Data-driven model reduction and fault diagnosis for an aero gas turbine engine

Abstract

In this paper, an aero gas turbine engine with three shafts are investigated. By employing data-driven method, a reduced-order model is obtained, which has the close output performance as the 14 ^th -order full-order model. Based on the reduced-order model, a fault detection filter is designed to detect actuator faults and sensor faults for the system subjected to input and output noises. Genetic optimization algorithm is used to design the filter gains such that the residual signal is sensitive to the faults, but robust to process and sensor noises. Simulated results demonstrate the efficiency of the present algorithm.