| In this work, a fault detection and isolation strategy based on parameter estimation, a widely used method in system identification, is employed for a servo-positioning hydraulic system. Model selection criteria for fault detection and isolation purposes are presented and elaborated. The batch least squares method, well known for linear time-invariant system identification, is employed to estimate the coefficients of the selected model. The relationship between the estimated model coefficients and the actual physical parameters is explored empirically. Direct threshold checking is used as the basic fault detection and isolation logic. The performance of this strategy is evaluated based on fault conditions due to incorrect supply pump pressure and changes in equivalent viscous damping coefficient. It is shown that fault detection and isolation for faults due to a change in a single physical parameter can be achieved. However, complete fault detection and isolation is both difficult and conditional for faults originating from changes in more than one physical parameter. |
