Robust identification and residual generation application to a turbofan engine

摘要

We describe an application of a model-based fault detection and isolation technique to a turbofan aircraft engine. Our approach is based in H/sub /spl infin//-optimization, a frequency domain technique that explicitly addresses robustness. The engine model is obtained through a robust identification approach combining the classical time-data system identification via prediction error models and the so-called model error model theory. The time-data series are provided by one of Honeywell's airline customers. The model error model approach provides guidelines for the characterization of the model uncertainty stemming from the nominal system identification. The H/sub /spl infin// filter is used to detect faults in the engine's high-pressure turbine while minimizing disturbances and commands effects.