|
The extended Kalman filter (EKF) is one of the most popular
model-based techniques for fault detection and diagnosis.
Although its effectiveness has been widely recognized, the
practical applications of EKFs are still very limited. This is
due to the fact that the estimates of EKF are often biased in a
system with multiple faults. Thus, one of the objectives of
this work is to explore the feasibility of reducing or even
eliminating the chance of bias by properly selecting a set of
measurement variables and EKF parameters. In this study, we
have extended the findings of our previous research on fault
observability and diagnostic resolution of a set of parallel
single-parameter EKFs (Chang et al., 1993) to the multiple-
parameter EKFs which are designed to locate more than one fault
origin. Specifically,the problems in implementing EKFs , i.e.
misdiagnosis due to biased estimates and heavy computation load
due to the parallel configuration, have been partially solved
with a selection strategy for the best combinations of sensor
locations and parameters in the EKF models. In addition, the
scope of fault monitoring has been expanded to include the
possibility of simultaneous occurrence of several faults. More
importantly, a simple procedure has been developed to quickly
evaluate the performance of any given system. As a result, it
becomes feasible to construct an optimum fault identification
scheme without extensive computational effort. Finally, it
should be emphasized that reliability of the proposed approach
has been confirmed in numerous simulation studies without
exceptions.
|