Axisymmetric Analysis and Design of Ducted Rotors

Mark Drela
1
Harold Youngren
2
Draft 8 Dec 05
Summary
This report describes the analytical formulation used for the design and analysis of a ducted propul-
sor with single or multiple blade rows. It is based on a lifting-line representation of the rotor blade
together with an axisymmetric panel representation of the duct and centerbody. The induced ve-
locities associateed with blade-row loading are represented by vortex sheets shed into the oweld.
Blade element models are used for blade row sections using two-dimensional lift, moment and prole
drag characteristics to account for loading and viscous losses.
The modeling delity of the present approach lies between classical vortex/blade-element methods
of Betz [1] and Glauert [2], and a general 3-D vortex-lattice or panel method. Even the general 3-D
formulations would have to make the same time-averaging assumptions for the unsteady counter-
rotating ow as the present method, and hence would not be more sophisticated or more accurate in
this regard. The chief advantage of the present method is that it is extremely fast computationally
and is has simple inputs, making it ideal for interactive design work.
The formulation described here derives in part from the existing propeller design/analysis code
XROTOR, which combines classical propeller theory with some simple models for the eects of a
duct and centerbody. XROTORs simple duct treatment ignores the details of the duct velocity
eld and the eect of losses in the duct owpath, both of which have a strong inuence on ow
conditions on the rotor. In contrast, the present formulation includes a more detailed treatment of
the following eects.
Reformulation to allow arbitrarily large induced velocities relative to the freestream velocity.
This is necessary to treat the hovering case where the freestream velocity vanishes.
Incorporation of a shrouded tip and center body into the self-induced velocity formulation.
Inclusion of loss eects and non-uniform loading eects in the duct oweld calculation.
Nomenclature
r, , x cylindrical coordinates
, grid coordinates (in grid section)
, , helical coordinates (in blade geometry section)
m meridional coordinate: dm
2
= dx
2
+ dr
2
m

stretched meridional coordinate: dm

= dm/r
density
p static pressure
p
o
total pressure
H total enthalpy
s entropy
S kinematic entropy ( s p/ )
V

freestream speed along rotor axis

1
MIT Aero & Astro Dept.
2
Aerocraft, Inc.
1
v self-induced velocity
u externally-induced velocity

V total velocity relative to duct

W total velocity relative to blade

body surface tangential vortex sheet strength
wake tangential vortex sheet strength (

m
wake meridional vortex sheet strength
vorticity
B number of blades per rotor
c(r) blade chord
c

(r) blade section lift coecient

c
d
(r) blade section prole drag coecient

o
blade airfoil zero-lift angle
(r) blade angle
(r) net ow angle relative to blade
(r) blade circulation
(r) apparent source due to prole drag
rotor rotational speed
1 Geometry
Figure 1 shows the assumed blade row geometry for rotor or stator, dened in lefthanded cylindrical
coordinates. The angle is positive about the x direction, so a conventional righthanded rotor
rotates in the + direction.
m
x
r

V
vorticity
m
blade lifting line
with circulation,
profile drag
source ring
(object with profile drag)
W

Real Viscous Flow (RVF)

Equivalent Inviscid Flow (EIF)
EIF
RVF
Figure 1: Ducted propulsor geometry with blade lifting line, vortical wake, and prole drag source.
The velocities computed by the present approach will correspond to the Equivalent Inviscid Flow
(EIF), which is a locally irrotational continuation into a viscous wake, as shown in Figure 1. This
concept allows modeling of drag-producing objects by simple source elements such as a source ring
or axisymmetric source panel. The Real Viscous Flow (RVF) velocity, also shown in Figure 1, can
be reconstructed if necessary using convected quantities, as will be described later.
2

W
r
m
V
V
V
v
V
V
m
V

u
Figure 2: Velocity triangle at one radial blade station and velocity decomposition.
2 Velocity Decomposition
Figure 2 shows the velocity triangle seen by a blade section at radius r. The local velocity

V relative
to the duct is the vector resultant of the freestream velocity V

, the induced velocity of the blade

row and bodies v, and any external velocity u. The latter might be due to an upstream rotor or
obstruction.

