Rotorcraft Aerodynamics Models for a Comprehensive Analysis

Wayne Johnson
Johnson Aeronautics
Palo Alto, California

Recent developments of the aerodynamics models for the comprehensive analysis CAMRAD II
are described, particularly the unsteady aerodynamic models and dynamic stall models, and the
free wake geometry calculation. Three models for the unsteady aerodynamic loads in attached
flow are implemented: from incompressible thin-airfoil theory, from ONERA EDLIN, and from
Leishman-Beddoes. Five dynamic stall models are implemented: from Johnson, Boeing,
Leishman-Beddoes, ONERA EDLIN, and ONERA BH. A key feature of the implementation of
these models is revisions allowing the retention of airfoil tables for static loads in all cases.
Results are presented for a two-dimensional airfoil, a three-dimensional wing, and rotors.
Extensions of the CAMRAD II free wake method to include hover and ground effect are
described, including hover performance correlation.

αd delayed angle of attack

Notation . αz zero lift angle of attack
A rotor disk area, πR2 Γ bound circulation
cdz,cmz section drag and moment coefficients at zero lift µ advance ratio, (flight speed)/ΩR
cl,cd,cm section lift, drag, and moment coefficients ρ air density
cn section normal force coefficient σ rotor solidity, (blade area)/A
CL wind axis rotor lift force, L /ρA(ΩR)2 τ wake age
CP rotor power, P /ρA(ΩR)3 τK extent of initial convection (wake age)
CT rotor thrust, T /ρA(ΩR)2 Ω rotor rotational speed
CX wind axis rotor drag force, X /ρA(ΩR)2
D wake geometry distortion
f trailing-edge separation point (fraction chord Introduction
from leading edge)
CAMRAD II is an aeromechanical analysis of
k reduced frequency, frequency × semichord/speed helicopters and rotorcraft that incorporates a combination
M Mach number of advanced technology, including multibody dynamics,
M at advancing tip Mach number, (1+µ)Mtip nonlinear finite elements, and rotorcraft aerodynamics. For
M tip tip Mach number, ΩR/(speed of sound) the design, testing, and evaluation of rotors and rotorcraft;
r blade radial station at all stages, including research, conceptual design, detailed
design, and development; it calculates performance, loads,
rBetz initial radial station of tip vortex, Betz rollup
vibration, response, and stability; with a consistent,
rTV initial radial station of tip vortex balanced, yet high level of technology in a single computer
R blade radius program; applicable to a wide range of problems, and a
t time wide class of rotorcraft. Such capability is essential for
v conv wake convection velocity helicopter problems, which are inherently complex and
α angle of attack multidisciplinary.
____________ A comprehensive helicopter analysis must calculate
.Revised performance, loads, vibration, response, and stability. The
version of paper presented at the American
multidisciplinary nature of helicopter problems means that
Helicopter Society 54th Annual Forum, Washington, DC,
similar models are required for all of these jobs. It follows
May 20–22, 1998. Copyright © 1999 by Wayne Johnson.
that a comprehensive analysis must have a rotor wake
All rights reserved.

1
model; account for drag and stall of the rotor blades; in a wind tunnel; and an N-bladed rotor, with an articulated,
include nonlinear dynamics of the rotor and airframe; and hingeless, teetering, gimballed, or bearingless hub; perhaps
model the entire aircraft. The analysis must perform trim, with a swashplate. The aerodynamic model includes a wake
transient, and flutter tasks. The trim task finds the analysis to calculate the rotor nonuniform induced-
equilibrium solution (constant or periodic) for a steady velocities, using rigid, prescribed or free wake geometry.
state operating condition. The operating condition can be CAMRAD II is described in references 1 and 2.
free flight (including level flight, steady climb or descent, Flexibility and generality of the system configuration
and steady turns), or constrained (such as a rotor in a wind are obtained by assembling standard components with
tunnel, with typically the thrust and flapping trimmed to standard interfaces, and solving the system using standard
target values). It is usually necessary to identify the control procedures. Components perform most computations
positions and aircraft orientation required to achieve the associated with the physics of the model of the system. So
specified operating condition. The transient task components are the focus of modelling issues, including
numerically integrates the equations in time (from the trim the empiricism and approximations needed for a practical
solution), for a prescribed excitation. The flutter task model of many real systems. Development of an improved
obtains and analyzes differential equations for the system, model requires the development of a new component,
linearized about trim (probably by numerical perturbation). which will fit into the existing analysis framework.
A modern comprehensive analysis must be able to This paper describes recent developments of the
analyze arbitrary configurations — whatever the designers aerodynamics models for the comprehensive analysis
can invent. The system configuration must be defined and CAMRAD II. The focus is on features of the wing
changed by input to the analysis; it should not be component: unsteady aerodynamic models (thin-airfoil
necessary to change the code as long as the required physics theory for loads at low angle of attack), and dynamic stall
are available. The definition of the solution procedure must models; and the wake geometry components. Several
be just as flexible as the definition of the configuration. models are implemented for the unsteady aerodynamic loads
The solution procedure must be defined and changed by in attached flow: from incompressible thin-airfoil theory,
input to the analysis; it should not be necessary to change from ONERA EDLIN, and from Leishman-Beddoes.
the code as long as the required methods are available. Several dynamic stall models are implemented: from
CAMRAD II uses a building-block approach to achieve Johnson, Boeing, Leishman-Beddoes, ONERA EDLIN, and
flexibility in the model of the dynamic and aerodynamic ONERA BH. Extensions of the CAMRAD II free wake
configuration, and in the solution procedure. The method to include hover and ground effect are described.
mathematical model of the kinematics, dynamics, and
response allows nonlinearities (structural, aerodynamic, and Wing Component
kinematics); and arbitrary large motion, including rigid
body motions and large rotations of components relative to The CAMRAD II wing component is based on lifting-
each other. Hence CAMRAD II can model the true line theory, using steady two-dimensional airfoil
geometry of a rotorcraft, including multiple load paths characteristics and a vortex wake. Lifting-line theory
(such as a swashplate and control system, lag dampers, assumes that the wing has a high aspect ratio, or more
tension/torsion straps, and bearingless rotors); vibration generally that spanwise variations of the aerodynamic
control devices (such as pendulum absorbers or active environment are small. This assumption allows the
control); arbitrary elastic axis and arbitrary hinge order; problem to be split into separate wing and wake models,
drooped and swept tips; and dissimilar blades. The which are solved individually and combined. Viscous and
building-block approach, separating the specification of the compressibility effects are included by using experimental
configuration, the aeromechanical model, and the solution data for the two-dimensional airfoil characteristics that are
procedure, is essential for expandability of the analysis. the foundation for the wing solution. Corrections for
Otherwise the smallest change involves the entire analysis, yawed and swept flow are introduced, and an estimate of the
and growth becomes increasingly harder as each new feature spanwise drag. For low angles of attack, thin-airfoil-theory
is added. The building-block approach also leads naturally results are used to calculate the unsteady loading. For high
to more general and more rigorous models. For ease of use, angles of attack, an empirical dynamic stall model can be
a shell is provided to build typical rotorcraft and rotor used.
models, while the core input capability always gives
complete flexibility to define and revise the model. The Unsteady Airfoil Motion
system pieces (building blocks) constitute the core Unsteady lift and moment in attached flow are calculated
analysis. The rotorcraft shell constructs the core input for based on thin-airfoil theory. Models from incompressible
an aircraft with one or two or more rotors; in free flight or thin-airfoil theory, ONERA EDLIN, and Leishman-

2
Beddoes are implemented. The equations for these models by exponential functions of time. In this form, the
are given in Appendix A. Often only the unsteady, equations for the loads can be transformed from indicial
noncirculatory terms are required, since steady loads are response to Laplace domain, and thence to state equations
obtained from the airfoil tables and lift deficiency function (ordinary differential equations in time). The impulsive
effects are usually accounted for in the wake-induced indicial response is derived using piston theory, which is
velocity. The following modifications are introduced for all valid for nonzero Mach number and small enough time.
three models: (a) The steady loads are excluded. (b) The Thus while giving nonsingular results at zero Mach
moment about the quarter chord (theoretical aerodynamic number, this theory does not include the incompressible
center) is used for the moment about the actual limit exactly. The amplitude of the circulatory indicial
aerodynamic center. (c) The loads are corrected to get the response is obtained from the quasistatic incompressible
real lift-curve slope. (d) The expressions are extended to response. An additional modification is made here: for an
reverse flow. airfoil with a trailing-edge flap, the expressions are
The incompressible unsteady loads are derived following extended to include the effects of aerodynamic balance, and
reference 3. For an airfoil with a trailing-edge flap, the to include the flap hinge moment produced by pitch and
unsteady loads are derived following references 3 to 5. heave. The resulting unsteady loads are obtained from first-
Optionally the flap can be aerodynamically balanced, with order differential equations for both the impulsive and the
an open or sealed gap. circulatory terms (see Appendix A). This theory includes
the effects of the airfoil shed wake, but not entirely in the
The ONERA EDLIN (Equations Differentielles "circulatory" terms. Care must be taken with a vortex wake
Lineaires) theory for the unsteady loads is presented in or dynamic inflow model that the shed wake effects are
reference 6. The "extended model" of Petot includes the neither omitted nor duplicated.
effects of heave as well as pitch, and time-varying free
stream. The "pitch model" (see also refs. 7 to 9) is not Dynamic Stall
appropriate since at high frequency it either gives a lift
deficiency function approaching zero instead of one-half, or Dynamic stall is characterized by a delay in the
it neglects a lift-from-pitch term (depending on the occurrence of separated flow produced by the wing motion,
interpretation of the terms in the equation). In the absence and high transient loads induced by a vortex shed from the
of stall, Petot found that thin-airfoil-theory results leading edge when stall does occur. Dynamic stall models
compared well with measured behavior. To include the from Johnson, Boeing, Leishman-Beddoes, and ONERA
effects of compressibility, Küssner's coefficients are used, EDLIN, and ONERA BH are included. The equations for
as tabulated by van der Vooren (ref. 10) and curve-fit by these models are given in Appendix B. As implemented,
Petot (ref. 6). An additional modification is made here: a the dynamic stall models still use the airfoil table for
constant is changed so the incompressible circulatory loads steady characteristics, evaluated at an angle of attack that
match the lift deficiency function value of C = .5 at high includes the dynamic stall delay. Retaining the use of
frequency. Then compared with the incompressible airfoil table data is considered essential, both to provide the
unsteady loads result, the ONERA EDLIN model (see basic characteristics associated with airfoil shape, and to
Appendix A) introduces factors that are functions of Mach isolate the effects of the dynamic stall model. In addition to
number; and adds a first-order differential equation for a lift the delayed angle of attack, coefficient increments are
increment, to account for the airfoil shed wake effects (lift defined by the dynamic stall model. All the models are
deficiency function). These factors give a good extended to reverse flow. Let α d be the delayed angle of
representation of Küssner's coefficients, except that the attack, calculated from the angle of attack α . Then the
moment produced by heave is always real, when it should corrected coefficients are:
exhibit a phase shift for nonzero Mach number. For an α−αz
cl = c (α ) + ∆cl DS + cl US
airfoil with a trailing-edge flap, the effects of α d−α z l 2D d
compressibility are approximated by using these same
factors in the incompressible results. No shed wake terms cd = (ααd−−ααzz)2 (cd 2D(αd) − cdz) + cdz + ∆cdDS + cdUS
are included in the unsteady loads for flap lift or flap
moment. α−αz
cm = (c (α ) − cmz) + cmz + ∆cmDS + cmUS
α d−α z m 2D d
The Leishman-Beddoes theory for unsteady loads in
attached flow is presented in references 11 to 13. The where αz is the zero-lift angle of attack, and cdz and cmz
theory is based on the indicial response of a thin airfoil in are the corresponding drag and moment. The form of the
compressible flow to heave, pitch, and flap motions. The lift and moment corrections ensures that the coefficients
indicial response is a combination of impulsive (small below stall are unchanged. The ∆ c DS 's are increments
time) and circulatory (long time) terms, each approximated defined by the dynamic stall model, generally attributed to

