Unforeseen failures of hydraulic power generation systems can cause costly operation interruptions of aircraft. To overcome this, a health monitoring of the system components is desirable but not yet available. In this paper, a method is presented that allows a model-based development of such a capability.
This method uses physics-based models of system components in order to derive their behavior at considered fault modes with several extents. The gained knowledge is transferred into a model of the entire hydraulic system. This model is used to determine the fault effects on the system states in order to find candidate positions for sensors. The method allows finding an optimal set of sensors that enables the detection and isolation of system component faults. The system model is used further for the creation of a simplified steady-state nominal model, which is required for feature generation. On the basis of these features, support vector machines are trained for fault detection and isolation as well as the diagnosis of the actual fault extent. The developed health monitoring functions are validated on a test rig.