Optimization of observer design for sensor fault detection of nonlinear systems

This paper presents an optimization methodology for the design of an observer-based sensor fault detection scheme for a class of nonlinear systems. Taking into account bounded system disturbances and measurement noise, we design an observer aiming at maximizing the set of faults that are guaranteed to be strongly detectable. Strong fault detectability conditions are derived based on the limit sets that bound the residual under healthy and faulty conditions. A novel optimization method is designed based on the separation of the healthy and faulty limit sets. The distance between these sets represents the trade-off between robustness and sensor fault sensitivity. Simulation results are used to show the effectiveness of the proposed methodology applied to a simple example of a flexible link robot.