3
the leading-edge vortex. Note that for use in the wing cd 2D = cdz + (1 − η KD(f)) (α−αz)2
component of CAMRAD II, corrections for yawed flow are
required as well. The unsteady loads in attached flow are The Kirchhoff expression KN = (1/4)(1 +  √ f )2 is used;
included in the above expressions (cUS). Since these loads several functional forms of KM and KD are found in the
are calculated based on linear (thin-airfoil) theory, they can literature. In unsteady flow, a delayed separation point fd is
be evaluated separately and added to obtain the total loads. calculated from f, and then the loads evaluated using fd
Optionally the unsteady loads can be set to zero for stalled (Leishman and Beddoes use the notation f'″ instead of fd);
flow; however, most of the dynamic stall models have and terms for attached flow and the leading edge vortex
been developed assuming that the attached flow terms are added. Since fd corresponds to a delayed angle of attack, it
active at high angle of attack as well. is possible to replace the analytical functions K with the
loads from airfoil tables:
The Johnson dynamic stall model (adapted from refs. 14
to 17) uses an angle of attack delay proportional to α·, plus cn = cnα(M) KN(fd) (α−αz) + ∆cnDS + cnUS
impulsive lift and moment increments from the leading- α−αz
edge vortex. Thus the model has first-order differential = c (M) KN(fd) (αd−αz) + ∆cnDS + cnUS
α d − α z nα
equations for the delayed angle of attack, and algorithms to
evaluate the load increments produced by the leading-edge α−αz
= c (α ) + ∆cnDS + cnUS
vortex. α d−α z n 2D d
The Boeing dynamic stall model (developed in refs. 18
to 20) uses an angle of attack delay proportional to the cm = cmz + cn KM(fd) + ∆cmDS + cmUS
square-root of α· , which produces the basic hysteresis
α−αz
effects. The coefficient increments produced by the leading- = cmz + (c (α ) − cmz) + ∆cmDS + cmUS
α d−α z m 2D d
edge vortex are not used in this model. The delayed angle
of attack is calculated directly from the current α and α·
values; or using a first-order differential equation. cd= cdz + (1 − η KD(fd)) (α−αz)2 + ∆cdDS + cdUS
The Leishman-Beddoes dynamic stall model (refs. 21 to
23) uses a delayed angle of attack, plus lift and moment = cdz + (ααd−−ααzz)2 (cd 2D(αd) − cdz) + ∆cdDS + cdUS
increments from the leading-edge vortex. This model
characterizes the airfoil static stall behavior by the trailing-
edge separation point f (fraction of chord from leading No change to the model is implied for lift and moment,
edge), and a critical lift coefficient cl CR at the separation since these analytical functions are intended to be
onset boundary (leading-edge separation at low Mach equivalent to the airfoil table data. A change is implied for
number, shock reversal at high Mach number). The airfoil the drag equation only if KD depends on other variables
data for lift are used to identify constants s1 , s2 , and α s besides f. There will be some changes in the resulting
that generate f(α) as follows: loads however, when the K functions do not give a good
representation of the airfoil table data. At this point the lift
(|α − α z|− α s)/s1
f = 1. − .3 e for |α − α z| ≤ α s rather than the normal force can be used. While the
(α s−|α − α z|)/s2 function f(α ) should be identified from the normal force
f = .04 + .66 e for |α − α z| > α s data, there is usually little difference if it is identified from
the lift data instead. Similarly, cnDS can be used for cl DS;
Then |α − α z | = α s or f = .7 is taken as the definition of and cdDS derived from cnDS. Further modifications of the
stall. The parameters cl CR, s1, s2, and αs are required as a model are required because the above expression for f(α)
function of Mach number, for positive and negative angle does not distinguish between positive and negative angle of
of attack, normal and reverse flow, at each span station. attack. In order to handle oscillations through α z , a
The Leishman-Beddoes model for unsteady flow is based on continuous monotonic function of α is needed. The
fd = f(αd) at the delayed angle of attack. function f is also modified in order to handle large angle of
Here the Leishman-Beddoes model is modified to use attack. This modification to the definition of f does not
the static loads directly from the airfoil tables, instead of affect the model for attached flow or around stall, but with
fitting the static loads to analytical functions. Leishman it the delayed angle of attack behaves reasonably at very
and Beddoes (ref. 21) write the static normal force, large angles.
moment, and drag as functions of f: In the Leishman-Beddoes dynamic stall model, the
cn 2D = cnα(M) KN(f) (α−αz) delayed angle of attack α d is calculated including: static
hysteresis around stall; a lag in the leading-edge pressure
cm 2D = cmz + cn KM(f) relative cl; and an additional lag in the boundary layer

4
response. There are separate α d equations for lift and incompressible thin airfoil
moment, to allow different behavior during reattachment. ONERA EDLIN
Vortex lift accumulation begins at the onset of stall, driven 1.2 Leishman-Beddoes
by the difference between the linear and nonlinear lifts, cv. static
1.0
The vortex loads ∆ cDS are obtained from cv with a time

lift coefficient
lag. Thus the model has first-order differential equations for 0.8
the delayed angle of attack and the leading-edge vortex lift.
0.6
The ONERA EDLIN (Equations Differentielles
Lineaires) dynamic stall model (ref. 6) uses a stall delay, 0.4

plus lift, drag, and moment increments calculated from 0.2

second-order differential equations. The "extended model" of
Petot includes the effects of heave as well as pitch, and 0.0
0. 2. 4. 6. 8. 10.
time-varying free stream. Generalizations based on Petot's angle of attack (deg)
"pitch model" (see ref. 7) lead to more complicated

moment coefficient
equations. The load is divided into two parts. The first part
0.03
is the load in the absence of stall, which here gives the
unsteady load for attached flow. The second part of the load 0.00
is driven by the difference between the linear load
extrapolated to the unstalled domain, and the real nonlinear -0.03
0. 2. 4. 6. 8. 10.
static load. Tests show that dynamic stall occurs at a angle of attack (deg)
higher angle of attack than does static stall. The absence of Figure 1. Two-dimensional airfoil lift and moment, at k =
stall is preserved in the model by forcing the difference .188 and M = .5; oscillating in pitch, with circulatory
between the linear and nonlinear loads to be zero for a time terms.
τd after exceeding the static stall angle. Here the model is
modified in several ways. The static and unsteady terms are
separated from the dynamic stall effects. A pitch rate term
in the lift, that reference 6 associates with attached flow incompressible thin airfoil
unsteady loads, is here contained in the dynamic stall load. ONERA EDLIN
The loads are written in terms of α· . Reference 6 uses the 1.2 Leishman-Beddoes
· in order to include the effects of static
upwash rate of change w
1.0
time-varying free stream, but for a three-dimensional wing
α· also includes the wake-induced velocity. For all the
lift coefficient

0.8
dynamic stall models, the option is available to evaluate
α· from w· . 0.6

The ONERA BH (Bifurcation de Hopf) dynamic stall 0.4

model (ref. 24) uses a delayed angle of attack, plus lift and 0.2
moment increments calculated from first-order and second-
order differential equations. The Hopf bifurcation model 0.0
0. 2. 4. 6. 8. 10.
replaces the time-invariant equilibrium state of flow by a angle of attack (deg)
time-varying equilibrium state, as the angle of attack
moment coefficient

exceeds a critical value. The load is divided into two parts, 0.03
a "steady" part (static plus attached flow unsteady) and an
"unsteady" part (dynamic stall). The ONERA EDLIN 0.00
theory can be used for the unsteady load in attached flow.
For time-varying airfoil motion, the loads are evaluated at
-0.03
a delayed angle of attack that is calculated as in the 0. 2. 4. 6. 8. 10.
Leishman-Beddoes dynamic stall model. There are also angle of attack (deg)
dynamic stall load increments, driven by the pitch rate and Figure 2. Two-dimensional airfoil lift and moment, at k =
pitch acceleration. Here the model is modified by .188 and M = .5; oscillating in pitch, without circulatory
separating the static and unsteady terms from the dynamic terms.
stall effects.

5
 incompressible thin airfoil
ONERA EDLIN
Solving the State Equations
1.2 Leishman-Beddoes Some of the above models introduce ordinary
static differential equations for aerodynamic state variables. These
1.0
differential equations can be solved in CAMRAD II along
lift coefficient

0.8 with the structural dynamic equations. There are a large

number of states however, and the equations can be
0.6
nonlinear as well as time-varying. So an implicit solution,
0.4 implemented within the wing component, is more useful
for the trim and transient tasks. The equations must be
0.2 formulated as differential equations in order to be linearized
0.0
in the flutter task, although a linearized solution of the
0. 2. 4. 6. 8. 10. highly nonlinear aerodynamics involved in dynamic stall is
angle of attack (deg) not entirelyc onsistent. Two implicit methods are
implemented in CAMRAD II to solve the differential
moment coefficient

0.03
equations: a finite-difference solution, based on trapezoidal
integration; and a sampled-data solution, based on the
0.00
convolution integral or Duhamel's integral (ref. 21).

