FDI BY EXTENDED KALMAN FILTER PARAMETER-ESTIMATION FOR AN INDUSTRIAL ACTUATOR BENCHMARK

The extended Kalman filter (EKF) is formulated as a parameter estimato r and used to estimate position sensor bias and actuator current bias signals for the industrial actuator benchmark system. These bias estim ates are compared instantaneously to a threshold for fault detection a nd identification (FDI). The paper reports results for applying this m ethod to given benchmark data. The FDI performance is good for detecti ng position sensor and actuator current faults in the presence of unmo deled nonlinear dynamics and an unmodeled load change for small-amplit ude signal conditions when the EKF implementation assumes parameter ps eudonoise and a slow decay in the parameter dynamics. For large-amplit ude signals, the results are reasonably good, but they suggest that a more accurate model for a saturation nonlinearity could improve the me thod's FDI performance.