INDUSTRIAL APPLICATIONS OF A FAULT DETECTION
TOOL BASED ON INTERVAL MODELS
Joaquim Armengol, Josep Veh¶³, Miguel ¶Angel Sainz, Pau Herrero
Institut d'Informµatica i Aplicacions. Universitat de Girona.
Campus de Montilivi. E-17071 Girona, Catalonia, Spain
(armengol, vehi, pherrero)@eia.udg.es, sainz@ima.udg.es
Keywords: Fault detection, Intervals, Uncer-
tain dynamic systems, Redundancy, Processes.
Abstract
One of the techniques used for the detection of faults
in dynamic systems is the analytical redundancy.
An important di±culty to apply this technique to
real systems is taking into account the uncertainties
associated with the system itself and with the
measurements. In this paper, this uncertainty is
taken into account via the use of intervals for the
parameters of the model and for the measurements.
The proposed method analyzes the consistency
between the system's behaviour, obtained from the
measurements, and the model's one: if they are
inconsistent then there is a fault.
The behaviour of the model is obtained by sim-
ulation. This problem of simulation is reformulated
as a range computation problem, which is a hard
problem but can be softened using error-bounded
estimations. To carry out the range computation,
Modal Interval Analysis is used. It provides powerful
tools to extend the calculations over real functions to
the intervals. To improve the results of the detection
of the faults, the simultaneous use of several sliding
time windows is proposed.
This method, which guarantees the absence of
false alarms, is currently being used to detect faults
in real processes. Some of these applications us-
ing real data are being developed within the Eu-
ropean project CHEM (Advanced Decision Support
System for Chemical/Petrochemical Manufacturing
Processes). One of the main tasks of the project
is the model-based situation assessment of uncertain
processes. The fault detection technique presented
in this paper is part of this task. Among these ap-
plications there are the °exible chemical pilot plant
PROCEL, owned by the Universitat Politµecnica de
Catalunya and situated in Barcelona (Spain), the
steam generator pilot plant owned by the Labora-
toire d'Automatique et d'Informatique Industrielle de
Lille (France) and the FCC (Fluid Catalytic Crack-
ing) plant owned by the French Institute of Petroleum
(IFP) and situated in Lyon (France). This paper
presents some of the results obtained in these pro-
cesses and describes the method that has been used
to obtain them.
1 INTRODUCTION
Analytical redundancy is a method to detect faults
that consists in comparing the behaviour of a real
system and a reference one obtained from a model
of the system. A fault is detected when they are
inconsistent (Reiter 1987). The main problem is that
these two behaviours are seldom the same because
the model is, by deÆnition, inaccurate, i.e. it is an
approximate representation of the system. This is
the consequence of the uncertainties of the system
and the procedure of systems' modelling.
The uncertainty of the system can be considered
in the modelling procedure. A usual model consisting
of functions with real-valued parameters is precise but
inaccurate as it does not include uncertainty. In this
paper, the parametric uncertainties associated to the
systems are represented by interval models, i.e. mod-
els in which the values of the parameters are intervals.
Therefore, an interval model is imprecise but may be
accurate.
2 FAULT DETECTION
BASED ON INTERVAL
MODELS
The simulation of a real-valued model produces a tra-
jectory for each output variable which is a curve rep-
.....