-0.03
0. 2. 4. 6. 8. 10.
Unsteady Load Examples
angle of attack (deg) Figures 1 to 4 compare the lift and moment calculated
Figure 3. Two-dimensional airfoil lift and moment, at k = by the three unsteady aerodynamic models, for a two-
.188 and M = .5; oscillating in heave, with circulatory dimensional NACA 0012 airfoil oscillating in pitch or
terms. heave. The ONERA EDLIN and Leishman-Beddoes models
give somewhat different lift deficiency functions at this
condition (figures 1 and 3; the incompressible model does
not include the lift deficiency function). The "circulatory"
incompressible thin airfoil terms of the Leishman-Beddoes model are not just shed
ONERA EDLIN wake effects, so omitting these terms gives different results
1.2 Leishman-Beddoes than for the ONERA EDLIN model (figures 2 and 4).
static
1.0 Figures 5 and 6 compare the aerodynamic models for a
three-dimensional semispan wing in a wind tunnel,
lift coefficient

0.8
oscillating in pitch about the quarter chord. The measured
0.6 data are from reference 25. Two cases are considered. The
case with mean angle of attack of 11 deg is used to
0.4
compare the models for unsteady loads in attached flow.
0.2 The case with mean angle of attack of 15 deg is used to
compare the models for dynamic stall. The analysis
0.0
0. 2. 4. 6. 8. 10.
considered a full span wing of aspect ratio 10, with 25
angle of attack (deg) spanwise collocation points. The circulatory terms were
included in the unsteady aerodynamic models, so the wake
moment coefficient

0.03 model included only trailed vortex elements, omitting shed

vortex elements. A transient calculation was performed,
0.00 using 25 to 40 time steps per cycle; the loads converged
after three cycles. The wing had an NACA 0015 airfoil
-0.03
section. Static airfoil data were obtained from a two-
0. 2. 4. 6. 8. 10. dimensional test of this wing. Span stations are measured
angle of attack (deg) from zero at the full-span wing centerline, to one at the
Figure 4. Two-dimensional airfoil lift and moment, at k = wing tip.
.188 and M = .5; oscillating in heave, without circulatory
Figure 5 shows the lift and moment at 25% span
terms.
station for the mean α = 11 deg case, comparing the three
unsteady aerodynamics models. Although the oscillation is
at low angle of attack for this case, the dynamic stall

6
 measured measured
1.5
incomp. thin airfoil (static stall) Johnson
1.4
ONERA EDLIN (static stall) Boeing
Leishman-Beddoes (static stall) 1.3

lift coefficient
1.2
lift coefficient

1.1

1.0
0.9

0.8
0.7
10. 12. 14. 16. 18. 20.

0.6 pitch angle (deg)

6. 8. 10. 12. 14. 16.
0.15
pitch angle (deg)

moment coefficient
moment coefficient

0.06

0.04 0.00
0.02

0.00
-0.15
-0.02 10. 12. 14. 16. 18. 20.
6. 8. 10. 12. 14. 16.
pitch angle (deg) pitch angle (deg)

Figure 5. Three-dimensional wing oscillating in pitch, at k measured

= .096 and M = .287; lift and moment at 25% span station Leishman-Beddoes
1.5
(static stall). ONERA EDLIN
ONERA BH
1.3
lift coefficient

models would improve the correlation above α = 14 deg.

The difference between measured and calculated mean 1.1
moment is found along the entire span, so it should not be
an effect of three-dimensional aerodynamics.
0.9
Figure 6 shows the lift and moment at 47.5% span
station for the mean α = 15 deg case, comparing the five
0.7
dynamic stall models. The reduced frequency of the 10. 12. 14. 16. 18. 20.
oscillation is typical of the once-per-revolution angle of pitch angle (deg)
attack variation of a helicopter blade, but the 4 deg 0.15
moment coefficient

amplitude is relatively small, resulting in a low pitch rate

(maximum α·c/V = .0053). The differences between the
dynamic stall calculations are no more than should be 0.00
expected from empirical models. The correlation may be
judged remarkably good considering the simplicity of some
of the dynamic stall models; and also exhibiting significant -0.15
errors, as expected with empirical models. 10. 12. 14. 16. 18. 20.
pitch angle (deg)
A Puma helicopter with experimental swept-tip rotor
Figure 6. Three-dimensional wing oscillating in pitch, at k
blades was flight tested by the Royal Aeronautical
= .038 and M = .287; lift and moment at 47.5% span
Establishment at Bedford (ref. 26). The flight test case at station, with dynamic stall model.
high speed has a large value of cyclic pitch, hence a large
contribution from the unsteady aerodynamic loads. The
calculations were performed for an elastic rotor, trimming with rigid wake geometry. Figures 7 and 8 show the
thrust and flapping with collective and cyclic pitch. The normal force and moment for an advance ratio of µ = .38,
thrust, flapping, and shaft angle were obtained from the advancing tip Mach number Mat = .86, and thrust CT/σ =
flight test data. Since the loading at high speed is negative .080; at 92% radius (on the swept tip). The flight test
on the advancing tip, the dual-peak wake model was used, measurements are compared with results using the three

7
 measured, r/R = .92 measured, r/R = .92
ONERA EDLIN ONERA EDLIN
incompressible incompressible
0.3 Leishman-Beddoes Leishman-Beddoes
without circulatory terms 0.015 total, without circulatory terms
section lift, M2 cn

0.2

M2 c m
0.000

0.1
-0.015
0. 90. 180. 270. 360.
0.0 azimuth (deg)

0.015
unsteady, without circulatory terms
-0.1
0. 90. 180. 270. 360.

M2 c m
azimuth (deg) 0.000
0.3
with circulatory terms -0.015
0. 90. 180. 270. 360.
section lift, M2 cn

0.2
azimuth (deg)

0.015
total, with circulatory terms
0.1

M2 c m 0.000
0.0

-0.015
-0.1 0. 90. 180. 270. 360.
0. 90. 180. 270. 360.
azimuth (deg)
azimuth (deg)
Figure 7. Flight test of Puma with swept tip rotor blade, 0.015
unsteady, with circulatory terms
at µ = .38, Mat = .86, and CT /σ = .080; total section lift
and unsteady lift.
M2 c m

0.000

-0.015
models for unsteady aerodynamics. Shown are both the 0. 90. 180. 270. 360.
total load, and just the unsteady term. The correlation azimuth (deg)
between measurement and calculation is good. The Figure 8. Flight test of Puma with swept tip rotor blade,
circulatory terms of the ONERA EDLIN and Leishman- at µ = .38, Mat = .86, and CT / σ = .080; total section
Beddoes models are best left out, with the shed wake effects moment and unsteady moment.
accounted for in the vortex wake model. Without these
circulatory terms, the three unsteady aerodynamics models
give comparable results. primarily of the noncirculatory, unsteady term (figure 8),
which does not vary much with tip span station.
The correlation between measured and calculated normal
force is similar at 95% and 97.8% radius to that at 92% The McDonnell Douglas Advanced Bearingless Rotor
radius. The correlation for pitching moment is fair at 92% (MDART) was tested in the NASA Ames Research Center
radius (figure 8), but worse outboard. The lifting-line wing 40- by 80-Foot Wind Tunnel (refs. 27 and 28). Nearly
model of CAMRAD II is second-order for lift, which identical results for calculated lag damping and blade loads
contributes to the good correlation for lift; but it is still were obtained using the three unsteady aerodynamics
first-order for moments. The measured moments show a models. The unsteady loads are essential for the analysis,
variation with radial station at the tip that is probably since they provide the blade pitch damping.
associated with three-dimensional aerodynamics. The angle A Lynx hingeless-rotor helicopter was flight tested by
of attack varies radially at the tip, but is small enough so Westland Helicopters up to and beyond the stall boundary
at these Mach numbers (effectively reduced by the tip (ref. 29). The flight test included speed sweeps at low and
sweep) the static pitch moment is zero (the airfoil is high gross weight, extending into regimes with significant
symmetric). Thus the calculated pitching moment consists dynamic stall. Figure 9 compares the measured power with

8
 measured (mean shaft torque) important in rotor performance, yet their calculation in
measured (corrected engine power) practical analyses must rely on empirical models.
static stall
dynamic stall (5 models) Free Wake Geometry
0.010
The rotor vortex wake is an important factor in most
0.008
problems of helicopters, including poor performance, high
rotor power, CP/ σ

blade loads, high vibration, and high noise levels. An

0.006 accurate calculation of the wake-induced nonuniform inflow
and the resulting blade airloads is needed in order to predict
0.004 rotor behavior. Below an advance ratio of about µ = 0.20,
blade-vortex interaction is particularly strong and therefore
0.002 an accurate wake geometry is needed as well. The general
free wake geometry calculation for wings and rotors that is
0.000 implemented in CAMRAD II was described in reference
0.0 0.1 0.2 0.3 0.4
advance ratio, µ 31. The method gave good performance and airloads
correlation at advance ratios of 0.05 and above, with
Figure 9. Lynx flight test, at weight CT /σ = .07 and .11; reasonable computation speed. The wake geometry
rotor power (no blade lag or torsion motion). distortion can be calculated for multiple wings, multiple
rotors, and non-identical blades; for all wake structures,
dynamic stall (5 models) including multiple rolled-up trailed vorticity and inboard
static stall sheets as well as tip vortices; using the same wake model
no yawed flow correction as the induced velocity calculations; for transients as well
0.10 and no radial drag as the trim solution.
rotor equivalent drag, CD/ σ

Recent developments have extended this free wake

0.08 method to include hover and ground effect. Application to
hover required improvements in the wake geometry
0.06
extrapolation method, and a model for the tip vortex
formation process (initial radial contraction and initial
0.04
vertical convection). Both for convergence in hover and for
efficiency, the capability to require the distortion to be
0.02
identical for different wings (perhaps with a time shift) is
needed. The influence of the ground was introduced by
0.00
0.09 0.10 0.11 0.12 0.13 using image vortex elements in the wake geometry and
rotor lift, CL/σ induced velocity calculations. These developments are
Figure 10. Lynx rotor, at µ = .25 and shaft angle of attack documented here.
−4 deg; influence of aerodynamic model on calculated rotor The distortion is evaluated at time t and wake age τ, and
drag (no blade lag or torsion motion).
may be required at an age beyond which it has been
calculated. The time and age increment in the wake
calculations using the five dynamic stall models. This geometry calculation is ∆t. Let τlast be the maximum age
figure is just a reality check rather than correlation, since of the available distortion. The distortion is extrapolated by
the calculations were performed without elastic lag or assuming that the vortex element is convected for the time
torsion motion of the blades (to isolate the effects of the interval (τ−τlast) by a constant velocity:
different aerodynamic models), and because the static airfoil D(t,τ) = D(t−(τ− τlast),τlast) + (τ− τlast) vconv
data were not measured much beyond the onset of stall.
Figure 10 shows the influence of the aerodynamic model The distortion is used at a constant value of (t−τ), which is
on the calculated rotor drag as a function of rotor lift. the time when the wake element was created. The free
Calculations are presented using the five dynamic stall distortion is calculated for wake ages up to τ max . The
models, as well as with and without the yawed flow accuracy of the wake geometry is generally improved by
corrections. This is not a new result (see ref. 30), except not using the last part of the calculated distortion, hence
for the presence of five dynamic stall models instead of just starting the extrapolation at τ last less than τ m a x .
one. But figure 10 captures the state-of-the-art well: the Typically the last revolution of the calculated distortion is
effects of yawed flow and dynamic stall are extremely not used in the extrapolation, particularly for hover and

