Aircraft Design 3 (2000) 129}149

Aerodynamic optimisation of wings in multi-engined tractor

propeller arrangements
L.L.M. Veldhuis*, P.M. Heyma
Aerodynamics Laboratory, Department of Aerospace Engineering, Delft University of Technology,
Kluyverweg 1, 2629 HS Delft, Netherlands

Abstract

Aerodynamic optimisation of wings in multi-engined tractor propeller arrangements is discussed and

analysed with a fast calculation based on a Tre!tz-plane analysis where the conservation laws of mass,
momentum and energy are ful"lled in a control volume surrounding the con"guration. The paper discusses
the formulation of the optimisation algorithm based on augmented Lagrange integrals. The e!ect of viscous
e!ects is incorporated in the calculation process. The method was implemented in a computer program
which enables the user to "nd the optimum lift distribution for minimum drag for any tractor propeller/wing
arrangement of arbitrary shape. As input for the slipstream data the user can either select input of
experimental data or generate arti"cial data using a simple slipstream model based on the well-known blade
element theory with Prandtl tip loss factor. Some numerical studies show that optimisation of a modern
medium speed turboprop aircraft leads to performance increase by adapting the wing shape.  2000
Elsevier Science Ltd. All rights reserved.

1. Introduction

The era of modern turboprop, which started in the 1980s with the Dash-8, BAe ATP and Fokker
50 and in the 1990s with Dornier 328, Saab 2000 and BAe Jetstream 41, initially showed successful
programs. During this period, however, the market transformed by the arrival of the new-
generation regional jet. Although the turbofan powered aircraft like the Fokker 70/100 , Bombar-
dier CRJ and BAe 146 gained much popularity it is doubtful whether any regional aircraft has so
far shown to be really pro"table.

* Corresponding author. Tel.: 31-15-278-2009.

E-mail address: l.l.m.veldhuis@lr.tudelft.nl (L.L.M. Veldhuis).

1369-8869/00/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved.
PII: S 1 3 6 9 - 8 8 6 9 ( 0 0 ) 0 0 0 1 0 - 0
130 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149

Nomenclature

a axial velocity factor in the propeller slipstream x, y, z coordinates in streamwise, spanwise and
6
A wing aspect ratio vertical direction, respectively
b wing span a geometrical angle of attack

c local chord a zero lift angle of attack

C intersection curve in the Tre!tz plane a induced angle of attack
2
C local drag coe$cient o air density

C local lift coe$cient C bound circulation strength

C total lift coe$cient of the con"guration 1 local dihedral angle
*
C normal force coe$cient s airfoil camber
,
C propeller torque coe$cient g dimensionless spanwise coordinate
/
D drag force, propeller diameter u propeller rotational speed
e span e$ciency factor
h speci"c enthalpy
J propeller advance ratio Subscripts
n propeller speed, rev. per second b body
n unit normal vector c.l. centre line
N number of propellers d downstream

p pressure i induced
P power k propeller index
Q torque n normal direction
QQ heat added to the control volume opt optimum
R propeller radius p propeller
S wing area ref reference value
¹ thrust force, temperature s slipstream
u, v, w x, y and z component of the velocity vector tot total value of the complete con"guration
v axial velocity increase inside the slipstream u upstream
v tangential velocity increase in the slipstream v viscous/vortex sheet

< undisturbed #ow velocity w wing

= Q machine work performed by the control x, y, z x, y and z components, respectively
volume

Jet aircraft indeed show advantages in some directions, it is however justi"ed to acknowledge
that there will always be a fair number of short regional routes on which propeller-driven
aircraft will have a cost advantage. In this respect an indication for a typical trip length may
be found in US domestic #ight data were an average regional airline trip of 230 nm is found.
As operators rediscover the advantages of turboprops on shorter routes the forecast
(Fig. 1) given by Butterworth}Hayes [1] , stating that the propeller-driven aircraft are likely to
hold at least 40% share of the market expressed in dollar value in future, may well become
reality.
In this changing market it is worthwhile to investigate the technology regarding propeller
propulsion versus jet propulsion to see whether new concepts are within reach to optimise current
designs. In the following paragraphs some important aspects of propeller}wing interaction will be
discussed as part of this propulsion technology review.
For multi-engined propeller-powered aircraft one of the important points of concern is the
interaction between the propeller slipstream and the wing. Modern aircraft concepts, like the
 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149 131

Fig. 1. History and forecast of market share of propeller aircraft versus jet aircraft [1].

