Project Number : PS 3.1
Unsteady, Turbulent, Separated Flow
Around Helicopter Fuselages
PENNSTATE
1 8 5 5
PI:
Prof. Lyle N. Long
tel : (814) 865-1172
Email: lnl@psu.edu
Web: http://www.personal.psu.edu/lnl/
Comanche
Graduate Student:
Emre Alpman (PhD 2005)
Bell 214
2005 RCOE Program Review
May 3, 2005
Technical Barriers
European Helifuse
investigation found that
turbulence models such as k-,
k-, Baldwin-Lomax were not
able to accurately predict lift
and drag on complex helicopter
geometries.
RANS-based CFD methods
cannot accurately predict the
unsteady turbulent flow around
rotorcraft fuselages.
Objectives:
•
Develop better numerical methods for flow around helicopter fuselages and for
drag prediction
Approach:
•
•
•
Unstructured grid CFD methods on inexpensive parallel computers
Validate code on simple shapes such as spheres and ellipsoids
Make detailed comparisons between experimental data and numerical predictions
for flow around helicopter fuselages
Expected Research Results or Products:
•
•
Better numerical algorithms and understanding of unsteady separated flows
Efficient parallel CFD codes
Very
Complex
Geometries
PUMA2 Flow Solver
• Finite volume ANSI C++ parallel program
• Message Passing Interface (MPI) used for inter-processor
communication
• Unstructured grids to handle very complex geometries
• Runge-Kutta for time-accurate runs
• SSOR for steady-state runs
• Turbulence:
• Large Eddy Simulation (LES) with wall function
• Reynolds Stress Model (RSM)
• Runs on any Beowulf cluster or parallel computer
Turbulence Models
Approximate
Equations
Time
Average
Do not use
Boussinesq
Reynolds Stress
Model
(7 new PDE’s)
Exact
Equations
More
Physics
DNS
Unsteady,
Spatially
Filter
Use
Boussinesq
assumption
LES
2 Equation
Models
(K- & K-)
These are about as good as they are going to get
--and they are not good enough for rotorcraft !!
DES
combines
these
Less
CPU
Time
1 Equation
Models
(SpalartAllmaras)
Algebraic
Models
(e.g. BaldwinLomax)






~
 u 'i u ' j 
 u 'i u ' j Vk
t
xk


advection
Reynolds
Transport Exact
Equations
& RSM
Model
12 nonlinear coupled PDE’s:
- 6 Re Stress eqtns
- 1 Turb. Dissipation eqtn
- 5 Navier-Stokes Equations
Launder, B. E., Reece, G. J., Rodi W., Journal
of Fluid Mechanics, vol.68, part 3, 1975.
Wilcox, D. C., "Turbulence Modeling for
CFD", DCW Industries Inc.




  u 'i u ' j u 'k  u 'i  ' jk u ' j  'ik
xk



diffusion
 u 'i u ' j 2 u 'k 

 p' 



