Sei sulla pagina 1di 227

An Integrated Navier Stokes - Full Potential -

Free Wake Method for Rotor Flows




Ph. D Work by

Mert Enis Berkman


Advisors: Prof. S. M. Ruffin & Prof. L. N. Sankar



Georgia Institute of Technology
School of Aerospace Engineering
OUTLINE
Review of Rotorcraft CFD Techniques
Why is a hybrid approach more favorable ?
Hybrid Solver
Navier-Stokes Zone
Full Potential Zone
Boundary and Interface Conditions
Wake Model
Results
Hover Analysis
Two-Bladed, UH-60A and Tapered Tip Rotors
Forward Flight Analysis
Two-Bladed, UH60A and H-34 Rotors
Conclusions

Performance of rotary wings is limited by:
transonic flow (on advancing blade)
stall (on retreating blade)
operation under its own wake.

The flow field is 3-D, unsteady, viscous and compressible.

Rotor wake is a distorted, skewed helix that stays in the
vicinity of the rotor and affects entire flow field.

The rotor wake structure determines performance, vibratory
airloads and acoustics.

Modeling the wake and its effects remains a very challenging
task.
ROTARY WING AERODYNAMICS
Lifting - Line (- Surface) Methods

Blades are modeled as a lifting-line (or -surface).

Wake is represented by a network of vortex filaments.

Routinely used in industry. They need small CPU time, thus
easily incorporated into comprehensive codes as aerodynamics
modules.

They require table look up for airfoil load data, and are often
quasi-steady.

They are loaded with empirical corrections.
ROTORCRAFT CFD
Finite-Difference Methods: (Potential, Euler and N-S)

a) Finite-Difference Methods with External Wake Model:

The flow field is solved near the blade; the effects of the far
wake is modeled. solved

They can handle compressible flows.

They require external coupling with
a wake model to account for far wake.

modeled



ROTORCRAFT CFD






b) Wake Capturing Schemes

This class of methods attempt to capture the far wake as a
part of the solution.

They provide high quality detailed flow field solutions.

They require enormous computer time since they need to
resolve the tip vortex adequately.

They diffuse the tip vortex too rapidly due to the dissipative
nature of Euler/N.-S. schemes.

Higher order schemes, overset and/or unstructured grids were
used to conserve vorticity without significant success.
ROTORCRAFT CFD
Vorticity Embedding Technique

A unique finite difference technique that eliminates wake diffusion.

Vortex sheets are basically embedded inside a potential flow field
and their effect is confined to a small region.

The wake is tracked by a Lagrangean approach using vorticity
markers.

The technique is gaining popularity in helicopter industry due to
is efficiency and success in predicton of hovering rotor loads.

It lacks viscous features.

It cannot model BVI, dynamic stall and tip vortex generation well.


ROTORCRAFT CFD
c) Hybrid Schemes

They integrate different methods in different flow regions to
improve solution quality without a big penalty in computer time.

A hybrid rotor solver developed by Berezin and Sankar uses N.-S.
equations near the blades and full potential equations elsewhere
since viscous effects are negligible away from the blades.

The method typically shows 40% reduction in CPU time without loss
in accuracy compared to full blown N-S solution.

This method is available for hovering as well as advancing rotors.

This scheme requires coupling to a comprehensive code for account
for far wake and trim effects.
ROTORCRAFT CFD

Moulton and Caradonna integrated the Vorticity Embedding
Technique with a Navier-Stokes solver.

The resulting method enjoys advantages of high order Navier-
Stokes methods and wake treatment of vortex embedding scheme.

This scheme can freely convect wakes without diffusion and
account for viscous effects over the blade.

However, the method is limited to steady-state analysis, it can not
be used to analyze advancing rotors.

ROTORCRAFT CFD
ROTORCRAFT CFD
Current operational Current research Proposed research
methodology methodology methodology
First Generation Second Generation Hybri d Approach
Free Wake
Model
Compute Inflow
Potential,
Euler or N-S
Loads
Euler or N-S
Loads
Potential
Flow
Free
Wake
N-S
Loads
Trim
Trim
Trim
HYBRID SOLVER VERSUS OTHERS
A hybrid technique offers the following capabilities:

Capture viscous phenomena efficiently
Eliminate tip vortex preservation problem
Avoid external wake models
Be applicable to hovering and advancing rotors
Offer high order accuracy


No other existing CFD technique combines ALL of the
properties listed above !
OBJECTIVES OF THIS RESEARCH

1. Develop a new hybrid technique that will feature the
capabilities listed.

2. Validate the method by comparison with experimental or
flight test data available for realistic helicopter rotor
configurations.

3. Study the effects of wake model, grid density and spatial
accuracy on the solution quality.
HYBRID SOLVER
Three Modules:

Navier-Stokes Zone
Full Potential Zone
Lagrangean Wake
N-S zone
FPE zone
Lagrangean Wake
HYBRID SOLVER
Navier-Stokes (Inner) Zone
Viscous features are captured, including separation.
The near wake is captured as a part of the solution.
Far wake effects are felt through interface boundaries.

Potential (Outer) Zone
Viscous effects are negligible away from the blades.
The inner wake structure is not modeled.
An induced vortical velocity field due to a concentrated
tip vortex is generated.

Lagrangean Wake
The tip vortex emanating from each blade is represented
by a series of piecewise linear elements.
The tip vortex may deform based on local flow.
HYBRID SOLVER
N. -S.
FP
Kmatch
block 2
block 1
1
2
3
A cut in the radial plane

Inner and outer zones

Finite volume technique that uses Reynolds Averaged Navier-
Stokes equations.

Third order or fifth order accurate terms for inviscid fluxes
crossing cell faces.

2nd order accurate modeling of viscous terms.

MUSCL scheme with Roe averaging.

Baldwin-Lomax turbulence model.


INNER ZONE

In transformed coordinates, the N.-S. equations may be be written
in differential form after nondimensionalization as

INNER ZONE
Finite volume and finite difference fluxes in generalized coordinates
are related via


q
t
+F

+G
q
+H

=
1
Re
R

+S
q
+T

( )
S n E T
S n E H
S n E S
S n E G
S n E R
S n E F
V
I
V
I
V
I
A - =
A - =
A - =
A - =
A - =
A - =

for cell faces (i+1/2, j, k) and (i-1/2, j, k)




for cell faces (i, j+1/2, k) and (i, j-1/2, k)


for cell faces (i, j, k +1/2) and (i, j, k-1/2)


Treatment of Inviscid Fluxes

INNER ZONE
( )
( )
Vol
S n E
T S R RHSV
Vol
S n E
H G F RHSI
RHSV RHSI
dt
dq
V
I

A -
= + + =
A -
= + + =
+ =

q
q
o o o
o o o
Re
1
Roes approximate Riemann solver is used to calculate inviscid
fluxes

F
i+1/ 2, j,k
= 0.5 F
L
+F
R
( ) C q
L
q
R
( )
| |

Treatment of Inviscid Fluxes (cont.)

INNER ZONE
A third order spatial accuracy is obtained with a MUSCL scheme
i-1,j,k i,j,k L R i+1,j,k

i-1/2,j,k i+1/2,j,k

q
L
= q
i
+1/ 3(q
i+1
q
i
) + 1/ 6(q
i
q
i1
)
q
R
= q
i
1/ 3(q
i +1
q
i
) 1/ 6(q
i +2
q
i +1
)
or as an option a fifth order scheme which uses information from
two additional nodes is available.

Treatment of Viscous Fluxes

The viscous fluxes are also calculated with a finite volume scheme.
INNER ZONE
An eddy viscosity class of models is used to model Reynolds
stresses and turbulent transport of heat flux.

=
L
+
T


Pr = Pr
L
+ Pr
T

Temporal Discretization and Diagonal ADI Factorization

A first order semi-implicit scheme is used in the study.

The viscous fluxes are lagged by a time step.

Beam and Warmings linearization is done.

A diagonal ADI factorization proposed by Pulliam and Chaussee
is employed to solve the resulting system of equations.
INNER ZONE
The unsteady full potential equations are solved in the outer
zone away from the rotor blade.

Consider the continuity equation
OUTER ZONE
0 ) ( = - V + V
t


w
V V

+ V = |
The second term is a superimposed vortical velocity field that is
induced by the rotor wake.

The velocity consists of two parts
Along with energy equation and isentropic gas relation, and
after manipulations the governing equations form a second
order hyperbolic PDE as,







This system is converted in a set of equations to be solved for
perturbation potential.

A three factor ADI scheme is used to solve the system.
OUTER ZONE

a
2
|
\
|
.
|
tt
+ |
x
|
xt
+ |
y
|
y t
+ |
z
|
zt
| |
= |
x
( )
x
+ |
y
( )
y
+ |
z
( )
z
COUPLING OF THE TWO ZONES
Acoustic waves
V
n
+a
V
n
-a
N.-S. FP
Vorticity waves V
n
vortex filaments
Entropy wave V
n

Flow Information Transfer from Inner to Outer Zone

The outer zone requires specification of velocity potential along
all zonal interfaces that separate the two zones.

The normal component of velocity at the interface is passed to
the outer zone.

This condition is used to match the normal derivative of the
potential at the interface.

This type of a condition assures a smooth informtion passage
from inner to outer zone.
COUPLING OF THE TWO ZONES
Flow Information Transfer from Outer to Inner Zone

The velocity components at the interfaces are found by addition of
potential and wake induced velocities as
COUPLING OF THE TWO ZONES
k V w
j V v
i V u
w z S N
w y S N
w x S N

- + =
- + =
- + =

|
|
|
FP
N-S
Temperature and density are found from isentropic gas law
at the interface.
Flow Information Transfer from Outer to Inner Zone

Although these relations do not account for information from the
inner zone, they have been shown to pass information
accurately enough in the past for most cases.

In this study characteristic based interface conditions coded by
Mello is also available as an option.

These non-reflecting boundary conditions use Riemann invariants
so that flow information from both sides are used in subsonic
flows.
COUPLING OF THE TWO ZONES

Computatinal domain covers
1 -2 blade radius above and below the rotor disk
0.6 - 1 blade radius beyond the blade tip

BOUNDARY CONDITIONS
Inner zone (3 boundaries)

On blade surface: no-slip or flow tangency
Inboard: extrapolation from interior
Outboard: extrapolation from interior

Outer zone (6 boundaries)

Inboard: extrapolation from interior
Outboard: set to free stream conditions
Plane above rotor disk: set to free stream conditions
Plane below rotor disk: set to free stream conditions
Upstream plane: set to free stream conditions
Downstream plane: set to free stream conditions

BOUNDARY CONDITIONS
BOUNDARY CONDITIONS
Only one of the rotor blades is resolved in the hybrid technique.

For hover analysis since the flow field is periodic about each
blade only one blade needs to be resolved by CFD methods.

In forward flight analysis, all blades need to be resolved in
typical CFD methods.

In the hybrid method, the far wake of each blade is modeled (not
solved for), therefore the influence of the other blades are
accounted via their wake.

Significant reduction in computer time and memory requirements
since only one blade is resolved.

WAKE MODEL
Wake shed from the blade is captured inside the Navier-Stokes
zone in the near field.

The inner wake is neglected in the Full Potential zone once it
leaves the Navier-Stokes zone.

Each individual blades tip vortex is modeled by a series of
piecewise linear elements.

