Symbolic identification for fault detection in aircraft gas turbine engines

Abstract

This article presents a robust and computationally inexpensive technique of component-level fault detection in aircraft gas-turbine engines. The underlying algorithm is based on a recently developed statistical pattern recognition tool, symbolic dynamic filtering (SDF), that is built upon symbolization of sensor time series data. Fault detection involves abstraction of a language-theoretic description from a general dynamical system structure, using state space embedding of output data streams and discretization of the resultant pseudo-state and input spaces. System identification is achieved through grammatical inference based on the generated symbol sequences. The deviation of the plant output from the nominal estimated language yields a metric for fault detection. The algorithm is validated for both single- and multiple-component faults on a simulation test-bed that is built upon the NASA C-MAPSS model of a generic commercial aircraft engine.