University of Liege

Turbomachinery group

Olivier LEONARD, Professor Tel : +32 (0)4 366 91 87Institut de Mecanique et Genie civil (B52) Fax : +32 (0)4 366 91 36Chemin des chevreuils 1 O.Leonard@ulg.ac.beB-4000 Liege Belgium http ://www.ulg.ac.be/turbo

1 Brief description of the group

The Turbomachinery group is part of the Department of Aerospace and Mechanical En-gineering which involves 23 professors and about 130 researchers. The staff of the Turbo-machinery group involves 1 professor, 4 research engineers and 1 technician.

The group is devoted to teaching in the fields of turbomachinery (compressors, gas andsteam turbines, hydraulic machines) and aerospace propulsion (turbojets and rocket en-gines). Its research activities are organized along 3 major axis :

– The numerical simulation of flows in turbomachines for design purpose, withthe development of a throughflow finite-volume computer code and a multi-stage, bladerow per blade row, finite-volume mean-line code.

– The development of optimization techniques using genetic algorithms with ap-plication to the optimal design and operation of pumps, heat pipes, fans or blowers,

– The development of methods for measurements validation, health monitoringand model-based adaptive control of gas turbine engines.

The staff of the Turbomachinery Group involves 1 professor, 5 research engineers and 1technician :

Olivier ADAM (o.adam@ulg.ac.be) PhD StudentSebastien BORGUET (s.borguet@ulg.ac.be) Assistant - PhD StudentWenhai DU (wenhai.du@gmail.com) PhD StudentVincent KELNER (v.kelner@ulg.ac.be) PhD StudentOlivier LEONARD (o.leonard@ulg.ac.be) ProfessorJean-Philippe THOMAS (jp.thomas@ulg.ac.be) PhD StudentRichard LABENDA (R.Labenda@ulg.ac.be) Technician

The group is devoted to teaching in the fields of turbomachinery (compressors, gas andsteam turbines, hydraulic machines) and aerospace propulsion (turbojets and rocket en-gines). Its research activities are organized along 3 major axis :

– The numerical simulation of flows in turbomachines for design purpose, withthe development of a throughflow finite-volume computer code and a multi-stage, bladerow per blade row, finite-volume mean-line code.

– The development of optimization techniques using genetic algorithms with ap-plication to the optimal design and operation of pumps, heat pipes, fans or blowers,

– The development of methods for measurements validation, health monitoringand model-based adaptive control of gas turbine engines.

The Turbomachinery group has developed relationships with the von Karman Institute,CENAERO, ONERA, Ecole Centrale de Lyon, National Technical University of Athens,Chalmers, Snecma, Techspace Aero, Fluorem, Rutten s.a. and others.

Before heading the group, Olivier LEONARD carried out the main part of his researchactivities at the von Karman Institute, within the Turbomachinery Department.

Commercial codes such as FLUENT, FINE/TURBO and ECOSIMPRO are used withinthe group for flow simulations. A test bench for hydraulic pumps and a small jet engineequipped with data acquisition, fuel control and variable nozzle area are available for theresearch activities.

2 Health Monitoring and Model-Based Control

of Gas Turbine Engines

The first objective of this project is to develop and validate parametric and adaptivemodels for the health monitoring of gas turbine engines, so as to enable condition-basedmaintenance. The identification of the health parameters of the engines is based on mea-surements taken on the process and includes their validation. It makes it possible to followthe evolution with respect to time (a possible degradation) of the condition of the engine.The adaptation of the non linear model and the estimation of the health of the engine arebased on the use of Kalman filters, modified to take into account the nonlinear characterof the operation of the gas turbine engines and to allow the detection of erroneous mea-surements. The most recent versions of the developed algorithms are based on dynamicmodels and can benefit from the transients followed by the engines.

These tools can also provide an estimate of non measurable variables which are fundamen-tal for control such as maximum temperature or surge margin. The second objective ofthe project is to use these adaptive models and the diagnosis strategy to carry out actionsof control and adaptation of the gas turbine operation which would take into account itsreal, modified or degraded condition.