These elements are introduced at the inner/outer zone boundary
behind the blade trailing edge and extends for several revolutions
below the rotor plane outside the computational domain.

WAKE MODEL

The wake elements that lie in the computational domain are
allowed to move with the local flow for hover analysis.

In forward flight analysis, the wake elements remain fixed at
their predefined position, i.e. they are not allowed to move.

The induced velocity field due to these vortex filaments is
calculated by Biot-Savart law.

Two parameters are needed by Biot-Savart law to generate the
wake induced velocity field in the outer zone:
- vortex strength
- wake shape
WAKE MODEL
The Biot-Savart law is

I
=
elements
w
r
r l d
V
3
4

t
P
r
dl
WAKE MODEL
The tip vortex strength is taken to be either
- the bound circulation at 97-99% blade, or
- the peak bound circulation

The bound circulation at a blade section is calculated by the Kutta-
Joukowski theorem.
L = VI

In hover analysis, the tip vortex strength is updated at each iteration
with the change in blade loading.

In forward flight analysis, the tip vortex changes at every time level
based on current lift variation over the blade.

WAKE MODEL
To start the solution process either
- a non-contracting classical wake, or
- a contracting prescribed wake
structure is assumed.

Computational domain
Tip vortex
trajectory
Wake element
Wake markers
In the rigid wake option this wake shape remains unchanged.

In the free wake option, the wake markers are allowed to move
freely with the flow.

Inside the computational domain these markers are tracked using
a Lagrangean technique. The following procedure is followed
every time level in the free wake option:

1. determine in which cell each wake marker lies,
2. calculate the local flow speed at these positions,
3. move each marker to its new position

The remaining wake markers are attached to the last free marker
appropriately.
WAKE MODEL
Since the wake shape deforms at each time step, ideally, the
wake induced velocity coefficients should be updated at each
time step too.

However, calculation of these coefficients is computationally
intense. Therefore, only after each 10 degrees of blade rotation
the update is performed based on the latest wake structure.

In hover analysis this delay does not cause any problems since
a single steady-state solution is sought.

In forward flight, these updates need to be done at every time step
rendering the free wake option impractical.
WAKE MODEL
VALIDATION STUDIES
Two-Bladed Rotor in Hover
UH-60A Rotor in Hover
Tapered Tip Rotor in Hover
Non-Lifting Two-Bladed Rotor in High Speed Forward Flight
UH-60A Rotor in Forward Flight
H-34 Helicopter in Forward Flight
HOVERING TWO-BLADED ROTOR
Two-bladed rotor tested by Caradonna and Tung
Rectangular planforn, no twist, NACA 0012 sections, AR = 6

Non-Lifting Case
(Collective pitch = 0
o
, Tip Mach No. = 0.52, Reynolds No. = 2 Mil.)

r/R = 68% r/R = 96%
HOVERING TWO-BLADED ROTOR
Non-Lifting Case





upper half: hybrid solver




lower half: full Navier-Stokes


Hybrid solver is twice
as fast as the full Navier-
Stokes solver !

Density
contours





EFFECT OF WAKE MODEL
Two-Bladed Rotor: (Collective pitch = 5
o
, Tip Mach No. = 0.794)

r/R = 68% r/R = 96%





Two-Bladed Rotor:
Collective pitch = 8
o
, r=50%
Tip Mach No. = 0.439

Full N.-S. : 1 full revolution
Hybrid : 1/3 revolution

r=80% r=96%
FULL N.-S. VS. HYBRID SOLVER





EFFECT OF NUMBER OF WAKE ELEMENTS
Two-Bladed Rotor: (Collective pitch = 8
o
, Tip Mach No. = 0.612)

r/R = 50%





EFFECT OF WAKE STRUCTURE
Two-Bladed Rotor: (Collective pitch = 8
o
, Tip Mach No. = 0.612)

r/R = 68%





EFFECT OF WAKE STRUCTURE
Two-Bladed Rotor: (Collective pitch = 8
o
, Tip Mach No. = 0.612)

r/R = 96%





TIP VORTEX POSITION
Two-Bladed Rotor: (Collective pitch = 8
o
, Tip Mach No. = 0.612)

Radial Position of the Tip Vortex Variation with Iteration
0.8
0.85
0.9
0.95
1
1.05
0 50 100 150 200 250 300 350 400
r
T
V
/
R
Vortex Age (deg.)
i terati on





FIFTH ORDER VS. THIRD ORDER
Two-Bladed Rotor: (Collective pitch = 8
o
, Tip Mach No. = 0.612)

r/R = 68% r/R = 89%
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Experiment
3rd order
5th order
C
p
x/c
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Experiment
3rd order
5th order
C
p
x/c





HOVERING UH-60A ROTOR
Four twisted blades with rearward swept tip and two
different airfoil sections, AR = 15.3.

Collective pitch=10
o
, Tip Mach No. = 0.628, Reynolds No.=2.5 Mil.

A two block H-O grid with 90 chordwise, 43 spanwise and 80
normal nodes.

Approximately 37% of the nodes lie inside the Navier-Stokes zone.

Free wake option with 10 wake revolutions modeled.

Initial wake is non-contracting and updated every 10
o
of blade
rotation.

Peak bound circulation used as the tip vortex strength.






HOVERING UH-60A ROTOR
A total of 2850 iterations were enough to reach steady state.

Experimental Numerical
C
T
/o=0.085, C
Q
/o=0.0070 C
T
/o=0.086, C
Q
/o=0.0074

Sectional Thrust
Coefficient Variation





HOVERING UH-60A ROTOR
Chordwise Pressure Coefficients
r/R=40% r/R=67%
r/R=87%
r/R=99%





HOVERING TAPERED TIP ROTOR
Four dual linearly twisted blades, with taper starting at 82% radius.

Two different airfoil sections and AR=15.3.

Collective pitch=8.6
o
, Tip Mach No. = 0.628, Reynolds No.=2.5 Mil.

Two block H-O grid with 90 chordwise, 43 spanwise and 80
normal nodes.

Approximately 37% of the nodes lie inside the Navier-Stokes zone.

Rigid wake option with 10 wake revolutions modeled.

Initial wake shape is based on the K-T prescribed wake.






HOVERING TAPERED TIP ROTOR
Collective pitch = 8.6
0
, Tip Mach No. = 0.628



r /R= 77.5% r/R = 94.5%





ADVANCING UH-60A ROTOR
Tip Mach No. = 0.628, Advance Ratio = 0.3, Reynolds No.=2.5 Mil.

The blade motion is prescribed by a table, u = u(r ,+).

Unstable behavior is observed with this blade deformation scheme.

The blade pitching motion approximated as




Rigid wake option with 10 wake revolutions modeled.

Peak bound circulation used as the tip vortex strength.


u = 8.57.5cos +1.84sin





ADVANCING UH-60A ROTOR
r/R=77.5%
Pressure Coefficients at = 0
o
r/R=67.5%
r/R=94.5%
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
Ex p-u ppe r
Ex p-l ower
Hy b-u ppe r
Hy b-l ower
C
p
x /c
r/R=86.5%
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c





ADVANCING UH-60A ROTOR
r/R=77.5%
Pressure Coefficients at = 90
o
r/R=67.5%
r/R=94.5%
r/R=86.5%
-2
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c
-2
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c





ADVANCING UH-60A ROTOR
r/R=77.5%
Pressure Coefficients at =18 0
o
r/R=67.5%
r/R=94.5%
r/R=86.5%
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c
-2
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c





ADVANCING UH-60A ROTOR
r/R=77.5%
Pressure Coefficients at = 270
o
r/R=67.5%
r/R=94.5%
r/R=86.5%
-4
-3
-2
-1
0
1
2
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c
-4
-3
-2
-1
0
1
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c
-4
-3
-2
-1
0
1
2
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
Exp-upper
Exp-lower
Hyb-upper
Hyb-lower
C
p
x/c





CONCLUDING REMARKS
The hybrid solver is a very efficient new method for
prediction of complex viscous unsteady flows over
isolated helicopter rotors.

It requires only half to the CPU time compared to a full blown
Navier-Stokes solver over the same grid per time level.

It converges in less number of iterations for hover cases.

Only one blade needs to be resolved for advancing rotors.

The hybrid solver does not need fine grid away from the blades,
has no wake diffusion in the far field due to its unique wake
treatment.

A single method for rotors in hover and forward flight.





RECOMMENDATIONS FOR FUTURE STUDIES
Inclusion of inner wake

Newton sub-iterations

Turbulence modeling

Adaptive stencil

Parallel processing

Coupling with a structural dynamics solver

Overset grids

A Hybrid Flow Analysis for Rotors
in Forward Flight
A Ph.D Thesis Presentation


Zhong Yang
Advisor: L. N. Sankar

School of Aerospace Engineering
Georgia Institute of Technology
June 26, 2000
Georgia Tech
School of Aerospace Engineering
Technical Barriers Limiting Rotor Performance
in Forward Flight
Overview of Current Research Methods
Hybrid Methodology and Numerical Procedure
Implementation Details
Results and Discussion
UH-60A in high speed flight
AH-1G in low speed descent
Conclusions and Recommendations
Georgia Tech
School of Aerospace Engineering
Outline
Modeling forward flight phenomena requires
detailed modeling aerodynamics (transonic
flow, dynamic stall, blade vortex interaction),
elasticity, blade dynamics and pilot input.
First-principles based aerodynamics analyses
(N-S solver) have been available to the
industries, but are computationally expensive.
In some studies, an open-loop coupling
between CFD solver and the comprehensive
analysis are done.
Georgia Tech
School of Aerospace Engineering
Background
Georgia Tech
School of Aerospace Engineering
Helicopter Aerodynamic Environment
High speed forward flight:
transonic flow, dynamic stall effects
Low advanced ratios:
strong tip vortices, BVI
Flow asymmetry:
caused by complex blade dynamics, bending
and torsional deformation
Problem is multidisciplinary:
aerodynamics, elasticity, blade dynamics and
trim
Georgia Tech
School of Aerospace Engineering
Technical Barriers in Forward Flight
e.g. CAMRAD/JA, 2GCHAS, RDYNE,
UMARC, COPTER

blade element theory

Various wake models

Can handle trim, and elastic effects

These methods are not general enough to model
nonlinear and unsteady effects, except in an
empirical fashion (curve fits or synthesis of airfoil
load tables).
Georgia Tech
School of Aerospace Engineering
Comprehensive Codes
e.g. Caradonna, Chattot, RFS2 by Prichard and
Sankar, FPR by Strawn, HELIX by Steinhoff

limited to weak shock waves and inviscid flow

Far wake is modeled as inflow corrections supplied
from an external wake model (free wake model or
a prescribed wake model)

Vortex embedding techniques are sometimes used.