9
low advance ratio. Extrapolation uses the convection rB ∂ Γ rB
rC (ΓA−ΓB) = Γ Γ
velocity vconv , obtained from the average distortion ∫ ∂r r dr = B r B − A r A − ∫ Γ dr
increment at τlast: rA rA
1 1 = ΓBrB − ΓArA − (rA−rB)ΓM
vconv = ∑ ( D(t,τ last) − D(t− ∆ t,τ last− ∆ t))
T ∆t
t in terms of the mean bound circulation ΓM; hence
(in inertial axes, for each distorted structure). This
convection velocity is recalculated at the end of each rC = (1−w) rA + w rB
revolution of the wake geometry algorithm, and the final with w = (ΓM−ΓB)/(ΓA−ΓB). The spanwise displacement
value is saved for use when the wake geometry is calculated by the Betz rollup model is multiplied by an
evaluated. Optionally, the convection velocity from the input factor; and for the wing tips, added to the input
rigid or prescribed model can be used to extrapolate the free displacement. At the wing tips, it is assumed that the
distortion. For the transient task, the velocity is filtered vorticity that rolls up is between the nearest bound
over the latest period; or the trim convection velocity can circulation peak and the tip. At inboard stations, it is
be used. assumed that the vorticity that rolls up is between the
A subset of the wings can have identical trim distortion nearest peak in the panel to the left and the nearest peak in
(in an appropriate frame, with a time shift). Then it is only the panel to the right. With the single-peak wake model,
necessary to calculate the distortion for one wing of the this peak is at the maximum bound circulation; with dual
subset (the parent wing). The distortion of the child wing peaks in the circulation distribution, the right or left peak
is evaluated from the parent: is used. The mean bound circulation Γ M is evaluated
assuming piecewise-linear variation of the bound
D child(tchild,τ) = Dparent(tparent,τ)
circulation over the span.
where tparent = tchild + l shift ∆ t, and l shift is an input With a three-dimensional wing, the Kutta condition
time shift. For example, with N blades uniformly spaced requires that the wake leave the trailing-edge tangent to the
over the period T = J ∆t, the time shift for the m-th blade wing surface (ref. 32). In the absence of a calculation of the
is lshift = (m/N)J (which is an integer if J is a multiple of detailed flow field near the wing, this requirement can be
N). If lshift is not an integer, then the parent distortion is satisfied by using an initial convection velocity qK = Γ/πc,
interpolated over t and (t−τ) to obtain the child distortion. where Γ is the section bound circulation and c is the chord.
The tip vortex forms on the generating wing at a span This result is obtained using the zero-lift chord line for the
station inboard of the wing tip. The location of the tip trailing-edge bisector, and a lift-curve slope of 2π . The
vortex at the trailing edge can influence the loading and velocity direction is assumed to be perpendicular to the
near wake. This effect is implemented in the wing plane defined by the bound vortex and the trailed vortex
component, by assuming that the bound circulation is zero element to the collocation point. The initial velocity qK is
outboard of the location of the tip vortex at the trailing used for collocation points at wake age τ = 0, and the
edge. For a highly tapered wing tip, the tip vortex can actual wake self-induced velocity q at age τ ≥ τK, so
form 4–6% span inboard of the tip, and the effect on the (1 − τ/τK) qK + (τ/τK) q
loading is significant. For a rectangular planform, the
effect is small and can be ignored. Further entrainment replaces q for τ < τK. The age τK must be selected based
occurs in the wake, so the tip vortex effectively forms at on correlation with measured wake geometry and
an initial span station rTV inboard of the tip. The wake performance. Optionally this initial convection velocity
geometry components implement the effect of this initial can be ignored (qK never replacing q). Hover performance
contraction by assuming that the tip vortices and sheet calculations require τ K > 0, but in forward flight best
edges are described by trailed lines arising from span results are obtained with a much smaller value of τK. Thus
stations r T V rather than from the tips. In the wake for hover, the initial convection velocities are defined such
geometry components, rTV is an input parameter, or it is that the wake leaves the wing tangent to the wing surface.
calculated assuming that the centroid of the trailed vorticity The EHPIC analysis (refs. 33 and 34) has a similar feature,
is conserved (Betz rollup). implemented differently: the first element of the trailing
The initial span station of the tip vortex can be vortices is constrained to leave the rotor blade parallel to
obtained from Betz rollup as follows. Consider the bound the lifting surface.
vorticity from rA to rB , rolling up into a trailed line. The Generally it is only the distorted geometry of the tip
centroid of the trailed vorticity is at rC: vortices that is calculated. Prescribed geometry can be used
for the inboard vorticity sheets.

10
 Table 1. Tip vortex formation in hover.
rotor UH-60A tapered rect. ogee S-76 XV-15
number of blades 4 4 2 2 4 3
radius R (ft) 4.683 4.683 3.43 3.43 3.5 12.5
solidity σ .0825 .0825 .0464 .0396 .0748 .089
chord c/R or ctip/R .065 .022 .073 .073 .059 .093
2.5(c/R or ctip/R) .16 .05 .18 .18 .15 .23
τK .15 .05 .15 .15 .15 .23
prescribed rTV .99 .97 .96 .91 .99 .98
Betz rollup rTV rBetz rBetz−.01 rBetz−.02 rBetz−.02 rBetz .4rBetz

wake geometry at first encounter with following blade (vertical convection z/R and radial contraction r/R)
CT/σ .0856 .1069 .0773 .1014 .099 .103 .078 .1018
measured z/R .0285±.005 .0405±.005 .013±.005 .022±.005 .065 .069 .045±.015 .060±.015
prescribed .030 .046 .009 .019 .066 .070 .027 .067
Betz rollup .029 .047 .012 .020 .065 .070 .029 .068
measured r/R .925±.007 .921±.007 .907±.007 .907±.007 .865 .820 .915±.015 .890±.015
prescribed .946 .940 .915 .908 .867 .824 .947 .918
Betz rollup .948 .938 .912 .905 .870 .821 .943 .916

The influence of the ground can be included in the performance can be achieved only on the basis of
wake-induced velocity calculation and in the wake correlation with measured wake geometry, airloads, and
geometry calculation, through the use of image elements performance. Progress in the analysis of hovering rotors is
in the wake model. For inviscid flow, the boundary here measured by a narrowing focus of empiricism: from a
condition imposed by the ground is zero normal wake- factor on the uniform induced-velocity for momentum
induced velocity at the ground plane. This constraint is theory; to parameters defining the wake geometry for
satisfied by introducing an image element for every vortex prescribed methods; to the initial convection and initial
element in the wake model. The image element is created contraction, wake extent, and core size for free wake
from the vortex element by changing the sign of the methods based on inviscid aerodynamics.
strength, and reflecting the position across the ground
plane. A wake element can be too near or below the ground Hover Correlation
plane. This can occur with rigid or prescribed geometry Figure 11 shows an example of the calculated free wake
that does not include the effect of the ground; or because of geometry in hover. Figure 12 shows the calculated wake
numerical effects in the free geometry calculation. Here geometry in ground effect; the advance ratio is µ = .02, and
"near" the ground plane is defined by a distance δ above the height of the hub above the ground is 50% the rotor
ground level ( δ could be zero or even negative). diameter.
Optionally, the analysis neglects any part of a vortex
element (and the corresponding part of the image) that is The tip vortex formation process (initial radial
below the height δ . It is found that the solution for the contraction and initial vertical convection) was calibrated
distortion places the wake elements completely above the using measured wake geometry at the first encounter with a
ground plane (as long as the elements are not too long). following blade, and measured blade airloads. Table 1
Thus this feature improves convergence of the distortion summarizes the results. The required extent of the initial
calculation, without actually affecting the position of the convection is approximately τ K = 2.5c/R revolutions,
distorted elements above the ground plane. which is typically about 0.15 revolutions. With taper over
a large span range, the tip chord should be used to estimate
With this inviscid and steady free wake geometry, it is τK. The initial span station rTV depends on blade number
an idealized problem that is solved, not the actual viscous (table 1 shows rTV = .99R, .98R, .96R for 4, 3, 2 blades
and unsteady flow field of a hovering rotor. So the respectively) and on planform taper. Betz rollup typically
objective of a consistent and reliable calculation of hover

11
 τK = .05, rTV = .97
τK = .05, rTV = rBetz-.01
τK = .05, rTV = 1.
measured, tapered tip; CT/σ = .077
0.28 measured, tapered tip; CT/σ = .101

0.24

section lift, d(CT/σ)/dr

0.20

0.16

Figure 11. Calculated hover wake geometry (S-76 rotor at 0.12

CT/σ = .08). 0.08

0.04

0.00
0.2 0.4 0.6 0.8 1.0
radial station, r/R
0.28
Re correction for lift
0.24 τK = .15, rTV = .97

section lift, d(CT/σ)/dr

0.20

0.16
Figure 12. Calculated free wake geometry in ground effect
0.12
(S-76 rotor at CT/σ =.08, µ = .02, height above ground =
50% rotor diameter). 0.08

0.04

0.00
0.2 0.4 0.6 0.8 1.0
τK = .15, rTV = .99 radial station, r/R
τK = .15, rTV = rBetz
Figure 14. Blade section lift in hover for model tapered
τK = .15, rTV = 1.
rotor (CT/σ = .077 and .101, Mtip = .63).
measured, swept tip; CT/σ = .086
0.28 measured, swept tip; CT/σ = .107
Figures 13 to 15 compare the calculated loading and
0.24
section lift, d(CT/σ)/dr

wake geometry with measurements on a 1/5.73-scale

0.20 model UH-60A rotor and a corresponding tapered-tip rotor
0.16 (refs. 35 to 37). The UH-60A model had a 20-deg swept tip
beginning at .93R, and approximately the full-scale UH-60
0.12
twist distribution. The tapered model had 1:3 taper
0.08 beginning at .82R, no sweep, and linear twist. The cases
0.04
shown are those presented in reference 37. A Reynolds
number correction has been applied to the drag coefficient
0.00
0.2 0.4 0.6 0.8 1.0 (based on a 1/5-th power scaling of the drag), but not to
radial station, r/R the maximum lift. Figure 13 shows the loading for the
Figure 13. Blade section lift in hover for model UH-60A UH-60A. A .01R inboard shift of the initial span station
rotor (CT/σ = .086 and .107, Mtip = .63). of the tip vortex gives a small improvement of the peak
loading calculation. Figure 14 shows the loading for the
tapered tip. A .03R inboard shift of the initial span station
of the tip vortex gives a significant improvement of the
requires an additive or multiplicative correction, but has the peak loading calculation; and the extent of the initial
advantage that it can be easily applied to multiple rollup convection scales with the tip chord in this case. Applying
cases. These results for rTV give good correlation for the the Reynolds number correction to the maximum lift does
tip loading, although in some cases the measured wake not give the correct peak loading. Figure 15 compares the
geometry at the first encounter with a following blade is calculated wake geometry with vertical and radial positions
perhaps .02R further inboard than the calculated position. measured at the first encounter with a following blade.

