Sébastien Borguet

Comparison of adaptive filters for gas turbine performance monitoring

Kalman filters are widely used in the turbine engine community for health monitoring purpose. This algorithm has proven its capability to track gradual deterioration with a good accuracy. On the other hand, its response to rapid deterioration is either a long delay in recognising the fault, and/or a spread of the estimated fault in several components. The main reason of this deficiency lies in the transition model of the parameters that assumes a smooth evolution of the engines condition. The aim of this contribution is to compare two adaptive diagnosis tools that combine a Kalman filter and a secondary system that monitors the residuals. This auxiliary component implements on one hand a covariance matching scheme and on the other hand a generalised likelihood ratio test to improve the behaviour of the diagnosis tool with respect to abrupt faults.

The Fisher Information Matrix as a Relevant Tool for Sensor Selection in Engine Health Monitoring

Engine health monitoring has been an area of intensive research for many years. Numerous methods have been developed with the goal of determining a faithful picture of the engine condition. On the other hand, the issue of sensor selection allowing an efficient diagnosis has received less attention from the community. The present contribution revisits the problem of sensor selection for engine performance monitoring within the scope of information theory. To this end, a metric that integrates the essential elements of the sensor selection problem is defined from the Fisher information matrix. An example application consisting in a commercial turbofan engine illustrates the enhancement that can be expected from a wise selection of the sensor set.

A Generalized Likelihood Ratio Test for Adaptive Gas Turbine Performance Monitoring

Kalman filters are widely used in the turbine engine community for health monitoring purposes. This algorithm has proven its capability to track gradual deterioration with good accuracy. On the other hand, its response to rapid deterioration is a long delay in recognizing the fault and/or a spread of the estimated fault on several components. The main reason for this deficiency lies in the transition model of the parameters that is blended in the Kalman filter and assumes a smooth evolution of the engine condition. This contribution reports the development of an adaptive diagnosis tool that combines a Kalman filter and a secondary system that monitors the residuals. This auxiliary component implements a generalized likelihood ratio test in order to detect and estimate an abrupt fault. The enhancement in terms of accuracy and reactivity brought by this adaptive Kalman filter is highlighted for a variety of simulated fault cases that may be encountered on a commercial aircraft engine.

A sensor-fault-tolerant diagnosis tool based on a quadratic programming approach

Kalman filters are widely used in the turbine engine community for health monitoring purpose. This algorithm gives a good estimate of the engine condition provided that the discrepancies between the model prediction and the measurements are zero-mean, white random variables. However, this assumption is not verified when instrumentation (sensor) faults occur. As a result, the identified health parameters tend to diverge from their actual values, which strongly deteriorates the diagnosis. The purpose of this contribution is to blend robustness against sensor faults into a tool for performance monitoring of jet engines. To this end, a robust estimation approach is considered and a sensor-fault detection and isolation module is derived. It relies on a quadratic program to estimate the sensor faults and is integrated easily with the original diagnosis tool. The improvements brought by this robust estimation approach are highlighted through a series of typical test cases that may be encountered on current turbine engines.

A Methodology to Improve the Robustness of Gas Turbine Engine Performance Monitoring Against Sensor Faults

For turbine engine performance monitoring purposes, system identification techniques are often used to adapt a turbine engine simulation model to some measurements performed while the engine is in service. Doing so, the simulation model is adapted through a set of so-called health parameters whose values are intended to represent a faithful image of the actual health condition of the engine. For the sake of low computational burden, the problem of random errors contaminating the measurements is often considered to be zero mean, white, and Gaussian random variables. However, when a sensor fault occurs, the measurement errors no longer satisfy the Gaussian assumption and the results given by the system identification rapidly become unreliable. The present contribution is dedicated to the development of a diagnosis tool based on a Kalman filter whose structure is slightly modified in order to accommodate sensor malfunctions. The benefit in terms of the diagnostic reliability of the resulting tool is illustrated on several sensor faults that may be encountered on a current turbofan layout.

