Symbolic transient time-series analysis for fault detection in aircraft gas turbine engines

This paper presents a data-driven symbolic dynamics-based method for detection of incipient faults in gas turbine engines of commercial aircraft. Detection of incipient faults in such engines could be significantly manifested by taking advantage of transient data (e.g., during takeoff). From this perspective, the fault detection and classification algorithms are built upon the recently reported work on symbolic dynamic filtering. In particular, Markov model-based analysis of steady state data is extended by taking advantage of the available transient data. The fault detection and classification procedure has been validated on the NASA C-MAPSS transient test case generator.

I. Introduction

Several researchers have addressed performance monitoring of aircraft gas turbine engines based on quasi-stationary steady-state data during cruise conditions. However, transient operational data (e.g., from takeoff, climb, or landing) can be gainfully utilized for early detection of incipient faults. Usually engines operate under much higher stress and temperature conditions under transient operations compared to those under steady-state conditions. Signatures of certain incipient engine faults (e.g., bearing faults, controller miss-scheduling, and starter system faults) tend to magnify during the transient conditions, and appropriate usage of transient data could enhance the probability of fault detection [1]. Along this line, several model-based and data-driven fault diagnosis methods have been proposed by making use of transient data. For example, Srendar and Ganguli [2] used an adaptive Myriad filter to improve the quality of transient operations for gas turbine engines. However, model-based diagnostics may suffer from inaccuracy and loss of robustness due to low reliability of the transient models.