Computers & Fluids Vol. 27, No 5-6, pp.

651-661, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0045-7930/98 $19.00 + 0.00
AN INTERACTIVE BOUNDARY-LAYER METHOD
FOR MULTIELEMENT AIRFOILS
TUNCER CEBECI, ERIC BESNARD and HSUN H. CHEN
Aerospace Engineering Department
California State University, Long Beach
1250 Bellflower Blvd, Long Beach, CA 90840, U.S.A.

Abstract. A calculation method for the aerodynamic prediction of multielement airfoils based on an interactive
boundary-layer approach using an improved Cebeci-Smith eddy viscosity formulation is described. Results are first
presented for single airfoils at low and moderate Reynolds numbers in order to demonstrate the need to calculate
transition for accurate drag polar prediction and the ability of the improved Cebeci-Smith turbulence model to
predict flows with extensive separation, and therefore to predict maximum lift coefficient. Results, in terms of
pressure distributions and lift, drag and moment coefficients, are presented for a series of multielement airfoils with
flaps and slats. The importance of the compressibility effects and the turbulence model on stall, and, again, the need
to calculate the onset of transition, are demonstrated. Finally, recommendations are made for the preferred approach
to predicting the aerodynamic performance of multielement airfoils for use as a practical and efficient design tool.

1. INTRODUCTION
In recent years, there have been significant accomplishments in computational fluid dynamics. Whereas in
the early 1960’s calculations performed for airfoil flows with panel methods and boundary-layer methods
were under development for simple flows, today the calculations are being performed routinely with
Navier-Stokes methods not only for airfoil flows but also for complex aircraft configurations. Our
capabilities in aerodynamic flows have reached levels which were difficult to imagine in the early 1960’s.
Progress has been so great and so rapid that one may even say that computational fluid dynamics has
reached, or is very near to reaching, its maturity.
Despite these advances, there is still more to be done in developing design methods for high lift
configurations. The presence of high and low Reynolds number flows on various components of airfoils,
including slat and flap coves and significant regions of flow separation near stall conditions as well as
possible merging of shear layers, offer significant challenges to code developers. The required generality
and accuracy of the code and, equally important, its efficiency as a design tool, introduce additional
challenges.
The current development of design algorithms for high-lift devices follows two approaches. One
approach pursues the solution of the Navier-Stokes equations with structured and unstructured grids.
Some of the Navier-Stokes methods used for this purpose are based on the solutions of the incompressible
flow equations [1, 2] while others are based on the solutions of the compressible flow equations [3]. The
incompressible form of the equations employing the numerical procedure of Rogers et al. [1] provides
accurate results and allows the calculations to be performed efficiently. The solution of the compressible
equations, on the other hand, takes a considerable amount of computer time; their convergence rate is
slower and, at this time, they are in the research stage.
The other approach for developing methods for high lift configurations is based on the interactive
boundary layer theory, which involves interaction between inviscid and boundary-layer equations [4, 5]. In
this approach the inviscid flow is often computed by a panel method with or without compressibility
corrections, and the viscous flow is computed by a compressible boundary-layer method. This approach,
though not as general as the Navier-Stokes approach, may provide a good compromise between the
efficiency and accuracy required in a design process such as the one cited in [6].
Regardless of which approach is used to develop a computational tool for high-lift configurations, it is
necessary to include the compressibility effect if the calculation method employs incompressible flow
equations, and to calculate the onset of transition. Both can strongly influence the accuracy of the
calculation method. Experiments on single or multielement airfoils show that the compressibility effect,
even at Mach numbers around 0.30, has a pronounced effect on the maximum lift coefficient. Similarly
the predicted location of transition in the calculation method is important in properly identifying the
effects of wind tunnel and flight Reynolds numbers. Individual components of multi-element airfoils at
wind tunnel Reynolds numbers can experience relatively lower Reynolds numbers than the main airfoil.
At chord
 Computers & Fluids Vol. 27, No 5-6, pp. 651-661, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0045-7930/98 $19.00 + 0.00

Reynolds numbers less than 500,000, the components can have large separation bubbles, with the onset of
transition occurring inside the separation bubble [7]. As a result, the behavior of the flow can be
significantly different from the behavior of the flow on the main airfoil at higher Reynolds numbers.
Furthermore, the transition location can influence the drag coefficient, and therefore determining the
onset of transition is crucial for predicting drag polars, which are of major interest when designing high
lift systems for take-off and climb requirements [6].
Section 2 describes an interactive-boundary-layer approach to the calculation of high lift
configurations in two-dimensional flows. It is applied to a variety of high lift systems, including airfoils
with flaps and slats. Results are presented and discussed in Section 3. Finally, in view of the present
results, a preferred approach for predicting the flow about multielement airfoils is discussed in Section 4.

