Much of the earlier work presented in the area of computerised on line
fault diagnosis, has been based on heuristic principles, where knowle
dge has been represented in some kind of rule structure. This paper pr
esents an overview of the possibilities and pitfalls of different mode
lling techniques in fault diagnosis. Experiences from a large industri
al project, KSM, are presented, where results from risk identification
are used in on line fault diagnosis in two industrial plants. The nov
elties of this system is the introduction of characterisations in orde
r to evaluate the most probable cause among several possible causes. A
nother feature is the support system for detecting possible operator e
rrors after a deviation has occurred. This facilitates the decision of
which information should be given to the operator from the support sy
stem. Future development, where qualitative mathematical modelling is
used in order to find the most probable cause among several possible,
is also discussed.