Fault detection and isolation in aircraft gas turbine engines. Part 1: Underlying concept

Abstract

Degradation monitoring is of paramount importance to safety and reliability of aircraft operations and also for timely maintenance of its critical components. This two-part paper formulates and validates a novel methodology of degradation monitoring of aircraft gas turbine engines with emphasis on detection and isolation of incipient faults. In a complex system with multiple interconnected components (e.g. an aircraft engine), fault isolation becomes a crucial task because of possible input-output and feedback interactions among the individual components. This paper, which is the first of two parts, presents the underlying concepts of fault detection and isolation (FDI) in complex dynamical systems. The FDI algorithms are formulated in the setting of symbolic dynamic filtering (SDF) that has been recently reported in literature. The underlying concept of SDF is built upon the principles of symbolic dynamics, statistical pattern recognition, and information theory. In addition to abrupt large faults, the SDF-based algorithms are capable of detecting slowly evolving anomalies (i.e. deviations from the nominal behaviour) based on analysis of time series data of critical process variables of different engine components. The second part, which is a companion paper, validates the concept, laid out in the first part, on the simulation test bed of a generic two-spool turbofan aircraft engine model for detection and isolation of incipient faults.

Anomaly Detection on Gas Turbine Fuel System Using a Sequential Symbolic Method

“…The symbolic dynamic filtering (SDF) [30] was proposed and yield good performance in anomaly detection for robustness in comparison to other method such as principal component analysis (PCA), ANN and Bayesian approach [31] as well as proper accuracies. Gupta et al [32] and Sarkar et al [33] presented a SFD-based model for detecting fault of gas turbine subsystem and used it to estimate multiple component faults. Sarkar et al [34] proposed an optimized feature extraction method under the SDF framework [34].…”

Implementation of fault detection and prediction capabilities in a multilevel current source inverter

“…Sensor redundancy and analytic measurements are examples of these approaches that have been widely researched during the last 40 years. Examples of redundant sensor installations in complex engineering applications are: (a) inertial navigation sensors in both tactical and transport aircraft for guidance and control [1,2], (b) power, temperature, pressure, and flow sensing in nuclear and fossil-fueled power plants for health monitoring and feedforward-feedback control [3,4,5], and (c) pressure, temperature and shaft speed sensors for detection and isolation of incipient faults in one or more components of an aircraft gas turbine engine [6,7]. Sensor redundancy is often augmented with analytical measurements obtained from physical characteristics and/or physical model of the plant dynamics in combination with other available sensor data [8].…”

“…Since then there has been an extensive evolution reflected in the literature [11,12,13,14] and summarized in [15]. These FDII procedures have been successfully applied to complex engineering systems (e.g., nuclear reactors [11,12,13,16], fossil-fueled power plants [14,17,18], and aircraft gas turbine engines [6,7]). The technological platforms on which FDII algorithms were implemented also show a sharp evolution, starting with minicomputers [16] in the late 70s, to microprocessor boards in the 80s [3], going to microcontrollers in the 90s [4].…”

“…electronic circuits [20] and fatigue damage monitoring in polycrystalline alloys [21]). Recently an SDF-based algorithm for detection and isolation of aircraft gas turbine engine subsystem faults (specifically, faults that cause efficiency degradation in engine components) has been reported [6,7]. However, the patterns generated from the time series of a single sensor may not carry sufficient information to correctly diagnose an evolving fault.…”

In-Flight Isolation of Degraded Engine Components by Shape Comparison of Transient Outputs

“…They are then contrasted for component isolation with the characteristic signatures of individual components on the outputs. Direct solutions to engine health monitoring generally require training to determine the component characteristic signatures; e.g., [13][14][15]. This has deterred development of direct solutions for engine health monitoring.…”