2. CALCULATION METHOD
2.1 Inviscid Method
The inviscid flow field is computed by the Hess Smith panel method [8]. Three options were used in the
multielement panel method. In one option the vorticity strength was taken to be constant on all panels,
and a single value was adjusted to satisfy the condition associated with the specification of circulation.
Multielement configurations computed with the calculation method employing the multielement panel
method with viscous corrections showed that, while the results were in good agreement with experimental
data for conventional airfoils, the results for airfoils with thin trailing edges and/or supercritical airfoils
were not. To improve the results, two additional options for the vorticity distribution were incorporated
into the panel method. One option assumes that vorticity varies quadratically with the surface distance and
the second option assumes that it varies quadratically in a small region near the trailing edge. The
compressibility corrections depend upon the linearized form of the compressible velocity potential
equation and is based on the assumption of small perturbations and thin airfoils. The correction used in
the present panel method is based on the Prandtl-Glauert formula as described in [8].
2.2 Inverse Boundary Layer Method
Boundary layer equations
The compressible boundary layer equations (mass, momentum and energy) for laminar and turbulent
flows are well known and, with the algebraic eddy viscosity (ε m ) and turbulent Prandtl number (Pr t )
formulation of Cebeci and Smith [8], they can be expressed as follows:
∂ ∂
(ρu) + (ρv) = 0 (1)
∂x ∂y
∂u ∂u du ∂  ∂u 
ρu + ρv = ρe u e e +  (µ + ρε m )  (2)
∂x ∂y dx ∂y  ∂y 
∂H ∂H ∂  ε ∂T ∂u 
ρu + ρv =  ( k + c pρ m ) + u(µ + ρε m )  (3)
∂x ∂y ∂y  Prt ∂y ∂y 
where T is the temperature, H is the total enthalpy given by
u2
H = cpT + (4)
2
and
ρv = ρv + ρ' v' (5)
In the absence of mass transfer, the boundary conditions for an adiabatic surface are
∂H
y = 0, u = 0, v = 0 , =0 (6a)
∂y
y → ∞, u → u e ( x ) , H → H e (6b)
In the wake, where a dividing line at y = 0 is required to separate the upper and lower parts of the
inviscid flow, in the absence of normal pressure gradient, the boundary conditions at y = 0 are
 Computers & Fluids Vol. 27, No 5-6, pp. 651-661, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0045-7930/98 $19.00 + 0.00

∂u
y = 0, = 0, v = 0 (7)
∂y
Solution procedure
The above equations are first expressed in transformed coordinates. Two sets of transformed
coordinates are used, one for the direct problem when the equations are solved for the prescribed pressure
distribution, and the other for the inverse problem with the external velocity updated during the iterations.
The Falkner-Skan transformation is used in the direct mode and a modified version of this transformation
is used in the inverse mode. The transformed equations and their solution procedure are presented and
discussed in detail in [8]. The airfoil is divided into upper and lower surfaces. For each surface, the
calculations start at the stagnation point and proceed in the standard mode up to a certain specified x-
location. Then, the inverse calculations begin from that location and continue up to the far wake.
Interaction law
The boundary layer equations become singular at flow separation and do not allow the calculation of
separated flows for a prescribed velocity distribution. However, when the external velocity is treated as an
unknown, this difficulty can be overcome as discussed in [8]. The external velocity is represented by
u e ( x ) = u e ( x ) + δu e ( x )
o
(8)
where u eo is the inviscid velocity computed by the panel method and δu e is the perturbation velocity due to
viscous effects, given by the Hilbert integral
xb
dσ
δu e ( x ) =
1
∫ d
π dσ
( u . δ ).
e
*
x −σ
(9)
xa