Rotor is trimmed, and elastic deformations
accounted for, using a comprehensive code
Georgia Tech
School of Aerospace Engineering
Potential Flow Methods
e.g. Wake and Sankar , Srinivasan (TURNS),
Ahmad, Duque and Strawn (OVERFLOW),
Banglore and Sankar, Hariharan and Sankar
Most methods can capture wake as a part of the
solution.
Calculations are limited value because of:
- Excessive numerical viscosity
- Significant computational memory and time
- Rotor not trimmed; blade dynamics and aeroelasticity
inadequately modeled.
Smith, Bauchau and Ahmad coupled NS solvers to
structural dynamics codes.
Georgia Tech
School of Aerospace Engineering
Euler/Navier-Stokes Methods
Berezin (GT), Berkman (GT), Moulton,
Caradonna, and Bangalore (US Army)
Use the most appropriate models in different flow
regions to retain solution quality
Large savings in computer time compared to NS
methods
Berezin coupled hybrid solver to RDYNE to
account for the far wake and trim effects.
Berkman modeled the entire wake from first
principles, and obtained good results in hover.
Georgia Tech
School of Aerospace Engineering
Hybrid Methods
Moulton and Caradonna coupled HELIX to
TURNS for modeling rotors in hover.
Bangalore and Caradonna extended Moultons
work to advancing rotor flows through overset
grids.
Trim and elastic effects were not accounted for in
most of these calculations, excpet in Berezins
work (via RDYNE) and Moultons work (via
CAMRAD).
Georgia Tech
School of Aerospace Engineering
Hybrid Method (continued)
Develop innovative methods , which combine
solution efficiency and accuracy, for modeling
rotors in forward flight.
Validate the methods by comparisons with
experimental and flight test data for realistic
rotor configurations.
Investigate the capabilities and limitations of
the numerical models in capturing flow field
physics.
Georgia Tech
School of Aerospace Engineering
Research Objectives
Navier-Stokes solver:
modeling the viscous
flow and near wake
Potential flow solver:
modeling the inviscid
isentropic flow
Lagrangean approach:
convecting vortex
filaments without
diffusion in the FPE zone
and the far field
Georgia Tech
School of Aerospace Engineering
Hybrid Methodology
N-S zone
FPE zone
Lagrangean Wake
Solves the 3-D Reynolds-Averaged Navier-
Stokes equations.
Scheme is first or second order accurate in
time, third or fifth order in space.
Numerical viscosity provided through Roe
upwind scheme.
Effects of turbulence are modeled with a
Baldwin-Lomax eddy viscocity model.

Georgia Tech
School of Aerospace Engineering
Navier-Stokes Solver
Georgia Tech
School of Aerospace Engineering
Full Potential Solver
The time-marching scheme is fully implicit,
first order accurate in time, second order
accurate in space except in supersonic regions.
0 = V + V
t


wake
V V

+ V = |
Solves the continuity equation in finite volume
form
The velocity field is made of
Georgia Tech
School of Aerospace Engineering
Boundary and Interface Conditions
Georgia Tech
School of Aerospace Engineering
Boundary and Interface Conditions
N.-S.
FP
Kmatch
block 2
block 1
1
2
3
n V V
wake NS n


= ) ( |
wake NS
V V

+ V = | , e from isentropic flow
Georgia Tech
School of Aerospace Engineering
Computational Domain
Conventional N-S Solver Hybrid Method
blade
Domain 1
Domain 2
3
4
Georgia Tech
School of Aerospace Engineering
Implementation Details
CPU time was reduced by performing hybrid
analysis for a single blade.

The other blades are seen by the analysis as
a collection of bound and tip vortices.

There is no more need to match and patch the
grids around multiple moving, deforming
blades.

This allows pitching and flapping motion to be
modeled rapidly without need for inter-blade
grid continuity.
Georgia Tech
School of Aerospace Engineering
Blade Dynamics
A module for computing the rigid blade
motions in flap and pitch, and the complex
blade deformation due to aeroelastic effects has
been developed.

For rigid blades, the (x,y,z) positions in space
at any instance in time may be transformed
using Eulerian angles:

| || || |
old old new
x C B A x T x

= =
Georgia Tech
School of Aerospace Engineering
Blade Dynamics (continued)
If the blade is not rigid, the grid motion should
include additional rotations in twist, and
bending deformations:
bending
r
torsion
n deformatio elastic
dz
dy
dx
x x
(
(
(

+
(
(
(

=

u u
u u
cos 0 sin
0 1 0
sin 0 cos
_
( ) ( ) ( )
(

A + O ~
elastic surface b
z
dx
d
V r n v v sin

Transpiration boundary condition: used in present
calculation.
Exact Approach:
In forward flight, the helical vortices are
carried downward by the induced velocity and
rearward by free-stream velocity.
Prescribed wake model:
Inflow is computed from
Glauerts theory.
Wake markers are located
as follows:




Georgia Tech
School of Aerospace Engineering
Wake Model
| |
| | |
| | |
z z
y y
x x
R R z z
R R r y
R R r x
+ =
+ + =
+ + =
0 marker
marker
marker
) sin(
) cos(
Wake marker
Georgia Tech
School of Aerospace Engineering
Wake Geometry
Non-distorted wake
Distorted free wake
Georgia Tech
School of Aerospace Engineering
Wake Model (continued)
Free wake model:
It can model the distortion from the basic helix.
One option in the code is to search for the computational
cell each wake marker lies in, and calculate local velocities
of the markers by trilinear interpolation. This search can be
costly.
Another option is to directly use the Biot-Savart Law to
evaluate the self induced velocity.
This option was used here.
}
I
= A
3
4
1
r
l d r
v

t
Georgia Tech
School of Aerospace Engineering
Tip Vortex Strength
An automated procedure is employed for an
initial wake geometry and tip vortex strength.

The wake geometry and induced flow are
automatically updated at user-specified intervals
(e.g. every 5 degree azimuth)

As the blade rotates, the bound vortex
circulation is computed from the flow solver by
Kutta-Joukowski theorem:
l T T
cC U U
2
2
1
= I
Georgia Tech
School of Aerospace Engineering
Tip Vortex Strength
found as the peak bound circulation
Positive circulation
distribution
Negative circulation
distribution near the tip
Tip Vortex Strength
Tip Vortex Strength
Georgia Tech
School of Aerospace Engineering
Rotor Trim
User supplies C
T
/o , o
TPP
and desired moments
(usually zero).
The supplied tip path plane angle is used to set
blade flapping motion.
The desired C
T
/o and moments are achieved
through the adjustment of the collective and
cyclic pitch.
) , , (
) , , (
) , , (
1 1 0
1 1 0
1 1 0
s c M M
s c M M
s c T T
x x
y y
c c
c c
c c
u u u
u u u
u u u
=
=
=
Georgia Tech
School of Aerospace Engineering
Rotor Trim (continued)
A Newton-Raphson iterative method is
employed to compute the control settings
changes, and obtain a new guess.
) (
) 0 (
1
1
0
) 0 (
1 1 0
1 1 0
1 1 0
) 0 ( d
M
M
T
s
c
s
M
c
M M
s
M
c
M M
s
T
c
T T
M
M
T
x
y
x x x
y y y
x
y
c
c
c
c c c
c c c
c c c
c
c
c

A
A
A
(
(
(
(
(
(
(

c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
c
+

u
u
u
u u u
u u u
u u u
) 0 (
1
1
0
) 0 (
1
1
0
) 1 (
1
1
0
(
(
(

A
A
A
+
(
(
(

=
(
(
(

s
c
s
c
s
c
u
u
u
u
u
u
u
u
u
Georgia Tech
School of Aerospace Engineering
Validation Studies
2-D code Validation Studies (not documented)

UH-60A Model Rotor in High Speed Forward
Flight

AH-1G Flight Test Rotor in Descent

OLS Model Rotor in Descent
Georgia Tech
School of Aerospace Engineering
UH-60A in High Speed
Forward Flight
Tested in the DNW tunnel in Holland

Complex aerodynamic design: nonlinear twist,
several asymmetric airfoils, and swept tip

Nonlinear elastic deformations
Georgia Tech
School of Aerospace Engineering
UH-60A in High Speed
Forward Flight
Validation case:
Advance ratio =0.3
Tip Mach number M
tip
=0.628, C
T
/o= 0.082
The blades were trimmed to eliminate one-per-rev
flapping.
H-O multi-block grid: After grid sensitivity
studies, a 90x44x80 (NS zone: 62x44x44) grid was
chosen for optimum combination of accuracy and
computational efficiency.
Georgia Tech
School of Aerospace Engineering
Measured Torsional Deformations
) , ( sin 5 . 7
cos 84 . 1 5 . 11
u
u
r
elastic
+
+ =


Modeled
using
Transpiration
Velocity
Georgia Tech
School of Aerospace Engineering
Grid around the UH-60A Blade
Georgia Tech
School of Aerospace Engineering
C
P
at =0
0

(r=67.5%R and 94.5%R)
r/R=67.5%
-2
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
r/R=94.5%
-1.5
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
x/c
C
p
Georgia Tech
School of Aerospace Engineering
C
P
at =120
0

(r=67.5%R and 94.5%R)
r/R=94.5%
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
r/R=67.5%
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
C
p
*

Georgia Tech
School of Aerospace Engineering
C
P
at =270
0

(r=68%R and 94.5%R)
r/R=94.5%
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
r/R=67.5%
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
Georgia Tech
School of Aerospace Engineering
Mach Number Contour at r=96%R
(Blade at +=90
0
)
Georgia Tech
School of Aerospace Engineering
Sectional Normal Load at r=78%R
(with and without elastic deformation)
r/R=78%
0
0.3
0.6
0.9
1.2
1.5
0 45 90 135 180 225 270 315 360
C
n
hybrid without elastic
hybrid with elastic
experiment

Georgia Tech
School of Aerospace Engineering
Sectional Normal Load at r=92%R
(with and without elastic deformation)
r/R=92%
-0.2
0
0.2
0.4
0.6
0.8
1
0 45 90 135 180 225 270 315 360
C
n
hybrid without elastic
hybrid with elastic
experiment

Georgia Tech
School of Aerospace Engineering
Sectional Normal Load at r=78%R
(Comparison with Bangalore and Caradonna)
r/R=78%
0
0.3
0.6
0.9
1.2
1.5
0 45 90 135 180 225 270 315 360
C
n
hybrid with elastic
experiment
Results of Bangalore et al

Georgia Tech
School of Aerospace Engineering
Sectional Normal Load at r=92%R
(Comparison with Bangalore and Caradonna)
r/R=92%
-0.2
0
0.2
0.4
0.6
0.8
1
0 45 90 135 180 225 270 315 360
C
n
hybrid with elastic
experiment
Results of Bangalore et al

Georgia Tech
School of Aerospace Engineering
AH-1G in Low Speed
Descending Flight
Flight test done at NASA Ames
Two-bladed rectangular planform rotor
Linear twist is -10
0

Several calculations have been reported:
Hernandez used FPR coupled with
CAMRAD/JA
Ramchandran applied HELIX-II
Ahmad used Chi-TURNS with Chimera grid
Georgia Tech
School of Aerospace Engineering
AH-1G in Low Speed
Descending Flight (continued)
Validation case:
Advance ratio =0.19
Tip mach number M
tip
=0.65, C
T
/o =0.0713,
o
TPP
= -1.87
0

The first blade harmonics from flight test:
u
0
u
1s
u
1c
|
1s
|
1c
6.0
0
-5.5
0
1.7
0
-0.15
0
2.13
0
u
0
u
1s
u
1c
|
1s
|
1c
8.0
0
-6.5
0
2.5
0
-0.15
0
2.13
0
H-O multi-block grid: 90x43x80
After trimming:
Georgia Tech
School of Aerospace Engineering
C
P
at r=91%R (=90
0
)
-2
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
=90
0
(results of Ahmad et al)
=90
0
(hybrid method)
Due to input airfoil imperfections
Georgia Tech
School of Aerospace Engineering
C
P
at r=91%R (=105
0
and

