Fault diagnosis based on algebraic identification assisted by extended state observers

This paper explored different alternatives for fault diagnosis based on algebraic identification methods. As a novelty, the use of Extended State Observers (ESO) is proposed to estimate time-derivatives related to nonlinear complex terms, which have a difficult analytical handling from the perspective of the classic algebraic framework. In this context, three fault diagnosis schemes focused on residual generation, fault identification and parameter estimation are presented and compared. These strategies estimate as few parameters as possible, making their implementation simpler. To validate the proposed approach, a twin-rotor aerodynamic system affected by multiple, intermittent and simultaneous actuator faults was considered as case study. Finally, experimental results that show the effectiveness and robustness of the proposed strategies are provided.