European FLA project, exhibit high disk loading and an increased number of (swept) blades to
enable high cruising speed.
The generation of strong swirl velocities in the slipstream, however, generates a considerable
deformation of the lift distribution, which has an impact on the aerodynamic behaviour and
performance of the wing.
The propulsion}airframe installation problem and the experimental and numerical techniques
to analyse this topic have been EC-funded since 1990 through programs like Gemini I & II,
APIAN, DUPRIN I & II and ENIFAIR. Within these research projects speci"c attention was paid
to analysis of current concepts rather than investigation of possible new design strategies.
From earlier investigations it is known that both the position of the powerplant with respect
to the wing and the propeller angle of attack play an important role. Carefully designed
con"gurations may reveal some performance bene"ts when the propeller and the wing are closely
coupled.
Although the rotational kinetic energy in the wake of the conventional single-rotating propul-
sion systems is usually lost and does not contribute to the thrust, analyses by Kroo [2], Miranda
[3] and Veldhuis [4] have indicated that signi"cant wing drag reduction can be obtained for
propeller/wing interaction. Rather than manipulate wing geometry to approach two-dimensional
#ow, it would seem logical to use some energy source for the task of directing the #ow so that lower
induced drag is produced. Although the rotational component in the slipstream has always been
considered to represent lost energy it is in fact available for amplifying or attenuating the wing
vortex with a possible reduction of induced drag. In fact the wing acts as a stator vane that recovers
some of the swirl loss caused by the propeller.
To derive a better understanding of the interactive #ows causing the slipstream/wing interference
an optimisation program was developed in which the slipstream velocity e!ects are incorporated.
132 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149

The phenomena that play a role with respect to the interaction between the propeller and the
wing have been described by Kroo [2], Heyma [5] and Veldhuis [4] and many others. In the
subsequent paragraphs a description will be given of the optimisation program.

2. Theoretical approach

2.1. Introduction

In the last decade CFD programs have become available in which the interaction of propeller
and wing can be analysed based on a solution of the Navier}Stokes equations [6,7]. Although the
results of these codes contribute without doubt to a better understanding of the problem of
propulsion integration they generally lack one important feature : an optimisation algorithm.
Optimisation programs are important in the preliminary design phase when the main design
parameters like the location of the propulsion system have to be chosen. Since these methods
should be fast enough to quickly access the consequences of changes in the global shape of the
con"guration they are often based on elementary momentum considerations and relatively simple
numerical schemes. In the following paragraphs the theoretical approach which is the basis for
the optimisation program pwopt2 is described.

2.2. Drag minimization

In our analysis the so-called Tre!tz plane method is applied to the propeller wing con"guration.
The method, based on Munk's stagger theorem [8], was originally applied to wing tail
con"gurations but may be extended to propeller}wing con"gurations as well, as indicated by
Kroo [2].
Initially the in#uence of the wing on the propeller is neglected which implies that the propeller is
far upstream or downstream of the wing. Hence, the total downwash at in"nity, used in the
optimisation process, is a superposition of the downwash due to all lifting surfaces and the
downwash generated by the propeller at in"nity.
An inverse numerical optimisation is applied here, based on a relatively simple #ow solver which
utilises a lifting line theory to model the lifting surfaces combined with a Tre!tz plane analysis. The
bound circulation is modelled using a quadratic distribution function similar to the method of
Kuhlman [9].
Examples of comparable optimisation codes were already described by Miranda and Brennan
[3], Kroo [2] and Veldhuis [4]. These approaches di!er from the current analysis in the way the lift
distribution is modelled. The authors mentioned above, for example, proposed a technique where
the lift distribution is given as a Fourier sine series combined with axial and tangential velocities
induced by the propellers.

2.3. Trewtz plane analysis

The drag can be calculated by considering the momentum balance of a control surface S which
surrounds the con"guration (Fig. 2).
 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149 133

Fig. 2. Layout of the control volume and the control surface as used in the Tre!tz plane analysis.