=270
0
)
-2
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
-3
-2
-1
0
1
2
0 0.2 0.4 0.6 0.8 1
x/c
C
p
=105
0
=270
0
Georgia Tech
School of Aerospace Engineering
Sectional Normal Load at r=97%R
(with and without trimming)
r/R=97%
0
0.1
0.2
0.3
0.4
0.5
0.6
0 90 180 270 360
C
n
Flight test data
Hyb. (trimmed)
Hyb. (no trimming)

Georgia Tech
School of Aerospace Engineering
Sectional Normal Load at r=99%R
(Comparison with Ahmad and Duque)
r/R=99%
-0.1
0
0.1
0.2
0.3
0.4
0.5
0 90 180 270 360
C
n
Results of Ahmad et al.
Flight test data
Hyb. Method (trimmed)

Georgia Tech
School of Aerospace Engineering
Sectional Normal Load at r=99%R
(with different wake models)

r/R=99%
-0.1
0
0.1
0.2
0.3
0.4
0.5
0 90 180 270 360
C
n
Flight test data
Hyb. (f ree wake)
Hyb. (prescribed)
Hyb. (unif orm)
Georgia Tech
School of Aerospace Engineering
Sectional Normal Load at r=99%R
(with different parameters of free wake model)

r/R=99%
-0.1
0
0.1
0.2
0.3
0.4
0.5
0 90 180 270 360
C
n
Flight test data
Hyb. (0.2, 5 degree)
Hyb. (0.2, 10 degree)
Hyb. (0.1, 5 degree)
Georgia Tech
School of Aerospace Engineering
Wake Shed from Blade at +=270
0


-10
-8
-6
-4
-2
0
2
4
6
8
10
-10 0 10 20 30 40 50
X
Z
free wake
prescribed wake
Rotor disk
-5
-4
-3
-2
-1
0
1
2
3
4
5
-15 -10 -5 0 5 10 15
y
z
free wake
prescribed wake
Rotor Disk
-25
-20
-15
-10
-5
0
5
10
15
20
25
-10 0 10 20 30 40 50
X
Y
free wake
prescribed wake
Georgia Tech
School of Aerospace Engineering
Tip Vortex Visualization
Blade at
+=270
0
Free wake
model
Georgia Tech
School of Aerospace Engineering
Tip Vortex Visualization


Blade at
+=270
0
Prescribed
wake model
Georgia Tech
School of Aerospace Engineering
OLS 1/7 Scale Model Studies
The 1/7 scale model AH-1 rotor experiment
was tested by Splettstoesser et al.

Two-bladed rectangular planform rotor

Linear twist is -8.2
0
.
Georgia Tech
School of Aerospace Engineering
OLS 1/7 Scale Model Studies (continued)
Validation case:
Advance ratio =0.164
Tip mach number M
tip
=0.664, C
T
/o=0.0783,
o
TPP
= -1
0

The first blade harmonics from Strawn et al:



Coarse H-O multi-block grid: 90x43x80 (NS
zone: 62x43x44)
u
0
u
1s
u
1c
|
1s
|
1c
|
0
6.14
0
-1.39
0
0.9
0
0.0
0
-1.0
0
0.5
0
Georgia Tech
School of Aerospace Engineering
C
P
at r=95.5%R (=0
0
and

=45
0
)
Chordwise Pressure for BVI Case (r/R=0.955, Psi=0)
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
Hybrid Method (lower)
Hybrid Method (upper)
Experiment
Chordwise Pressure for BVI Case (r/R=0.955, Psi=45)
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
Hybrid Method (lower)
Hybrid Method (upper)
Experiment
=0
0
=45
0
Georgia Tech
School of Aerospace Engineering
C
P
at r=95.5%R (=90
0
and

=135
0
)
Chordwise Pressure for BVI case (r/R=0.955, Psi=90)
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
Hybrid Method (lower)
Hybrid Method (upper)
Experiment
Chordwise Pressure for BVI Case (r/R=0.95, Psi=135)
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
Hybrid Method (lower)
Hybrid Method (upper)
Experiment
=90
0
=135
0
Georgia Tech
School of Aerospace Engineering
Conclusions
A combination of Navier-Stokes, potential flow and
free wake methods can be used to model rotors in
forward flight, with input from an elastic analysis.

Inclusion of torsional deformations was found to be
extremely important. The AH-1G study, where elastic
deformation was not available, gave less satisfactory
results.

At the advance ratios considered ( >0.16), free wake
and prescribed wake based inflow models gave
comparable results, even though the vortex geometry
was entirely different and BVI phenomena were
present.
Georgia Tech
School of Aerospace Engineering
Conclusions (continued)
Measured data regarding blade dynamics is often
inaccurate, or simply not available. A trim analysis
should always be done as part of any forward flight
analysis, based on user supplied C
T
/o, o
TPP
and rolling
moment information.

In this work, methods have been developed for
handling all of these important aspects of the analysis.
Georgia Tech
School of Aerospace Engineering
Recommendations
Transpiration BC is very approximate and can handle
only small deformations. Improved grid deformation
algorithms are needed.

The Baldwin Lomax model is inadequate for
modeling dynamic stall. Improved turbulence and
transition models must be developed and validated.

Biot-Savart law will occasionally produce velocity
spikes when a marker is very close to a computational
node. This can lead to unrealistically high velocity, and
low density values. Alternate approaches for
computing the rotational component of velocity must
be explored.
FIRST-PRINCIPLES BASED HIGH ORDER
METHODOLOGIES FOR ROTORCRAFT
FLOWFIELD STUDIES

Nathan Hariharan
CFD Research Corporation
Huntsville, AL

Lakshmi Sankar
Georgia Institute of Technology
Atlanta, GA

AHS 55th Annual Form & Technology Display
Montreal, Canada
May 27, 1999
CFD Research Corporation
215 Wynn Dr. , Huntsville, AL 35805 (256) 726-4800 FAX: (256) 726-4806 www.cfdrc.com
OUTLINE
Background

High Order Methods
- Fifth Order ENO
- Discontinuous Galerkin (DG) Scheme
- Seventh Order ENO

Overset Refinement

Results and Discussions

Conclusions and Recommendations

OUTLINE
Background

High Order Methods
- Fifth Order ENO
- Seventh Order ENO

Overset Refinement

Results and Discussions

Conclusions and Recommendations

ROTORCRAFT FLOWFIELD SIMULATION
First-Principles Based Methods
- High Order Methods
- Vorticity Confinement


Analytical Methods
- HELIX /Hybrid analysis
- Design Codes, such as CAMRAD


ROTORCRAFT FLOWFIELD SIMULATION
Direct Solution of Euler/NS Equations Including Vorticity


Vortex Formation, Convection, and Interaction


Necessity of First-Principles for Next Generation Rotorcraft
- Advanced Tip Shapes
- Active Devices to Enhance Aerodynamic Efficiency,
Decrease Acoustic Signal
- Blade-Vortex Interaction, Vortex Miss-Distance, etc.
First-Principles Based Methods
OBJECTIVES
Examine the ability of first-principles Euler/NS methodologies to capture
rotor-blade tip vortices.

Develop 3rd,5th spatial order, compact stencil, Discontinuous Galerkin
methodology.

Compare the vortex diffusion characteristics of DG with ENO for a 3D
rotor computation. Analyze the relative computational speed, memory
overheads.

Develop vortex tracking grids for blade tip vortices with unsteady the
vortex grids embedded inside the main grid.

Capture the tip-vortex for the first 180 degrees with less than 10-20%
diffusion, enabling truly first-principles based blade vortex interactional
studies.
OBJECTIVES
Examine the ability of first-principles Euler/NS methodologies to
capture rotor-blade tip vortices.

Develop vortex tracking grids for blade tip vortices with
unsteady the vortex grids embedded inside the main grid
(overset refinement).

Compare the vortex diffusion characteristics of DG with ENO for
a 3D rotor computation. Analyze the relative computational
speed, memory overheads.

Capture the tip-vortex for the first 180 degrees with less than 10-
20% diffusion, enabling truly first-principles based blade vortex
interactional studies.
ROTOR IN HOVER
Ceradonna-Tung Rotor Blade
ROTOR IN HOVER
Ceradonna-Tung Rotor Tip Vortex
Top View Side View
ROTOR IN HOVER
Tip Vortex Capture (Fifth Order)
ROTORCRAFT FLOWFIELD SIMULATION
Fifth Order ENO ( H-H grid, Caradonna-Tung Rotor)









Vorticity Iso-Surfaces Showing the Tip Vortex

Compact Scheme Issues
First-Principles Based Methods
Top View
Side View
Fifth Order Stencils for Computing Left and Right Primitive Variables
HIGH ORDER NAVIER-STOKES FORMULATION
NS Equations in Moving Finite Volume Framework




Roe Scheme for Inviscid Fluxes




Baseline Fifth Order ENO Scheme (Third Order Temporal Accuracy,
Newton Iterative Scheme)

( ) ( ) ( ) S F S F q V V q
t
6
1 i
v
6
1 i
G F J
A

= A

+ + O
c
c
= =

( ) ( )
( )
L R
R 1 L 1
q q A
2
q F q F
F
+
=
Fifth Order Stencils for
Computing Left and Right
Primitive Variables
HIGH ORDER NAVIER-STOKES
FORMULATION (DG)




Assume an Approximate Form of Local Solution, let



where a
i,j
are the moments
b
j
= {b
0
, b
1
, b
2
, } are some basis functions
n=3 is the third order accurate approximation
O
i
as in the below figure for 1D case




Using the Classical Galerkin Technique, One Minimizes the Error by,

0
x
u
t
u
=
c
c
+
c
c
( ) ( ) ( ) x b t a t , x V u
j
3 n
0 i
j , i i
i
=
=
=
O
x
i-1
x
i-1/2
x
i+1
x
i+1/ 2
O
i
x
i
x
ci
( ) 0 dx
x
u
t
u
x b
i
k
=
c
c
+
c
c
}
(

O
Stencil for DG Method
DISCONTINUOUS GALERKIN FORMULATION



Quadrature-free approach (Atkins and Shu)

3-D Euler Equation
0 F U
t
= V +

( )

} }
= + O V O
}
=
O O c

O
N
0 i
i
R
i
1
i k i i
1
i k i j
t
j , i k
i i i
0 s d J F J b d J F J b d J b v b

( ) I i 0 , N k 0 s < s s
( )
( ) , q c
c

, ,
z , y , x
J
i
DISCONTINUOUS GALERKIN FORMULATION
Three One-Dimensional Solutions in psi, eta, and tau
Directions













Third Order Explicit Runge-Kutta Time Stepping
0
F
t
v
=
c
c
+
c
c

0
F
t
v
=
q c
q c
+
c
c
0
F
t
v
=
, c
c
+
c
c
,
3 2
k
, , , 1 b =
3
3
2
2 1 0
a a a a v + + + =
Fourth Order (in 1-D Sense) Solution Requires Storage Of:
- a0-a9 - 10 for each independent variable
- 5 variable - 50 for each time-level stored
- 3 time level - 150 elements for each cell

