Available online at www.sciencedirect.

com
Procedia
Engineering
Procedia Engineering 00 (2009) 000000

Procedia Engineering 6 (2010) 293302 www.elsevier.com/locate/procedia

www.elsevier.com/locate/procedia
IUTAM Symposium on Computational Aero-Acoustics
for Aircraft Noise Prediction

Aerodynamic and Aeroacoustic Simulations

of a Two-wheel Landing Gear
Taro Imamura a,b,*, Tohru Hirai c, Kazuhisa Amemiya d
Yuzuru Yokokawa b, Shunji Enomoto b, Kazuomi Yamamoto b
a
The University of Tokyo, 7-3-1, Bunkyo-ku Hongo, Tokyo, 113-8656, JAPAN
b
Japan Aerospace Exploration Agency, 7-44-1, Jindaiji-higashi, Chofu, Tokyo, 182-8522, JAPAN
c
Ryoyu Systems, Ryoyu Systems, 2-19-13, Takanawa, Minato-ku, Tokyo 108-0074, JAPAN
d
JAST, 1-16-6, Izumi-chuo, Izumi-ku, Sendai, Miyagi, 981-3133, JAPAN

Abstract

Landing gear noise is one of the dominant noise sources from an aircraft during approach phase due to low engine rate setting.
Two different computational fluid dynamics (CFD) codes are used to understand the flow physics around a two-wheel main
landing gear. First, flow solver for unstructured grid is used to simulate the steady-state flow field around a complex landing
gear. The basic flow pattern around the landing gear is investigated and the potential noise sources, such as separation, wake and
accelerated regions are observed. Additionally, the effects of the wind tunnel wall, both closed and open test sections, are
investigated by including the tunnel wall in the computation. Secondly, high-order structured grid solver is used to understand the
unsteady flow field on the simplified geometry. The far-field sound pressure level is evaluated using Ffowcs Williams and
Hawking method. In combination, flow field around the two-wheel main landing gear will be discussed.

c 2009
2010Published
Publishedbyby Ltd.Ltd. Open access under CC BY-NC-ND license.
Elsevier
Elsevier

Keywords: Landing gear; Computational Fluid Dynamics; Computational Aeroacoutics

1. Introduction

Due to the continuous increase in air travel, attention to the environment around an airport must be paid, and
noise reduction is one of the key topics. Especially, during the approach phase when the engines are at low thrust,
engine noise and airframe noise become comparable. In order to accommodate with the regulation which is
becoming more and more stringent, the development of airframe noise reduction technology is necessary. The
airframe noise mainly comes from high-lift device and landing gear [1], and we will focus on landing gear noise in

* Corresponding author. Tel.:+81-3-5841-1838; fax:+81-3-5841-8560.

E-mail address: timat@mail.ecc.u-tokyo.ac.jp

1877-7058 c 2010 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
doi:10.1016/j.proeng.2010.09.031
294 T. Imamura et al. / Procedia Engineering 6 (2010) 293302
Taro Imamura/ Procedia Engineering 00 (2010) 000000

this study. There are many studies devoted to landing gear noise problem [2-7], but unfortunately, most of the data
are confidential. Additionally, two-wheel main landing gear is a niche in a sense that it only appears on relatively
small aircrafts, such as regional jets. Therefore, a non-sensitive landing gear configuration was designed and several
wind tunnel experiments are being performed by Japan Aerospace Exploration Agency (JAXA) [2,3]. In this study,
numerical simulations, which were performed parallel to the experimental studies, are described. Two different CFD
codes are used to understand the flow physics around the landing gear model. First, flow solver for unstructured grid
is used to solve steady state flow field around the complex landing gear geometry. In order to estimate the absolute
sound pressure level, unsteady flow simulations is necessary. However, even from the steady-state flow simulations,
flow related to the noise generation, such as accelerated region or wake interaction can be investigated.
Additionally, the effects of the wind tunnel wall, both closed and open test sections, are investigated including the
tunnel wall in the computation. Secondly, flow solver for structured grid is used to simulate unsteady flow field
around the landing gear. Far field sound pressure level is calculated using the near field results. Structured grid
approach is selected since high order scheme can be used relatively easily compared with unstructured grid.
However, it is difficult to generate grid around a complex geometry. Furthermore, smooth grid is necessary in order
to use high order scheme and this makes grid generation more difficult. Thus, the model is simplified and only the
flow around the tire, axle, cylinder & piston will be focused in this simulation.
In the following section, a landing gear model will be introduced. In section 3, steady state flow simulation for
the complex landing gear model will be described. Unsteady flow simulations results will be shown in section 4.
Finally, section 5 concludes this paper.