A detailed mathematical description of the lift and the drag analysis is beyond the scope of this
paper. Therefore, only the most important calculation and derivation results will be summarized
here. A more detailed analysis is presented by Van den Dam [10], Middel [11], Heyma [5] and
Veldhuis [4]. In the following analysis the e!ects of compressibility will be neglected since most
propeller aircraft operate in the (low) subsonic speed regime.
For a steady, incompressible, inviscid #ow the conservation equations of mass, momentum and
energy, applied to control volume < surrounded by control surface S become

 o(V ) n) dS"0,
1

 oV(V ) n) dS"!
 (pn) dS#F, (1)
1 1

 oh(V ) n) dS"QQ !=Q .

1
The second equation states that the time rate of change of linear momentum is equal the sum
of the forces acting on the control volume. Here F is the force acting on the #uid inside the
control volume (generally zero), o is the density, V is the velocity vector, n is the outward pointing
normal vector to the control surface and h is the stagnation enthalpy. Since for the moment the
viscous forces acting on a body with surface S are neglected the only resulting force in x-direction


will be the induced drag, D . The control surface S consists of several sub-surfaces, as indicated

in Fig. 2: S"S #S #S #S #S #S #S . We now can apply some known boundary
  
  
conditions
E On S : V ' n"0,


E On S : V ' n"0 and n "0,
 V
E On S : V ' n"0 and p is continuous (vortex sheet fully developed),

E On S : V ' n"0 and p is continuous (stream surface).

134 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149

Hence the surface integrals (1) can be reduced to

0"
 o(V ) n) dS,
1 >1

D "!

(pn #ou(V ) n) dS,
V
1 >1 >1
  

QQ !=Q "
 oh(V ) n) dS. (2)
1 >1 >1.
The propeller thrust is de"ned by

¹ "!
  (pn #ou(V ) n)) dS.
V
(3)
1.
Thus, the second integral of Eq. (2) may be rewritten as

D "!

(pn #ou(V ) n)) dS#¹.
V
(4)
1 >1
When there is an active propeller with surface S there will be an energy increase which may be
.
regarded as negative dissipation. Introducing the velocity potential U the induced drag can be
written as



dU
D "! pn #o (
U ) n) dS#¹. (5)
V dx
1 >1
Introducing the velocity potential U"; x#u , Eq. (5) can be further simpli"ed based on the
 V
so-called Tre!tz plane analysis. In the Tre!tz plane the #ow is fully developed which means that no
velocity gradient and pressure gradient exist in streamwise direction. The "nal result becomes


o
D" +!u#u#u, dS #¹ (6)
2 V W X 
1
which can be further simpli"ed accepting the fact that u"0 in the Tre!tz plane. It should be
V
noted that in this analysis , based on inviscid theory, the velocity increase or defect that normally
occurs in the vortex cores is neglected.


o
D" +u#u, dS#¹. (7)
2 W X
12
 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149 135

Now exclude the change in the thrust force from Eq. (7) since the interest is focused on the
optimisation of the induced drag for a given propeller condition ("xed thrust force and torque).
This simpli"cation is allowed as long as the thrust force and the propeller torque do not change
signi"cantly with adaptation in the geometry of the wing behind the propeller. In general, this is the
case for all practical designs where the distance between propeller and wing is not small.
The perturbation potential u is built up of a disturbance due to the wing/body (index b) and the
potential due to the propeller (index p). Neglecting Ru /Rx (the #ow is fully `developeda), the


induced drag (without the thrust contribution) becomes

 
o
D "  (uW #uX ) dS#o (u W u W #u X u X ) dS
2




12 12


o
#  (uW #uX ) dS. (8)
2  
12
Using Green's identity the integral over S is transformed into an integral over the curve(s)
C formed by the intersection of the wake and the Tre!tz plane.
2
The "rst integral of Eq. (8) is rewritten in the form




   

Ru  Ru  Ru Ru Ru
# dS" u dC! u # dS (9)
Ry Rz Rn Ry Rz
!2
12 12
while the second integral becomes

  

Rv Rw
(V )
u ) dS" u(V ) n) dC! u  #  dS, (10)

 Ry Rz
!2
12 12
where V "(Ru /Rx, Ru /Ry, Ru /Rz) is the velocity vector induced by the propeller.
   