in the interaction region (a, b). Its evaluation is described in detail in [8].
Once the boundary layer development is computed, the blowing velocity distribution is determined
from
d
vn = ( u e . δ* ) (10)
dx
and used as a boundary condition in the inviscid method. This interactive procedure is repeated until
convergence of the lift coefficient is achieved.
Turbulence model
An improved version of the algebraic eddy-viscosity formulation of Cebeci-Smith is used here. The
eddy viscosity distribution across the boundary layer is defined by two separate expressions,
   − y   ∂u
2

( ε m ) i =  0.4 y 1 − exp   . . γ tr 0 ≤ y ≤ yc
   A   ∂y
εm =  (11)
∞

 ( ε m ) o = α∫ ( u e − u) dy . γ tr . γ yc ≤ y ≤ δ
 0

where
1 1
0.0168 ν ρ 2  τ 2
α= , A = 26   , u τ =   (12)
F1.5 u τ  ρw   ρ  max
where F is related to the ratio of the product of the turbulence energy by normal stresses to that by shear
stress evaluated at the location where the shear stress is maximum. As discussed in [8], it is given by
∂u ∂x
F =1−β (13)
∂u ∂y
where the parameter β is a function of R t = τ w ( −ρu ' v ') max , which, for τ w ≥ 0 , is represented by
 Computers & Fluids Vol. 27, No 5-6, pp. 651-661, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0045-7930/98 $19.00 + 0.00
 6
 1 + 2R (2 − R ) R t ≤ 1.0
 t t
β= (14)
 1 + R
R t ≥ 1.0
t

 R t
For τ w ≤ 0, R t is set equal to zero.

Also, whereas in the original Cebeci-Smith eddy viscosity formulation the intermittency expression was
valid only for zero pressure gradient flows, the expression in the improved formulation is applicable for
flows with favorable and adverse pressure gradients as well as zero pressure gradient flows. It is based on
Fiedler and Head’s correlation [9] and is given by
1 (y − Y) 
γ = 1 − erf  (15)
2 2σ 
where Y and σ are general intermittency parameters with Y denoting the value of y where γ = 0.5 and σ
denoting the standard deviation. For details, see [8].
The condition used to define y c is the continuity of the eddy viscosity, so that ε m is defined by ( ε m ) i
from the wall outward (inner region) until its value is equal to that given for the outer region by ( ε m ) o .
The expression γ tr models the transition region and is given by
 x 
 dx 
γ tr = 1 − exp −G ( x − x tr )
 u e 
∫ (16)
 x tr

Here x tr denotes the onset of transition and G is defined by
3
3 ue
G=
−1. 34
R (17)
ν x C
2 2 tr

where C is 60 for attached flows and the transition Reynolds number is R x tr = ( u e x / ν) tr . In the low
Reynolds number range from R c = 2 × 10 5 to R c = 6 × 10 5 , the parameter C is given by
(
C = 213 log R x tr − 4.7323
2
) (18)
The corresponding expressions for the eddy-viscosity formulation in the wake are
−(x − x 0 )
ε m = (ε m ) w + [(ε m ) t .e. − (ε m ) w ] exp (19)
l w δ t .e .
where δt.e. is the boundary layer thickness at the trailing edge, lw = 20, ( ε m ) t .e. is the eddy viscosity at the
trailing edge computed from its value on the airfoil, and ( ε m ) w is the eddy-viscosity in the far wake given
by the larger of
y min

( ε m ) w = 0.064
l
∫ (u e − u ). dy (20a)
−∞
and
∞

(ε u
m )w = 0.064 ∫ (u e − u ). dy (20b)
y min

with y min denoting the location where the velocity is minimum.

