Robust fault detection in systems with uncertainties

A new methodology for instrument failure detection in linear dynamical systems subject to plant parameter variations or uncertainties is presented. The proposed fault-tolerant control system is based on a single robust estimator that can simultaneously estimate the unmeasurable state variables of the system, for the purpose of control, and provides necessary information to a detection logic device capable of detecting and isolating actuator and sensor failures. In addition, it is possible to identify the shape and magnitude of the failures with this approach. The algorithm allows for fast and accurate fault detection, and can account for structural uncertainties and variations in the parameters of the system´s dynamical model. At the heart of this approach is the robust estimator, for which the necessary and sufficient conditions to its existence are given. The overall fault-tolerant control system strategy proposed is verified through simulation studies performed on the control of a helicopter in the vertical plane