Fault diagnosis in a momentum wheel system based on unknown input observer

The momentum wheel in the satellite is a complex non-linear closed-loop system with unknown input. The unknown input and the feedback increase the difficulty of its fault diagnosis. Firstly, this paper proposes a method in theory to detect and recognize the fault in a closed-loop system. This method diagnoses the different faults through an unknown input observer (UIO) as well as a common observer which is set behind the control mechanism. Then a model of the momentum wheel in the satellite is built and the previous theory is applied to the model. In the end, the experimental data demonstrates that this method can successfully detect and recognize the sensor failure, the control mechanism failure and the system parameter failure. The method proposed in this paper has reference value for other non-linear closed-loop systems with unknown input as long as they meet all the conditions.