Adaptive Estimation Algorithm for Aircraft Engine Performance Monitoring

In the frame of turbine engine performance monitoring, system identification procedures are often used to adapt a simulation model of the engine to some observed data through a set of so-called health parameters. Doing so, the values of these health parameters are intended to represent the actual health condition of the engine. The Kalman filter has been widely used to achieve the identification procedure in real-time onboard applications. However, to achieve a proper filtering of the measurement noise, the health parameters are often assumed to vary in time relatively slowly, preventing any abrupt accidental events from being tracked effectively. This contribution presents a procedure called adaptive filtering. Based on a covariance-matching method, it is intended to automatically release the health parameters once an accidental event is detected. This enables the Kalman filter to deal with both continuous and abrupt fault conditions.

Aircraft Engine Gas Path Diagnostic Methods: Public Benchmarking Results

Abstract Recent technology reviews have identified the need for objective assessments of aircraft engine health management (EHM) technologies. To help address this issue, a gas path diagnostic benchmark problem has been created and made publicly available. This software tool, referred to as the Propulsion Diagnostic Method Evaluation Strategy (ProDiMES), has been constructed based on feedback provided by the aircraft EHM community. It provides a standard benchmark problem enabling users to develop, evaluate and compare diagnostic methods. This paper will present an overview of ProDiMES along with a description of four gas path diagnostic methods developed and applied to the problem. These methods, which include analytical and empirical diagnostic techniques, will be described and associated blind-test-case metric results will be presented and compared. Lessons learned along with recommendations for improving the public benchmarking processes will also be presented and discussed. Introduction

ON-LINE TRANSIENT ENGINE DIAGNOSTICS IN A KALMAN FILTERING FRAMEWORK

A common assumption made in the performance assessment of a turbine engine for aircraft propulsion consists in restricting the data processing to steady-state data. This especially holds for onboard performance monitoring of a commercial aircraft which spends up to 90% of the time in cruise flight where such conditions are satisfied. The present contribution is intended to investigate the ability of a diagnosis method to process unsteady data rather than steady-state data. The aim of this unsteady approach is to strongly reduce the time and the efforts spent to obtain a reliable diagnosis. In order to assess the improvements in terms of diagnosis efficiency and engine operability, the resulting diagnostic method is tested for different degradations that can be expected on commercial turbofans. The results are also compared to those obtained from cruise flight steady-state data in order to balance the two approaches.Copyright

A Sparse Estimation Approach to Fault Isolation

Least-squares-based methods are very popular in the jet engine community for health monitoring purposes. In most practical situations, the number of health parameters exceeds the number of measurements, making the estimation problem underdetermined. To address this issue, regularization adds a penalty term on the deviations of the health parameters. Generally, this term imposes a quadratic penalization on these deviations. A side effect of this technique is a relatively poor isolation capability. The latter feature can be improved by recognizing that abrupt faults impact at most one or two component(s) simultaneously. This translates mathematically into the search for a sparse solution. The present contribution reports the development of a fault isolation tool favoring sparse solutions. It is very efficiently implemented in the form of a quadratic program. As a validation procedure, the resulting algorithm is applied to a variety of fault conditions simulated with a generic commercial turbofan model.

Ten Years of Experience With a Small Jet Engine as a Support for Education

In 1997 the Turbomachinery Group of the University of Liege decided to acquire a small jet engine to illustrate the courses in propulsion and to provide the students with the opportunity to get some experience on data measurement, acquisition, and interpretation. Among others, the SR-30 engine from Turbine Technology Ltd. Chetek, WI was chosen. It consists of a single spool, single flow engine with a centrifugal compressor, a reversed combustion chamber, an axial turbine, and a fixed convergent nozzle. This engine was installed on a test bench allowing for manual control and providing fuel and oil to the engine. The original setup included measurements of intercomponent pressure and temperatures, exhaust gas temperature, and rotational speed. Since then both the engine and the test bench have been deeply modified. These modifications were led by a triple objective: the improvement and the enrichment of the measurement chain, the widening of the engines operational domain, and, last but not the least, the wish to offer appealing hands-on projects to the students. All these modifications were performed at the University of Liege and were conducted by the students as part of their Master theses. Several performance models of the engine were developed to support data validation and engine condition diagnostic. This paper summarizes the developments conducted with and by the students, and presents the experience that was gained by using this engine as a support for education.