At the Tre!tz plane both the second terms on the right-hand side of Eqs. (9) and (10) vanish.
The jump in the velocity across the wake is equivalent to the local circulation. Hence with
*u"C and Ru/Rn"v , the induced drag in the Tre!tz plane becomes
L

 
o o
D "  C (v #2(v ) n #w ) n )) dC#  (v #w) dS. (11)
2

 W  X 2  
!2
1.
Three terms can be distinguished


o
A"  C v
dC (12)
2

! 2

which is the induced drag without the presence of a slipstream

A "o
.   ! 2
C (v n #w n ) dC"o

 W  X   ! 2
C v  dC


(13)
136 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149

which is the propeller-wing interaction e!ect. And "nally


o
A"  (v #w) dS (14)
 2  
1
which is the `swirl lossa of the propeller.
Now Eq. (11) can be written as a `draga coe$cient


1 ,
C " C(v #2(v n #w n )) dC# CH (15)
" < S   W  X I
  !2 I
with


1
CH" (v#w) dS. (16)
I < S  
  1
Without further proof it is stated that in an analogue way the following relation can be derived for
the lift coe$cient:


2
C " C ((< #u )cos f) dC, (17)
* < S
 
  !2
where u is the axial velocity increase due to the propeller and f is the wing dihedral angle.

Since we will only deal with the C-distribution of the `bodya the subscript b will be omitted in the
following analysis.

2.4. Viscous drag

An optimisation in which only the induced drag is minimised may result in a lift distribution that
di!ers considerably from the one in which the total drag is minimised. The wing pro"le drag
contribution can found from


1
C " C  (v #u )c dS. (18)
" < S   
  !2
The pro"le drag coe$cient can be determined either through the input of two-dimensional airfoil
characteristics or alternatively by using empirical relationships.
The pro"le drag coe$cient is estimated using a simple quadratic relation
C  "f (C )"C  #f C #f C (19)
      
with the local lift coe$cient being


1 2 (< #u )cos f C
C"   C " (20)
 c < c

the viscous drag becomes


1
C " (< #u )c (a #a C#a C) dS. (21)
" < S     
  !2
 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149 137

Expressions (15), (17) and (21) may now be employed to "nd an optimum load distribution
(C"C(y)) in which for a given lift coe$cient and propeller setting the drag coe$cient is minimised.

2.5. Optimisation formulation

The minimisation technique that is used in the program is based on the augmented lagrange
integral method [10,11].
The optimisation algorithm minimises the object function given by
I"C #C  #j (C !C  ) (22)
" " * *
which includes the constraint for the lift coe$cient C  .
*
All parts of this object function can be written in terms of the circulation distribution combined
with the unknown Lagrange multipliers j.
To solve the minimisation problem the projection of the wing geometry onto the Tre!tz plane,
formed by curve C is divided into linear elements (panels). Thus, a system of equations can be set
2
up which should satisfy the optimisation constraints.
This system of linear equations is solved for the unknown values of the local circulation strength
with the method of variational calculus.
Summation over the complete con"guration then yields the downwash at all spanwise stations.
From these values the total lift coe$cient and the total drag coe$cients are found.

2.6. Generation of the optimised geometry

After the discussion about the acquisition of the optimum lift distribution for a given
propeller/wing combination the logical step is to de"ne how the required distribution can be
obtained.
To get some understanding of the magnitude of the changes to an initial wing design, a simple
analysis is suggested here.
From Eq. (20) we may write for the spanwise position y
C
C (y)(a (y)!a (y)!a (y))" (y). (23)
?   c
Since the induced angle of attack a is "xed by the given vorticity distribution along the span, the

user may specify one of the following parameters to ful"l the requirement of Eq. (23)
E the two-dimensional lift curve slope, C ,
?
E the geometrical airfoil angle of attack which is determined by the twist distribution, a ,

E the zero lift angle of attack of the airfoil, a ,

E the chord of the airfoil.
It is clear that the wing design of an optimum wing with installed propeller is complicated due to
the di!erence between the optimum lift distribution and the `normala elliptic-like distribution,
especially at transonic speeds. For high-speed aircraft the pro"le shape should be modi"ed to
prevent unwanted viscous and compressibility e!ects in the part of the wing that is immersed in the
slipstream.
138 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149

The leading edge and the trailing edge of the wing are normally kept straight. However, to attain
the optimum wing shape for example the twist distribution or the camber distribution should be
adapted.
Due to deformation of the wing shape as a result of wing twist, a combination of twist adaptation
and variation of section shape will in practice be employed. In subsequent calculation examples
only the e!ect of wing twist will be discussed.
The solution converges to the theoretical optimum value of 1.0 when more panels are used. With
a model containing 25 panels the induced drag is already computed with an accuracy of 0.2%.