For high Reynolds numbers, if the flow is attached, the onset of transition is determined by Michel's
criterion as described, for example, in [8]. At high angles of attack, the flow separates downstream of the
pressure peak before Michel's criterion can be satisfied. Therefore, the onset of transition is chosen to
coincide with laminar separation.
At Reynolds numbers less than 10 6 , where large separation bubbles may be present even at low
angles of attack, it is necessary to calculate the onset of transition by the en-method as discussed in [8]. For
high lift configurations, flaps usually have low Reynolds numbers. Fortunately, the pressure distribution
on the
 Computers & Fluids Vol. 27, No 5-6, pp. 651-661, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0045-7930/98 $19.00 + 0.00

flap upper surface does not vary greatly with angle of attack, and therefore the computation of the
transition location does not have to be performed at each angle of attack.

3. RESULTS AND DISCUSSIONS

3.1 Single Airfoils
Before we present a sample of results for high lift configurations, it is useful to discuss the role of
transition by examining the results shown in Figs. 1 and 2 for two airfoils at high and low Reynolds
numbers. The results in Fig. 1 are for the NACA 0012 airfoil at a chord Reynolds number of 3 x 106. As
can be seen, there is a significant difference between the calculated lift coefficients in which the transition
location was fixed near the stagnation point for all angles of attack or computed. A similar observation
can be made for the drag coefficients. Those calculated with the transition computed show a much better
agreement with experimental data than those in which the transition was fixed.
2.0 0.040

1.5 0.030

Cl 1.0 cd 0.020

0.5 0.010

0.0 0.000
0.0 5.0 10.0 15.0 20.0 0.0 0.5 1.0 1.5
(a) α (b)
Cl
Fig. 1. Comparison of measured (symbols) and calculated (lines) (a) lift and (b) drag coefficients for the NACA 0012 airfoil. Continuous
lines denote calculations with computed transition, and dotted lines calculations with transition near the stagnation line.

Fig. 2 shows the results for the Eppler airfoil at a chord Reynolds number of 200,000. Here, the
location of transition does not play an important role in predicting lift coefficients (Fig. 2a). However, the
calculations with a fixed transition location can be performed until α = 13.5o. As before, the drag
coefficients obtained with the transition location computed show better agreement with data than those
with a fixed transition location (Fig. 2b).
1.5 0.10

0.08

1.0
0.06
Cl cd
0.04
0.5
0.02

0.0 0.00
-2.0 3.0 8.0 13.0 18.0 -5.0 0.0 5.0 10.0 15.0
(a) α (b) α
Fig. 2. Force coefficients for the Eppler airfoil at Rc = 200,000, (a) lift and (b) drag. Circles denote experimental data, continuous lines
denote calculations with computed transition, and dotted lines denote calculations with transition set at the leading edge.

3.2 Two-element Airfoils

Figs 3 and 4 show the results for the NLR 7301 supercritical airfoil/flap configuration. The
experimental data of Van den Berg and Oskam [10] include pressure distributions at α = 6o and 13.1o, lift
curve and drag polar for an airfoil with a flap of 32% chord deflected at an angle of 20o and with a gap of
2.6% chord. The experimental free stream Mach number was M∞ = 0.185, and the chord Reynolds
 Computers & Fluids Vol. 27, No 5-6, pp. 651-661, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0045-7930/98 $19.00 + 0.00
number was 2.51 x 106. Fig. 3 shows a comparison between measured and computed pressure
distributions at α =

6.0o and α = 13.1o, and Fig. 4 shows a similar comparison for the lift coefficient and drag polar. The
viscous flow calculations were performed with the onset of transition location calculated on the main
element and flap. As can be seen from the results in Fig. 4a, the calculated results, including drag, agree
well with experimental data.
8.0 12.5

10.0
6.0

7.5
4.0
-Cp -Cp 5.0
2.0
2.5
0.0
0.0

-2.0 -2.5
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
(a) x/c (b) x/c
Fig. 3. Comparison of measured (symbols) and calculated (continuous lines) pressure distributions for the NLR 7301 supercritical
airfoil/flap configuration for (a) α = 6o and (b) α = 13.1o.
3.5
0.10

3.0 0.08

0.06
cl 2.5 cd
0.04
2.0
0.02

1.5 0.00
0.0 5.0 10.0 15.0 20.0 1.5 2.0 2.5 3.0 3.5
(a) α (b) cl
Fig. 4. Comparison of measured (symbols) and calculated (continuous lines) (a) lift and (b) drag coefficients for the NLR 7301
supercritical airfoil/flap configuration