V = v + u + V

x (1)
The local velocity

W relative to the blade also includes the tangential component from rotor
rotation.

W =

V r

(2)
The individual velocity components are therefore
V
x
= v
x
+ u
x
+ V

V
r
= v
r
+ u
r
(3)
V

= v

+ u

W
x
= v
x
+ u
x
+ V

W
r
= v
r
+ u
r
(4)
W

= v

+ u

r
The meridional velocity

V
m
shown in Figure 1 is dened to be tangent to the mean streamline in
the meridional x-r plane,

V
m
=

W
m
= V
x
x + V
r
r (5)
so that the total velocity vectors can be given only in terms of the m and components, as pictured
in Figure 2.

V =

V
m
+ V

(6)

W =

V
m
+ W

(7)
3
3 Circulation and Tangential Velocity
Conservation of circulation around a streamtube gives the tangential velocity v

explicitly in terms
of the net circulation

of all upstream blade rows intersecting that streamtube, as shown in
Figure 3. The circulation contributed by one blade row with B blades each having circulation ,
is simply B.
v

2r
(8)

upstream
B (9)
In some situations it is necessary to obtain the tangential velocity on the blade row itself, e.g. in
x
r
B
B

~
Figure 3: Field circulation

is sum of circulations of upstream blade rows.
order to relate the blade loading to the local ow angle and blade section properties. In this case
that blade row contributes half of its own circulation to its own tangential velocity.
(v

)
row
=
1
2r
_

+
1
2
B
_
(10)
4 Vorticity Representation
At any blade location where the circulation changes by (including the root and tip), a helical
lament of strength will be shed into the ow. To be force-free, the laments must be parallel
to the total velocity

W relative to the blade. To allow an axisymmetric treatment, these laments
are smeared into tangential vortex sheets of strength

, and meridional vortex sheets of strength

m
, as shown in Figure 4. These strengths are simply the circulation/length densities in the two
directions. The subscript on

will be dropped for convenience.

m
=
B
2r
(11)

=
B
2r
W

W
m
(12)
For a eld location with some number of upstream blade rows, such as the one shown in Figure 3,
the more general form of relation (11) uses the local total circulation jump across the sheet.

m
=

2r
(13)
The general expression for

is obtained from pressure relations, developed in the following sections.

4

m
2
B

m
2
B

m
B
2r
B
2r

Vortex filaments Vortex sheet approximation

W

W
W
m
W
m
W
W
m
W
=

=
Figure 4: Blade circulation change producing helical vortex laments on an m- surface. The
laments are represented by equivalent tangential and meridional vortex sheets

and
m
. A
negative tangential relative velocity W

is shown.
5 Thermodynamic relations
5.1 Total pressure
A rotor, stator, or other obstruction will result in downstream changes in total enthalpy H and
entropy s. These must be accompanied by changes in the total pressure p
o
, according to the
following denition of entropy.
s s

= ln
_
(H/H

)
/(1)
p
o
/p
o
_
(14)
Dening changes relative to the far-upstream values,
p
o
p
o
p
o

H H H

s s s

relation (14) becomes

1 +
p
o
p
o
=
_
1 +

H
H

_
/(1)
exp ( s) (15)
For low Mach numbers, it can be assumed that
p
o
/p
o
1

H/H

1 s 1
which simplies (15) to the following.
p
o

_

H

S
_
(16)

S
p

s (17)
At low Mach numbers p/ is very nearly constant, in which case it is valid to treat

S as the
convected quantity rather than s.
5
5.2 Disk jump conditions
After the periodic ow about the rotor blades is averaged in , there will generally be a change
in the total enthalpy H across the rotor disk as a result of the work done on the uid. Similarly,
there will be a change in entropy across a friction-producing disk, ideally modeled as a screen with
friction coecient C
f
. Figure 5 shows two points immediately in front and behind the rotor or
friction disk.
r
V
m
V

x
H
disk
r
V
m
V

disk
S
C
f

B
Figure 5: Jumps across a loaded rotor disk and resistive screen.
The jump relations are
H
disk
= (rV

) =
B
2
(18)
S
disk
=
1
2
V
2
m
C
f
(19)
and the values

H and

S at any eld point are the accumulated jumps of all the upstream blade
rows or obstructions. These are computed the same way as the circulation, as shown in Figure 3.