2.7. Generation of propeller slipstream data

To "nd the correct interference e!ect that the slipstream imposes on the wing the axial and the
tangential #ow velocities, which together form the velocity vector (u , v , w ), must be incorporated
  
as stated in the equations for the lift and the drag coe$cient. In the current program it is possible to
import experimental slipstream data or calculate the propeller-induced velocities with a simpli"ed
slipstream model. In the latter case (u , v , w ) is calculated with the vortex tube model.
  
In this model, the vorticity in the slipstream originally concentrated on a "nite number of helical
vortex sheets, is replaced by two superimposed continuous distributions of vorticity: axial and
tangential vorticity.
Using a Biot}Savart calculation algorithm the propeller-induced velocity vector is then found in
the Tre!tz plane. In case the propeller is put at a positive angle of attack to the #ow, the induced
velocity distribution should be given according to the increased values at the down-going blade
side and vice versa. If only data are available for the setting a "03 at the position of the propeller

plane instead of the wing location, the program will calculate the propwash data according to the
following scheme.
The axial velocity increase at a certain distance x behind the propeller is given by a "y /< .
 V  
According to a velocity potential formulation as described by Smelt and Davies [12], that is based
on an analogy with magnetic shell theory, the axial velocity increase, as a function of x, can be
estimated by


1
a "a 1# , (24)
V ((R/x )#1

where a"(u /< ), is the axial velocity factor at the position of the propeller plane (x "0).
  
The #ow velocity in the direction of the thrust line is given by < cos a whereas the velocity
 
perpendicular to this line is < sin a . The slipstream centreline will therefore deviate from the
 
thrust line, by an angle a . given by
 
< sin a tan a
tan a "   "  . (25)
  < (1#a ) cos a (1#a )
 V  V
To account for the `upwash e!ecta given by a the local swirl velocity at the wing position is
 
corrected through
a  "tan (a #a !a #arctan (a (y))), (26)
     
where a (y)"v (y)/< , is the uncorrected swirl velocity factor.
  
 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149 139

2.8. Propeller incidence relative to the wing

When the direction of the air#ow is not perpendicular to the propeller disk, the blades will be
subject to alternating loads with a period equal to the time of revolution of the propeller. The
correct setting for minimum alternating loads on the blades depends typically on the streamwise
position of the propeller with reference to the wing. It should be noted that for closely coupled
propeller}wing con"gurations the wing-induced angle of attack is substantial. Therefore, the wing
e!ects on the propeller are taken into account making use of a simple lifting line mode of the wing.
In this #ow model the e!ects of the nacelles are neglected although it is clear that they may
introduce signi"cant #ow non-uniformity.
The propeller at angle of attack will, in addition to moments, produce a thrust and a normal
force. The true propeller tilt down-e!ect on the performance of the propeller/wing con"guration is
calculated through usage of either an estimated propeller normal force gradient from ESDU [13]
or alternatively experimental values when they are available.
Conversion of the normal force coe$cient from propeller to wing gives
dC  dC p D dC 
, " , " , g. (27)
da da 4 S da
   
The primary (direct) contributions of the propeller to the lift and the drag of the propeller}wing
con"guration are given by
*C "(C  cos (a )#¹ sin (a ))g,
* ,    (28)
*C "(C  sin (a )!¹ cos (a ))g,
" ,   
where, the propeller geometrical angle a is de"ned by

a "a #a . (29)
  
In the following paragraph the span e$ciency factor e, de"ned as
C
e" * (30)
pAC
"

will be used to analyse the performance of the propeller}wing con"guration.

3. Calculation results

3.1. Optimisation of a tractor propeller wing conxguration

To investigate the propeller slipstream e!ect on the wing performance a con"guration similar to
that of the Fokker 50 was selected (Fig. 3).
Two typical cases were investigated: the so-called `high-speed casea and the `low-speed casea
(Table 1).The slipstream data for these cases, which were taken from Janssen [14], are symmetrical
with respect to the propeller axis; i.e. the propeller is positioned perpendicular to the local #ow.
Hence, the up#ow in front of the wing is compensated for.
140 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149

Fig. 3. Fokker F50 con"guration investigated in the optimisation process.

