LAUSR

This study investigates the problem of reconstructing actuator faults for descriptor systems via a PD-type learning observer. By synthesising the derivatives of the output estimation error into the P-type learning law, a novel PD-type learning observer is established to simultaneously reconstruct original system states and actuator faults. Stability analysis of the PD-type learning observer is explicitly provided. A systematic design method is also suggested based on a linear matrix inequality technique. Further, a robust PD-type learning observer is designed against process disturbances and measurement noises. At last, a simulation example is used to demonstrate the effectiveness of the proposed fault-reconstructing method.