Landing Gear Noise Modelling - ONERAmusaf2016.onera.fr/sites/musaf2016.onera.fr/files/aeroacoustics_2... · [Xu/Sagaut_JCP2013] [Ricot_2013] • Malaspinas & Sagaut « Consistent

Landing Gear Noise Modelling – MUSAF III Colloquium

Landing Gear Noise Modelling Progresses in LBM industrial use

Aloïs SENGISSEN

with contributions from Jean-Christophe GIRET, Christophe COREIXAS & Thomas ASTOUL

© AIRBUS Operations S.A.S. All rights reserved. Confidential and proprietary document.

JET TURBINE CORE

FORWARD FAN

REARWARD FAN

AIRFRAME ~50%

Introduction & stakes

Typical noise contribution breakdown

for a long range Aircraft at approach :

SLATS

FLAPS

CLEAN WING

HTP

NOSE L/G MAIN L/G

APPROACH – AIRFRAME only


Page 2 / 36

Noise generation mechanisms

way more complex to capture

than noise propagation to farfield

Focus on sources


Enablers of Landing Gear noise High Fidelity simulation


Solver Scalability

Maintenance, adaptability

Advanced Boundary Conditions

Numerical properties

Solver core Performance

Coupling / far-field acoustic radiation

Turbulence modelling

Wall modelling

Abilities on very complex Geometries

Physical mechanisms

understanding

Assessed with step by step validation

Assessed with seamless

end to end workflow

Page 3 / 36


Outline on Simulation accuracy


Numerical approach overview

Lattice-Boltzmann Solver LaBS

Coupling with FWH solver

Numerical Setup & Grids

Accuracy assessment

Mean & RMS Velocity fields comparison

CL and CD distributions

Wall pressure spectra comparison

Far field PSD & OASPL comparison

Sensitivity analysis to num. parameters

Grid convergence

Subgrid scale model influence

Wall law model components

Relevant test case : simplified LGs « LAGOON »

Full study available in

[AIAA2015–2993]

Page 4 / 36


Outline on Industrial readiness


Physical mechanisms understanding

Wheels inner cavity modes – LAGOON1

Tow bar vortex shedding – LAGOON2

Wheel rim caps removal – LAGOON3

Corner stones towards industrial cases

Core performance & scalability

Flexible & Seamless meshing

of complex configurations

Perspectives of applications

on industrial configurations

Relevant test case : real A/C applications

Page 5 / 36


Collaboration Airbus-ONERA (2006-2010) [Manoha_AIAA2008, Manoha_AIAA2009]

Highly detailed experimental data

Aerodynamic Measurements (F2 Wind Tunnel)

Acoustic Measurements (CEPRA19 Wind Tunnel)

LAGOON 1: Two wheels + Axle + Main Leg LAGOON 2: + Tow bar + Lights + Steering actuator LAGOON 3: + Torque link – Rim periphery caps

3 configurations of increasing complexity

LAGOON1 disclosed to NASA BANC III workshop, for accuracy & sensitivity assessment [Manoha_AIAA2015]

LAGOON2 & 3 used to cross compare geom. effects


Page 6 / 36

LAGOON geometries & objectives


Numerical Approach


Page 7 / 36


• Latt & Chopard, “Lattice Boltzmann method with regularized pre-collision distribution functions," Math. & Comp. in Sim., 2006

• Augier & Dubois « Isotropy properties for lattice boltzmann schemes » ICMMES conference ,2012

• Marié, Ricot & Sagaut « Comparison between lattice Boltzmann method and Navier–Stokes high order schemes for

computational aeroacoustics», J. Comput. Phys.228, 1056-1070, 2009


Page 8 / 36

Numerical Approach : LaBS solver

Lattice Boltzmann Solver (LaBS)

2010 – 2013 French collaborative project

Led by Renault, Airbus, CS, Univ. Paris VI, ECL

http://www.labs-project.org/

LaBS general numerical method

Classical LB approach, fully transient

D3Q19 BGK collision scheme, improved

with regularization [Latt_MCS2006] [Ricot_2013]

Octree mesh refinement,

specific treatment at resolution interface

Meshing embedded in the solver (and parallel!)

