Output estimator based fault detection for a class of nonlinear systems with unknown inputs

Fault detection problem is studied using output estimator design rather than observer design for a class of nonlinear systems with unknown inputs. In order to carry out the output estimator design, an input-output relation is derived from the original state space model, which removes the effect of the unknown inputs on fault detection completely. Based on the input-output relation, a necessary and sufficient condition is derived for the existence of an output estimator which is invariant to the unknown inputs. An output estimator is designed based on the input-output relation for the purpose of fault detection, and a novel fault detection strategy using output estimator design is proposed which neither has the relative degree restriction, nor requires the system under consideration to be detectable. The efficacy of the proposed fault diagnosis strategy is tested on a single-link flexible robot manipulator model thorough computer simulations.