2. Two-Wheel Main Landing Gear Geometry

A non-sensitive two wheel main landing gear CAD model was designed to promote research for wind tunnel
experiment as well as CFD. This geometry is called Landing gear noise Evaluation Geometry (LEG) [2-4]. The
assumed aircraft is 100 PAX civil aircraft (two-engine wing mount), and the main landing gear is two-wheel landing
gear which is typical for this size of aircraft. Figure 1 shows the CAD model of the LEG viewing from the
downstream side. Major components of the landing gear are also described in this figure. The CAD model consists
of large components (tires, cylinder & piston, doors, side brace) and detail components (torque link, door rod, un-
lock actuator, uplock-pin, hydraulic and electrical lines).

Figure 1. CAD model of Landing gear Noise Evaluation Geometry (LEG).

3. Steady-state Computations around a Complex Geometry

The purpose of steady-state simulations is to investigate the mean flow patterns around the landing gear. The
calculated model is based on a 40% scale landing gear noise measurement model for the wind tunnel experiment
[2,3]. First, the basic flow patterns are investigated under uniform flow condition. Additionally, the effects of the
wind tunnel wall, both closed and open test sections, are investigated by including the tunnel wall in the
 T. Imamura et al. / Procedia Engineering 6 (2010) 293302 295
Author name / Procedia Engineering 00 (2010) 000000

computation. This is performed to check the difference in the local flow speed among the different experimental
conditions [2,3].

3.1. Flow Solver & Grid generator

The flow solver is J-TAS code which was originally developed at Tohoku University [8-10]. In this code, 3D
Reynolds-averaged Navier-Stokes equations are solved on unstructured grid by a cell-vertex finite volume method
and Spalart-Allmaras one equation model is chosen as a turbulence model. MEGG3D is an unstructured mesh
generator with graphical user interface (GUI) tools [11-14]. It can generate triangular surface mesh with the
advancing front method and tetrahedral volume mesh using Delaunay tetrahedral meshing, as well as hybrid volume
mesh composed of tetrahedrons, prisms, and pyramids for viscous flows with high Reynolds number.

3.2. Computational model, Flow Conditions & Computational Grid

The computational model for the simulations is shown in Fig. 2. All of the major components, such as tires,
cylinder & piston, doors, side brace, and torque link are included in the computational model, but the detail
components, such as door rod, un-lock actuator, uplock-pin, hydraulic and electrical lines are omitted. In order to
generate grid around such a complex geometries, unstructured grid approach is chosen.
Figure 3 shows the computational grid around the landing gear. The basic flow field is analyzed under uniform
flow condition and the grid size is 7.1 million. Also, Fig. 4 shows the cases with the wind tunnel wall. Since the
wind tunnel experiments were performed using closed and open test section, both conditions are simulated. The
closed test section corresponds to the low speed wind tunnel (2m x 2m) of JAXA, and open wind tunnel case
represents the large scale low speed wind tunnel owned by Railway Technical Research Institute of Japan (RTRI).
The grid sizes for closed and open cases are 6.9 million and 10.2 million, respectively. The mean flow velocity is 68
m/s which is a typical approach speed of a civil aircraft and angle-of-attack and slip-angle are set to 0 degrees. The
Reynolds number based on the diameter of the tire (0.417m for 40% scale model) and the freestream velocity is
1.8106.

(a) Span constant (b) Chord constant

Figure 2. Computational model for Figure 3. Computational grid for RANS simulation.
steady state flow simulation.