Figures 5 to 9 show the results for two NASA high lift configurations tested by Omar et al. [11] for a
free stream Mach number of M∞ = 0.201 and chord Reynolds number of 2.83 x 106. The compressibility
corrections were made to the inviscid flow using the Prandtl-Glauert formula.
Figure 5 shows the pressure distributions for an airfoil/flap configuration at α = 0.01o and 8.93o. It is
seen that the calculated results are in good agreement with experimental data, although there is some
discrepancy on the upper surface of the flap. This discrepancy is due to two phenomena, the merging of
shear layers and the inaccurate wake center line prediction, as shown in Figure 6. It is seen that in the
experiments, merging occurs at the trailing edge of the flap for both α = 0o and α = 8o, and that it is not
predicted by the present method. Also, in the present method, the wake center line is assumed to be the
dividing streamline originating at the main element trailing edge and is closer to the flap upper surface
than the measured one (thick line in Figure 6), which causes a greater predicted flow acceleration on the
flap upper surface than the actual acceleration.
 Computers & Fluids Vol. 27, No 5-6, pp. 651-661, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0045-7930/98 $19.00 + 0.00
6.0 20.0
Critical Pressure Coefficient
16.0
4.0
12.0

-Cp 2.0 -Cp 8.0

4.0
0.0
0.0

-2.0 -4.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
(a) x/c (b) x/c
Fig. 5. Comparison of measured (symbols) and calculated (continuous lines) pressure distributions for the NASA airfoil/flap configuration
at (a) α = 0.01o and (b) α = 8.93o.

wake center lines

shear layer edges

measured merging
measured merging
(a) (b)
Fig. 6. Comparison of measured and computed shear layers for the NASA airfoil/flap configuration, (a) α = 0o and (b) α = 8o. Shaded
areas denote computed shear layers and thick lines denote edges of measured shear layers

Figure 7 shows the lift and drag coefficients for the same configuration. The lift coefficient is slightly
over predicted due to the discrepancies on the flap upper surface, and similarly the drag coefficient is
slightly under predicted. Also, the calculated incompressible stall angle is rather different from the
calculated compressible one. The critical pressure coefficient for M∞ = 0.201 is indicated in Fig. 5b. At α
= 8.93o, the measured stall angle, there exists a small region of supersonic flow and a shock may occur.
Even though this shock cannot be predicted by the panel method, the compressibility corrections,
theoretically applicable at lower angles of attack only, provide a significant improvement to the stall
prediction. However, a truly compressible method such as the one of [12] would be preferable.
3.0 0.15

2.0 0.10

cl cd
1.0 0.05

0.0 0.00
-10 -5 0 5 10 15 0.0 1.0 2.0 3.0
(a) α (b) l c
Fig. 7. Comparison of measured (symbols) and calculated (lines) (a) lift coefficients, and (b) drag polars for the NASA airfoil/flap
configuration. Continuous lines denote incompressible calculations and dotted lines compressible calculations.

Figures 8 and 9 show the results for a slat/airfoil configuration. Figure 8 presents a comparison
between measured and calculated pressure distributions at 26.88o. While the agreement is excellent for the
slat, the pressure peak on the main airfoil is slightly under-predicted. Figure 9 shows the lift and drag
coefficient comparisons for the same configuration. The agreement with data is excellent up to stall for the
lift coefficient and the drag polar is predicted satisfactorily. The computed maximum lift coefficient and
 Computers & Fluids Vol. 27, No 5-6, pp. 651-661, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0045-7930/98 $19.00 + 0.00

stall angle are 2.88 and 27.5o, respectively, and the measured ones are 2.90 and 26.9o. This gives a 0.7%
error for the maximum lift coefficient and an error of 2.2% for the stall angle.
18.0

14.0

10.0
-Cp
6.0

2.0

-2.0
-0.2 0.0 0.2 0.4 0.6 0.8 1.0
x/c
Fig. 8. Comparison of calculated (continuous lines) and measured (symbols) pressure distributions for the NASA slat/airfoil configuration
at α = 26.88o.
3.0 0.150

Total 0.125

2.0 0.100
cl Main cd 0.075
Element

1.0 0.050
Slat
0.025

0.0 0.000
10.0 15.0 20.0 25.0 30.0 0.0 1.0 2.0 3.0
(a) α (b) cl
Fig. 9. Comparison of measured (symbols) and calculated (continuous lines) (a) lift coefficients, and (b) drag polars for the NASA
slat/airfoil configuration.