H =

upstream
H
disk
(20)

S =

upstream
S
disk
(21)
5.3 Static pressure Bernoulli equation
According to (16), the total pressure p
o
at any point is equal to the freestream value p
o
, plus
upstream work and loss contributions. The static pressure then follows from the Bernoulli equation.
p
o
= p
o
+ (

H

S) (22)
p = p
o

1
2
V
2
+
_

H

S
_
(RVF velocity) (23)
As indicated, (23) assumes that V is the velocity of the Real Viscous Flow. Obtaining this RVF
requires representation of all the vorticity present, including that in viscous wakes. An alternative
and simpler approach is to treat the Equivalent Inviscid Flow velocity, by using a source sheet at
the drag element itself, with no trailing vortex sheets. Figure 6 compares the two approaches. To
obtain the pressure in the EIF model, the entropy is simply excluded from the total pressure.
p = p
o

1
2
V
2
+

H (EIF velocity) (24)
6
C
f
r
x

C
f
r
x

V
m V
m
EIF
RVF
S S
p = 0 p = 0
= 0
= 0
Figure 6: Two alternative representations of a drag-producing object.
Either approach will correctly predict a zero pressure jump, p = 0, across the vortex sheet or
bounding streamline.
6 Rotor vortex sheet jump relations
6.1 Velocity jumps
Figure 7 shows velocities on the two sides of a vortex sheet trailing from one rotor radial station.
The sheet strengths are dened in terms of these velocities, using relations shown in Figure 4,
and the swirl/circulation relation (8), with

= B for the single rotor. Were assuming that no
additional vorticity is created by blade prole drag, so that the velocities in this case represent EIF.

m
= V

1
V

2
=
B(
2

1
)
2r
(25)

= V
m
2
V
m
1
=
B(
2

1
)
2r
W
avg
W
mavg
(26)
x
r

1
2
V
m
2
V
m
1
V
2

1
2
1
Figure 7: Velocities on the two sides of a vortex sheet shed by rotor.
The sheet-average velocities are dened as follows.
W
avg

1
2
(W

1
+ W

2
) =
1
2
(V

1
+ V

2
) r (27)
W
mavg

1
2
(W
m
1
+ W
m
2
) =
1
2
(V
m
1
+ V
m
2
) (28)
7
Combining equations (25) and (26) gives
W
mavg

W
avg

m
= 0 (29)
V
mavg

_
V
avg
r
_

m
= 0 (30)
1
2
_
V
2
2
V
2
1
_
=
B(
2

1
)
2
(31)
where V
2
= V
2
m
+V
2

is the total velocity magnitude. We also note that the sheet strengths can be
dened as

m
= W

1
W

2
(32)

= W
m
2
W
m
1
(33)
which can be combined with (29) to give
1
2
_
W
2
2
W
2
1
_
= 0 (34)
|W
2
| = |W
1
| (35)
6.2 Sheet pressure jump
Using the Bernoulli equation (23), the static pressure jump across the sheet is then given by one
of the following relations.
p
2
p
1
=
1
2

_
V
2
2
V
2
1
_
+
_

H
2

H
1


S
2
+

S
1
_
(RVF) (36)
p
2
p
1
=
1
2

_
V
2
2
V
2
1
_
+
_

H
2

H
1
_
(EIF) (37)
Combining this with the

H denitions (18) and (20), and the sheet velocity jump condition (31),
gives
p
2
p
1
= 0 (38)
as expected from physical considerations.
7 Tangential Vortex Sheet Strength
The tangential vortex sheet strength immediately behind a rotor is dened by relation (12). This
relation is not appropriate for using at an arbitrary location in the eld, since the r component
of W

is then ill-dened. Instead, we will use the alternative and equivalent condition that the
sheet pressure jump as given by (37) is zero. Assuming the EIF representation (37) for illustration,
setting p
2
p
1
= 0 gives
1
2
_
V
2
2
V
2
1
_
=