12
 swept tip, CT/σ = .086 τK = .15, rTV = .96
swept tip, CT/σ = .107 τK = .15, rTV = rBetz-.02
tapered tip, CT/σ = .077 0.024 τK = .15, rTV = 1.
tapered tip, CT/σ = .101 measured

bound circulation, Γ/ΩR2

0.00
measured
0.018

0.05

0.012
0.10
z/R

0.15 rectangular tip

0.006
0.20

0.25 0.000
0. 90. 180. 270. 360. 0.2 0.4 0.6 0.8 1.0

wake age (deg) radial station, r/R

1.00 τK = .15, rTV = .91
τK = .15, rTV = rBetz-.02
0.95
0.024 τK = .15, rTV = .95
measured

bound circulation, Γ/ΩR2

0.90
r/R

0.85 0.018

0.80
0.012
0.75
0. 90. 180. 270. 360.
ogee tip
wake age (deg) 0.006

Figure 15. Hover wake geometry for model UH-60A and

tapered rotors (vertical convection z/R and radial 0.000
contraction r/R). 0.2 0.4 0.6 0.8 1.0
radial station, r/R
Figure 16. Blade bound circulation in hover for model
Figures 16 and 17 compare the calculated circulation rotors with rectangular and ogee tip planforms (CT /σ =
and wake geometry with measurements on a model two- .099 and .103, Mtip = .22).
bladed rotor, for rectangular and ogee tip planforms (ref.
38). A .04R inboard shift of the initial span station of the
tip vortex gives an improvement of the peak circulation motion. Figure 21 shows the performance of the AH-64
calculation. Note that for the ogee tip, the blade lift and and UH-60 rotors. The aircraft performance was measured
circulation are suppressed outboard of .95R; the in flight by the USAAEFA (refs. 43 and 44); the data were
measurements showed near the blade trailing edge the tip corrected to just main rotor performance by Baserga and
vortex forming at about .94R. Figure 17 compares the Scully of the USAATCOM (ref. 45). The calculated
calculated wake geometry with the vertical and radial performance is improved for the AH-64 using an elastic
positions measured at the first encounter with a following instead of a rigid blade model. Figure 22 shows the
blade. Figure 17 includes results for the S-76 and XV-15 performance of the XV-15 rotor. The measured data are
rotors as well; the measured data are from references 39 and from reference 46.
40 respectively. The measured hover performance and wake geometry of
Figures 18 to 22 compare measured hover performance the XV-15 rotor with a subwing are presented in references
with free wake geometry calculations for several rotors. 47 and 40. The subwings had approximately 6 inch span
Power is presented instead of figure of merit for the usual and 3 inch chord; and were tested at high and low incidence
reason: the correlation appears much better. Figure 18 relative the tip. Figures 23 and 24 show the calculated
shows the performance of the 1/5.73-scale model UH-60A wake geometry with the subwing; and figure 25 compares
rotor and the tapered-tip rotor; the measured performance the measured and calculated performance, with and without
data are from references 36 and 37. Figures 19 and 20 show the subwing. All the normalized results in figures 24 and
the performance of the S-76 and Bo-105 rotors. The 25 are based on the radius and solidity of the blade without
measured data are from references 41 and 42 respectively. the subwing. The geometry of the primary tip vortex and
The calculations do not show a large effect of blade elastic the influence of the subwings on the rotor performance are

13
 rectangular, CT/σ = .099 whirl tower test
ogee, CT/σ = .101 rigid blade, rTV = .99
S-76, C T/σ = .078 0.015 elastic blade, rTV = .99
XV-15, CT/σ = .102 rigid blade, rTV = rBetz
measured 0.012

rotor power, CP/ σ

0.00

0.05 0.009

0.10
0.006
z/R

0.15
0.003
0.20
0.000
0.25 0.04 0.06 0.08 0.10 0.12 0.14
0. 90. 180. 270. 360.
rotor thrust, CT/σ
wake age (deg)
Figure 19. Hover performance of S-76 rotor (Mtip = .605).
1.00

0.95

measured (α s = -10)
0.90
measured (α s = 0)
r/R

0.85 0.015 calculated, elastic blade

calculated, rigid blade
0.80 0.012
rotor power, CP/ σ
0.75
0. 90. 180. 270. 360. 0.009
wake age (deg)
Figure 17. Hover wake geometry for model rotors with 0.006
rectangular and ogee tip planforms, model S-76 rotor, and
full scale XV-15 rotor (vertical convection z/R and radial 0.003
contraction r/R).
0.000
0.04 0.06 0.08 0.10 0.12 0.14
rotor thrust, CT/σ
measured, swept tip
0.012 measured, tapered tip Figure 20. Hover performance of Bo-105 rotor (Mtip =
free wake geometry .65).
rotor power, CP/ σ

0.009

AH-64 calculated, rigid blade

0.006 AH-64 calculated, elastic blade
AH-64 measured
0.003
UH-60 calculated, rigid blade
UH-60 calculated, elastic blade
0.012 UH-60 measured
0.000
0.03 0.05 0.07 0.09 0.11
rotor power, CP/ σ

rotor thrust, CT/σ 0.009

Figure 18. Hover performance of model UH-60A and

tapered rotors (Mtip = .63). 0.006

0.003
calculated reasonably well. In this thrust range, the
performance calculations without the subwing compare
well with the OARF test data (ref. 46, see figure 22), 0.000
0.06 0.07 0.08 0.09 0.10 0.11 0.12
while the wind tunnel measurements (ref. 47) show rotor thrust, CT/σ
somewhat higher power. So it is the change in the power
produced by the subwings that should be compared in Figure 21. Hover performance of AH-64 and UH-60 rotors
(measured performance derived from flight tests).
figure 25. Figure 24 shows the influence of rollup model,

14
 rigid blade, rTV = .98 OARF test, without subwing
elastic blade, rTV = .98 without subwing
0.020 rigid blade, rTV = .4rBetz low-incidence subwing
0.013
measured high-incidence subwing
0.016
measured
rotor power, CP/ σ

0.012

rotor power, CP/ σ

0.012
0.011

0.008
0.010

0.004
0.009

0.000
0.03 0.05 0.07 0.09 0.11 0.13 0.15 0.17 0.008
0.10 0.11 0.12 0.13
rotor thrust, CT/σ
rotor thrust, CT/σ
Figure 22. Hover performance of XV-15 rotor (Mtip =
without subwing
.69). 0.013 low-incidence subwing
high-incidence subwing
0.012

rotor power, CP/ σ

calculated
0.011

0.010

0.009

0.008
0.10 0.11 0.12 0.13
rotor thrust, CT/σ
Figure 25. Hover performance of XV-15 rotor with and
without subwing (Mtip = .69).

specifically the initial span station. The calculations

Figure 23. Calculated hover wake geometry for XV-15
rotor with subwing (CT/σ =.102, Mtip = .69). produce 90 deg rotation of the subwing vortex about the
primary vortex at wake ages between 35 and 90 deg,
depending on the rollup model; the measurements show 90
deg rotation at about 45 deg wake age (ref. 40). Thereafter
the calculated subwing vortex does not rotate about the
primary vortex as fast as measured: 180 deg rotation is
1.2
found at 140 to 210 deg wake age, instead of at 75 deg
1.0 wake age as measured. This difference reflects the absence
0.8 of a model of the entrainment and combination of the
subwing and primary vortex into a single vortex, as
z/R or r/R

0.6 primary rTV = .98

rTV = .4rBetz observed in the measurements.
0.4
primary rTV = 1.
0.2 Concluding Remarks

0.0 Recent developments of the aerodynamics models for

the comprehensive analysis CAMRAD II have been
-0.2
0. 60. 120. 180. 240. 300. 360. described, specifically the unsteady aerodynamic models and
wake age (deg) dynamic stall models. Three models for the unsteady
Figure 24. Influence of rollup model on calculated hover aerodynamic loads in attached flow are implemented: from
wake geometry of XV-15 rotor with subwing, at CT /σ incompressible thin-airfoil theory, from ONERA EDLIN,
=.102 (vertical convection z/R and radial contraction r/R). and from Leishman-Beddoes. Five dynamic stall models are