First the e!ect of the rotation direction on the optimal lift distribution was investigated. In fact
three di!erent cases were examined:
E inboard up rotating propellers,
E outboard up rotating propellers,
E co-rotating propellers.
The optimum bound circulation distribution given in Fig. 4 di!ers signi"cantly from the
elliptical clean wing con"guration due to the action of the propeller. The local circulation strength
 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149 141

Table 1
Flight conditions for the low-speed and the high-speed case of the Fokker 50

Parameter High speed Low speed

Density, q (kg/m) 0.934 0.872

Temperature, T (K) 264 261
True airspeed, TAS (m/s) 119.5 75
Revolutions per second (Hz) 20 20
Blade angle (deg) 37.61 33.78
Advance ratio 1.633 1.025
C ("ctitious) 0.6 1.2
*
¹ 0.10 0.60


Fig. 4. Optimal spanwise bound circulation distribution for di!erent rotation directions; C  "0.6.
*

shows a jump going from the inboard to the outboard side of the wing. This could be expected since
the magnitude of the lift vector tilted forward is bigger than the one tilted backward. The propeller
produces both positive and negative contributions to the drag induced at the spanwise locations.
The negative contribution at the up going blade (UBS) side exceeds the positive contribution at the
downgoing blade side (DBS) resulting in the reduction of the overall induced drag.
As described in Eq. (23) the sectional shape can now be changed to realize the optimum lift
distribution. In Fig. 5 the optimum twist distribution is presented when the local chord and airfoil
shape are left unaltered. The resulting maximum twist angle is about 43. Changing the local chord
with unchanged twist distribution produces unrealistic chord values hence twist and camber
adaptation seem to be the only acceptable design variables.
For the low-speed case the magnitude of the slipstream velocities has increased. This makes the
results somewhat less accurate since in the Tre!tz plane analysis the perturbation velocities were
142 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149

Fig. 5. Optimal twist distribution for di!erent rotations directions at C  "0.6.
*

assumed to be small. Of course, the shape of the optimum circulation di!ers from the high-speed
case. This signi"es that the wing can only be optimised for a speci"ed #ight condition (generally
cruise). Due to the higher slipstream velocities the wing-induced drag in the low-speed case with
inboard up rotation 60 counts lower than for outboard up rotation. For the high-speed case this
in#uence is 10 counts in favour of the inboard up rotation case.

3.2. Propeller at angle of attack

As indicated by experiments on a generic propeller/wing model, performed earlier by van Es

[15], the wing performance may signi"cantly be in#uenced by the propeller angle of attack with
respect to the wing reference chord line. The so-called PTD-con"guration (propeller-tilt-down)
used by Veldhuis [4] suggests a reduction of overall wing-induced drag. This e!ect of the propeller
angle of attack was investigated again for the given F50-like con"guration. The normal force
gradient of the six-bladed propeller was taken from windtunnel tests on a generic propeller-wing
model (APERT-JR01), data of which were published by Kusomo et al. [16].
To express the performance of the propeller}wing con"guration incorporating the direct forces
acting on the propeller we use
¹  "!(C #*C  #*C ) (31)
 " " "
with
*C "(C ) -(C ) ,
" "   "   
2C *C #*C
" * * * , (32)
pAe
where index p denotes the e!ect due to the propeller.
 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149 143

Fig. 6. E!ect of propeller angle of attack on the wing span e$ciency factor e.

Here the pro"le drag is ignored in the optimisation process. Fig. 6 shows the e!ect of propeller
angle of attack on the wing e$ciency. The highest `e!ective thrusta is found at relatively large
negative angles of attack. Apparently the bene"ts of the wing drag reduction due to the presence of
the propeller increase with negative propeller-angle of attack, a . It should be remarked that, in

contrast with what one would expect, the reduction of wing lift and e!ective thrust, through direct
forces acting on the propeller, are smaller than the bene"ts in the sense of reduction of the overall
induced drag.
The phenomenon sketched above can be explained as follows. When a becomes more