3.3. Computational Results

Figure 5 shows the Mach number distribution at several different cross sections, parallel to the upper wall. Since
this computation is steady-state flow simulation, direct noise prediction is not possible but still we can observe
potential noise sources. Those are, accelerated flow region, wake, and shear layer. Around the root of the landing
gear, interaction between the cylinder and side-brace is observed. Around the tire, large separated region is
observed. Due to the existence of piston and torque link, the flow between the tires is blocked. Accelerated and
decelerated region co-exists in narrow region which indicates the existence of strong shear layer. Therefore, this part
of the flow could be the potential noise source of a two wheel landing gear.
296 T. Imamura et al. / Procedia Engineering 6 (2010) 293302
Taro Imamura/ Procedia Engineering 00 (2010) 000000

Figure 6 compares the effect of the wind tunnel wall. The plot is the normalized Mach number along the red line
in the right figure. The difference is observed at the peak where the flow is accelerated near the stagnation of the
axle. Since the dominant noise source seems to exist between the tires, the difference in the mean flow velocity
might result in the difference in the sound pressure level. Current result of the closed test section case indicates
approximately 4% increase at the peak. This is due to the blockage effect of the model and this difference might
cause difference around 1 dB if we assume that sound power scales to the power of 6 to the mean flow velocity. On
the other hand, the difference between uniform flow and open test section is negligible.

(a) Simulation with closed wall test section (Same setup as JAXA-LWT2 wind tunnel)

(b) Simulation with open test section (Same setup as RTRI wind tunnel)

Figure 4. TAS simulations including wind tunnels.

(a) Cross section locations (b) 20% Zmax (c) 30% Zmax

(d) 70% Zmax (e) 80% Zmax (f) 90% Zmax

Figure 5. Steady-state Mach number distribution at different cross sections.
 T. Imamura et al. / Procedia Engineering 6 (2010) 293302 297
Author name / Procedia Engineering 00 (2010) 000000

Figure 6. Comparison of Mach number along the line near the axle.

4. Unsteady Flow Simulations around a Simplified Geometry

The main objective of unsteady flow simulation is to show the potential of our current approach by comparing
the far field sound pressure level with the experimental results. The experiment [2] was performed using 40% scale
model, and the largest noise source was observed around the piston, tires and axle. Therefore, the computational
model is simplified and flow field around that particular region will be focused.

4.1. Flow Solver

Numerical methods used in the calculation are briefly described. In this study, a flow solver called UPACS-LES
is used. This flow solver is developed based on UPACS code which is a in-house CFD code in JAXA. The
governing equations are the 3D compressible Navier-Stokes equations. The flow solver is based on a cell-centered
finite-volume method on multi-block structured grids. The code is parallelized by a flexible domain decomposition
concept and MPI. The convection terms are discretized using 6th-order compact scheme developed by Kobayashi
[15]. The viscous terms are discretized using 2nd-order central scheme. To avoid numerical instability, 6th-order
filter developed by Gaitonde-Visbal [16] is used. Considering the stability and the accuracy, filter constant is
defined as 0.45. Second order time integration is performed by MFGS [17] with 5 sub-iterations. The non-
dimensionalized time step using the tire diameter and sonic speed as reference values is set to t=2.5e-3. This values
is chosen to satisfy CFL condition (CFL<1) in the wake region of the landing gear, excluding the boundary layer.
This constraint must be satisfied to prevent the numerical dissipation of the vorticity and acoustic waves inside the
large eddy simulation (LES) regions. The standard Smagorinsky model with Smagorinsky constant of 0.2 is used as
a sub-grid scale model. At the wall boundary of the LES region, van Driest damping is applied.

4.2. Computational model, Flow Conditions & Computational Grid

The experimental results [2,3] indicate the largest noise source around the tires. Thus, this computation will focus
on this region and the geometry is simplified as shown in Fig. 7. The size of the model is the same as the experiment
(40% scale), which makes the comparison with the experiment straightforward. The freestream velocity is 68 m/s
and AoA or slip angles are set to 0 degrees. The Reynolds number based on the diameter of the tire (0.417m) and the
freestream velocity is 1.8106.
298 T. Imamura et al. / Procedia Engineering 6 (2010) 293302
Taro Imamura/ Procedia Engineering 00 (2010) 000000