H
2

H
1
(39)
V
2
m
2
V
2
m
1
+ V
2

2
V
2

1
= 2
_

H
2

H
1
_
(40)
V
2
m
2
V
2
m
1
=
_
1
2r
_
2_

2
2

2
1
_
+ 2
_

H
2

H
1
_
(41)
8
Equation (41) can be used in two ways to determine the vortex sheet strength from a specied

and

H eld. If V
m
2
is known, we can determine V
m
1
from (41), and the sheet strength then follows.
= V
m
2
V
m
1
(42)
Equations (41) and (42) can be marched radially inward at one streamwise location. The march is
started just outside the outermost sheet, where V
m
2
is known (e.g. V
m
V

). For each sheet, V

m
1
and are computed. This V
m
1
is then also assumed to equal the V
m
2
value for next inside sheet.
In eect, the V
m
velocity prole is assumed to be constant between vortex sheets.
If instead the sheet V
mavg
is known, equation (41) can be recast as an explicit expression for .
V
mavg
=
1
2
(V
m
1
+ V
m
2
) (43)
V
2
m
2
V
2
m
1
= 2V
mavg
(44)
=
1
V
mavg
_

1
2
_
1
2r
_
2_

2
2

2
1
_
+

H
2

H
1
_
(45)
8 Vortex Sheet Discretization
The induced velocity v is the result of the duct and center bodies, the trailing vorticity, and viscous
displacement eects. The eect of the bodies is represented by tangential vortex sheets of strength
placed on the body surfaces (the subscript on is omitted for convenience). The eect of
the trailing vorticity is represented by the tangential and meridional vortex sheets of strengths
and
m
, as shown in Figure 4. Viscous displacement is represented by source sheets of strength .
Figure 8 shows the various sheets on the geometry.
m
x
r

n
^

i
k
k
Figure 8: Vortex sheets used to represent meridional velocity

V
m
.
The continuous sheet strengths will now be approximated via discrete panel-node values
i
,
i
, and

i
. The induced velocity components at any location i are then given via inuence matrices.
v
x
i
= a
x
ij

j
+ a
x
ij

j
+ b
x
ij

j
(46)
v
r
i
= a
r
ij

j
+ a
r
ij

j
+ b
r
ij

j
(47)
9
The tangential induced velocity is given directly in terms of the local circulation of all upstream
blade rows.
v

i
=

i
2r
i
(48)
9 Body vorticity elimination
As in any panel method, the body vortex sheet strengths
i
are constrained by the ow tangency
boundary condition.

V
i
n
i
= 0 (49)
This can be written as
a
ij

j
+ a
ij

j
+ b
ij

j
+ V

n
x
i
= 0 (50)
where
a
ij
= a
x
ij
n
x
i
+ a
r
ij
n
r
i
(51)
a
ij
= a
x
ij
n
x
i
+ a
r
ij
n
r
i
(52)
b
ij
= b
x
ij
n
x
i
+ b
r
ij
n
r
i
(53)
Factoring the a
ij
matrix then gives explicitly as follows.

i
=
o
i
V

+ a

ij

j
+ b

ij

j
(54)

o
i
a
1
ij
n
x
j
(55)
a

ij
a
1
i
a
j
(56)
b

ij

b
1
i
b
j
(57)
The freestream-inuence vector
o
i
and source-inuence matrix b

ij
depend only on the body and
source panel geometry, and hence can be precomputed and stored. The wake-inuence matrix a

ij
depends also on the wakes constant-m surface shapes, which in turn depend on the freestream
speed as well as the rotor loading.
The meridional induced velocities can now be given only in terms of the freestream speed V