negative the local blade angle of attack of the downgoing blade increases while that of the
upgoing blade becomes smaller. This results in an asymmetrical slipstream. Since the lift vector
at the side of higher local lift coe$cient is tilted forward and the lift vector at the other side is
tilted backward a net drag reduction remains. Accordingly, a swirl velocity distribution with
lower values at the DBS and higher values at the UBS should increase the wing e$ciency
factor.
Besides this, the whole slipstream will be placed at an angle which attenuates the forward
tilting process (resulting in increased leading edge suction) at both sides of the nacelle. Again
the con"gurations with inboard up rotating propellers demonstrate a superior performance
compared to the outboard up rotating cases. In Fig. 7 the optimum lift distribution is presented
for several propeller angles of attack. Apparently the propeller angle of attack strongly in#uences
the form of the optimum lift distribution. An inevitable disadvantage of a con"guration with
negative propeller angle of attack (PTD) is of course the emergence of cyclic blade loading. The
resulting variation in blade stresses with azimuth position and the possible increase in noise
level are problems to be addressed before a PTD-con"guration can be practically implemented
(Fig. 8).
144 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149

Fig. 7. Optimum spanwise bound circulation distribution for di!erent propeller angles of attack.

Fig. 8. Propeller angle of attack e!ect on the `e!ective thrust coe$cienta of the con"guration.

3.3. Ewect of proxle drag

To investigate the e!ect of the pro"le drag on the optimum wing loading distribution NACA
64 A015 airfoil were used. At low lift coe$cients, between !0.3 and 0.3, the C ? -curve is rather
 
 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149 145

Table 2
In#uence of the pro"le drag distribution on the wing drag coe$cient of a low aspect ratio wing at C "0.9 [15]
*
Con"guration Minimisation C C  C 
" " "
Inb. up i 0.043711 0.010669 0.054380
Inb. up i#v 0.043732 0.010642 0.054374
Outb. up i 0.046287 0.010531 0.056818
Outb. up i#v 0.046299 0.010522 0.056821

i"induced drag; v"viscous drag.

Fig. 9. Higher-order panel model of the initial geometry without nacelle.

#at. This means that an optimisation with a constrained value in that lift range will hardly result in
a changed optimal lift distribution.
In Table 2 the optimisation result with and without the pro"le drag contribution of a simple
rectangular wing of low aspect ratio (A"5.3) is given. This wing was used during experiments by
van Es [15]. Here, the constrained value of the design lift coe$cient is equal to 0.9 to demonstrate
the in#uence of the pro"le drag into the optimisation results. When both drag contributors are
used in the optimisation process, there is only a minor change in the lift distribution. This agrees
with the result of Middel [11].
The total values of the drag coe$cients of the wing are given in Table 2 for both the inboard-up
as well as the outboard-up rotating propeller. Although the di!erences are small for the current test
case, it can be seen that for the inboard-up rotating propeller the induced drag is greater when the
sum of the induced drag and the pro"le drag is minimised than when the induced drag alone is
optimised. However, the pro"le drag has become smaller towards a more optimised value. The
total drag, as a result of the optimisation of the sum of the induced and viscous drag, is less than
it would have been if only the induced drag was minimised and the viscous drag was added
afterwards.
146 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149

3.4. Comparison of optimisation and analysis

The performance of the optimised propeller}wing con"guration was further analysed with
a higher-order panel method called FASD [17].

Fig. 10. Panel model with pressure distribution at the upper side of the initial geometry under in#uence of an inboard-up
rotating propeller at M"0.35.

Fig. 11. Panel model of the optimised geometry of the wing con"guration at C  "0.6. Optimised for inboard-up
*
rotation.
 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149 147

Here only the contra-rotating propeller case was calculated with the same slipstream input as
used in the optimisation process.
The high-speed case was calculated as M"0.35 and C "0.6 while the low-speed case was run
*
at M"0.22 and C "1.2. In Figs. 9 and 10 the initial panel geometry is presented. When the wing
*
geometry is adapted, as presented in Fig. 11 according to the optimum twist distribution (Fig. 12b)
the resulting induced drag is indeed reduced as can be seen in Table 3. The e!ect for the
symmetrical slipstream that was used here is however very small.

Fig. 12. Twist distribution of the initial and optimised wing at C  "0.6; high-speed case; (a) prop-o!, (b) inboard-up
*
rotation.