Optimal Implementation Depends on Machine/Environment

Current Implementation
- for each grid
-- read data from disk
-- update a0-a9 for all time level
-- write data to disk

- next grid
-- (similar to OVERFLOW)
DISCONTINUOUS GALERKIN FORMULATION
Memory Management
UNSTEADY OVERSET FRAMEWORK
3D, Unsteady, Overset Solver
Instantaneous Cp Distribution on the Crownline of Airframe when Rotor is at an Azimuth of PSI = 156
-4
-3
-2
-1
0
1
2
3
Cp
4
0.1 0.325 0.55 0.775 1
x/R
PSI=156
Experiment
Euler_516
Euler_696
Experiment
Euler-First Rev
Euler- 1/2Rev
Later
+=156
-8
-6
-4
-2
0
2
4
6
8
Cp
0.1 0.325 0.55 0.775 1
x/R
PSI=6
Experimental
Euler_366
Experiment
Euler
+ =6
OVERSET REFINEMENT
Wing-Vortex System


Vortex Grid Adaptation



Additional Overset Grids






Combination of Both. Provide Enough Points by Oversetting.
Wing Components Across
Vortex Grid System
OVERSET REFINEMENT
Vortex Grid Adaptation
Schematic of Unsteady Vortex Grid System
Top View
Initial
Final
Side View
Tip Vortex
Movement of a Streamwise Plane
Wing-Vortex Grid (Vortex Grid 100*30*30)
OVERSET REFINEMENT
Vortex Grid Adaptation
Wing Vortex Grid System
Vorticity Iso-surface
OVERSET REFINEMENT (cont.)
Vortex Grid Adaptation
Top View of the Tip Vortex
Side View of the Tip Vortex
MECHANISM OF 3D VORTEX STRUCTURE
VORTEX CONVECTION -
DISCONTINUOUS GALERKIN
Vortex Grid 100*30*30, Mach No = 0.4
a. x/c=0.1 b. x/c=0.5 c. x/c=0.9
Retains Vortex Up to Four Chord Lengths for the Given Grid
VORTEX CONVECTION -
DISCONTINUOUS GALERKIN
Vortex Grid 100*30*30, Mach No = 0.4
a. x/c=1.0 b. x/c=2.0 c. x/c=3.0
Cannot Use DG Methods Like ENO Methods (for high order
projection only)


The Mass-Matrix Dependency of Solution on its High Order
Moments is a Central Characteristic of this Method
VORTEX CONVECTION -
DISCONTINUOUS GALERKIN
Vortex Grid 100*30*30, Mach No = 0.18,
Eighteen Chord Lengths
Streamwise Momentum Contours Across Spanwise Section of the Vortex Grid
Original and Adapted Vortex Grid
VORTEX CONVECTION - FIFTH ORDER
ENO
VORTEX CONVECTION - FIFTH ORDER
ENO (cont.)
Comparison of Axial (black) and Tangential (red) Momentum Variation Across the Vortex
SEVENTH ORDER ENO
Seven Point Stencil










One Sided Stencil Near Boundaries. Uniform High Order
Accuracy

Third Order Temporal Accuracy
Seventh Order Stencil for Smooth Flow Conditions
VORTEX CONVECTION - SEVENTH
ORDER ENO (cont.)
Stability
- Same as Fifth Order ENO - One Chord Length Every
(50/Mach No.) Iterations


CPU Requirements








27% More CPU for Seventh Order
SCHEME
CPU/TIME
ITERATION
Fifth ENO
Seventh ENO
13.5 seconds
17.2 seconds
VORTEX CONVECTION - SEVENTH
ORDER ENO (cont.)
Vortex Grid 100*30*30, Mach No = 0.18, Eighteen Chord
Lengths (~180 degrees of revolution for a rotor of AR=6)
U-Momentum Contours Across a Spanwise Section of the Vortex Grid
U, W Momentum Variation Across the Vortex at Various Streamwise Sections
VORTEX CONVECTION - SEVENTH
ORDER ENO (cont.)
Vortex Grid 100*30*30, Eighteen Chord Lengths,
Skewed Grid
M
x
=0.18
M
x
=0.04
U-Momentum Contours Across a Normal Plane of the Vortex Grid
U, W-Momentum Variation Across the Vortex at Various Streamwise Stations
VORTEX CONVECTION - SEVENTH
ORDER ENO (cont.)
Vortex Grid 100*30*30, Eighteen Chord Lengths,
Skewed Grid
U-Momentum Contours at Several Streamwise Stations
Grid-1
Grid-2
Hole Boundary for Grid-1
Interface_In
Interface_Out
Tip Vortex
Trajectory
VORTICITY TRANSFER ACROSS
OVERSET GRIDS
(ENO Schemes)
High Order Near Boundaries (one-sided stencils)

Reduced Order Near Hole Boundaries (...can be high order)
Schematic of the Vortex Trajectory through the Main and Vortex Grid
VORTICITY TRANSFER ACROSS
OVERSET GRIDS (cont.)
Interface_In
-1.5
-1
-0.5
0
0.5
1
Vz/Vinf
-1 -0.5 0 0.5 1 1.5
y/c
Grid-2
Grid-1
x/c ~ 0.1
0.8
1
1.2
1.4
1.6
1.8
Vx/Vinf
-1 -0.5 0 0.5 1 1.5
y/c
Grid-1
Grid-2
x/c ~ 0.1
VORTICITY TRANSFER ACROSS
OVERSET GRIDS (cont.)
Interface_Out
Grid-1
Grid-2
Hole Boundary for Grid-1
Interface_Out
Tip Vortex
Trajectory
x
y
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
Vz/Vinf
-1 -0.5 0 0.5 1 1.5
y/c
Grid-1
Grid-2
x/c ~ 4.0
0.95
1
1.05
1.1
1.15
1.2
1.25
1.3
Vx/Vinf
-1 -0.5 0 0.5 1 1.5
y/c
Grid-2
Grid_1
x/c ~4.0
VORTEX CONVECTION - SEVENTH
ORDER ENO
Vortex Grid 300*30*30, Fifth Chord Lengths
(~ 3-4 half-revolutions for a rotor of AR=6)
Axial-Momentum Iso-Surface Showing the Vortex. Graphs Show Axial (black) and
Tangential (red) Momentum Distribution across Vortex
x/c=2
x/c=25
x/c=50
SUMMARY
A fourth order spatial and third order temporal, 3-D,
Discontinuous Galerkin scheme was implemented in unsteady
overset settings and was proven feasible.

The efficiency of self-adaptive vortex-grid vortex grid
technology was proven using tip vortex generated over a wing.

Vortex capturing using a wing grid-vortex grid overset system
was undertaken using a baseline fifth order spatial/third order
temporal ENO scheme.

The behavior of three dimensional vortices was analyzed in
detail and the importance of capturing the axial momentum
variation of the tip vortex was elucidated. Vorticity transfer
characteristics between overset grids in a high order setting
were analyzed.
SUMMARY (cont.)
A seventh order ENO methodology was implemented in
extension to a baseline 5th order scheme.


An ambitious objective of identifying and proving a high order
overset method to capture blade tip vortices over 180 degrees
with less than 10% dissipation has been achieved and
bettered in a demonstration of capturing the wing tip vortex
over fifty chord lengths with negligible dissipation of the vortex.
CONCLUSIONS
3D Euler/NS Vortex Capturing has to Resolve the Axial
Component Across the Vortex. Otherwise the Captured Vortex
Resorts to a Wake-Like Mode Triggered by Numerical
Dissipation.


A Combination of High Order Method and Overset
Refinement/Adaptation is Highly Efficient and Ideal for
Rotorcraft Tip Vortex Resolution.


It is Possible to Capture Tip Vortices up to 3-4 Half Revolutions
(or even 5-6 revolutions) with Grid Sizes Small Enough to Run
with Workstations.
CONCLUSIONS (cont.)
First-principles Based Euler/NS Simulations with Overset
Refinement have Reached a Point Where They Can be Used
in Routine Rotorcraft Computation.


ENO-based Methods are Ideal for Rotorcraft Computations
such as Forward Flight BVI, Evaluation of Tip Shapes, Active
Devices etc. (relatively well known vortex structure).

L. N. Sankar
School of Aerospace Engineering
Georgia Institute of Technology

* This work was funded by the National Rotorcraft Technology
Center (NRTC) and RITA
Http://www.ae.gatech.edu/~lsankar
Georgia Tech
School of Aerospace Engineering
Rotorcraft Research
at
Georgia Tech
Overview of all the Rotorcraft Center Tasks
Rotorcraft CFD Research by the Present
Investigator and Coworkers being funded under
NRTC and RITA

Related research activities
Wind Turbines
Compressor Flow Control
Circulation Control
Georgia Tech
School of Aerospace Engineering
Outline
Project Title : Active Rotorcraft Blade Tips for Tip Vortex Core Modifications Project Number: GT 1.1
PI: N. Komerath nkomerath@ae.gatech.edu, S. Dancila sdancila@ae.gatech.edu, L. Sankar lsankar@ae.gatech.edu
Technical barriers/problems :
Rotor blade tips substantially affect blade
performance.
Passive and active modifications may be necessary to
improve the performance and noise characteristics
of rotors.
Objectives :
Systematically study through experiments a number
of passive and active tip shapes.
Perform computational studies to further understand
the flow physics near the blade tip, and how/if active
and passive control strategies may be beneficial.
Develop and demonstrate innovative active and
passive control methods for modifying flow field in the
vicinity of the rotor tip, and in the wake.
Key Milestones
milestones 01 02 03
01 Accomplishments :
Deatiked wake measurements were done for a
baseline rotor.
Algorithmic improvements were done to the
TURNS analysis, in collaboration with Task 1.2
Preliminary numerical results were obtained for
the wake structure behind a baseline rotor.

02 Plans :
PIV measurements will be done for new baseline
and rounded-tip rotors in forward flight
Detailed Analysis, modeling and implementation
of piezoelectric actuation for blowing modulation
on blade tips will be done.
CFD studies done in support of the experiment.
Detailed wake field measurements for an
advancing rotor.
Computational and experimental modeling
of several passive control concepts
Preliminary CFD studies of active control
concepts
Project Title : First-Principles based Modeling of Rotors in Hover, Forward Flight, and Maneuver Project Number: GT 1.2
PI / tel /e-mail L. Sankar, S. Ruffin, D. Peters lsankar@ae.gatech.edu, 404-894-3014; sruffin@ae.gatech.edu 404-894-8200
Technical barriers/problems :
First principles based models of rotors in hover
and forward flight suffer from numerical
errors such as dispersion, and diffusion.
Physics of the flow (e.g. tip vortex formation)
can not be adequately modeled until these errors are
minimized.
Strategies are needed for tightly coupling
these methodologies to trim and aeroelastic
models.

Objectives :
Develop spatially, and temporally accurate algorithms.
Develop embedded and adaptive grid
based methods for tracking vortices.
Validate methodology with data for rotors in hover
and forward flight.
Key Milestones
milestones 01 02 03
01 Accomplishments :
4
th
, 6
th
, and 8
th
order algorithms were evaluated for
rotors in hover.
Preliminary calculations were done for UH-60A
rotor.
Two papers were published; improved methods
were made available to industries.