Partners : National Technical University of Athens, Chalmers, Techspace Aero

Recent related publications :

– Robust Validation of Measurements on Jet Engines, P. Dewallef, O. Leonard, EuropeanJournal of Mechanical and Environmental Engineering, Vol 46, No 4, 2001

– On-Line Validation of Measurements on Jet Engines Using Automatic Learning Me-thods, P. Dewallef, O. Leonard, Proceedings of the XV International Symposium onAirbreathing Engines, Bangalore, 2001

– On-Line Measurement Validation and Performance Monitoring Using Robust KalmanFiltering Techniques, P. Dewallef, O. Leonard, Proceedings of the 5th European Confe-rence on Turbomachinery Fluid Dynamics and Thermodynamics, Prague, 2003

– On-Line Performance Monitoring and Engine Diagnostic Using Robust Kalman Filte-ring Techniques, P. Dewallef, O. Leonard, ASME Paper GT-2003-38379, 2003

– On-Line Aircraft Engine Diagnostic Using a Soft-Constrained Kalman Filter, P. De-wallef, O. Leonard, K. Mathioudakis, ASME Paper GT2004-53539, 2004

– On-Line Transient Engine Diagnostic in a Kalman Filtering Framework, S. Borguet, P.Dewallef, O. Leonard, ASME Paper GT2005-68013, 2005

– Application of the Kalman Filter to Health Monitoring of Gas Turbine Engines - ASequential Approach to Robust Diagnosis, P. Dewallef, PhD Thesis, ISSN 0075-9333,D/2005/0480/44, 2005.

– Combining Classification Techniques With Kalman Filters for Aircraft Engine Diagnos-tics, P. Dewallef, O. Leonard, K. Mathioudakis, Journal of Engineering for Gas Turbinesand Power, Vol 128, No 2, pp 281-287, 2006

– A Way to Deal with Model-Plant Mismatch for a Reliable Diagnosis in Transient Ope-ration, S. Borguet, P. Dewallef, O. Leonard, ASME Paper GT2006-90412, 2006

engine to monitorcommandparameters

externaldisturbances

vk

uk

observedmeasurements

estimatedmeasurements

residualsrk

engine performancesimulation model

xk wk

vkestimatedexternal

disturbances

estimatedhealth

parameters

estimatedstate

variables

yk

yk

-

0 100 200 300 400 500 600 700−0.5

0

0.5

% o

f dev

iatio

n fr

omno

min

al v

alue

0 100 200 300 400 500 600 700−1.5

−1

−0.5

0

0.5

time [s]

% o

f dev

iatio

n fr

omno

min

al v

alue

FAN capacity

FAN efficiency

LPC capacity

LPC efficiency

HPC capacity

HPC efficiency

HPT capacity

HPT efficiency

LPT capacity

LPT efficiency

Nozzle area

– On Inverse Problems in Turbine Engine Parameter Estimation, M. Henriksson, S. Bor-guet, T. Gronstedt, O. Leonard, ASME Paper GT2007-27756, 2007

– A Sensor-Fault-Tolerant Diagnosis Tool Based on a Quadratic Programming Approach,S. Borguet, O. Leonard, ASME Paper GT2007-27324, 2007

– An Adaptive Estimation Algorithm for Aircraft Engine Performance Monitoring, O.Leonard, P. Dewallef, S. Borguet, submitted to the AIAA Journal of Propulsion andPower, 2007

– Coupling Principal Component Analysis and Kalman Filter Algorithms for On-lineAircraft Engine Diagnostics, S. Borguet and O. Leonard, Proceedings of the XVIIIInternational Symposium on Airbreathing Engines, Beijing, 2007

– A Quadratic Programming Framework for Constrained and Robust Jet Engine HealthMonitoring, S. Borguet, O. Leonard, Proceedings of the 2nd European Conference onAerospace Sciences, Brussels, 2007