Table 3
E!ect of the optimisation on the wing drag coe$cients at C  "0.6, as calculated with FASD. High-speed case
*
Initial design
Con"guration C C
* "
No propeller 0.598 0.0101
Inboard-up 0.599 0.0088
Outboard-up 0.600 0.0093

Optimised design
No propeller 0.599 0.0091
Inboard-up 0.599 0.0078
Outboard-up 0.600 0.0083
148 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149

4. Conclusions

The analysis of the tractor propeller con"guration that was optimised leads to the following
conclusions:

E A simple and fast optimisation program for the optimisation of propeller}wing con"gurations in
the preliminary design phase was developed which enables investigations of the e!ects of:
propeller position, slipstream velocity distribution and adaptations to conventional wing de-
signs.
E Calculations show that the propeller}wing con"guration may bene"t from a relatively large
negative propeller angle of attack (PTD-con"guration). This e!ect is mainly due to the asym-
metry in the slipstream velocity distribution that is introduced. However, in an analysis
including viscous e!ects in high subsonic #ow there will be a maximum in the attainable e!ective
e$ciency since the drag rise with Mach number and separation e!ects may nullify the local
leading edge suction.
E It is shown that the inboard-up rotating propeller exhibits a superior performance compared to
the conventional co-rotating or the outboard-up rotating case. Therefore, further investigation
of this con"guration to arrive at a practical implementation is recommended.

References

[1] Butterworth-Hayes P. Aerospace America, January 1999.

[2] Kroo I. Propeller/wing integration for minimum induced loss. Journal of Aircraft 1986;23(7):561}5.
[3] Miranda LR, Brennan JE. Aerodynamic e!ects of wing tip mounted propellers and turbines. AIAA 86-1802, 1986.
p. 221}8.
[4] Veldhuis LLM. Experimental analysis of tractor propeller e!ects on a low aspect ratio semi-span wing. Second
Paci"c International Conference on Aerospace Science and Technology & the Sixth Australian Aeronautical
Conference, vol. 2, Melbourne, 20}23 March 1995. p. 491}8.
[5] Heyma PM. Analysis and optimization of a tractor propeller}wing con"guration. TWAIO-thesis, Dept. of
Aerospace Engineering, Delft University of Technology, June 1996.
[6] LoK tstedt P. Accuracy of a propeller model in inviscid #ow. Journal of Aircraft 1995;32(6).
[7] Colin P, Moreux V, Barillier A. Numerical study of high speed propeller engine integration on transport aircraft.
ICAS-96-4.10.1, Sorrento Italy, September 1996.
[8] Munk MM. The minimum induced drag of aerofoils. NACA Report 121, 1921.
[9] Kuhlman JM. Higher order far"eld drag minimisation for a subcritical wing design code. Journal of Aircraft,
1980;17(9).
[10] Dam RF. van den. Samid, an interactive system for aircraft drag minimization; Mathematical models and methods.
NLR TR 880714, 1998.
[11] Middel J. Development of a computer aided toolbox for aerodynamic design of aircraft at subcritical conditions
with application to three-surface and canard aircraft. Dissertation, Delft University Press, 1992.
[12] Smelt R, Davies H. Estimation of increase in lift due to a slipstream. ARC R&M 1788, 1937.
[13] ESDU. In-plane forces and moments of installed propellers at low speeds, ESDU 89047, ISBN 0 085679 721-9,
ESDU International plc, 1989.
[14] Janssen M. An evaluation of methods for the calculation of the #ow around propeller at angle of attack. MSc thesis,
Dept. of Aerospace Engineering, Delft University of Technology, 1991.
 L.L.M. Veldhuis, P.M. Heyma / Aircraft Design 3 (2000) 129}149 149

[15] Es GW. van. Experimental investigation of the e!ect of propeller angle of attack and position on the aerodynamic
characteristics of a tractor}propeller wing combination. Thesis, Dept. of Aerospace Engineering, Delft University of
Technology, 1993.
[16] Prijo Kusumo J, Gayus LGM, Custers LGM, de Haij LH, Veldhuis LLM. Experimental investigation on propeller
slipstream e!ects on a swept wing at low speeds. ICAS-96-4.10.2, Sorrento Italy, September. 1996.
[17] Hoeijmakers HWM, Bosse S. Private communications.