, the
source sheet strengths
i
, and the wake vortex sheet strengths
i
.
v
x
i
= a
x
ij

o
j
V

+
_
a
x
i
b

j
+ b
x
ij
_

j
+
_
a
x
i
a

j
+ a
x
ij
_

j
(58)
v
r
i
= a
r
ij

o
j
V

+
_
a
r
i
b

j
+ b
r
ij
_

j
+
_
a
r
i
a

j
+ a
r
ij
_

j
(59)
It should be pointed out that the cost of evaluating the sums in (58) and (59) strongly depends on
the evaluation order of the matrix-matrix-vector products. For example, the last such term in (58)
should be evaluated left to right,
a
x
i
a

j

j
= a
x
i
a
1
i
a
j

j
= a
x
i
_
a
1
i
_
a
j

j
_
_
(60)
which requires only matrix-vector products. A right to left evaluation order would require vastly
more expensive matrix-matrix products.
10
10 Blade Load Representation
The wake vortex sheet strength is related to the circulation and total enthalpy jumps across the
sheet, via relation (45).

i
=
1
V
m
i
_

1
2
_
1
2r
_
2

2
_
+
_

H
i
_
_
(61)
Both

and

H are functions of the individual blade station circulations
k
. Substitution into
equations (46) and (47) then gives the meridional induced velocities only in terms of V

, the drag
sources
i
, and the blade circulations
k
.
v
x
i
= v
xo
i
V

+ B
x
ij

j
+ A
x
ik

j
(V

,
j
,
k
) (62)
v
r
i
= v
ro
i
V

+ B
r
ij

j
+ A
r
ik

j
(V

,
j
,
k
) (63)
v
xo
i
= a
x
ij

o
j
(64)
v
ro
i
= a
r
ij

o
j
(65)
B
x
ij
= a
x
i
b

j
+ b
x
ij
(66)
B
r
ij
= a
r
i
b

j
+ b
r
ij
(67)
A
x
ij
= a
x
i
a

j
+ a
x
ij
(68)
A
r
ij
= a
r
i
a

j
+ a
r
ij
(69)
The nonlinear
j
(V

,
j
,
k
) function is given by equation (61).
The remaining tangential induced velocity dened by equation (48) can be put into the same form,
except that it has no V

or
j
dependence.
v

i
= A

ik

k
(70)
A

ik

_
_
_
B
4 r
i
, [i i(k), i(k1)]
0 , [i i(k), i(k1)]
(71)
11 Viscous Drag Equivalent Source
11.1 Blade prole drag
The rotor blade prole drag/span (r) is given by
=
1
2
Wc c
d
(72)
This is represented by an axisymmetric source sheet placed on the rotor disk. The sheet strength
at some radius r is obtained by smearing the total source/span strength of all the blades around
the local circumference.
=
B
2r
=
B
4r
Wc c
d
(73)
11
11.2 Stationary object prole drag
The source strength needed to represent the displacement eect of an objects viscous wake is
related to the drag by
D = V
m
A
where A is the area of the source sheet. The drag is most conveniently represented by a drag area
A
D
, so that
=
D
V
m
A
=
1
2
V
m
A
D
A
(74)
The object being modeled is assumed to be axisymmetric, so that discrete objects such as cylinder
heads must be represented by an equivalent axisymmetric obstruction of similar total A
D
. This
model also assumes that the mass defect of the trailing wake does not vary signicantly downstream
of the object, which is a good approximation if the streamwise pressure gradients along the wake
are modest. The worst case situation would be a massive blockage due to viscous wake bursting
in a strong adverse pressure gradient. Modeling this would require placing source panels all along
the wake, and varying their strength in accordance with the wake displacement area as governed
by strongly interacting boundary layer equations. This level of blockage modeling is beyond the
scope of the present analysis approach.
In the source strength expressions (73) and (74), the drag coecient c
d
and drag area A
D
can
usually be specied explicitly, but the local velocities W and V
m
will depend on the overall ow
solution. Hence, will depend not only on V

, but also on the entire rotor circulation distribution

j
. The most reliable solution approach is to include as an unknown in the overall Newton
system.
12 Blade Section Relations
The rotor blade is represented as a lifting line, with a circulation (r).
=
1
2
Wc c

(75)
The section lift coecient c

depends on the local aerodynamic angle of attack, as shown in Figure 9.