– A Study on Observability and Sensor Selection for Efficient Jet Engine Health Moni-toring, Sebastien Borguet and Olivier Leonard, to be presented at the ISROMAC-12,Honolulu, 2008

– A Generalised Likelihood Ratio Test for Adaptive Engine Health Monitoring, SebastienBorguet and Olivier Leonard, ASME Paper GT2008-50117, 2008

3 Optimization using genetic algorithms

The purpose of this project is to develop robust metaheuristic methods for solving optimi-zation, design or process control problems which utilizes simultaneously binary, discreteand continuous parameters. This situation results in a great number of configurationsto be considered, which must satisfy a important number of constraints. Moreover “realworld” problems lead to the optimzation of several but contradictory objectives.

This project brings a response to this type of problems, by combining the advantagesof genetic algorithms and nonlinear mathematical programming. The genetic algorithmsallow a broad and systematic exploration of the design space, the various possible solutionsbeing subjected to a Darwinian natural selection process. The nonlinear programmingbrings the efficiency of a local and fast exploration around a promising configuration. Theresulting optimization tool can be coupled to any metamodel of the application to beoptimized (data bases, response surfaces, neural networks, analytical models, parametricmodels...). The multiplicity of the objectives is addressed following the Pareto approach.

Up to now these optimization tools were applied to several problems such as the optimalsizing of lubrication pumps for turbojets, the optimization of pump scheduling, the designand the operation of blowers and the optimization of heat pipes geometries.

Partners : Partners : CENAERO, Fluorem, Euro Heat Pipes, Techspace Aero, Valeo

Recent related publications :

– Application of Genetic Algorithms to Lubrication Pumps Stacking Design, V. Kelner,O. Leonard, Journal of Computational and Applied Mathematics, Vol 168/1-2 pp 255-265, 2003

– Optimal Pump Scheduling for Water Supply Using Genetic Algorithms, V. Kelner andO. Leonard, Proceedings of the 5th International Conference on Evolutionary Compu-ting for Industrial Applications - EUROGEN’03, Barcelona, 2003

– An Hybrid Optimization Technique Coupling Evolutionary and Local Search Algo-rithms, V. Kelner, F. Capitanescu, O. Leonard, and L. Wehenkel, Proceedings of the3rd International Conference on Advanced Computational Methods in Engineering -ACOMEN’05, Ghent, 2005

– Multi Objective of a Fan Blade by Coupling a Genetic Algorithm and a ParametricFlow Solver, V. Kelner, G. Grondin, O. Leonard, and P. Ferrand, Proceedings of the6th International Conference on Evolutionary Computing for Industrial Applications -EUROGEN’05, Munich, 2005

– Robust Design of a Fan Blade by Coupling Multi Objective Genetic Optimization andFlow Parameterization, V. Kelner, G. Grondin, O. Leonard, P. Ferrand, Proceedings ofthe International Congress on Fluid Dynamics Applications in Ground Transportation,Lyon, 2005

– Geometric Optimization of Grooved Heat Pipes by a Genetic Algorithm Technique, C.Goffaux, S. Pierret, S. Rossomme, V. Kelner, S. Van Oost and L. Barremaecker, Procee-dings of the 6th International Conference on Heat Pipes, Heat Pumps and Refrigerators,Minsk, 2005

Fig. 1 – Otimizing a jet engine lubrication pump stacking

Fig. 2 – Optimizing the pressure jump and the pressure losses of blower blades

4 Throughflow simulations in turbomachines

4.1 High-order throughflow, deterministic and circumferentialstresses

The objective of this project is to introduce non stationary effects due to the rotor-stator interactions into a throughflow model for compressors and turbines. Following theapproach of Adamczyk, several averages of the unsteady 3-D Navier-Stokes equations areperformed to end up with a steady axisymmetric set of equations. In this final set ofequations, different stresses and forces appear. They introduce the (mean) effect of theflow phenomena that have been averaged ; in that sense they are similar to the Reynoldsstresses. The inclusion of these deterministic and circumferential stresses in a stationarymodel makes it possible to predict the radial mixing process observed in experiments andin non stationary simulations, but which are not reproducible in stationary simulationsusing a mixing plane.