3.3 Slat-airfoil-flap configurations

The previous section demonstrated the ability of the method to compute multielement airfoil lift and
drag coefficients accurately, including stall and post-stall, for slat-airfoil and airfoil-flap configurations.
For these cases, stall is mostly due to flow separation occurring on the main element and is well predicted
by the present method. For well-optimized slat-airfoil-flap configurations, however, stall often occurs
without flow separation on the main element. Instead, the slat wake merges into the main element
boundary layer or wake. Due to the adverse pressure gradient induced by the presence of the flap, this
complex shear layer may drastically thicken downstream of the main element trailing edge, thus reducing
the overall circulation and causing a reduction in lift.
To illustrate this phenomenon and evaluate the present method ability to predict such flows, the three-
element airfoil of [13] was considered for a free stream Mach number of about M∞ = 0.2 and chord
Reynolds number of 3.52 x 106. The slat is deflected at 25o and the flap at 20o. Figure 10 shows a
comparison of the calculated and measured pressure distributions at α = 4o and α = 22o. Figure 11 shows
a similar comparison for the integrated lift, drag and moment coefficients. Agreement is generally good
up to stall. Discrepancies exist in the flap well on the main element and near the leading edge of the flap
because current results were obtained by using a fairing of the flap well. Work is in progress to improve
the flap well procedure of [14] in order to refine predictions in this region. It is conjectured that the
discrepancies in lift, drag and moment coefficients are partially due to this phenomenon. Also, as
previously mentioned, the merging of shear layers was not accounted for and may have contributed to the
differences between experimental data and calculations.
 Computers & Fluids Vol. 27, No 5-6, pp. 651-661, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0045-7930/98 $19.00 + 0.00
4.0 18.0

3.0
14.0
2.0
10.0
-Cp 1.0 -Cp
6.0
0.0

-1.0 2.0

-2.0 -2.0
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
(a) x/c (b) x/c
Fig. 10. Comparison of calculated (continuous lines) and measured (symbols) pressure distributions
for a three-element airfoil at (a) α = 4o and (b) α = 20o.

5.0 0.40

4.0 0.30

cl cd
3.0 0.20

2.0 0.10

1.0 0.00
0.0 10.0 20.0 30.0 0.0 10.0 20.0 30.0
(a) α (b) α

-0.20

-0.30

-0.40
cm
-0.50

-0.60

-0.70
0.0 10.0 20.0 30.0
(c) α

Fig. 11. Comparison of calculated (lines) and measured (symbols) (a) lift, (b) drag, and (c) moment coefficients. Continuous lines denote
original wake turbulence model and dotted lines modified wake turbulence model.

Also, Figure 11 shows that, while agreement is excellent up to α = 22o, using the original turbulence
model in the wake does not allow to capture stall. However, when the constant lw of Eq. (19) is increased
to 100, as shown in dotted lines in Fig. 11, the stall behavior of this multielement airfoil is captured
appropriately. Figure 12, which presents the variation of displacement thickness along the chord for each
element at α = 22o, illustrates the growth of the displacement thickness in the main element wake induced
by the presence of the flap. Work is in progress to incorporate a turbulent wake formulation similar to that
 Computers & Fluids Vol. 27, No 5-6, pp. 651-661, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0045-7930/98 $19.00 + 0.00

used on the body to improve current predictions. More importantly, these results demonstrate the need for
accurately modeling the turbulent shear layers developing in wakes in order to calculate multielement
airfoil stall reliably.
0.10

0.08

0.06 Main element

δ∗/c
0.04
slat
0.02
flap
0.00
0.5 1.0 1.5 2.0
x/c
Fig. 12. Displacement thickness distribution at α = 22o for the three-element airfoil. Continuous lines denote calculations performed with
the original wake turbulence model and dotted lines with the modified turbulence model.