c

= m
2D
(
o
) (76)
= arctan
W
m
W

= arctan
V

+ v
m
+ u
m
r v

(77)
Application of equation (75) to all blade radial stations i = 1 . . . N, and elimination of v
m
and v

in (77) in terms of
j
, produces an N N system for all the
i
strengths. If source sheets are
present, the system can be augmented by including the strengths
i
as additional unknowns, with
(73) or (74) for each source sheet node being the corresponding equations. If
i
are expected to
have a small eect, they can be set explicitly after each Newton iteration, although this will of
course degrade the Newton convergence of
i
by some degree.
13 Axisymmetric wake geometry
The vortex laments and vortex sheets shown in Figure 4 lie on axisymmetric streamsurfaces
parallel to the meridional velocity

V
m
. One approach to computing these surfaces would be to
12
r
V

v
v
c
zerolift line
+
+
v
Sign convention

W
m
m
u
u
u

m m
v u
Figure 9: Blade-section velocities and angles
integrate the streamlines dened by

V
m
. However, in the context of a panel method, this is an
expensive proposition. An alternative approach, which will be taken here, is to use an elliptic grid
generation method.
13.1 Axisymmetric streamfunction
First we consider the axisymmetric streamfunction (x, r), which denes the velocity components
as
V
x
=
1
r

r
(78)
V
r
=
1
r

x
(79)
where is the air density. These velocity components identically satisfy the continuity equation.
(rV
x
)
x
+
(rV
r
)
r
= 0
Evaluation of the circumferential vorticity gives

V
x
r

V
r
x
=
1
r
(
xx
+
rr
) +
x
_
1
r
_
x
+
r
_
1
r
_
r
(80)
which can be written as a Poisson equation for .

xx
+
rr
=
1
r

r
+ r(

V
m
)
_

_
+ r

(81)
The steady Crocco relation is

V =

S
V
m

m
=


H
n



S
n
(82)
13
We note that
V

2r

m
=
1
r
(rV

)
n
=
1
2r

n
V

m
=
1
2
_
1
2r
_
2
(

2
)
n
(83)
so that (82) becomes
V
m

=
1
2
_
1
2r
_
2
(

2
)
n
+


H
n



S
n
(84)
Then using relation (84) to replace the

term from (81) gives the nal Poisson equation governing

(x, r).

xx
+
rr
= Q
0
+ Q
1
(85)
Q
0
=
1
r

r
(86)
Q
1
= r
_

V
m

1
2
_
1
2r
_
2

V
m
(

2
)
V
2
m
+

V
m

H
V
2
m

V
m

S
V
2
m
_

_
(87)
The rst source term Q
0
is simply from the axisymmetry of the problem. The four terms in the
second source term Q
1
represent the density dilatation eect of compressibility, the transverse
circulation gradient, the transverse work gradient, and the transverse loss gradient, respectively.
In low speed ow, is negligible, and the following Q
1
denition can then be employed.
Q
1
=
r
V
2
m
_
1
2
_
1
2r
_
2

V
m
(

2
) +

V
m

H

V
m

S
_

_
(88)
Note that

points out of the x-r plane.

13.2 Grid equations
The grid shown in Figure 10 is dened parametrically by the functions
x(, ) , r(, ) or (x, r) , (x, r)
where is constant along each quasi-radial line, and is constant along each streamline. The
functions are chosen to be governed by Poisson equations.

xx
+
rr
= P (89)

xx
+
rr
= Q (90)
By setting
Q = Q
0
+ Q
1
(91)
we make (90) identical to (85), so that = and the constant- lines of the grid correspond to
physical streamlines. The coordinate has no physical interpretation, and its source term P is
somewhat arbitrary. It is workable to simply set P = 0.
14
x
r

Figure 10: Grid and grid parameters.