This set of equations represent the ultimate throughflow model, provided that one couldfind a way to model the stresses and forces. This throughflow model differs from classicalstreamline curvature method as it is directly based on the Navier-Stokes equations solvedby finite volume techniques. In a first step, the results of unsteady 3-D simulations wereused to evaluate the effects of the different stresses and forces (and their modelization)inside the throughflow environment. In a second step, the circumferential stresses and theblade forces are being modelled using an harmonic approach.

Partners : ONERA, Ecole Centrale de Lyon

Recent related publications :

– A Throughflow Analysis Tool Based on the Navier-Stokes Equations, J.-F. Simon, O.Leonard, Proceedings of the 6th European Conference on Turbomachinery Fluid Dy-namics and Thermodynamics, Lille, 2005

– Modeling of 3-D Losses and Deviations in a Throughflow Analysis Tool, J.-F. Simon,O. Leonard, Journal of Thermal Science, Vol 16, No 3, pp 208-214, 2007

– Contribution to Throughflow Modelling, J.-F. Simon, PhD Thesis, University of Liege,2007

– On the Role of the Deterministic and Circumferential Stresses in Throughflow Calcu-lations, Jean-Francois Simon and Olivier Leonard, ASME Paper GT2008-50119, 2008

– Investigating Circumferential Non-Uniformities in Throughflow Calculations using anHarmonic Reconstruction, Jean-Philippe Thomas, Jean-Francois Simon and OlivierLeonard, ASME Paper GT2008-50328, 2008

4.2 Fast quasi-one-dimensional simulations in multistage turbo-machines

The performances of a gas turbine engine must be analyzed well before the engine istested on the bench or in flight. This study provides the designer with guidelines for thechoice of the many design parameters and for optimizing the final configuration of the jetengine. It also allows to test (in a virtual way) the correct operation of the engine duringcritical manœuvres. The present project is aimed at providing a quasi-1D modern CFDtool for the numerical modeling of the operation of a compressor and a whole jet engine.

Thanks to the progress of the simulation methods and computing power, it is now possibleto develop and to use models of jet engines based on the application of the laws of thefluid mechanics to a great number of cells. This approach makes it possible to describewith a high degree of accuracy the exchanges of mass, energy and momentum within themachine, while reducing to the bare minimum the quantity of information of empiricalnature. It takes advantage of the high precision and efficiency of the CFD methods forspace and time discretization, with CPU times of a few seconds.

This quasi-1D tool tool may be applied to the a large number of problems such as waterand hail ingestion, bleeds and cooling flows, mechanical and thermal transients, reversedflows and surge, characteristic map extrapolation, building global compressor model fromstage/partial results, coupling with cycle calculations and global modeling of a jet engine.

Recent related publications :

– Explicit Thermodynamic Properties Using Radial Basis Functions Neural Networks, O.Adam, O. Leonard, Proceedings of the 2nd SIAM International Conference on DataMining, Arlington, 2002

– A Quasi-One-Dimensional Model for Axial Compressors, O. Adam, O. Leonard, Pro-ceedings of the XVII International Symposium on Airbreathing Engines, Munich, 2005

– A Quasi-One-Dimensional Model for Axial Turbines, O. Adam and O. Leonard, Pro-ceedings of the XVIII International Symposium on Airbreathing Engines, Beijing, 2007

– A Quasi-One-Dimensional CFD Model for Multistage Turbomachines, O. Leonard andO. Adam, submitted to the Journal of Thermal Sciences

Fig. 3 – Blade force and circumferential stress distribution in a compressor stage

Fig. 4 – Comparison of the quasi-1D and experimental results on a 3-stage compressor

Fig. 5 – Comparison of the quasi-1D and experimental results on a 2-stage cooled turbine