02 Plans :
Complete hover studies; implement higher order
metrics and load integration.
Improve UH-60A results with measured blade
dynamics
Begin adaptive grid based vortex tracking.
Hover methodology development.
Forward flight method development,
and validation against UH-60A airloads.
Adaptive grid method development
0 0.02 0.04 0.06 0.08 0.1 0.12
CT/o
CQ/o
Experiment
TURNS-STVD6-WENO5
Objective(s):
2001 Accomplishments:


01 02 03 Tasks (CY)
LES code extension and
validation
Extend RANS codes to include
variety of turbulence models
Evaluate steady semi-infinite
wing
Evaluate unsteady semi-infinite
wing
Evaluate steady finite wing

Project Title: Simulations of Unsteady Flow-Rotor Interactions to Predict Dynamic Loading in a Turbulent Environment Proj No. GT 1.3
Extend and validate LES methods for unsteady flows of
interest.
Compare RANS, LES, and experimental data for steady and
dynamic stall situations
Use LES to determine how to modify RANS turbulence
models
LES code extended to include BC/IC for wing geometry
Ability of different grid-types: H,C,O and embedded
investigated for application to LES of pitching wings
RANS turbulence model study conducted, RANS codes
underway to update turbulence models, RANS steady
testing begun.
2002 Plans
Technical Barriers, Problems:
Modern rotorcraft rotor and airframe loading is not well-
predicted by RANS methods.
Existing turbulence models developed for steady flows
LES provides the opportunity to investigate turbulence
models in unsteady flows at the small-scale level where
experimental methods cannot provide data
Compare LES with RANS and experiments for steady semi-
infinite wing..
Begin evaluation for dynamic stall of semi-infinite wing..

PI M. Smith, S. Menon marilyn.smith@ae.gatech.edu, 404894-3065 suresh.menon@ae.gatech.edu, 404-894-9126
0
0.2
0.4
0.6
0.8
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Y
o
Velocity
Velocity Profiles at X=1.6, Z=0
Baldwin-Lomax
Degani-Schiff
Johnson-King
Spalart-Allmaras
Project Title : Efficient and Affordable Joining of Composites Project number: GT 2.1
PIs : Armanios, (404) 894 8202, erian.armanios@ae.gatech.edu, Dancila (404) 894 8197, stefan.dancila@ae.gatech.edu
and Makeev (Delta), 404) 714 3655, andrew.makeev@delta.com
Technical barriers/problems :
Premature failure of joints
- Boundary layer distribution of interlaminar stresses
- Presence of peel stress
Lack of understanding of interacting failure mechanisms
- retro-fit fixes
Lack of reliable tools for failure prediction

Objectives :
Development of affordable composite joint concept
Isolate mechanism driving premature failure
Development of associated predictive failure models
Stress-based
Strain energy release based

Key Milestones
milestones 01 02 03
01 Accomplishments :
Nested-overlap concept developed
-achieved compressive peel
Corrugated interface
- testing
Edited special volume on bonded joints
- FAA, ASTM
02 Plans :
Validation of nested-overlap concept
-Manufacturing and testing
Improved corrugated concept
Edit follow-up volume on bonded joints
Development and validation of analytical
Models for interlaminar stresses
Performance of parametric studies
Manufacturing and testing under monotonic
And fatigue loading
Identification of associated damage growth/
Crack resistance mechanisms
Nested overlap concept leading to compressive peel stress
Objective(s):
2001 Accomplishments:


01 02 03 Tasks (CY)
Unstructured & Cartesian
Fuselage Analysis
Incorporation of rotating force
field model
Replacement of body force with
individual blades
Couple Fuselage method
w/CSM
Evaluate other features (vortex
modeling, turbulence. modeling,
etc.)

Project Title: Phenomenological & First Principles Based Models for Complete Helicopters Project No. GT 3.1
Develop rapid first-principles based methods for modeling
complex fuselage shapes.
Develop hierarchy of methods for coupling rotor
aerodynamics to fuselage aerodynamics.
Use the computational tools to improve the aerodynamic
characteristics of rotorcraft.
Several unstructured grid methodologies, and an adaptive
Cartesian grid based method were evaluated for use in
rotor-airframe interactions.
A seventh order accurate method (developed under Task
1.2) was modified for modeling rotor-airframe interactions.
2002 Plans
Technical Barriers, Problems:
Modern rotorcraft have adverse rotor-airframe, and rotor-
empennage interactions, which are not clearly understood.
Existing methods based on panel methods and lifting line
theory can not model these interactions well.
In some instances, the unsteady airloads on the tail due to
the wake can cause tail fatigue, and/or loss of directional
control.
Apply the unstructured grid method, and the Cartesian grid
based method to ROBIN fuselage.
Couple the rotor solver to the fuselage analyses, obtain
preliminary results.

PI M. Smith, L. Sankar, S. Ruffin msmith@ae.gatech.edu, 404894-3065 lsankar@ae.gatech.edu, 404-894-3014; sruffin@ae.gatech.edu 404-894-8200
Project Title : Damage Tolerance Analysis of Stiffened Composites and Rotor Hubs Project Number: GT 5.2
Prof. E. Armanios (GT) / 404 894-8202 Dr. A. Makeev (Delta) / 404 714 3655 Prof. A. Badir (CAU) / 404 880 6900
Technical barriers/problems :
Absence of general damage tolerance analysis methodology
Absence of efficient and accurate models to predict interlaminar
stresses and energy release rates
Finite element based techniques for evaluation of energy release rate
components not convergent

Objectives :
Development of cost effective, reliable models for damage tolerance
analysis of rotorcraft composites


Key Milestones
milestones 01 02 03
01 Accomplishments :
Simple damage tolerance analysis methodology
identified
Simple boundary element models for composites
developed

02 Plans :
Incorporation of higher order elements
Large systems capability based on global/local
FEM/BEM
Selection of efficient and accurate fracture
mechanics analysis techniques for composites
Analysis for structures of simple
geometry
Extension of the analysis to nonlinear
tapered flexbeams
Development of a 3D model for
composites
-1
0
1
2
3
4
5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ABAQUS, 1098 variables
ABAQUS, 15972 variables
BEM, 144 variables
BEM, 896 variables
b
x
GPa ,
0
u
b t
y
Project Title : COMPOSITE BEAM CROSS-SECTIONAL OPTIMIZATION Project Number: GT 5.3
V. V. Volovoi & D. H. Hodges / 404-894-9811 & -8201 / Vitali.Volovoi@ae.gatech.edu & Dewey.Hodges@ae.gatech.edu
Technical barriers/problems :
Design space for composite rotor blades vast and mainly
unexplored
Traditional design based on evolutionary changes of existing
layouts; inherently leads to sub-optimal configurations
Industry is yet to exploit elastic couplings
Optimal design of rotor blades is a tightly coupled
interdisciplinary problem
3-D structural optimization of rotor blades is not feasible
Objectives :
Apply cross-sectional analysis based on a rigorous asymptotic
framework to design rotor blades with desired properties
Compliment evolutionary rotor blade design with
conceptually new designs
Develop numerical methods that produce models suitable
for practical rotor blade cross-sectional configurations
Key Milestones
01 Accomplishments :
Stand-alone parametric model of a rotor-blade has
been created
This model has been connected with VABS and and an
optimizer in an automatic fashion
Tools for convenient assessment of stress distribution
over the cross section were developed.
This optimization environment was tested on a simple
example
02 Plans :
Increase the fidelity of the model, consider several
other objective functions
Consider an example with discrete variables
Investigate the robustness of the solution

Baseline and
design variables
Optimized configuration
Stress field assessment
for various loads
Milestones
Tasks
01 02 03 04 05
Probabilistic optimization
Global optimization
Library of traditional
lay-outs
Parametric optimization:
sizing
Modeling Manufacturing
constraints
Proof of concept
Visualization/Stress recovery
Accumulating Database
UCAR Proto/ Demo Topological
Optimization
Industry survey
Accounting for
Stress distribution
Simple Example
Intermediate example
Robust Design
Aeromechanics example
Proof of concept
Formal Set of constraints
Usingas optimizationconstraints
Project Title : Wakes of Rotorcraft Maneuvering in Ground Effect Project Number: GIT 8.1
N.M. Komerath, GIT 404-894-3017 ; A.T. Conlisk, OSU 614-292-0808
Technical barriers/problems :
Multiple time scales of unsteadiness associated with
ground vortex phenomena.
Origin of the ground vortex.
Influence of flight condition on parameters such as
thrust and load.

Objectives :
Develop physically-based models for rotor wake
behavior in ground effect, with unsteadiness due to
maneuvers or gusts.
Understand time scales of unsteadiness, including
ground vortex, and vortex interactions.
Use findings to improve the aerodynamics in reduced-
order flight simulation models.

Key Milestones
milestones 01 02 03
01 Accomplishments :
Flow visualization experiments successful in
capturing steady (cleanly periodic) and unsteady
test cases
Computation modeled wake geometry in hover and
forward flight IGE.
Computation captures thrust variation due to
vortex interaction effects in inflow.

02 Plans :
Quantify flow field of rotor wake & ground vortex.
Quantify time scales of unsteadiness.
Fuselage loads experiment & computation
Experiments wake distortion
Experiments on time lag in inflows
and loads
Experiments: fuselage loads
Computation: loads and wake
deflection; reduced-order model
Project Title: Limit Detection and Limit Avoidance Methods for Carefree Maneuvering Project Number: GT 8.2
PIs: J.V.R. Prasad & A.J. Calise tel: (404) 894-3043, (404)894-7145
Technical Barriers/Problems:
Current limit prediction methods are based on the
availability of an accurate simulation model of the vehicle.
Dynamic trim based limit prediction is not
applicable to the transient limit parameter predictions.
It is not clear how to provide cues for multiple and
conflicting limit parameters in multiple control axes.
An effective limit cueing system will facilitate full
exploitation of flight envelope with reduced pilot
workload for highly reliable and safe operations of rotorcraft.
Objectives:
Develop adaptive algorithms for prediction of limit
parameters that reach limit boundaries during dynamic trim.
Develop algorithms (in collaboration with Prof. Horn of
Penn State) for prediction of limit parameters that reach
limit boundaries during transient part of the response.
Develop approaches to combine pilot cueing and limiting
using AFCS for envelope protection.
Carry out simulation and flight test evaluations in
collaboration with industry and government labs.
Investigate potential applications of the envelope limiting
algorithms to UAVs using the UAV test bed at Georgia Tech.
Key Mile Stones:
CY 01 Accomplishments:
Developed a neural net based adaptive limit
parameter prediction method
Carried out simulation evaluations of the adaptive
algorithms using the Generalized Tilt Rotor (GTR)
simulation model.
Developed a method for extraction of dynamic trim
maps directly from time response data.
Developed a method for limiting using the automatic
flight control system for UAV applications.
CY 02 Plans:
Adaptive algorithms using nonlinearly parameterized
neural networks
Combined pilot cueing and limiting through AFCS.
Flight test evaluations of envelope limiting using the
AFCS on our UAV helicopter test bed.