15
implemented: from Johnson, Boeing, Leishman-Beddoes, Recent developments have extended the CAMRAD II
ONERA EDLIN, and ONERA BH. free wake method to include hover and ground effect.
For attached flow unsteady aerodynamics, the ONERA Application to hover required improvements in the wake
EDLIN and Leishman-Beddoes models are approximations geometry extrapolation method; a model for the tip vortex
to the results of compressible thin-airfoil theory. They formation process (initial radial contraction and initial
give comparable results for calculations of rotor behavior. vertical convection); and the capability to require the
Care must be taken with a vortex wake or dynamic inflow distortion to be identical for different wings. A model for
model that the shed wake effects are neither omitted nor the tip vortex formation is required in the absence of
duplicated. Correlation with the Puma measured airloads detailed calculations of the flow field at the wing tip and
shows that the circulatory terms of these theories are best near wake. Such detailed calculations are much needed.
left out, with the shed wake effects accounted for in the With this inviscid and steady free wake geometry, it is an
vortex wake model. Without the circulatory term, the idealized problem that is solved, not the actual viscous and
ONERA EDLIN model is simplest, just introducing unsteady flow field of a hovering rotor. Correlation with
factors that are functions of Mach number, without any measured wake geometry, airloads, and performance has
state equations. These factors give a good representation of demonstrated the capability to calculate hover performance.
Küssner's coefficients, except that the moment produced by
heave is always real, when it should exhibit a phase shift References
for nonzero Mach number. The Leishman-Beddoes method 1) Johnson, W. "CAMRAD II, Comprehensive Analytical
seems to offer a sounder basis for approximating the loads, Model of Rotorcraft Aerodynamics and Dynamics." Johnson
Aeronautics, Palo Alto, California, 1992-1997.
and it may be possible to use a quasistatic solution for the
2) Johnson, W. "Technology Drivers in the Development of
impulsive terms. While giving nonsingular results at zero CAMRAD II." American Helicopter Society, Aeromechanics
Mach number, this theory does not however include the Specialists Meeting, San Francisco, January 1994.
incompressible limit exactly. 3) Johnson, W. Helicopter Theory. Princeton University
The dynamic stall models have been revised as required Press, Princeton, New Jersey, 1980.
to use the airfoil tables for steady characteristics (evaluated 4) Küssner, H.G., and Schwarz, L. "The Oscillating Wing with
Aerodynamically Balanced Elevator." NACA TM 991, October
at an angle of attack that includes the dynamic stall delay). 1941.
Retaining the use of airfoil table data is considered 5) Theodorsen, T., and Garrick, I.E. "Nonstationary Flow
essential, both to provide the basic characteristics About a Wing-Aileron-Tab Combination Including Aero-
associated with airfoil shape, and to isolate the effects of dynamic Balance." NACA Report 736, 1942.
the dynamic stall model. Among the dynamic stall models, 6) Petot, D. "Differential Equation Modeling of Dynamic
the Boeing model is the simplest, requiring only three Stall." La Recherche Aerospatiale, Number 1989-5
parameters. However, the Boeing model does not include (corrections dated October 1990).
the loads produced by the leading-edge vortex; and the form 7) Peters, D. "Toward a Unified Lift Model for Use in Rotor
Blade Stability Analyses." Journal of the American Helicopter
of the angle of attack delay is much different from that of Society, Volume 30, Number 3, July 1985.
the other models. The ONERA EDLIN and ONERA BH 8) Barwey, D.; Gaonkar, G.H.; and Ormiston, R.A.
models require a large number of parameters. The "Investigation of Dynamic Stall Effects on Isolated Rotor
Leishman-Beddoes model is attractive because the Flap-Lag Stability with Experimental Correlation." Journal of
parameters defining trailing-edge separation point and the American Helicopter Society, Volume 36, Number 4,
critical lift coefficient are obtained from the static airfoil October 1991.
characteristics. However, the time constants required by the 9) Barwey, D., and Gaonkar, G.H. "Dynamic-Stall and
Structural-Modeling Effects on Helicopter Blade Stability with
Leishman-Beddoes model do not seem to have universal Experimental Correlation." AIAA Journal, Volume 32, Number
values, but depend on the airfoil shape. The differences 4, April 1994.
obtained using these dynamic stall models are no more 10) van der Vooren, A.I. "The Theodorsen Circulation
than should be expected from empirical methods. The Function and Aerodynamic Coefficients." AGARD Manual on
correlation for the oscillating three-dimensional wing may Aeroelasticity, Volume VI, January 1964.
be judged remarkably good considering the simplicity of 11) Leishman, J.G. "Validation of Approximate Indicial
some of the models; and also exhibiting significant errors, Aerodynamic Functions for Two-Dimensional Subsonic
Flow." Journal of Aircraft, Volume 25, Number 10, October
as expected with empirical models. There is a very large 1988.
effect of the dynamic stall model as well as yawed flow 12) Hariharan, N., and Leishman, J.G. "Unsteady
corrections on calculated rotor performance in extreme Aerodynamics of a Flapped Airfoil in Subsonic Flow by
operating conditions, yet their calculation in practical Indicial Concepts." Journal of Aircraft, Volume 33, Number 5,
analyses must rely on empirical models. September-October 1996.

16
13) Leishman, J.G., and Nguyen, K.Q. "State-Space 31) Johnson, W. "A General Free Wake Geometry Calculation
Representation of Unsteady Airfoil Behavior." AIAA Journal, for Wings and Rotors." American Helicopter Society Forum,
Volume 28, Number 5, May 1990. May 1995.
14) Johnson, W. "The response and Airloading of Helicopter 32) Hess, J.L. "The Problem of Three-Dimensional Lifting
Rotor Blades Due to Dynamic Stall." Massachusetts Institute Potential Flow and Its Solution by Means of Surface
of Technology, ASRL TR 130-1, May 1970. Singularity Distribution." Computer Methods in Applied
15) McCroskey, W.J. "Recent Developments in Dynamic Mechanics and Engineering, Volume 4, Number 3, November
Stall." Symposium on Unsteady Aerodynamics, Tucson, 1974.
Arizona, March 1975. 33) Felker, F.F.; Quackenbush, T.R.; Bliss, D.B.; and Light,
16) Beddoes, T.S. "A Synthesis of Unsteady Aerodynamic J.S. "Comparisons of Predicted and Measured Rotor
Effects Including Stall Hysteresis." Vertica, Volume 1, Number Performance in Hover Using a New Free Wake Analysis."
2, 1976. Vertica, Vol. 14, No. 3, 1990.
17) Ham, N.D., and Garelick, M.S. "Dynamic Stall 34) Quackenbush, T.R.; Bliss, D.B.; Wachspress, D.A.; and
Considerations in Helicopter Rotors." Journal of the American Ong, C.C. "Free Wake Analysis of Hover Performance Using a
Helicopter Society, Volume 13, Number 2, April 1968. New Influence Coefficient Method." NASA CR 4309, July
18) Harris, F.D.; Tarzanin, F.J., Jr., and Fisher, R.K., Jr. 1990.
"Rotor High Speed Performance, Theory vs. Test." Journal of 35) Lorber, P.F.; Stauter, R.C.; and Landgrebe, A.J. "A
the American Helicopter Society, Volume 15, Number 3, April Comprehensive Hover Test of the Airloads and Airflow of an
1970. Extensively Instrumented Model Helicopter Rotor." American
19) Tarzanin, F.J., Jr. "Prediction of Control Loads Due to Helicopter Society Forum, May 1989.
Blade Stall." Journal of the American Helicopter Society, 36) Tung, C., and Ramachandran, K. "Hover Performance
Volume 17, Number 2, April 1972. Analysis of Advanced Rotor Blades." American Helicopter
20) Gormont, R.E. "A Mathematical Model of Unsteady Society Forum, June 1992.
Aerodynamics and Radial Flow for Application to Helicopter 37) Tung, C., and Lee, S. "Evaluation of Hover Performance
Rotors." USAAVLABS TR 72-67, May 1973. Prediction Codes." American Helicopter Society Forum, May
21) Leishman, J.G., and Beddoes, T.S. "A Semi-Empirical 1994.
Model for Dynamic Stall." Journal of the American Helicopter 38) Ballard, J.D.; Orloff, K.L.; and Luebs, A.B. "Effect of Tip
Society, Volume 24, Number 3, July 1989. Shape on Blade Loading Characteristics and Wake Geometry
22) Leishman, J.G., and Beddoes, T.S. "A Generalized Model for a Two-Bladed Rotor in Hover." Journal of the American
for Airfoil Unsteady Aerodynamic Behavior and Dynamic Stall Helicopter Society, Volume 25, Number 1, January 1980.
Using the Indicial Method." American Helicopter Society 39) Norman, T.R., and Light, J.S. "Rotor Tip Vortex
Forum, June 1986. Geometry Measurements Using the Wide-Field Shadowgraph
23) Leishman, J.G., and Crouse, G.L., Jr. "State-Space Model Technique." Journal of the American Helicopter Society,
for Unsteady Airfoil Behavior and Dynamic Stall." AIAA Paper Volume 32, Number 2, April 1987.
Number 89-1319, April 1989. 40) Lau, B.H.; Wadcock, A.J.; and Heineck, J.T. "Wake
24) Truong, V.K. "A 2-D Dynamic Stall Model Based on a Visualization of a Full-Scale Tilt Rotor in Hover." AHS
Hopf Bifurcation." European Rotorcraft Forum, Italy, Specialists' Meeting for Rotorcraft Acoustics and
September 1993. Aerodynamics, Williamsburg, Virginia, October 1997.
25) R.A. Piziali, "2-D and 3-D Oscillating Wing 41) Jepson, D.; Moffitt, R.; Kilzinger, K.; and Bissell, J.
Aerodynamics for a Range of Angles of Attack Including "Analysis and Correlation of Test Data from an Advanced
Stall." NASA TM 4632, September 1994. Technology Rotor System." NASA CR 3714, August 1983.
26) Bousman, W.G.; Young, C.; Gilbert, N.; Toulmay, F.; 42) Peterson, R.L. "Full-Scale Hingeless Rotor Performance
Johnson, W.; and Riley, M.J. "Correlation of Puma Airloads - and Loads." NASA TM 110356, June 1995.
Lifting-Line and Wake Calculation." NASA TM 102212, 43) Picasso, B.D., III; Downs, G.T.; Buckanin, R.M.; and
November 1989. Ottomeyer, J.D. "Airworthiness and Flight Characteristics
27) Nguyen, K.; McNulty, M.; Anand, V.; and Lauzon, D. Test, Part 2, YAH-64 Advanced Attack Helicopter." USAAEFA
"Aeroelastic Stability of the McDonnell Douglas Advanced Report 80-17-2, February 1982.
Bearingless Rotor." American Helicopter Society Forum, May 44) Nagata, J.I.; Piotrowski, J.L.; Young, C.J.; Lewis, W.D.;
1993. Losier, P.W.; and Lyle, J.A. "Baseline Performance
28) Nguyen, K.; Lauzon, D.; and Anand, V. "Computation of Verification of the 12th Year Production UH-60A Black Hawk
Loads on the McDonnell Douglas Advanced Bearingless Helicopter." USAAEFA Report 87-32, January 1989.
Rotor." American Helicopter Society Forum, May 1994. 45) Baserga, C. "Rotor/Propeller Axial Flight Performance
29) Lau, B.H.; Louie, A.W.; Sotiriou, C.P.; and Griffiths, N. Model." Advanced System Research and Analysis Office,
"Correlation of the Lynx-XZ170 Flight-Test Results Up To USAATCOM, Moffett Field, California, April 1994.
and Beyond the Stall Boundary." American Helicopter Society 46) Felker, F.F.; Betzina, M.D.; and Signor, D.B.
Forum, May 1993. "Performance and Loads Data from a Hover Test of a Full-Scale
30) Harris, F.D.; Tarzanin, F.J., Jr.; and Fisher, R.K., Jr. XV-15 Rotor." NASA TM 86833, November 1985.
"Rotor High Speed Performance, Theory vs. Test." Journal of 47) Light, J.S. "Results From an XV-15 Rotor Test in the
the American Helicopter Society, Volume 15, Number 3, July National Full-Scale Aerodynamics Complex." American
1970. Helicopter Society Forum, May 1997.