Following the procedure of Thompson [3], equations (89) and (90) are inverted to give
x

2x

+
1
r
(rx

1
r
r

+ J
2
rQ
1
x

= 0 (92)
r

2r

+
1
r
(rr

1
r
r

+ J
2
rQ
1
r

= 0 (93)
where
= x
2

+ r
2

(94)
= x
2

+ r
2

(95)
= x

+ r

(96)
J = x

(97)
The two dierential indentities of the transformation are
_
x

_ _
d
d
_
=
_
dx
dr
_
(98)
_

x

r

x

r
_ _
dx
dr
_
=
_
d
d
_
(99)
Inverting the rst identity (98) gives
_
d
d
_
=
1
J
_
r

_ _
dx
dr
_
(100)
and comparing this to the second identity (99) we see that

x
= r

/J (101)

r
= x

/J (102)

x
= r

/J (103)

r
= x

/J (104)
15
Hence the velocities can be computed as follows.
u =

r
r
=
x

rJ
(105)
v =

x
r
=
r

rJ
(106)
With a or value assigned to each grid line, equations (92) and (93) can then be solved for the
node x, r positions. The appropriate boundary conditions are
On solid walls: x, r xed (Dirichlet condition)
On streamlines: u
2
+ v
2
xed (Neumann condition)
On inlet/outlet planes: u
2
+ v
2
xed (Neumann condition)
SLOR is a fast and eective solution method. The grid streamlines then provide the geometry of
the wake vortex sheets needed to dene the overall velocity eld.
13.3 Source term evaluation
The source term Q
1
which appears in equations (92) and (93) is evaluated using its denition (88).
Figure 11 shows an element of the rotor imparting a change in total enthalpy H to its streamtube
of mass ow m
s
. We also allow for the possibility of rotor blade prole drag D
p
along the relative
ow velocity W (not shown), imparting an entropy increase S to the streamtube. A nonlifting
object with drag D, represented by the source ring, imparts only entropy. These quantities are
x
r

m
V
s
.
S H,
r

S
V
D
m
s
.
m
m
,
,
Figure 11: Total enthalpy and entropy source elements.
given by
m
s
= V
m
A
Q = V
m
Br r = V
m
AB/2
D =
1
2
V
2
m
A
D
D
p
=
1
2
W
2
Bc r c
dp
16
where A = 2rr is the area of the annular captured streamtube, and A
D
is the annular objects
prole drag area. The ow jumps produced by the object are given by
m
s
H = Q (107)
m
s
S = V
m
D + W D
p
(108)
There will also be some additional H and S if the object delivers heat, as in the case of a cylinder
head or radiator. But the eect of this heating on the oweld is not likely to be signicant in
typical applications, and will not be considered here. Substituting for m
s
, D, D
p
, and Q in (107)
and (108), gives the nal expressions for the convected quantities needed to evaluate the Q
1
source
term using (88).
H =
B
2
(109)
S =
1
2
V
2
m
A
D
A
+
1
2
W
3
V
m
Bc
2r
c
dp
(110)
14 Blade Geometry
14.1 Parameterization
The blade surface is dened using two parameters , as shown in Figure 12. The range of the
chordwise parameter is assumed to be
0 1
from leading edge to trailing edge. The range of the radial parameter is arbitrary. Two simple
and convenient denitions are
= r or = (r r
root
)/(r
tip
r
root
)
The blade airfoil sections are dened on constant-r (and constant-) cylinders. A third parameter
is used to dene the airfoil shape on this cylinder. The lines are helixes with a helix angle of
from the circumferential direction. The lines are the orthogonal helixes with a helix angle of
90

(not shown).
Both and have the same normalization with the local chord c(), so that a physical arc length
s on the unwrapped airfoil cylinder is
s = c
_

2
+
2
The increments do not have to be small.
14.2 Blade geometry specication
The blade surface shape is specied by the following quantities:
r() radius

() skew
x() rake
c() chord
() twist

() thread position
17

TE
LE
()
() r
()
() x

()
() c
=0
=1
=
helix
thread

x
z
y
n
Figure 12: Blade geometry parameterization.
There will also be an airfoil shape and corresponding camber surface dened in the - plane at
each station. This approach closely follows ship propeller denition techniques as outlined by
Kerwin [4].
It is convenient to also work in the cartesian coordinates (x, y, z) shown in Figure 12. Any given
point (, , ) on the blade has the following (x, y, z) location.
x(,,) = x + c
_
(