Block-structured one-to-one point-matched grids are generated with the commercial software Gridgen and the
grid topology is shown in Fig. 8. The overall grid points are approximately 26 million. As shown in Fig. 7, X-, Y-,
and Z-directions correspond to the streamwise, spanwise, and vertical directions, respectively. The simulations were
performed using characteristic boundary conditions along the far-field boundaries. Slip-wall boundary condition is
applied to the upper surface plane and no-slip boundary conditions are imposed at the landing gear surface. Near the
wall, sufficient number of grid points couldnt be used. The first grid point normal to the wall was located within
y+<1, which is sufficient for LES. But x+ and z+ that are on the plane parallel to the wall are on the order of 100 or
more. Since the Reynolds number for the current problem is quite high, it is impossible to resolve the boundary
layer with grid points on the order of 10 million which is typical number of grid points for practical LES
calculations. On the other hand, most of the grid points are concentrated around the wake of the landing gear. The
sizes in all directions around the landing gear are almost uniform as shown in Fig. 8. This allows simulating the
evolution of three-dimensional vortical flow structures at the wake region which is estimated to be important for
noise generation. The experimental results show that important frequency range is 100 to 1000Hz (40% scale model
in the experiment) component [2]. Thus, we would like to calculated unsteady features up to 1 kHz. In the red
region, vortical structure which corresponds to the 1 kHz unsteadiness will be calculated. The grid size is almost
uniform in 3 directions in order to allow 3D vortical evolution. In the cyan region, the grid size is defined so that
acoustic wave propagation will be calculated correctly.

(a) Span constant (b) Chord constant

Figure 7. Computational model
for steady state flow simulation. Figure 8. Computational grid for LES.

Based on the near-field computational results, far-field noise is estimated using Ffowcs Williams and Hawkings
(FW-H) method [18,19]. Closed surfaces around the noise sources (permeable surface) are defined and the time
series data from the LES calculation are used as an input. The data are taken every 16 step from the LES
computations. The upper surface where the landing gear is attached is treated as complete reflection wall. The
downstream surface is excluded from the permeable surface to remove artificial acoustic waves generated by density
fluctuation due to the vortex convection. Thus, the choice of the permeable surface has an influence to the far-field
noise prediction. Three different surfaces are selected as shown in Fig. 9 and far-field sound pressure levels are
estimated for each surface to check the sensitivity. The number of elements for each surface is 95320, 108488, and
89700.
 T. Imamura et al. / Procedia Engineering 6 (2010) 293302 299
Author name / Procedia Engineering 00 (2010) 000000

(a) Surface 1 (b) Surface 2 (c) Surface 3

Figure 9. Permeable surface for far-field sound pressure level calculations.

4.3. Computational Results

Figure 10 shows the time-averaged surface static pressure coefficient along the center of each tire (shown in red
lines) compared with the experimental results [3]. Also, the results of steady flow computation (Reynolds averaged
Navier-Stokes simulation using Spalart-Allmaras model) on the same grid are compared. 0 degree point corresponds
to the rear stagnation point, and tires contact with the ground at 270 degrees. The experimental and RANS results
show good agreement. The current LES computation is not resolving the boundary layer sufficiently. The separation
is earlier than the experimental results, which results in earlier separation and lower suction peaks around 90 and
270 degrees. Since RANS results shows good agreement, current unsteady results are estimated to improve by
choosing LES/RANS hybrid method. On the other hand, the LES result of the base pressure around 0 degree
compares well with the experiment. It is thought that most of the noise comes from the largely separated flow in the
wake. Therefore, further comparison will be preformed for unsteady component.
Far-field noise is calculated using the near field unsteady results. The computed results are compared with the
far-field noise spectrum obtained at large scale low speed wind tunnel. The experimental model was transformed
like as the simulation model, which has no doors, side-brace, torque link and other detail components. However, the
detail configuration is not completely the same. Since the focus for this calculation is to estimate the noise by the
major components of the landing gear, secondary effect caused by the detail components are beyond the scope at
this moment. It is known the detail components have an effect to the overall noise level, but this assumption is
reasonable since we are only focusing on the largest noise source between the tires.

Figure 10. Comparison of static pressure coefficient.

