric faults) and/or discrete behavior (discrete faults).

The general principle of diagnostic methods is based on the use of a model representing normal and faulty system behavior. Discrete model approaches ignore the continuous dynamics of the system, while continuous model approaches do not take into account changes in the discrete modes (configurations) of the system. They are therefore unable to correctly diagnose faults that could impact SDH operation, since they do not integrate both continuous and discrete dynamics into the model. For this reason, SDH diagnostic methods are based on the use of a hybrid model representing both continuous and discrete dynamics. Several SDH modeling tools have been proposed in the literature. These include :

• hybrid controllers ;

• hybrid Petri nets ;

• hybrid statecharts ;

• hybrid bond graphs, etc.

In this article, the best-known diagnostic methods for SDH are reviewed. They are classified into three categories:

• parametric fault diagnosis methods ;

• discrete fault diagnosis methods ;

• parametric and discrete fault diagnosis methods.

The general principle of the diagnostic methods in each of these categories is presented, and the advantages and disadvantages of these methods are illustrated and compared using a number of pedagogical examples....