LaBS numerical properties

Low Numerical dispersion & dissipation [Marié_JCP2009]

Excellent isotropy properties [Augier/Dubois_ICMMES2011]







7 points filter

3 points filter

5 p

oin

ts filte

r

In-Flow Turbulence modelling

LES-LBM modelling using 2 possible subgrid-scale models

High order filtering “Approximate Deconvolution Model”, [Malaspinas_PoF2011, Malaspinas_JFM2012, Touil_PoF2013]

Subgrid scale model “Shear Improved Smagorinsky Model” [Leveque_JFM2007, Touil_PoF2013]

Near-wall turbulence modelling

Immersed boundaries with Bounce-back principles [Chen/Doolen_IJNMF2014]

Wall Laws with pressure gradient effect [Afzal_IUTAM_Symposium1996][Malaspinas_JCP2014]

Inlet & Outlet Boundary Conditions

Velocity imposed inlet & Pressure imposed outlet

Buffer zone as non-reflecting Boundary conditions [Xu/Sagaut_JCP2013] [Ricot_2013]

• Malaspinas & Sagaut « Consistent subgrid scale modelling for Lattice Boltzmann methods », J. Fluid Mech. 700, 514‐542 (2012)

• Malaspinas & Sagaut « Advanced Large‐eddy simulation for Lattice Boltzmann methods: the Approximate Deconvolution Model » Phys. Fluids. 23, 105-103, 2011

• Malaspinas & Sagaut « Wall model for Large‐eddy simulation on based on Lattice Boltzmann methods », J. Comput. Phys .275, pp. 25–40, 2014

• Xu & Sagaut « Analysis of the absorbing layers for the weakly‐compressible lattice Boltzmann methods », J. Comput. Phys.245, 14‐42, 2013

• Levêque & Bertoglio, « Shear-improved Smagorinsky model for large-eddy simulation of wall-bounded turbulent Flows," J. Fluid Mech. , Vol. 570, 2007, pp. 491-502


Numerical Approach : Turbulence Modelling

Page 9 / 36


Numerical Approach : Grids

Setup

LAGOON LG as installed in anechoic WTT (no ceiling wall)

Constant velocity inlet & constant pressure outlet

Simple setup (meshing directives through GUI)

Meshes

Aim: Try to match the points distribution of a wall-modelled LES [Giret_AIAA2012] with fast turnover times

Wake region : 2mm to 4mm (based on simple shapes)

Near wall region : 0.4mm to 0.625mm (based on offsets of the geometry) ~ Y+ = 60

Rough mesh convergence study COARSE : 20M nodes MEDIUM : 40M nodes FINE : 80M nodes

CPU time on MEDIUM : 2 days on 360 core (0,32s physical time starting from scratch on intel XEON5-2697 @ 2,7Ghz)

10 / 36


Page 10 / 36


Accuracy Assessment



Mean Axial Velocity : Plane Z = 0mm

LaBS_COARSE LaBS_MEDIUM

EXPE (PIV) LES [Giret_AIAA2012]

Results provided to

NASA BANC III workshop

Shear Layer thickness

Too thick on COARSE

mesh

Ok on MEDIUM mesh

Recirculation zone

Size ok

Intensity ok

Wake shape

Wake dimensions

qualitatively satisfying

Better predicted by LBM

than by LES


Page 12 / 36


-0.2

-0.1

0.0

0.1

0.2

y (

mm

)

100806040200 Streamwise velocity (m/s)

PIV X=220 Z=0

LaBS COARSE LaBS MEDIUM LES AVBP

X=220 mm

-0.2

-0.1

0.0

0.1

0.2

y (

mm

)


PIV X=180 Z=0


X=180 mm

Mean Velocity profiles

Excellent quantitative agreement

Improvement with MEDIUM mesh in

plane Z = -104 mm

-0.2

-0.1

0.0

0.1

0.2

y (

mm

)


PIV X=160 Z=0


X=160 mm

-0.2

-0.1

0.0

0.1

0.2

y (

mm

)


PIV X=128 Z=-104


X=128 mm

13 / 36



RMS Axial Velocity : Plane Z = 0mm

LaBS_COARSE LaBS_MEDIUM

EXPE (PIV) LES [Giret_AIAA2012]

Shear on wheel flanks:

Size & intensity of

Shedding well predicted

on MEDIUM mesh

(X=0,2 ; Y=0,15)

Results in good

agreement

with LES (AVBP)