) sin cos

(111)
y(,,) = r cos
_

+
c
r
_
(

) cos + sin

_
(112)
z(,,) = r sin
_

+
c
r
_
(

) cos + sin

_
(113)
Dening

(; ) as the set of airfoil camberlines for all radii, the corresponding cartesian-coordinate
camber surface is
x(,) = x + c
_
(

) sin

cos
_
(114)
y(,) = r cos
_

+
c
r
_
(

) cos +

sin
_
_
(115)
z(,) = r sin
_

+
c
r
_
(

) cos +

sin
_
_
(116)
These equations can be used to construct the cartesian coordinates for any other set of (; )
curves, such as the upper and lower blade surfaces.
14.3 Normal-vector construction
For constructing a ow-tangency condition on the blade surface, it is typically necessary to dene a
vector n normal to the camber surface at some chosen chordwise location
n
. For the vortex lattice
18
method, this would be at the
n
= 3/4 position. We rst dene cartesian vectors u

and u

which
lie on the camber surface at the chosen
n
location, and point in the and directions.
u

() =
_
x

+
y

+
z

=n
(117)
u

() =
_
x

+
y

+
z

=n
(118)
The cartesian normal unit vector is then readily computed via a cross product and normalization.
n() = u

(119)
n() =
n
|n|
(120)
15 Overall Solution Strategy
The solution procedure for a typical rotor analysis problem is described below.
15.1 Inputs
Rotor blade geometry: c(r), (r)
Rotor aero properties:
o
(r), m
2D
(r), c
d
(r)
Duct and hub geometry: x(s), r(s)
Drag object properties: r, z, A
D
Freestream speed: V

Rotor rotational speed:

15.2 Initial Grid via Zero-thrust, Unit-V

solution
1. Dene paneling on duct and hub
2. Evaluate for panel control points: n
i
, a
x
ij
, a
r
ij
3. Assemble a
ij
, augment with Kutta condition
4. LU-Factor a
ij
a
1
ij
5. Back-substitute for
o
i
6. Set initial streamfunction grid, assign grid values
7. Evaluate v
x o
i
, v
r o
i
at grid boundaries
8. Use v
x o
i
, v
r o
i
to compute and assign grid
9. Set Q
1
= 0
10. Relax grid with SLOR, using v
x o
i
, v
r o
i
for Neumann boundary conditions
19
15.3 Rotor-On Solution
Iteration setup:
1. Dene paneling on grid streamsurfaces, dene source drag panels
2. Evaluate at body panels, rotor blade stations, source drag panels: a
x
ij
, a
r
ij
, a
ij
, b
x
ij
, b
r
ij
,
b
ij
3. Set initial guess for
k
4. Set corresponding

and

H elds
5. Set initial guess for
i
using (41) and (42)
6. Set initial
i
= 0
One Newton iteration:
1. Using current
k
,
i
, evaluate
i
, v
x
i
, v
r
i
, v

i
,

V
i
and derivatives w.r.t.
k
,
i
2. Evaluate residuals of equations (75), (73), (74), and derivatives w.r.t.
k
,
i
3. Solve Newton system for
k
,
i
4. Update
k
,
i
Streamline grid update (after Newton convergence)
1. Evaluate

V at grid cells, using grid derivatives
2. Evaluate

,

H,

S for all grid cells downstream of blade stations and drag-producing objects
(source panels)
3. Evaluate Q
1
for all grid cells
4. Relax grid via SLOR
5. Begin again at Iteration setup
References
[1] A. Betz. Airscrews with minimum energy loss. Report, Kaiser Wilhelm Institute for Flow
Research, 1919.
[2] H. Glauert. Elements of Airfoil and Airscrew Theory. Cambridge University Press, Cambridge,
1937.
[3] J.F. Thompson, F.C. Thames, and C.W. Mastin. Automatic numerical generation of body-
tted curvilinear coordinate system for eld containing any number of arbitrary two-dimensional
bodies. Journal of Computational Physics, 15:299319, 1974.
[4] J.E. Kerwin. Marine propellers. Annual Reviews of Fluid Mechanics, 18:367403, 1986.
20