Application of system identification techniques to aircraft gas turbine engines

Evans, C.; Fleming, Peter J.; Hill, David; Norton, J.P.; Pratt, I. H.; Rees, D.; Rodríguez-Vázquez, Katya

doi:10.1016/s0967-0661(00)00091-5

Cited by 42 publications

(19 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Multisines have been used to a large extent for system identification, see for instance [19,20] for some applications to turbine engines. A number of advantages explain their popularity:…”

Section: Time Domainmentioning

confidence: 99%

A Study on Engine Health Monitoring in the Frequency Domain

Borguet

Henriksson

McKelvey

et al. 2010

Volume 3: Controls, Diagnostics and Instrumentation; Cycle Innovations; Marine

View full text Add to dashboard Cite

Most of the techniques developed so far for module performance analysis rely on steady-state measurements from a single operating point to evaluate the level of deterioration of an engine. One of the major difficulties associated with this estimation problem comes from its underdetermined nature. It results from the fact that the number of health parameters exceeds the number of available sensors. Among the panel of remedies to this issue, a few authors have investigated the potential of using data collected during a transient operation of the engine. A major outcome of these studies is an improvement in the assessed health condition.The present contribution proposes a framework that formalises this observation for a given class of input signals. The analysis is performed in the frequency domain, following the lines of system identification theory. More specifically, the meansquared estimation error is shown to drastically decrease when using transient input signals. The study is conducted with an engine model representative of a commercial turbofan. Keywords: Gas path analysis, frequency domain, least-squares, system identification. N (m,C) the Gaussian probability density function with mean m and covariance matrix C NOMENCLATURE E(·) INTRODUCTIONPredictive maintenance aims at scheduling overhaul actions on the basis of the actual level of deterioration of the engine. The benefits are improved dispatch reliability and safety as well as reduced life cycle costs. Generating reliable information about the health condition of the gas turbine is therefore a requisite and has been the subject of intensive research in the community.In this paper, Module Performance Analysis is considered. Its purpose is to detect, isolate and quantify the changes in engine module performance, described by so-called health parameters, on the basis of measurements collected along the gas path of the engine [1]. Typically, the health parameters are correcting factors on the efficiency and flow capacity of the modules (fan, LPC, HPC, HPT, LPT) while the measurements are intercomponent temperatures, pressures, shaft speeds and fuel flow. Since the pioneering work by Urban [2], most of the literature on module performance analysis has considered the processing of data observed during steady-state operation of the engine.In this framework, the estimation of the health parameters can be cast as an optimisation problem which is characterised by a number of difficulties. The underlying process is non-linear, engine measurements are subject to noise and bias and the introduction of model inaccuracy is inevitable. Moreover, the number of sensors is usually smaller than the number of health parameters, making the problem underdetermined.One obvious way to tackle this non-uniqueness in the solution would be to complement the sensor set installed on the engine, but significant additional costs would be incurred. Hence, this solution on the hardware side is disregarded by engine manufacturers. Two roads of remedy on the software side have the...

show abstract

“…Multisines have been used to a large extent for system identification, see for instance [19,20] for some applications to turbine engines. A number of advantages explain their popularity:…”

Section: Time Domainmentioning

confidence: 99%

A Study on Engine Health Monitoring in the Frequency Domain

Borguet

Henriksson

McKelvey

et al. 2010

Volume 3: Controls, Diagnostics and Instrumentation; Cycle Innovations; Marine

View full text Add to dashboard Cite

show abstract

“…Wavelet coefficients are also considered to provide a certain degree of interpretability, as they were used in [44], where genetic algorithms were applied to select wavelet threshold parameters in an exchange-rate forecasting problem. Another approaches for nonlinear modeling use polynomial models, as can be seen in [12,39]. Many other different, problem specific, analytical modeling approaches had been developed.…”

Section: Evolutionary Transparent Modeling Of Chaotic Systemsmentioning

confidence: 99%

“…In some of our own previous works [13], we have pro-posed to use a linear combination of the quadratic error and the largest Lyapunov exponent for the fitness function, and have optimized it by means of a genetic algorithm. In [12,39] a Pareto-based approach is used instead of scalar functions, in combination with the MOGA algorithm described in [15]. There are some different Pareto-based multiobjective strategies that could also be applied for the same problem, as can be seen in [6].…”

Section: Evolutionary Transparent Modeling Of Chaotic Systemsmentioning

confidence: 99%

“…Chaotic signals are not an exception to this rule: we expect a technique that produces a black box from data [12,20,25,32,34,41] Luciano Sánchez and José R. Villar Computer Science Department, Universidad de Oviedo, Edificio Departamental 1, Campus de Viesques, 33213 Gijon (Spain), Tel. : +34 985182597; fax: +34 985 181 986., e-mail: villarjose@uniovi.es to produce more accurate results than other procedures that also gains insight into the block structure of the system.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Multiobjective Evolutionary Search of Difference Equations-based Models for Understanding Chaotic Systems

Sánchez

Villar

2008

Foundations of Generic Optimization

View full text Add to dashboard Cite

In control engineering, it is well known that many physical processes exhibit a chaotic component. In point of fact, it is also assumed that conventional modeling procedures disregard it, as stochastic noise, beside nonlinear universal approximators (like neural networks, fuzzy rule-based or genetic programming-based models,) can capture the chaotic nature of the process.In this chapter we will show that this is not always true. Despite the nonlinear capabilities of the universal approximators, these methods optimize the one step prediction of the model. This is not the most adequate objective function for a chaotic model, because there may exist many different nonchaotic processes that have near zero prediction error for such an horizon. The learning process will surely converge to one of them. Unless we include in the objective function some terms that depend on the properties on the reconstructed attractor, we may end up with a non chaotic model. Therefore, we propose to follow a multiobjective approach to model chaotic processes, and we also detail how to apply either genetic algorithms or simulated annealing to obtain a difference equations-based model.

show abstract

“…LTV identification with optimal smoothing is long established and has been found effective in a range of applications (Norton, 1975(Norton, , 1976Young et al, 1991;Evans et al, 2001;Chanat and Norton, 2003;Norton and Chanat, 2005). This paper examines modifications to extend its scope.…”

Section: Introductionmentioning

confidence: 99%

Identification of Time-Varying Modal Models

Norton

2005

IFAC Proceedings Volumes

Self Cite

View full text Add to dashboard Cite

Modal, decoupled models of linear systems have easily interpreted parameters: gains and poles for independent rate processes, hence time constants and damping ratios. Time-varying modal models can fit a wide range of time-invariant, non-linear systems, and the time variation shows the effects of non-linearity. Modal parameters can be estimated via an ARMAX model, but the calculation is ill conditioned, so direct estimation of time-varying modal parameters is examined. The modal components of the output cannot be observed so must be estimated along with the parameters, a nonlinear estimation problem. It is treated as state estimation, using the extended Kalman filter (EKF) and optimal smoothing. The capabilities and limitations of the EKF are investigated by simulation. With care, good results can be obtained even in difficult cases, e.g. with abrupt change in an input or output offset. Coypright 2005 IFAC

show abstract

Application of system identification techniques to aircraft gas turbine engines

Cited by 42 publications

References 12 publications

A Study on Engine Health Monitoring in the Frequency Domain

A Study on Engine Health Monitoring in the Frequency Domain

Multiobjective Evolutionary Search of Difference Equations-based Models for Understanding Chaotic Systems

Identification of Time-Varying Modal Models

Contact Info

Product

Resources

About