17
 APPENDIX A: Unsteady Airfoil Motion

This appendix presents the equations for the unsteady aerodynamics models of CAMRAD II. Models from incompressible
thin-airfoil theory, ONERA EDLIN, and Leishman-Beddoes are implemented. The unsteady loads depend on the upwash w
(the normal velocity of the wing relative to the air, measured at the quarter chord), and the upwash gradient along the chord
w
= ∂w/∂x (typically from a pitch rate). In the equations below, U is the resultant velocity of the airfoil section (always
positive); V is chordwise velocity of the wing relative the air (negative in reverse flow); M is the section Mach number; and c
is the chord. When implemented, the appropriate changes in reverse flow are included.

Incompressible Thin-Airfoil Theory

The incompressible unsteady loads are derived following reference 3. In coefficient form, the results for unsteady circulation,
lift, and moment are:
a 
c
cgU S = w
U 2
a  c c c 
cU S = 2 V w
+ ẇ + ẇ

U 2 4 4
  
a c c 3c
cmU S = (xAF − xAC )cU S + 2 −V w
− ẇ + ẇ

U 16 16 8
The following options are implemented for the lift-curve slope a: (1) constant; (2) Prandtl-Glauert; (3) Prandtl-Glauert with
lift divergence; (4) from the airfoil table at zero lift; (5) from the airfoil table at the local angle of attack; (6) secant slope
from the airfoil table. Different choices can be made for the circulatory lift, other lift, and moment. According to steady
thin-airfoil theory, both the lift and moment scale with the Prandtl-Glauert factor. The following options are implemented for
the aerodynamic center shift δx = xAF − xAC : (1) constant; (2) from the airfoil table at zero lift; (3) from the airfoil table at
the local angle of attack. Here xAF is the position of the reference axis of the airfoil coefficients, and xAC is the position of the
actual aerodynamic center; both measured as fraction of the chord aft of the leading edge (usually xAF = .25). These several
choices for a and δx are needed to accommodate the aerodynamic theories as originally developed.

ONERA EDLIN Theory

The ONERA EDLIN (Equations Differentielles Lineaires) theory for the unsteady loads is presented in reference 6. To
include the effects of compressibility, Küssner’s coefficients are used, as tabulated by van der Vooren (ref. 10) and curve-fit by
Petot (ref. 6). In coefficient form, the results are:
a 
c c c
 a
a 
c a
cU S = V w + ẇfL0 + ẇ fL1 + L1 cgU S = w + L1
U2 2 4 4 U U 2 U
    
a c c 3c c
cmU S = (xAF − xAC )cU S + 2 −V w
fM 0 − ẇfM 0 + ẇ
fM 1 L̇1 + λL1 = µ ẇ + ẇ

U 16 16 8 2

where fL0 = β[1 + 5(β .57 − 1)], fL1 = β[1 + 3.92(β − 1)], fM 0 = β[1 + 1.4M 2 ], fM 1 = β[−1.2625 + 1.5330 tan−1 (10.5 −
√
15M )]; and β = 1 − M 2 . These factors give a good representation of Küssner’s coefficients, except that the moment produced
by heave (cmh = π4 k 2 fM 0 ) is always real, when it should exhibit a phase shift for M > 0. The L1 term accounts for the airfoil
shed wake effects (lift deficiency function), with λ = (2U/c)λ0 (1. − .76M ) and µ = − 14 (3 − β). Petot (ref. 6) gives a value
of 0.17 for λ0 . The ONERA EDLIN theory as presented by Petot (ref. 6) uses the Prandtl-Glauert option for the lift-curve slope
a. The ONERA BH theory (ref. 24) implies using the lift-curve slope at the local angle of attack.

18
 Leishman-Beddoes Theory

The Leishman-Beddoes theory for unsteady loads in attached flow is presented in references 11 to 13. In coefficient form,
the results are:
a
cgU S = (LC1 + LC2 )
U
a aβ  
cU S = (LC1 + LC2 ) + 2 c c0α kα Lα + c2 c0q kq Lq
U U

L̇C1 + λ1 LC1 = λ1 A1 c∞q cw
− A1 c∞α ẇ L̇α + λα Lα = λα ẇ

L̇C2 + λ2 LC2 = λ2 A2 c∞q cw
− A2 c∞α ẇ L̇q + λq Lq = λq ẇ

λ1 = (2U/c)b1 β 2 λ2 = (2U/c)b2 β 2
λα = U/(cM kα ) λq = U/(cM kq )

a aβ  
cmU S = (xAF − xAC )cU S + MC1 + 2 c c0α kα (Mα3 + Mα4 ) + c2 c0q kq Mq
U U

ṀC1 + λ5 MC1 = λ5 c∞q cw
Ṁq + λq Mq = λq ẇ

Ṁα3 + λα3 Mα3 = λα3 b3 A3 ẇ Ṁα4 + λα4 Mα4 = λα4 b4 A4 ẇ

λ5 = (2U/c)b5 β 2 λq = U/(cM kq )
λα3 = U/(cM kα b3 ) λα4 = U/(cM kα b4 )

1−η   a   1  a 
cdU S = (c + cU S ) c + (LC1 + LC2 ) − c2 − (c + cU S ) (LC1 + LC2 )
cα U cα U

where A1 = .3, A2 = .7, A3 = 1.5, A4 = −.5, b1 = .14, b2 = .53, b3 = .25, b4 = .1, b5 = .5. The time constants are:
c0 κ
T = 2M k = 2M

c1 (1 − M ) + c∞ 2M 2 β (Ab)

where (Ab) is A1 b1 + A2 b2 = .413 for the lift; b5 for the moment kq ; and zero for the moment kα . The factors κ are
introduced to improve correlation with measured loads. Good results are obtained with κL = .75 for lift and κM = .80 for
moment. The constants c0 , c1 , and c∞ are given in the following table.

derivative load 2πc0 2πc1 2πc∞

α c 4 4 2π or cα
cm −1 −1 0 or cmα
q = α̇c/U c 1 1 π
cm − 12
7
− 54 − π8

The expressions for c1 are obtained from reference 12. The quasistatic terms of incompressible thin-airfoil theory give c∞ .
Optionally the α derivatives for c∞ can be evaluated from the airfoil tables (then a = 2π/β should be used). In cdU S , c is the
lift coefficient without the unsteady load or dynamic stall terms. The unsteady drag is based on approximating the steady drag
in attached flow as follows (ref. 21): cd = cdz + ((1 − η)/cα ) c2 , where η < 1 is the chord force recovery factor (typically

19
η = .95) and cdz is the drag at zero lift. The following options are implemented for the drag recovery factor η: (1) constant;
(2) from the airfoil table at small angle of attack; (3) from the airfoil table at the local angle of attack. The Leishman-Beddoes
theory as presented in ref. 11 uses the secant slope for a in the LC terms, and the Prandtl-Glauert option in the remaining terms;
and the aerodynamic center and drag recovery factor from the airfoil tables at the local angle of attack.

APPENDIX B: Dynamic Stall

This appendix presents the equations for the dynamic stall models of CAMRAD II. Dynamic stall models from Johnson,
Boeing, Leishman-Beddoes, ONERA EDLIN, and ONERA BH are included. Dynamic stall is characterized by a delay in the
occurrence of separated flow produced by the wing motion, and high transient loads induced by a vortex shed from the leading
edge when stall does occur. Let αd be the delayed angle of attack, calculated from the angle of attack α. Then the corrected
coefficients are:  
α − αz
c = c2D (αd ) + ∆cDS
αd − αz
 2  
α − αz
cd = cd2D (αd ) − cdz + cdz + ∆cdDS
αd − αz
 
α − αz  
cm = cm2D (αd ) − cmz + cmz + ∆cmDS
αd − αz
where αz is the zero-lift angle of attack, and cdz and cmz are the corresponding drag and moment. The form of the lift and
moment corrections ensures that the coefficients below stall are unchanged. The ∆cDS ’s are increments defined by the dynamic
stall model, generally attributed to the leading-edge vortex. The dynamic stall effects are washed out for angles of attack near
±90. When implemented, the appropriate changes for negative angle of attack and reverse flow are included in all models.

Johnson Model

The Johnson dynamic stall model (adapted from ref. 14) uses an angle of attack delay proportional to α̇, plus impulsive
lift and moment increments from the leading-edge vortex. The angle of attack is evaluated with a time delay ∆t = τd c/2U
that accounts for the hysteresis effects around stall: αd = α(t − ∆t) ∼
= α − ∆t α̇ = α − τd α̇c/2U . Alternatively, αd is the
solution of a state equation: α̇d + λd (αd − αz ) = λd (α − αz ), where λd = 2U/cτd . McCroskey (ref. 15) and Beddoes (ref. 16)
found that the dynamic stall delay correlates fairly well in terms of the normalized time constant τd . The values τL = 9.2 and
τM = 5.4 are typical. The equations for the loads include the increments ∆cDS , ∆cdDS , and ∆cmDS , which are produced by
the leading-edge vortex. When the blade section angle of attack reaches the dynamic stall angle αDS , a leading-edge vortex is
shed. As this vortex passes aft over the airfoil upper surface it induces large transient loads. The experimental data of reference
17 show that the peak incremental aerodynamic coefficients depend on the pitch rate at the instant of stall, α̇c/U . It is assumed
that the incremental coefficients caused by the shed vortex (∆cDS ) rise linearly to these peak values in the time increment
∆tDS = τv c/2U , and then fall linearly to zero in the time ∆tDS again. Hence the model involves impulsive lift and nose down
moment changes when dynamic stall occurs. After these transient loads decay, the wing section is assumed to be in deep stall,
and dynamic stall is not allowed to occur again until the flow has reattached. Flow reattachment takes place when the angle of
attack drops below the angle αRE . The dynamic stall angle αDS and reattachment angle αRE correspond to fd = f (αd ) = .7,
where f is the trailing-edge separation point of the Leishman-Beddoes model. The experimental data of reference 17 give
∆cLEV = 2.0 and ∆cmLEV = −0.65 for the peak loads at high pitch rate (α̇c/U at stall above .05). Typically τv = 3.6 to
5.6. Note that if the total rise and fall time 2∆tDS is interpreted as the time the leading-edge vortex takes to traverse the chord
(distance c), then the speed of the vortex is vvortex = U/τv . So τv = 4 implies the leading-edge vortex travels at one-fourth the
free stream speed.

20
 Boeing Model

The Boeing dynamic stall model (developed in refs. 18 to 20) uses an angle of attack delay proportional to the square-root
of α̇, which produces the basic hysteresis effects. The coefficient increments produced by the leading-edge vortex are not used

in this model. The delayed angle of attack is αd = α − τd |α̇c/2U | sign α̇. Alternatively, this can be considered a time lag

dependent on the pitch rate, giving a state equation for αd : α̇d + λd (αd − αz ) = λd (α − αz ), where λd = (2U/cτd ) |α̇c/2U |.
The time constant τd is a function of Mach number and the airfoil section, obtained from oscillating airfoil tests (ref. 20).