ij
 x


x
3

x
j
i
k

pressure redistribution
~
~

V j
Vi 

   u 'i u 'k
  u ' j u 'k



x

x
k
k

production
u ' j 

u 'i

    ' jk
  'ik
x
xk 

k

dissipation
Modelled
RSM Solution for a 6:1 Prolate
Spheroid






Re = 6.5x106
M = 0.1322
α = 30°
Turbulence intensity: 0.03%
Grid is composed of 5.1
million tetrahedral cells
Solution took 7 days on 30
2.4 GHz Xeon processors
6:1 Prolate Spheroid (RSM)
Lateral Skin Friction Comparison at x/L = 0.738
Re = 6.5x106, M = 0.1322,  = 30 deg
0.0030
0.0025
0.0020
Experiment
RSM Solution
0.0015
Cflat
0.0010
•Qualitative agreement
with experiment
•Experimental data also
contain some uncertainties
0.0005
0.0000
-0.0005
-0.0010
-0.0015
-0.0020
90
100
110
120
130
140
150
160
170
180
[deg]
Alpman, E., and Long, L. N., AIAA Paper 2005-1094, 2005
Experiment: Kreplin, H. P., Volmers H., Meier H. U., DFVLR Rept, IB 222-84 A 33, 1985.
6:1 Prolate Spheroid (RSM)
• Vorticity contours with surface
skin friction lines
• Asymptotic convergence of skin
friction lines means separation
• At the upper lee side of the body
a second separation line is also
observed
RSM Solution
Measurement
Circumferential Location of Primary
Separation [degrees]
~ 105
~ 108
Circumferential Location of Secondary
Separation [degrees]
~ 159
~ 156
RSM Solution for a 6:1 Sphere
= 1.14x106
 M = 0.1763
 Turbulence intensity:
0.45%
 Grid is composed of 3.8
million tetrahedral cells
 Solution took 6 days on
30 2.4 GHz Xeon
processors
 Re
RSM Solution & Experiment Sphere
Re = 1.14x106 M = 0.1763
Midplane Skin Friction Coefficient Distribution
Re = 1.14x106 , M = 0.1763
Circumferential Pressure Distribution of a Sphere
Re = 1.14x106
1.50
4.00
3.50
1.00
3.00
Cf*sqrt(Re)
Cp
0.50
0.00
-0.50
2.50
Experiment
RSM Solution
2.00
1.50
1.00
0.50
Experiment
RSM
-1.00
-1.50
0
20
40
60
80
100
 [deg]
120
140
160
180
0.00
-0.50
-1.00
0.00
30.00
60.00
90.00
120.00
 [deg]
Alpman, E., and Long, L. N., AIAA Paper 2005-1094, January, 2005
Achenbach, E., Journal of Fluid Mechanics, Vol. 54, No. 3, 1972, pp. 565 – 575.
150.00
180.00
Sphere
Re = 1.14x106 M = 0.1763
Normalized τxx contours
Normalized τxz contours
• In isotropic turbulence, normalized τxx and τxz take the values of 2/3 and 0
respectively
• Flow is highly anisotropic
• Anisotropic models (e.g. RSM) necessary for 3-D separated flows
Sphere Drag Prediction
Re = 1.14x106 M = 0.1763
Cd
Experiment
(Achenbach, JFM 1972)
0.13 ± 0.01
LES
(Jindal & Long, 2004)
0.141
RSM
(Alpman & Long, 2005)
0.141
RSM Solution for a Bell 214ST
Fuselage









