Model-based diagnosis using MathModelica
In industrial processes, sudden faults must quickly be determined to avoid
general failures. For this purpose model-based fault diagnosis can be used,
which requires a consistent model of the real process. The powerful
modelling tool MathModelica, based on the Modelica language, can be used
to accomplish this. Systems for fault diagnosis could be both time-
consuming and expensive if built manually. Instead, to automatically
generate a fault diagnosis system, based on models built with MathModelica,
would provide an efficient means of fault diagnosing. This thesis is about
an algorithm which puts this into practice.
MathModelica has previously not been used for fault diagnosis, which makes
this a pioneering work. Therefore, the algorithm is limited to consider
only static electrical circuits and to diagnose constant voltage sources
and linear resistors.
The algorithm takes a MathModelica model of a circuit and observations
from the corresponding real system as input. Then a fault diagnosis system
is generated and all possible diagnoses are obtained. The complexity of
generating the diagnosis system grows very fast when the number of
components is increased. Therefore, the capacity of the used computer puts
limitations on the algorithm. An interesting extension would be to make
the algorithm independent of the size of the circuit concerned, which
could be done by considering subsets of the circuit.
Senast uppdaterad: 2019-06-05