Leishman-Beddoes Model

The Leishman-Beddoes dynamic stall model (refs. 21 to 23) uses a delayed angle of attack, plus lift and moment increments
from the leading-edge vortex. This model characterizes the airfoil static stall behavior by the trailing-edge separation point f
(fraction of chord from leading edge), and a critical lift coefficient cCR at the separation onset boundary (leading-edge separation
at low Mach number, shock reversal at high Mach number). The airfoil data for lift are used to identify constants s1 , s2 , and αs
that generate f (α) as follows:

1. − .3 exp((|α − αz | − αs )/s1 ) |α − αz | ≤ αs
f=
.04 + .66 exp((αs − |α − αz |)/s2 ) |α − αz | > αs

Then |α − αz | = αs or f = .7 is taken as the definition of stall. The parameters cCR , s1 , s2 , and αs are required as a function of
Mach number, for positive and negative angle of attack, normal and reverse flow, at each span station. The Leishman-Beddoes
model for unsteady flow is based on fd = f (αd ) at the delayed angle of attack. Here the model is modified to use the static loads
directly from the airfoil tables, instead of fitting the static loads to analytical functions. Further modifications of the model are
required because the above expression for f (α) does not distinguish between positive and negative angle of attack. The model
is also extended to large angle of attack and reverse flow. In order to handle oscillations through αz , a continuous monotonic
function of α is needed: 
fz − f α − αz ≥ 0
f =
f − fz α − αz < 0
where fz = f (αz ) = 1. − .3 exp(−αs /s1 ); note that fz may not be the same for positive and negative angle of attack. In order
to handle large angle of attack, the function f is modified:

 1. − .3 exp((|α − αz | − αs )/s1 ) |α − αz | ≤ αs

f = .04 + .66 exp((αs − |α − αz |)/s2 ) |α − αz | > αs


fH (90 − |α − αz |)/(90 − αH ) |α − αz | > αH

Here αH = αs + Hs2 and fH = .04 + .66e−H , and H = 6 is used. This modification to the definition of f does not affect the
model for attached flow or around stall, but with it the delayed angle of attack behaves reasonably at very large angles.

The delayed angle of attack αd is calculated as follows. Static hysteresis around stall is modelled by using a smaller αs
when the angle of attack is decreasing: αs = αs input − ∆αs (1 − fd )1/4 (with fd from the last time step). There is a lag in the
leading-edge pressure relative c , so a lagged lift is used in the stall criterion:

L̇p + λp Lp = λp (α − αz + LC /A)

αp = Lp + LI /A + αz

where A = cα . The dimensionless time constant Tp gives λp = 2U/cTp . The unsteady lift cU S is split into LC (the LC1 + LC2
term in Leishman-Beddoes theory) and LI (the remaining terms), calculated using a = cα instead of the secant slope. The stall

21
criterion is based on cp = A (αp − αz ). If |cp | ≥ cCR , the critical condition for leading-edge or shock-induced separation
has been reached; while if |cp | < cCR , reattachment is allowed. This angle of attack gives a trailing-edge separation point
f p = f (αp ). There is an additional lag in the boundary layer response, modelled as a lag in f :
˙
f d + λf f d = λf f p
where λf = 2U σf /cTf . There are separate fd equations for lift and moment, to allow different behavior during reattachment
(implemented using different values of σf ). If fd is decreasing (∆fd < 0), the flow is separating; if fd is increasing, the flow is
reattaching. The difference ∆fd is calculated at the end of the procedure, for use during the next time step. Finally, the delayed
angle of attack αd is calculated from f d . Vortex lift accumulation begins at the onset of stall (indicated by |cp | = c ), driven
CR

by the difference between the linear and nonlinear lifts: cv = cL − c = A(α − αz ) − c . The leading-edge vortex reaches the
trailing edge at time τDS = Tvl , where tDS = τDS c/2U is the time since the onset of stall. The speed of the vortex implied is
vvortex = 2U/Tvl , or one-fourth the free stream velocity for Tvl = 8. The vortex loads are obtained from cv with a time lag:

L̇v + λv Lv = Dċv ∆cdDS = Lv tan(α − αz )

 
πτDS
∆cDS = Lv ∆c mDS = −x s 1 − cos ∆cDS
Tvl
where xs = .20 or .25 typically. The moment exists only until τDS = 2Tvl . The time constant gives λv = 2U σv /cTv . The
switch D (equal zero or one) turns the vortex lift accumulation on and off. With the switch D present, it is not possible to
formulate the equation for Lv in terms of cv instead of ċv , such that the CAMRAD II analysis can deal with the resulting
differential equation of motion. So an implicit solution is always used for Lv . Interactions between the mechanisms are
accounted for by modifying the time constants, using the parameters σf (for lift and moment) and σv .

An alternative form of the delayed angle of attack calculation uses a lagged lift coefficient cp to obtain the angle of attack
αp :
cC = A (α − αz ) + LC cp = Lp + LI
L̇p + λp Lp = λp cC αp = cp /A + αz
To match the implicit solution of the Leishman-Beddoes model in references 21 and 22, the following form is required:
˙
cC = A (α − αz ) + LC cp = cC − Lp + LI L̇f + λf Lf = f p
L̇p + λp Lp = ċC αp = cp /A + αz f d = f p − Lf
The time derivatives on the right hand side of these state equations are not available analytically, so in this form an implicit
solution must be used. In practice, identical results are obtained from all these forms of the equations.
√
The trailing-edge separation point f is related to the airfoil lift using the Kirchhoff expression: c = cα ((1+ f )/2)2 (α−
αz ) (ref. 21). Hence the airfoil table data for lift define f as a function of α. The parameter αs is given by f = .7; and then
s1 and s2 are identified by fitting the table data. The critical lift cCR should be identified from pressure data (ref. 21), but can
also be determined based on the break in chord force (loss of leading-edge suction) or the sudden increase in drag. The time
constants are determined by correlation with unsteady airfoil data (see, for example, ref. 22).

ONERA EDLIN Model

The ONERA EDLIN (Equations Differentielles Lineaires) dynamic stall model (ref. 6) uses a stall delay, plus lift, drag,
and moment increments calculated from second-order differential equations:
L̈2 + aL̇2 + bL2 = −bU ∆c − eU α̇

M̈2 + aṀ2 + bM2 = −bU ∆cm − eU α̇

D̈2 + aḊ2 + bD2 = −bU ∆cd − eU α̇ sign(α − αz )

22
with α in degrees here. These equations are driven by the difference between the linear and nonlinear loads: ∆c = cL − c =
cα (α − αz ) − c , ∆cm = cmL − cm = cmα (α − αz ) + cmz − cm , ∆cd = cdL − cd = cdz − cd ; where c , cm , and cd are here
the static coefficients, without the unsteady or leading-edge vortex terms. Then the load increments are
1
∆cDS = (L2 + U ∆c + dU α̇)
U
1
∆cmDS = (M2 + U ∆cm + dU α̇) + M3
U
1
∆cdDS = (D2 + U ∆cd + dU α̇ sign(α − αz ))
U
The stall delay is accounted for by setting the right-hand side of the differential equation to zero if τSS < τd , where tSS =
τSS c/2U is the time since the static stall angle was exceeded. The static stall angle corresponds to f (α) = .7, where f is
the trailing-edge separation point of the Leishman-Beddoes model. Petot has described a refined transition model, intended to
accommodate airfoils that exhibit larger nose-down pitching moments at dynamic stall. The refined transition model assumes
that the extra lift from dynamic stall is convected aft from the quarter chord after moment stall occurs, producing the extra
moment term M3 in ∆cmDS . Here this refined transition model is implemented by

M3 = −(τSS − τdM ) µ ∆cDS

for (τSS − τdM )µ = 0 to 1.5 (convection from quarter chord to trailing edge). This extra moment is turned off after lift stall,
by multiplying M3 by the factor (2 − τSS /τdL ) when τdL < τSS < 2τdL . The coefficients in these equations depend on the lift
difference ∆c :
 
a = (2U/c) a0 + a2 (∆c )2 
 2 (c/2U ) d1 |∆c | lift and moment
b = (2U/c)2 b0 + b2 (∆c )2 d=
(c/2U ) (d0 |α − αz | + d1 |∆c |) drag
e = (2U/c) e2 (∆c )2
The notation has been changed somewhat from the original ONERA notation. The parameters must be evaluated from data on
airfoils oscillating in the stalled flow regime (see, for example, ref. 6).

ONERA BH Model

The ONERA BH (Bifurcation de Hopf) dynamic stall model (ref. 24) uses a delayed angle of attack, plus lift and moment
increments calculated from first-order and second-order differential equations:

L̇1 + λL1 = −λU ∆c L̈2 + aL̇2 + bL2 = eU α̇ + dU α̈

Ṁ1 + λM1 = −λU ∆cm M̈2 + aṀ2 + bM2 = eU α̇ + dU α̈

with α in degrees here. Then the load increments are
1
∆cDS = (L1 + U ∆c + L2 )
U
1
∆cmDS = (M1 + U ∆cm + M2 )
U
The appropriate changes for reverse flow are included. The delayed angle of attack is calculated using a simplified version of
the Leishman-Beddoes model:
˙
α̇p + λp (αp − αz ) = λp (α − αz ) f d + λf f d = λf f p
f p = f (αp ) αd = α(f d )

23
where λp = 2U/cTp and λf = 2U/cTf . The coefficients in these equations depend on the load increments:

  e = (2U/c) ωs e0
a = (2U/c) ωs −a0 + a2 ∆c2
  d = ωs d0
b = (2U/c)2 ωs2 1 − b1 ∆c sign(α − αz ) − b2 ∆c2
λ = (2U/c) λ0

using ∆c = L2 /U and M2 /U in the lift and moment equations respectively (from the last time step if an implicit solution is
used). The notation has been changed somewhat from the original ONERA notation. The parameters have different values for
separating and reattaching flow: separating flow values are used if fp < fCR ; reattaching flow values are used if fp ≥ fCR
(note the use of fp rather than fd ). For lift, fCR = .7 is used. The critical angle of attack is ∆αm larger for moment than for
lift. So 
.04 + .66 exp(−∆αm /s2 ) if α increasing
fCR =
1. − .3 exp(−∆αm /s1 ) if α decreasing
is used for the moment. For reattaching flow, only a0 has a nonzero value, so the dynamic stall loads decay. The parameters
must be evaluated from data on airfoils oscillating in the stalled flow regime (see, for example, ref. 24).

24