Re = 1.5x106 per ft
M = 0.3322
α = -2.28°, ψ=0° (low angle of attack cruise condition)
α = 17.04°, ψ=0° (high angle of attack condition)
α = -1.6°, ψ=16.4° (high yaw angle condition)
α = -2.28°, ψ=0° (low angle of attack cruise condition with rotors
modeled using momentum theory with linear loading)
Turbulence intensity: 1%
Grid is composed of 2.9 million tetrahedral cells
Solution took 7 days on 30 2.4 GHz Xeon processors
Computational Mesh
BELL 214ST
y+ ~ 40
Low Angle of attack Cruise Condition
Re = 1.5x106 per ft
M = 0.3322
(without rotors)
Dorsal Centerline Pressure Distribution
Re = 1.5x106 ft-1, M = 0.2322,  = -2.28 deg.,  = 0 deg.
-1.50
-1.00
Cp
-0.50
0.00
0.50
1.00
RSM Solution
Experimental Data
1.50
0
2
4
6
8
10
12
14
x (m)
Good agreement with the measurements.
Surface Pressure Distribution
Alpman, E., and Long, L. N., AHS International 61st Annual Forum and Display, June, 2005
Experiment: Oldenbuttel, R. H., Report No. LSWT 554, Vought Corporation, 1978.
High Angle of Attack and High Yaw
Angle Conditions (without rotors)
Dorsal Centerline Pressure Distribution
Re = 1.5x106 ft-1, M = 0.2322,  = -17.0 deg.,  = 0 deg.
Dorsal Centerline Pressure Distribution
6 -1
Re = 1.5x10 ft , M = 0.2322,  = -1.6 deg., = 16.4 deg.
-2.00
-1.50
-1.50
-1.00
-1.00
-0.50
Cp
Cp
-0.50
0.00
0.00
0.50
0.50
1.00
RSM Solution
Experimental Data
1.00
0
2
4
6
8
10
12
x (m)
High Angle of Attack Condition
Good agreement even when the
expansions are quite abrupt
14
RSM Solution
Experimental Data
1.50
0
2
4
6
8
10
12
14
x (m)
High Yaw Angle Condition
Good agreement with the measurements
except around the tail boom.
Mainly due to differences between wind
tunnel and computational geometry
High Angle of Attack Condition
Normalized τxz contours
Computed Using RSM
Normalized τxz contours
Computed Using Boussinesq Hypothesis during post
processing
Reynolds stresses and mean strain rates are grossly misaligned.
Turbulence models based on the Boussinesq approximation might
perform poorly for this flow and warrants the use of RSM.
Simulation with Main and Tail Rotors
Vertical velocity contours
without rotors
Vertical velocity contours
T = 17500 lbs., Ttr = 1104 lbs
• Induced downwash velocities
• Tip vortices at the edge of rotor plane
Simulation with Main and Tail Rotors
Normalized τyz contours
without rotors
Normalized τyz contours
T = 17500 lbs., Ttr = 1104 lbs
Vortices generated by the main rotor affects downstream
turbulence structure
Bell 214ST Total Drag Predictions
Re = 1.5x106 per ft M = 0.3322 (without rotors)
 = -2.28 and  = 0
D/q (ft2) % Error
Wind Tunnel Data
(Oldenbuttel,1978)
4.596
?
LES prediction
(Souliez & Long, 2002)
6.225
35.4
RSM Simulation
(Alpman & Long, 2005)
5.405
17.6
RSM Simulation (with rotors)
(total drag)
5.547
N/A
Bell 214ST Drag Predictions
Re = 1.5x106 per ft
M = 0.3322
(without rotors)
 = -2.28 and  = 0
D/q
(ft2)
%
Err
or
?
Wind Tunnel Data
(Oldenbuttel,1978) (total drag)
4.596
Bell Simulations
(Narramore et.al. 1992) (pressure
drag)
5.466
18.9
RSM Simulation
(pressure drag)
4.356
5.22
RSM Simulation
(total drag)
5.405
17.6
RANS
Solution
Inaccurate
90% of
Total
Drag
Bell 214ST Drag Predictions
Re = 1.5x106 per ft
M = 0.3322
(without rotors)
 = 17 and  = 0
D/q (ft2)
% Error
wind tunnel data
(Oldenbuttel,1978) (total drag)
6.521
?
RSM Simulation
(total drag)
7.159
9.7
 = -1.6 and  = 16.4
D/q (ft2)
% Error
wind tunnel data
(Oldenbuttel,1978) (total drag)
15.058
?
RSM Simulation
(total drag)
12.886
14.4
Accomplishments

CY 2002 Accomplishments:
– Drag of Bell 214 compared to experiment
– Unsteady tail loadings predicted on Bell 214 ST
– Steady Comanche fan-in-fin simulations compared to experiment

CY 2003 Accomplishments
–
–
–
–
–

Comanche Fan-in-fin Simulations: Unsteady, rapid maneuvers
LES Wall function implemented on unstructured grids
RSM implemented on unstructured grids
LES & RSM Sphere Simulations:
LES & RSM Ellipsoid simulations:
CY 2004 Accomplishments
– Detailed comparisons between LES & RSM
– Bell 214ST RSM & LES simulations
– French fuselage simulations ?
Completed
Short Term
Long Term
Schedule / Milestones
Tasks
Comparisons to experimental data
- Cone & 3-D Cylinder
Generic Fuselage Simulations
- Robin Body w & w/o NLDE
R. Hansen Ph.D. Thesis
Bell 214ST grid & steady solution
Unsteady loads and drag
F. Souliez Ph.D. Thesis
Grid and viscous flow over ellipsoid
Re-Stress Model for turbulent flow
over ellipsoid
S. Jindal M.S. Thesis
Steady/unsteady Comanche flows
Detailed compare of RSM & LES
Re-Stress Model & LES for Bell 214
Helicopter drag and unsteady flows
E. Alpman Ph.D. Thesis
2001
2002
2003
2004
2005