300 T. Imamura et al. / Procedia Engineering 6 (2010) 293302
Taro Imamura/ Procedia Engineering 00 (2010) 000000

Figure 11 shows the results of 1/3 octave band spectrum. The measuring point is located 5.0 m from the upper
wall. The data length is 512 and Hanning window is used with 50% overlap for the FFT analysis. The experimental
results show that important frequency range is 100 to 1000Hz (40% scale model in the experiment) component [2].
Thus, the simulation is tailored to capture the fluctuation within this frequency range. The lower limit comes from
the simulated total time. The meaningful upper frequency for the simulation is around 1000 Hz and is governed by
the grid size (8mm) around the landing gear. From 100 to 1000 Hz, current prediction using LES with FW-H
method shows fair agreement with the experimental results. The differences among the different permeable surface
locations are small which indicates the reliability of our current estimation procedure. Also, fair agreements are
observed for different measuring points.

(a) Measuring points (b) Flyover point

(c) Upstream point (d) Downstream point

Figure 11. Far-field sound pressure level at three different locations.

 T. Imamura et al. / Procedia Engineering 6 (2010) 293302 301
Author name / Procedia Engineering 00 (2010) 000000

Figure 12 shows the instantaneous Q criteria around

the main landing gear. Behind the cylinder, vortical
structure is observed, but it is not a simple 2D Karman
vortex. Since the Reynolds number is on the order of 106,
the shear layer after the separation breaks up and three
dimensional structures are observed. Behind the tires,
there is also a large separation. Especially, strong vortical
structures are observed between the tires over the axle.
The separated flow from axle, piston and tires seems to be
interacting with each other and vortical structures in
streamwise direction are generated.
Figure 13 shows the time averaged turbulent kinetic
energy distribution at different cross sections parallel to
the upper wall. z/D is the distance from the upper wall
normalized by the tire diameter. At the root of the landing Figure 12. Instantaneous Q criteria around a simplified
gear, it is a simple cylinder, but as moving toward the main landing gear.
tires, the flow interaction is observed. The three figures on the right, show the TKE distribution at the axle and
z/D=1.7 and 2.5 are taken symmetrically regard to the axle (z/D=2.1). The only difference between z/D=1.7 and 2.5
is the existence of the piston between the tires. High TKE region appears due to the interaction of the wake of tires
and piston at the wake. Therefore, there is a possibility of noise reduction by changing the flow field between the
tires over the axle.

z/D=1.0 z/D=1.4 z/D=1.6 z/D=1.7

z/D=1.8 z/D=1.9 z/D=2.0 z/D=2.1

z/D

z/D=2.2 z/D=2.3 z/D=2.4 z/D=2.5

Figure 13.Turbulent kinetic energy distribution at different cross sections

302 T. Imamura et al. / Procedia Engineering 6 (2010) 293302
Taro Imamura/ Procedia Engineering 00 (2010) 000000

5. Conclusions

Two different CFD codes were used to understand the flow physics around a two-wheel main landing gear.
Unstructured grid flow solver was used to simulate the steady-state flow field around complicated landing gear. As a
result, large and small separations from various components and their interaction were clarified. The effect of the
wind tunnel wall was also investigated and its variation was acceptable. High-order structured grid flow solver was
used to understand the unsteady flow field on a simplified geometry. Based on the near field computation, far-field
noise was predicted in the combination of FW-H method. The estimated far-field noise shows reasonable agreement
with the experimental results. The differences among the different permeable surface locations are small which
indicates the reliability of our current estimation procedure. Additionally, complicated vortical structures generated
from the cylinder, piston, tires and axle has been visualized. Further studies regarding application of LES/RANS
hybrid approach and development of noise reduction devices will be performed in the near future.