Results in good agreement

with PIV despite some

measurement artifacts


Page 14 / 36


-0.2

-0.1

0.0

0.1

0.2

y (

mm

)

3020100 Rms Streamwise velocity (m/s)

PIV X=220 Z=0


-0.2

-0.1

0.0

0.1

0.2

y (

mm

)


PIV X=128 Z=-104


-0.2

-0.1

0.0

0.1

0.2

y (

mm

)


PIV X=180 Z=0


RMS Velocity profiles

-0.2

-0.1

0.0

0.1

0.2

y (

mm

)


PIV X=160 Z=0


X=220 mm X=180 mm X=160 mm X=128 mm

Fair quantitative agreement

Improvement with MEDIUM mesh

Subgrid scale contribution is NOT included



110

100

90

80

70

60

50

Pre

ssu

re P

SD

(d

B/H

z)

102

2 3 4 5 6 7 8 9

103

2 3 4 5 6 7 8 9

104

Frequency (Hz)

Kulite 1

LaBS COARSE LaBS MEDIUM

PSD at the wall : Wheel Rolling Band

Low freq. filter in the experiments (0 – 200Hz)

Tones @ 1KHz & 1.5kHz well captured

Low freq.

Filter in the

experiments


Page 16 / 36


110

100

90

80

70

60

50

Pre

ssu

re P

SD

(d

B/H

z)

102

2 3 4 5 6 7 8 9

103

2 3 4 5 6 7 8 9

104

Frequency (Hz)

Kulite 3 LaBS COARSE LaBS MEDIUM

Gradual increase of PSD levels at the wall

along with boundary layer development



Page 17 / 36


130

120

110

100

90

80

70

Pre

ssu

re P

SD

(d

B/H

z)

102

2 3 4 5 6 7 8 9

103

2 3 4 5 6 7 8 9

104

Frequency (Hz)




Page 18 / 36


130

120

110

100

90

80

70

Pre

ssu

re P

SD

(d

B/H

z)

102

2 3 4 5 6 7 8 9

103

2 3 4 5 6 7 8 9

104

Frequency (Hz)


K7 corresponds to the detachment point due to GradP

COARSE mesh predicts slightly earlier separation



Page 19 / 36


130

120

110

100

90

80

70

Pre

ssu

re P

SD

(d

B/H

z)

102

2 3 4 5 6 7 8 9

103

2 3 4 5 6 7 8 9

104

Frequency (Hz)


Flow fully detached at K8



Page 20 / 36


Physical Mechanisms Investigations



sdfsdf

Observation of cavity modes


FEM analysis conducted to cavity eigenmodes identification :

Mode @ 1048Hz

Mode @1529Hz

Superposition of these 2 modes observed in the simulations

Observation of cavity modes

y

x

y

x

Courtesy of EXA [Casalino_JSV_2014]

130

120

110

100

90

80

70

Pre

ssure

PS

D (

dB

/Hz)

102

2 3 4 5 6 7 8 9

103

2 3 4 5 6 7 8 9

104

Frequency (Hz)



Page 23 / 36


sdfsdf

Shedding of the Tow Bar

24 / 36

Shedding of the Tow Bar Landing Gear Noise Modelling – MUSAF III Colloquium

Page 24 / 36


Peak around 1100Hz on K1 & K2

Emerging even more behind tow bar

Shedding of the Tow Bar

25 / 36

Pre

ssure

PS

D (

dB

/Hz)

102

2 3 4 5 6 7 8 9

103

2 3 4 5 6 7 8 9

104

Frequency (Hz)

10

dB

Tow Bar Wake LaBS SISM_FINE

Pre

ssure

PS

D (

dB

/Hz)

102

2 3 4 5 6 7 8 9

103

2 3 4 5 6 7 8 9

104

Frequency (Hz)

10

dB

Kulite 1 LaBS SISM_FINE LaBS SISM_MEDIUM

Pre

ssure

PS

D (

dB

/Hz)

102

2 3 4 5 6 7 8 9

103

2 3 4 5 6 7 8 9

104

Frequency (Hz)

10

dB

Kulite 2 LaBS SISM_FINE LaBS SISM_MEDIUM



Corner stones towards industrial cases

26 / 36



Performance & Scalability


Numerical

Approach

NS – O(2) Centered + RK3

NS – O(3) DRP Tam + RK3

NS – O(6) Lele + RK6

LBM BGK scheme

Number of

Flop/iteration 711 2862 11295 588

Single core peak performance of LBM algorithm much faster than classical N-S :

Faster than any FD schemes, way faster than FV approaches (Courtesy of Sagaut et al.)

