Fault diagnosis using a parameter-estimation method is investigated in
this paper. A digital state variable filter is employed to obtain der
ivatives of the variables, and an interpolation technique is used to a
pproximate the values of the variables between samples. The method is
applied to a hydraulic test rig based on real data, and the simulated
faults - changes in the physical parameters - are diagnosed. The simul
ation results demonstrate that faults having the same direction in the
state space of the system model can be isolated using this technique,
which is impossible using any observer-based method or any parity equ
ation method. Fault size can also be approximately identified. Copyrig
ht (C) 1997 Elsevier Science Ltd.