01 02
03 04 05 Milestones
Adaptive limit prediction algorithms
Simulation evaluation of adaptive algorithms
Transient limit detection algorithms
Combined pilot cueing and limiting using AFCS
Piloted simulation and flight test evaluations
Potential UAV applications

Adaptive
Limit
Detection
Control Inputs
and
Vehicle States
Estimation of Future
Limit Variables

Limit
Avoidance

Adaptive Limit Detection
and Avoidance
Project Title: Deformable Wake Dynamics for maneuvering Flight Simulation Project Number: GT 9.1
PIs: J.V.R. Prasad & D.A. Peters tel: (404) 894-3043, (314)935-4337
Technical Barriers/Problems:
Wake distortion effects are the primary source
of the off-axis response behavior observed in maneuvering flight.
Finite state inflow models offer a viable alternative in terms of
accuracy and computational expense.
Development of accurate models to capture the essential
physics of wake distortion effects on the flow behavior
at and off the rotor are important for development and
evaluation of model decoupling flight control laws and for
effective use of piloted simulation for various aircraft
subsystem development and pilot training.
Objectives:
Development of inflow models to capture wake bending,
skew and spacing dynamics during transient maneuvers.
Development of inflow models for inflow off the rotor.
Integration of refined inflow models into a comprehensive
flight simulation program and carry out simulation evaluations.
Correlations with available wind tunnel and flight test data.
Transition of inflow models to govt. labs and industry.
Key Mile Stones:
CY 01 Accomplishments:
Formulated a reduced-order wake distortion
model for transitional flight from hover
Extracted time constants of the model using
results from vortex tube theory
Carried out model validations through
comparison of simulation predictions using
GENHEL with the Black Hawk flight test data
CY 02 Plans:
Development of reduced order wake distortion
models for forward flight
Development of reduced order models for
inflow off the rotor
Model validations using flight test data
Hover
Pitch Up
Forward Flight
Climb
Rotor Dynamic Wake
Distortions during
Transitional Flight From Hover



01 02
03 04 05 Milestones
Wake dynamics modeling in hover
Wake dynamics modeling in forward flight
Wake bending and skew coupling
Modeling of inflow off the rotor
Correlations with test data and model refinements
Transitions to industry and govt. labs

Project Title : Neural Network Based Adaptive Flight Control Project Number: 9.3
PIs: Prof. A.J. Calise and Prof. J.V.R. Prasad Tel: (404)894-7145, 3043
Technical barriers/problems :
Gain scheduled control designs are awkward and difficult
to apply to high bandwidth UAV control design
Adaptive control using neural nets offers a viable
alternative and is adaptive to parameter uncertainty
High bandwidth adaptive control will have to also
address more difficult issues related to time delays,
unmodeled dynamics and actuator saturation

Objectives:
Extend our current research to the case of adaptive output
feedback control, permitting robustness to unmodeled
dynamics
Develop and approach to directly deal with control limits
and in an adaptive control setting
Validation in both simulation and flight experiments,
Collaborative efforts with both industry and government
labs
Pursue technology transition opportunities
Key Milestones
01 Accomplishments :
Development and improvement to an approach
to output feedback adaptive control
Development of an adaptive approach for
vibration suppression
Implementations on the R-50 helicopter and
laboratory demonstration in vibration
suppression. We are getting to higher
bandwidths.

02 Plans :
Continue to refine NN based adaptive output
feedback control
Continue to develop our approach to vibration
suppression and its application
Continue Flight testing and higher bandwidths.
This will include closing outer loops that
control velocity, flight direction and position
Pursue technology efforts with industry in UAV
high bandwidth rotary wing flight control and
active vibration control.
Continue to aggressively pursue rapid
developing technology transfer opportunities,
and leveraging with other NASA/Air Force
/DARPA programs
01 02 03 04 05
Milestones
Basic research in limited authority adaptive output
feedback
Research in active rotor control (Boeing Mesa)
Applications of high bandwidth adaptive flight control
(Bell and NASA Ames)
Experimental demonstrations using the R-50
Linear
controller
Model
Inversion
Helicopter
Adaptive
NN/FL
Project Title : Elastically Tailored Smart Composite Rotor Blades Project Number: GT 5.1
PI: E Armanios earmanios@ae.gatech.edu, S Dancila sdancila@ae.gatech.edu, O Bauchau obauchau@ae.gatech.edu
Technical barriers/problems :
Change of linear twist of 20+ degrees required between cruise
and hover to optimize performance in both regimes on typical
tiltrotor configuration (Nixon et al. )
Closed cell composite beams - insufficient level of coupling
while meeting torsional stiffness stability requirements
Open cell composite beams appropriate level of coupling but
insufficient torsional stiffness
Need for a structural concept that provides adequate level of
coupling without weight or stability penalties
Need for efficient piezoelectric actuator amplification
strategies/mechanisms
Objectives :
Improve the performance of rotor blades by combining elastic
tailoring and piezoelectric actuation
Develop configurations functional and efficient on tailored active
rotor blades at full scale
Key Milestones
milestones 01 02 03
01 Accomplishments :
Systematic investigation of tailored star cross section
beams
Fundamental understanding tailoring the entire blade
structure not effective
New approach:
Untailored blade spar - high torsional stiffness
Extension-twist coupled deformation of trailing edge
blade section (flap)
02 Plans :
Modeling and analysis of flap hinge tailored beams
Modeling and analysis of hinge tension-torsion beam
warping actuation using piezoelectric stack actuators
Systematic investigation of star cross
section beams
Modeling and analysis of tailored beams
for blade flap hinge
Modeling and analysis of hinge tension-
torsion beam warping actuation using
piezoelectric stack actuators
CFD Activities by the
Present Investigator and Coworkers
Spatially High Accuracy Algorithms for
improved tip vortex modeling and
performance predictions.
Efficient Airloads Prediction methods for
Rotors in Forward Flight .
Modeling of Complete Rotor-Airframe
Configurations.
Rectangular plan form, Aspect Ratio=6
NACA 0012 airfoil sections
Untwisted rotor
Tip Mach No. 0.388, corresponding to a rotor rpm of
1100.
This rotor has been extensively tested by McAlister.
Wake survey LDV data are available.
Surface pressure and thrust data for a similar
configuration tested by Caradonna et al are also
available.
C
on
fig
ur
ati
on
St
ud
ie
d
Need for High Accuracy Algorithms
Many industry standard codes (e.g. TURNS,
OVERFLOW) have low order spatial
accuracy which leads to excessive numerical
diffusion, and dispersion.
A very fine grid, and large CPU resources are
needed to reduce these errors.
High order algorithms are an effective way of
reducing these errors and achieving accurate
solutions on moderately fine girds.

Problems with Existing Methods
Numerical dissipation
Dissipation causes a gradual decrease in the
amplitude of an acoustic wave or the
magnitude of the tip vortex as it propagates
away from the blade surface.

The computed vortical wake, in particular,
diffuses very rapidly due to numerical
dissipation

Problems with Existing Methods
Numerical dispersion

Dispersion causes waves of different wavelengths
originating at the blade surface to incorrectly
propagate at different speeds.

Because of dispersion errors, the waves may distort
in nonphysical manner as they propagate away from
the blade surface.
Model Problem
Propagation of a 1-D wave
Consider the following
simple PDE:

0 =
c
c
+
c
c
x
u
t
u
Initial Condition at t=0:
16
2
) 0 , (
x
e t x u

= =
The exact solution:
( )
16
2
) , (
t x
e t x u

=
Initial Solution
0
0.2
0.4
0.6
0.8
1
1.2
-100 -75 -50 -25 0 25 50 75 100
u
How well do the 3
rd
order schemes in
OVERFLOW, TURNS, and CFL3D do?
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
0 25 50 75 100
u
Upwind, T=50
Exact
Dissipation
Dispersion
What happens at later time levels?
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
50 75 100 125 150
x
u
Upwind, T=100
Exact
A Possible Cure
Compute the spatial derivatives with a
sufficiently high order finite difference
approximation (Text book solution)
Further optimize the coefficients in the finite
difference form by minimizing the dissipation
and dispersion errors (low dispersion schemes
by Tam; Nance and Sankar).
Performance of a Spatially Fourth Order Algorithm for this problem
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
0 25 50 75 100
u
STVD-4, T=50
Exact
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
100 125 150 175 200
x
u
STVD-4, T=150
Exact
Application to a Fixed Wing
Surface Pressure Distribution
89% Span
1
0
-1
-2
-3
-4
-5
C
p
0 0.2 0.4 0.6 0.8 1
Experiment
Fifth Order
x/c
% Semi -span = 89
Third Order
Surface Pressure Distribution
97% Span
1
0
-1
-2
-3
-4
-5
C
p
0 0.2 0.4 0.6 0.8 1
Experiment
Fifth Order
x/c
% Semi -span= 97
Third Order
Tip Vortex Velocity Field
Downwash one chord length downstream
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
V
z
/
V
i
n
f
-0.8 -0.6 -0.4 -0.2 0 0.2
y/c
x/c = 1.0
0.4 0.6 0.8
Downstream of 5 chord lengths, the vortex quickly
Diffuses, even with the 5
th
order scheme.
3
rd
order
MUSCL
Axial Velocity in the
Core of the Tip Vortex
0.6
0.8
1
1.2
1.4
1.6
V
x

/

V
i
n
f
-0.8 -0.6 -0.4 -0.2 0 0.2
y/c
x/c = 0.5
0.4 0.6 0.8
Axial velocity field is well predicted by the 5
th
order
Scheme, up to 1 chord length in the wake.
OVERSET REFINEMENT
Wing-Vortex System


Vortex Grid Adaptation



Additional Overset Grids






Combination of Both. Provide Enough Points by Oversetting.
Wing Components Across
Vortex Grid System
OVERSET REFINEMENT
Vortex Grid Adaptation
Schematic of Unsteady Vortex Grid System
Top View
Initial
Final
Side View
Tip Vortex
Movement of a Streamwise Plane
Wing-Vortex Grid (Vortex Grid 100*30*30)
OVERSET REFINEMENT
Vortex Grid Adaptation
Wing Vortex Grid System
OVERSET REFINEMENT
Before
Vortex Grid Adaptation
Top View of the Tip Vortex
Side View of the Tip Vortex
After..
VORTEX CONVECTION - SEVENTH
ORDER ENO (cont.)
Vortex Grid 100*30*30, Eighteen Chord Lengths,
Skewed Grid
U-Momentum Contours at Several Streamwise Stations
RITA and NRTC Activities on High Order Algorithm
Spatially high order algorithms (4
th
, 5
th
, 6
th
, 7
th
, 8
th
)
have been systematically implemented in public
domain codes such as TURNS, OVERFLOW.
In most instances, the change from the user
perspective is a simple flag in the make file.
Computer time does increase (per point per time step)
by 10% to 20% with the higher order methods,
compared to 3
rd
order MUSCL schemes.
This increase is offset by the ability obtain accurate
results on relatively coarse grids.
Performance of the UH-60A Rotor
Grid Size 149x89x61(808K)
Error of 0.01-0.02 in FM; well within 100 lb. or 200 lb. error in
thrust;considered very good by industry.
Tip loads still not satisfactory due to highly stretched grids. Work
must be done to improve performance of high order algorithms on
highly stretched grids. This is critical to rotor tip design.

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 0.05 0.1 0.15
CT/o
FM
Experiment
TURNS-STVD6-WENO5
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
0.01
0 0.02 0.04 0.06 0.08 0.1 0.12
CT/o
CQ/o
Experiment
TURNS-STVD6-WENO5
Efficient Methods for High Speed Forward
Flight
Modeling high speed forward flight phenomena requires detailed
modeling aerodynamics (transonic flow, dynamic stall), elasticity,
blade dynamics and pilot input.