Publications & Theses

2005:
–
–
–
–
–

Alpman, Long, “Separated Flow Simulations,” AIAA-2005-1094, January, 2005
Alpman Long, “Bell 214ST RSM Simulations”, AHS Annual Forum, June 2005.
Lee, Sezer-Uzol, Horn, and Long, “Ship Airwakes,” Jnl of Aircraft, 2005
Sezer-Uzol, PhD Thesis, 2005
Alpman, PhD Thesis, 2005
2004:
–
–
–
–
–
–
Corfeld, Strawn, and Long, “Martian Rotor,” AHS Journal, 2004
Jindal, Long, Plassmann, and Sezer-Uzol, “LES,” AIAA 2004-2228, 2004
Modi, Sezer-Uzol, Long, Plassmann, “Visualization,” Jnl of Aircraft, 2004
Alpman, Long, and Kothmann, “Comanche (steady),” Jnl. of Aircraft, 2004
Alpman, Long, and Kothmann, “Comanche (unsteady),” Jnl. of Aircraft, 2004
Jindal, MS Thesis, 2004
Publications & Theses (cont.)

2003:
– Alpman, Long, and Kothmann, “Comanche,” AHS Forum, 2003
– Lee, Sezer-Uzol, Horn, and Long, “Ship Airwakes,” AHS Forum, 2003
– Alpman and Long, “Comanche,” AIAA-2003-4231, CFD Conf., June, 2003.
 2002:
– Souliez, Long, Sharma, and Morris, Intl. Jnl. of Aeroacoustics,
– Corfeld, Long and Strawn, AIAA Paper, St. Louis Mtg., June, 2002
– Souliez, Long, Morris, and Sharma, AIAA 2002-0799, Reno, Jan., 2002
– Hansen and Long, AIAA 2002-0982, Reno, Jan., 2002
– Fred Souliez, Ph.D. Thesis (Unsteady CFD for Helicopter Fuselages) (at BMW)
– Anirudh Modi, Ph.D. Thesis (Computational Steering and Wake Vortices) (at Intel)
– Kelly Corfeld, M.S. Thesis (CFD for Martian Rotorcraft) (at Lockheed)
 2001:
– L. Long, P. Plassmann, and A. Modi, “Airport Capacity,” London, Sept., 2001
– Long and Modi, NCSA Linux Revolution Conference, Illinois, June, 2001.
– LTC Bob Hansen, Ph.D. Thesis (Unsteady CFD using unstructured grids)
– Nilay Sezer-Uzol, M.S. Thesis (CFD simulations of rotors)
– Anupam Sharma, M.S. Thesis (ship airwake simulations)
Technology Transfer:






Worked with Bruce Kothmann at Boeing Helicopter on Comanche fan-in-fin
Got Bell 214ST data from Jim Narramore
Working with Georgia Tech (joint DARPA project)
Very good relationship with West Point (USMA)
Graduate student (Kelly Corfeld) was in Co-op program with NASA Ames
Rotorcraft worked on Martian Rotorcraft with Dr. Roger Strawn (she is now at
Lockheed)
Working with Dr. Earl Duque who is using different CFD approaches
Leveraging or Attracting Other
Resources or Programs
• DARPA quiet helicopter project (joint Penn State, GATech, & NAU effort)
• NSF Center for Particle Methods (Monte Carlo, Molecular Dynamics, & Vortices)
• Army DURIP grant for computer hardware
• 120 processor Beowulf for RCOE center (with Prof. Brentner)
• Institute for Computational Science and Engineering (2004), wide-spread
•
financial support across Penn State
Grant from National Renewable Energy Lab for Wind Turbine Aeroacoustics
(with Morris and Brentner)
2005 Recommendations:
The task work is excellent. It is suggested to compare various
turbulence modelings and to contact Langley for Comanche tail
buffett data.
Response:
Have compared LES and RSM for same geometry. Have
also post processed results to see what 2-equation model
might yield. With cancellation of Comanche we decided to
focus on Bell 214ST and simple shapes with good
experimental data.