Concurrent Fault Diagnosis Based on an Extended Kalman Filter

This work introduces a novel methodology to detect, isolate, and estimate (i.e. to diagnose) simultaneous faults in a class of nonlinear systems, when faults are constant and they occur in the system's actuators. The methodology augments the system's model by adding
$k$
new states representing, each one, the dynamics of a fault. Then, it builds an extended Kalman Filter (EKF) to estimate the faults. An advantage of our approach lies in the fact that a single filter can be used to diagnose all the faults, even if they occur simultaneously, and thus, the designing process is significantly reduced. The fault estimation can be useful in the subsequent Fault Tolerant Control stages. In order to show the effectiveness of the proposed approach, experimental results computed from the actual data logged obtained from a real three-wheel omnidirectional mobile robot are shown.