First-principles based aerodynamic analyses (Navier-Stokes,
Potential Flow) have been available to the industries for some
time, but are computationally expensive, and require several
days of turn-around time.

In some studies, an open-loop coupling between aerodynamics
and the other effects are done.
Trim, elasticity, blade dynamics, wake are handled by the
comprehensive analysis
Viscous flow, transonic effects handled by CFD.
This method integrates the most appropriate models in
different flow regions to retain solution quality.
A large reduction in computer time is reached.
Related Prior Work
Berezin coupled hybrid solver to RDYNE to account for the far wake
and trim effects.
Berkman (under Sikorsky support) modeled the entire wake from first
principles, and obtained good results in hover.
Moulton and Caradonna coupled HELIX to TURNS for modeling
rotors in hover.
Bangalore and Caradonna extended Moultons work through overset
grid for advancing rotor flows.
Hybrid Methods for Rotors in
Forward Flight
Navier-Stokes solver for
modeling the viscous flow
and near wake
Potential flow solver for
modeling the inviscid
isentropic flow
Lagrangean approach for
convecting vortex
filaments without
diffusion in the potential
flow zone and the far field
Georgia Tech
School of Aerospace Engineering
Hybrid Methodology
N-S zone
FPE zone
Lagrangean Wake
Georgia Tech
School of Aerospace Engineering
Implementation Details
CPU time was reduced by performing hybrid
analysis for a single blade.

The other blades are seen by the analysis as
a collection of bound and tip vortices.

There is no more need to match and patch the
grids around multiple moving, deforming
blades.
Georgia Tech
School of Aerospace Engineering
Implementation Details (Cont.)
This allows pitching and flapping motion to be
modeled rapidly without need for inter-blade
grid continuity.

The solutions are manually trimmed outside
the flow solver, once every revolution.

Elastic deformations are included, where
available.
Georgia Tech
School of Aerospace Engineering
Blade Dynamics
A module to cope with the rigid blade motions in flap
and pitch, and the complex blade deformation due to
aeroelastic effects has been developed.

For rigid blades, the (x,y,z) positions in space at any
instance in time may be transformed using Eulerian
angles:


If the blade is not rigid, the grid motion should
include additional rotations in twist, and bending
deformations.
| || || |
old old new
x C B A x T x

= =
Georgia Tech
School of Aerospace Engineering
Wake Model
Wake markers may lie inside the grid, or outside.

Rigid wake model is used by default.

Free wake model that can model the distortion from a
basic helical shape is also available.

We use Biot-Savart Law to evaluate the self induced
velocity.

We have also programmed Steinhoffs Clebsch
formulation
Georgia Tech
School of Aerospace Engineering
UH-60A in High Speed Forward Flight
Validation case:
Advance ratio =0.3
Tip mach number M
tip
=0.628
The blades were trimmed to eliminate one-
per-rev flapping.


H-O multi-block grid: 90x44x80 (NS zone:
62x26x44)
u sin 5 . 7 cos 84 . 1 5 . 11
0 0 0
+ =
Georgia Tech
School of Aerospace Engineering
C
P
at =0
0
(r=68%R and 94.5%R)
r=68%R. -2
-1.5
-1
-0.5
0
0.5
1
1.5
-0.1 0.1 0.3 0.5 0.7 0.9 1.1
Hyb. Method (Lower)
Hyb. Method (Upper)
Exp. (Upper)
Exp. (Lower)
r=94.5%R.
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
-0.1 0.1 0.3 0.5 0.7 0.9 1.1
Hyb. Method (Lower)
Hyb. Method (Upper)
Exp. (Upper)
Exp. (Lower)
C
P

C
P

x/c
x/c
r/R=94.5%
-1.5
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1
x/c
C
p
C
P
at =120
0
(r=94.5%R)
Georgia Tech
School of Aerospace Engineering
C
P
at =270
0
(r=68%R and 94.5%R)
r=68%R.
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
-0.1 0.1 0.3 0.5 0.7 0.9 1.1
Hyb. Method (Lower)
Hyb. Method (Upper)
Exp. (Upper)
Exp. (Lower)
r=94.5%R.
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
-0.1 0.1 0.3 0.5 0.7 0.9 1.1
Hyb. Method (Lower)
Hyb. Method (Upper)
Exp. (Upper)
Exp. (Lower)
x/c
x/c
C
P

C
P

Georgia Tech
School of Aerospace Engineering
Mach Number Contour at r=96%R
(Blade at +=90
0
)
Georgia Tech
School of Aerospace Engineering
Sectional Thrust Coefficient at r=78%R
r/R=78%
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
0 90 180 270 360
hybrid, rigid
hybrid, elastic
experiments
C
n


Georgia Tech
School of Aerospace Engineering
Sectional Thrust Coefficient at r=92%R
r/R=92%
-0.2
0
0.2
0.4
0.6
0.8
1
0 90 180 270 360
hybrid, rigid
hybrid, elastic
experiments
C
n


Modeling complete helicopter configurations
Motivation:
Helicopter configurations are often
complex in shape, full of corners and
edges.
Present overset methods require
considerable CPU time, and a very large
number of overset blocks.
Design studies require quick turn around
time, and the ability to model local changes
to the geometry.
Prior Work
OVERFLOW
Ganesh Rajagopalans work.
Steinhoffs Cartesian grid based approach
Other Cartesian grid based approaches (e.g.
SPLITFLOW)
Georgia Tech CHIMERA approach
Proposed Approach
Model the fuselage using an unstructured
grid approach.
We are starting with USM3D and/or FUN3D
Model the main and tail rotors using
structured grid methods.
We will be using GT codes for start.
Tightly couple these two approaches using
Georgia Tech version of the CHIMERA
scheme.
-1.5
-1
-0.5
0
0.5
1
1.5
2
C
p
0.1 0.95 1.8 2.65 3.5
Experiment
x/R
Euler
Mean surface pressure
distribution along the crown
line of the airframe
-2
0
2
4
6
8
10
12
0.2
C
p
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/R
PSI=174
Experiment
Euler_354
Experiment
Euler
Instantaneous surface
pressure distribution along
the crown line
Wake Modeling for Hover, Forward Flight, maneuvers
This task is a continuation of prior work by
Nathan Hariharan (ENO, adaptive CHIMERA)
Ebru Usta (STVDx schemes)
Zhong Yang (Hybrid Approach)
Basic algorithm developments done under
the Center funding will feed into applied
work with industry partners.
Active Control of Rotors
Joint task with Dr. Komerath and Dr. Dancila
We will be computationally investigating
various tangential and normal jet concepts.
Starting point is the circulation control airfoil
research being done at Georgia Tech under
NASA support.
Related Research
Wind Turbine Aerodynamics
Compressor Flow Control
Circulation Control Airfoil Research

-10
-5
0
5
10
15
20
0 5 10 15 20 25
Wind Speeds[m/s]
G
e
n
e
r
a
t
o
r

P
o
w
e
r
[
k
w
]
NREL experiment
N-S Solver
Hybrid Code
Lifting Line results

-10
-5
0
5
10
15
20
0 5 10 15 20 25
Wind Speeds[m/s]
G
e
n
e
r
a
t
o
r

P
o
w
e
r
[
k
w
]
NREL experiment
N-S Solver
Hybrid Code
AeroDyn results
Results for the Phase II Rotor
Georgia Tech School of Aerospace Engineering

0
5
10
15
20
0 5 10 15 20
Wind Speed[m/s]
G
e
n
e
r
a
t
o
r

P
o
w
e
r
[
k
w
]
NREL Test data
AeroDyn
Present Hybrid code
RESULTS for the Phase III Rotor
Georgia Tech School of Aerospace Engineering
Georgia Tech School of Aerospace Engineering
The NREL Blind Run Comparison
The Phase VI Rotor
Full Scale Wind Tunnel Tests at NASA Ames
Chordwise pressure tap at 0.3, 0.47, 0.63, 0.8, 0.95R -0.5
0.5
0.8C
0.03m
C
Measured Point
Georgia Tech School of Aerospace Engineering
Blind Run Comparison (I)
Upwind Condition, Zero Yaw
0
0.5
1
1.5
2
2.5
3
5 10 15 20 25 30
Wind Speed (m/s)
9
5
%

S
p
a
n

N
o
r
m
a
l

F
o
r
c
e

C
o
e
f
f
i
c
i
e
n
t
NREL
Present Simulations
The 95%R Normal Force Coefficients
Georgia Tech School of Aerospace Engineering
Blind Run Comparison (II)
Upwind Configuration, Zero Yaw
0
1000
2000
3000
4000
5000
5 10 15 20 25 30
Wind Speed (m/s)
R
o
o
t

F
l
a
p

B
e
n
d
i
n
g

M
o
m
e
n
t

(
N
m
)
NREL
Present Methodologies
Flap Bending Moment for One Blade
Georgia Tech School of Aerospace Engineering
Blind Run Comparison (II)
Upwind Configuration, Zero Yaw
0
1000
2000
3000
4000
5000
6000
7000
8000
5 10 15 20 25 30
Wind Speed (m/s)
R
o
o
t

F
l
a
p

B
e
n
d
i
n
g

M
o
m
e
n
t

(
N
m
)
UIUC/Enron-C
UIUC/Enron-UIUC
ROTABEM - DTU
Loughborough University
Global Energy Concepts, LLC
Windward (1)
Windward (2)
Windward (3)
ECN
NASA Ames
Teknikgruppen AB
RISOE -- HawC
Risoe NNS
DTU1
Georgia Tech
Glasgow University
TU Delft
NREL
Circulation Control Concept
Advanced CCW Airfoil: 0 - 90 degree small CCW
flap
Maintain the high lift when taking off and landing with
large flap angle and jet blowing
Reduce the drag when cruise with 0 flap angle and
non-blowing.
The CCW Airfoil Shape
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Jet Slot Location
30 degree integral flap
To Maintain the high-lift characteristics of CCW
Airfoil while greatly reduce the drag and noise
compared to large angle flap
The Variation of Lift Coefficient with the Angle of Attack
0
1
2
3
4
-2 0 2 4 6 8 10 12
Angle of Attack
C
l
Cmu=0.1657
Cmu=0.111
Cmu=0.0566
Cmu=0.0
Leading Edge Stall
STEADY JET RESULTS
Computed vs. Measured Variations of Lift Coefficient with
Momentum Coefficient
0
1
2
3
4
5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
C
C
l
Cl, Computed
Cl, Measured
Angle of Attack 0 degree, Integral Flap 30 degrees
The Stream Function Contours for the
No-Blowing Case
The Stream Function Contours for the
Blowing Case, C=0.1657
Concluding Remarks
A snapshot of some of the ongoing CFD research
at Georgia Tech has been presented.
Sikorsky and United Technologies have been
partners in many of our efforts.
This has lead to fruitful interactions with Dennis
Hiff, Brian Wake, Chip Berezin, Mike Torok, Bob
Moffitt, Ebru Usta (UTRC Intern), Nathan
Hariharan, and Alan Egolf.
We look forward to continued collaboration with
Sikorsky researchers on areas of importance to
Sikorsky/United Technologies, and to NRTC.

Potrebbero piacerti anche