Structured hypothesis tests based diagnosis :
application to a common rail diesel injection system
Zahi SABEH, José RAGOT, Frédéric KRATZ
Delphi Diesel Systems, Centre Technique de Blois
9 boulevard de l’Industrie, 41042 Blois
zahi.sabeh@delphi.com
Centre de Recherche en Automatique de Nancy, INPL
2 avenue de la Forêt de Haye, 54516 Vandoeuvre les Nancy
Jose.Ragot@ensem.inpl-nancy.fr
Laboratoire de Vision et Robotique, Université d’Orléans, IUT de Bourges
63 avenue de Lattre de Tassigny, 18020 Bourges
Frederic.Kratz@bourges.univ-orleans.fr
Abstract : common rail injection system has been developed to increase diesel engine performances and to
reduce noise, emission and fuel consumption. Such goals are possible only if the whole system is perfectly
controlled. But, any system component failure can lead to significant engine performances decrease and
degraded emission control.
A faults diagnosis system based on structured hypothesis tests is proposed in this paper, in order to detect
and isolate different types of failures which are able to affect the pressure control loop in a common rail
diesel injection system.
Keywords : common rail injection system, diagnosis, faults detection, faults isolation, structured hypothesis
tests, modeling.
1 Introduction
The key feature of the common rail injection system
is that the injection pressure is generated
independently of engine speed and injected fuel
flow.
This characteristic property presents new horizons
for air/fuel mixture preparation and injection process
control because, from now on, the injection pressure
will be freely realizable in the cartography.
Several mechanical, electromagnetic and electronic
components (pump, electro valve, sensor…)
contribute to generate and control the injection
pressure which reaches around 1400 bars.
Any drift or failure in one of these components
implies predictably a modification of generated
pressure level.
This modification influences naturally the injection
process and move consequently the engine
performances and emission out of their optimal
areas.
Considering the previous effects, a diagnosis system
becomes essential in order to detect as early as
possible eventual failures in the pressure loop and to
suggest recovery actions.
Model based fault diagnosis has gained increased
interest during the last ten years thanks to
technological advances of the on-board electronic
control units, and increased demand on diagnosis
performance in many areas.
Many studies have been devoted to diagnosis theory
using models and analytical redundancy techniques
(Patton 1994) (Gertler 1991) (Ragot et al. 2000).
Several research works have been performed in this
field, on diesel engines in particular.
Some methods consist in training neural nets based
models (Gopinath 1994) (Chandroth et al. 1999) (Gu
F. et al. 1999) with cylinder pressure, vibration and
instantaneous speed sensors data of a diesel engine
in order to detect and identify several types of
failures having different signatures.
Interesting results have been also presented in
(Nyberg et al. 2001) (Truscott et al. 2000).
They processed different types of faults in the intake
and supercharging air paths of an automotive
turbocharged diesel engine.
This is the reason why we here suggest a model
based diagnosis method using structured hypothesis
tests.
Both the framework and the basic principles of the
method are summarized in Section 2.
Afterwards, we describe in Section 3 the process to
which the suggested method will be applied.
Finally, the diagnosis system construction as well as
some method application results will be shown in
Section 4......