The emergence of stronger safety and environmental

norms, the need for early decision mechanisms, together

with the widespread diusion of sensors of all kinds, result

in an increased need for research investigations in the eld

of fault detection and isolation (FDI). In particular, e-

cient and robust methods for fault and damage detection,

diagnostics and localization, are necessary for condition-

based maintenance and for fatigue and aging prevention.

Many approaches to FDI problems assume that a model of

the monitored system is available. This is a reasonable as-

sumption since many machines and processes rely on phys-

ical principles and models can be built on that basis (Ljung

(1999)). However, it is often argued that a major drawback

of model-based FDI methods lies in the requirement of a

precise model of the system to be monitored. In particular,

changes in the functioning modes and un