It is well known that there is an increasing demand for modern systems to become
safer and reliable. This real world’s development pressure has transformed fault
diagnosis, initially perceived as the art of designing a satisfactorily safe system,
into the modern science that it is today.
Indeed, the classic way of fault diagnosis boils down to controlling the limits of
single variables and then using the resulting knowledge for fault-alarm purposes.
Apart from the simplicity of such an approach, the observed increasing complexity
of modern systems necessitates the development of new fault diagnosis
techniques. Such a development can only be realised by taking into account the
information hidden in all measurements. One way to tackle such a challenging
problem is to use the so-called model-based approach. Indeed, the application
of an adequate model of the system being supervised is very profitable with respect
to gaining the knowledge regarding its behaviour. A further and deeper
understanding of the current system behaviour can be achieved by implementing
parameter and state estimation strategies. The obtained estimates can then be
used for supporting diagnostic decisions.