Reference

(1) Crighton, D. G., Airframe Noise, Aeroacoustics of Flight Vehicles: Theory and Practice, Volume 1: Noise Sources, NASA Reference
Publication 1258, Vol. 1, NASA, 1991, pp. 391447.
(2) Yokokawa, Y., Imamura, T., Ura, H., Kobayashi, H., Uchida, H., Yamamoto, K., Far-field Characteristics of JAXA Main Landing Gear
Noise Research Model, 41st Fluid Dynamics Conference of JSASS and the Aerospace Numerical Simulation Symposium 2009 (in
Japanese)
(3) H. Ura, Y. Yokokawa, T. Imamura, K. Yamamoto, An Investigation on Airframe Noise Sources of Landing Gear Model, Mechanical
Engineering Congress, 2009 Japan (in Japanese)
(4) T. Imamura, T. Hirai, Y. Yokokawa, M. Murayama, K.Yamamoto, Numerical Analysis of Steady Flow around a Landing Gear Noise
Measurement Model , Journal of the Japan Society for Aeronautical and Space Sciences Vol. 57 No. 671 pp.493-498, 2009 (in Japanese)
(5) Guo, Y., "A Study on Local Flow Variations for Landing Gear Noise Research", AIAA paper 2008-2915, 2008
(6) Lockard, D., Khorrami, M., Li, F., "High Resolution Calculation of a Simplified Landing Gear", AIAA paper 2004-2887, 2004
(7) Dobrzynski, W., Schning, B., Chow, L.C., Wood, C., Smith, M., Seror, C., "Design and Testing of Low Noise Landing Gears", AIAA
paper 2005-3008, 2005
(8) Nakahashi, K., Togashi, F., Fujita, T., and Ito, Y., Numerical Simulations on Separation of Scaled Supersonic Experimental Airplane from
Rocket Booster at Supersonic Speed, AIAA Paper 2002-2843, June 2002.
(9) Murayama, M., Togashi, F., Nakahashi, K., Matsushima, K., and Kato, T. Simulation of Aircraft Response to Control Surface Deflection
Using Unstructured Dynamic Grids, Journal of Aircraft, Vol. 42, No. 2, March-April 2005, pp. 340-346.
(10) Murayama, M. and Yamamoto, K., Comparison Study of Drag Prediction for the 3rd CFD Drag Prediction Workshop by Structured and
Unstructured Mesh Method, AIAA Paper 2007-0258, Jan. 2007.
(11) Ito, Y., Shih, A. M., Soni, B. K. and Nakahashi, K., Multiple Marching Direction Approach to Generate High Quality Hybrid Meshes, AIAA
Journal, Vol. 45, No. 1, January 2007, pp. 162-167
(12) Ito, Y., Shih, A. M. and Soni, B. K., Octree-Based Unstructured Hexahedral Mesh Generation, 3rd Asian-Pacific Congress on
Computational Mechanics (APCOM'07) in conjunction with 11th International Conference on the Enhancement and Promotion of
Computational Methods in Engineering and Science (EPMESC XI), Kyoto, Japan, December 2007
(13) Ito, Y., Shih, A. M. and Soni, B. K., Reliable Isotropic Tetrahedral Mesh Generation Based on an Advancing Front Method, Proceedings of
the 13th International Meshing Roundtable, Williamsburg, VA, September 2004, pp. 95-105
(14) Ito, Y., Murayama, M., Yamamoto, K., Shih, A. M. and Soni, B. K., Efficient CFD Evaluation of Small Device Locations with Automatic
Local Remeshing, AIAA Paper 2008-7180, 26th AIAA Applied Aerodynamics Conference, Honolulu, Hawaii, August 2008
(15) Kobayashi, M. H. On a Class of Pade Finite Volume Methods, J. Comp. Phys. 156 1999 , pp.127-180.
(16) Gaitonde, D. V., and Visbal, M. R., Pade-Type Higher Order Boundary Filters for the Navier-Stokes Equations, AIAA Journal, Vol.38,
No.11, 2000, pp2103-2112.
(17) Shima, E., A Simple Implicit Scheme for Structured/Unstructured CFD, Proceedings of 29th Fluid Dynamics symposium, Hokkaido,
Japan, 1997, pp.325-328. (in Japanese)
(18) Ffowcs Williams, J. E., and Hawkings, D. L., "Sound Generation by Turbulence and Surfaces in Arbitrary Motion," Phil. Trans. Roy. Soc.
(London), Ser. A, 264, 321-342. (Philosophical Transactions of the Royal Society of London, Vol. 264A, May 1969, pp. 321-342.)
(19) A.S.Lyrintzis, "Surface integral methods in computational aeroacoustics - From the (CFD) near-fiedl to the (Acoustic) far-field", Inter. J.
Aeroacoustics, Vol. 2, No. 2, 2003, pp.95-128.