Accuracy of LBM collision scheme [Ricot_JCP2009]

Almost equivalent to O(3) in terms of dispersion

Way better than any classical approaches, incl. O(6), in terms of dissipation

Local time step (gift from octree mesh) yields additional performance benefits without the difficult compromises from implicit schemes

Overall computational cost ~1µs/iteration/cell (7x to 10x faster than fastest NS explicit schemes)

Page 27 / 36


Performance & Scalability


Local algorithm, very suitable for massively parallel computing

Constant performance, and efficient speedup

~ 80% yet not perfect, due to the fast base algorithm

0

2

4

6

8

10

12

14

96 288 480 672 864 1056 1248

Speedup solver

Ideal speedup

number of cores

Page 28 / 36


Seamless meshing of complex configurations


Geometrical details play a significant

role in noise radiated (HF content).

Real life configurations for LG noise

are much more than complex !!!

Structured meshes have given up

Unstructured becomes very tricky

Immersed octree takes the lead

Page 29 / 36




Meshing done in parallel by the solver

Example without any wake refinement

Every second cell displayed

Page 30 / 36




Scaling up seamlessly

Every second cell displayed

Page 31 / 36


Perspective of applications


Page 32 / 36

For acoustics purpose :

Real configurations can be finally investigated without hazardous simplifications

(unknown impact in terms of physics)

Installation effects reachable

Velocity deficit due to wing circulation,

Interaction between LG, Flaps,

Influence of NLG wake…

For wider purpose :

Steady loads

Unsteady loads

…


Conclusions


Page 33 / 36


Conclusion & Lessons learned

Conclusions on accuracy of LBM

Level of agreement observed :

Very satisfying on both Mean & RMS velocity fields & integrated aero quantities (not shown today)

Very good agreement on wall PSD

Meaningful physical phenomena well captured

Gradual increase of wall PSD with BL development, then sudden change after detachment point

On LAGOON1, Cavity mode directly visible on time resolved wall pressure [Casalino2014] & [Giret2013]

On LAGOON2, Vortex shedding from tow bar perturbs this cavity mode and yields another peak (1100Hz)

However, everything is not yet perfect in current approach

Wall laws / near wall treatment

Parler des limites des transitions

Conclusions on industrial readiness

Engineer’s daily life game changer wrt unsteady well established NS-CFD methods

Real A/C geometry without simplification

Automatic/Flexible Billion cells mesh if needed

1000+ core scalable on HPC

Turnaround time breakthrough


Page 34 / 36

Next step

Take full benefit of LBM

techniques for other industrial

applications of interest.



© AIRBUS S.A.S. All rights reserved. Confidential and proprietary document. This document and all information contained herein is the sole property of AIRBUS S.A.S. No intellectual property rights are granted by the

delivery of this document or the disclosure of its content. This document shall not be reproduced or disclosed to a third party without the express written consent of AIRBUS S.A.S. This document and its content shall not be

used for any purpose other than that for which it is supplied. The statements made herein do not constitute an offer. They are based on the mentioned assumptions and are expressed in good faith. Where the supporting

grounds for these statements are not shown, AIRBUS S.A.S. will be pleased to explain the basis thereof.

AIRBUS, its logo, A300, A310, A318, A319, A320, A321, A330, A340, A350, A380, A400M are registered trademarks.

Thank you for your attention !

ACKNOWLEDGEMENTS :

E. Manoha (ONERA) & B. Caruelle (Airbus) for LAGOON initiative

D. Ricot (Renault), B. Gaston & R. Cuidard (CS) for support on LaBS

G. Rahier (ONERA) for providing KIM FW-H solver

HP for powering Airbus’ HPC resources

Page 35 / 36

Documents

Landing Gear Noise Modelling - ONERAmusaf2016.onera.fr/sites/musaf2016.onera.fr/files/aeroacoustics_2... · [Xu/Sagaut_JCP2013] [Ricot_2013] • Malaspinas & Sagaut « Consistent