4. CONCLUDING REMARKS
An interactive boundary-layer method with an improved Cebeci-Smith eddy viscosity formulation is
used to calculate the aerodynamic performance characteristics of high lift systems. Results show good
agreement with measurements for a variety of multielement airfoils comprising slats, main element and
flaps. In general, the predictions are excellent for relatively low angles of attack and very satisfactory up to
stall. The study shows that the onset of transition location plays a significant role in predicting drag and
that its calculation must be a part of the computational method. The study also shows that for some flows
merging of shear layers takes place and, since the present method does not account for this merging,
calculated results, while satisfactory, show some discrepancy with measurements.
The stall prediction is satisfactory for most cases. However, when the pressure peak on the main
element reaches values close to or greater than the critical pressure coefficient, a truly compressible
inviscid method, such as a full potential or Euler method, should be employed. Also, the need for accurate
turbulence models in wakes is stressed in order to compute stall with reliability.
In view of these remarks, it appears that a preferred approach for predicting the aerodynamic
performance of high lift systems would combine the present compressible boundary layer method,
enhanced by adding the merging of shear layers, an improved wake turbulence model, and a cove
calculation procedure, with an analytical Euler method similar to the one previously mentioned.
Work is in progress to incorporate these capabilities. Such a method would provide a tool combining
efficiency, practicality and accuracy for transport aircraft high lift system design and optimization.
 Computers & Fluids Vol. 27, No 5-6, pp. 651-661, 1998
 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0045-7930/98 $19.00 + 0.00

REFERENCES
1. Rogers, S.E., Wiltberger, N.L. and Kwak, D., “Efficient Simulation of Incompressible Viscous Flow
Over Single- and Multielement Airfoils,” AIAA Paper No. 92-0405, 1992.
2. Rogers, S.E., “Progress in High-Lift Aerodynamic Calculations,” AIAA Paper No. 93-0194, Jan.
1993.
3. Valarezo, W.O. and Mavriplis, D.J., “Navier-Stokes Applications to High-Lift Airfoil Analysis,”
AIAA Paper No. 93-3534, Aug. 1993.
4. Cebeci, T., “Calculation of Multielement Airfoils and Wings at High Lift,” AGARD Conference
Proceedings 515 on High-Lift Aerodynamics, Paper No. 24, Banff, Alberta, Canada, Oct. 1992.
5. Cebeci, T., Besnard, E. and Messié, S., “A Stability/Transition-Interactive-Boundary-Layer Approach
to Multielement Wings at High Lift,” AIAA Paper No. 94-0292, Jan. 1994.
6. Nield, N.B., “An Overview of the 777 High Lift Aerodynamic Design,” Proceedings of High Lift and
Separation Control Conference, Royal Aeronautical Society, March 1995.
7. Cebeci T, “Essential Ingredients of a Method for Low Reynolds-Number Airfoils,” AIAA Journal,
Vol. 27, pp. 1680-1688, 1983.
8. Cebeci, T., An Engineering Approach to the Calculation of Aerodynamic Flows, to be published,
1997.
9. Fiedler, H. and Head, M.R., “Intermittency Measurements in the Turbulent Boundary Layer,” J. Fluid
Mech., Vol. 25, Part 4, pp. 719-735, 1966.
10. B. Van den Berg and B. Oskam, “Boundary Layer Measurements on a Two-Dimensional Wing with
Flap and a Comparison with Calculations,” AGARD CP-271, Sept. 1979.
11. E. Omar, T. Zierten and A. Mahal, “Two-Dimensional Wind Tunnel Tests of a NASA Supercritical
Airfoil with Various High Lift Systems,” NASA CR-2215, 1977.
12. Verhoff, A., Chen, H.H., and Cebeci, T., “An Accurate and Efficient Interactive Boundary Layer
Method for Analysis and Design of Airfoils,” AIAA Paper No. 96-0328, 1996.
13. I.R.M. Moir, “Measurements on a two-dimensional Aerofoil with High-Lift Devices,” AGARD-AR-
303, Test Case A-2, 1994.
14. T. Cebeci, E. Besnard and H.H. Chen, “Calculation of Multielement Airfoil Flows, Including Flap
Wells,” AIAA Paper No. 96-0056, 1996.