Unified Diagnostic Filtering for Turbofan Engine Control Systems Based on Matrix Pencil Observers

Motivated by decoupling the adverse effect of exogenous disturbances on dynamic state filtering, unknown input observers (UIO’s) have received considerable attention in the fault diagnosis domain over the past decades. The model used in the UIO theory, however, does not include any disturbance item in system outputs. This can impair the robustness of diagnostic performance in the face of output disturbances. In this context, a unified filter, intended for both fault detection and isolation (FDI), is presented through the application of matrix pencil decomposition. First of all, a general fault model for the disturbed engine is formulated, with output disturbances considered. Then, this engine model is transformed by means of decomposing a matrix pencil derived from model matrices in a block triangular form. Based on this decomposition, a reduced-order subsystem can be obtained which is not affected by external disturbances. Finally, a single conventional filter with regional pole assignment constraints is designed to detect and isolate faults of an engine control system. Simulation results on one turbofan engine show the advantages of this given method over the UIO and Kalman filter.Copyright © 2012 by ASME