Sei sulla pagina 1di 29



92008

DIRECT PREDICTION OF A SEPARATION BOUNDARY FOR AEROFOILS USING A


VISCOUS-COUPLED CALCULATION METHOD

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

SUMMARY
The Data Item presents the principal features necessary in a viscous-coupled calculation method for it to
be capable of predicting transonic flow around an aerofoil where there is some boundary layer separation.
One widely used method (BVGK) which incorporates these features is a development of the VGK method
(Reference 1). The BVGK method employs a semi-inverse coupling of the viscous and inviscid flow
elements, and includes higher-order terms in the formulation of the boundary layer and wake. It can deal
with transonic flow with a subsonic free stream.
The Item gives an assessment of BVGK as a method of predicting separated transonic flows, by comparing
results obtained using a BVGK code with data from wind tunnel experiments corrected to free air conditions.
Attention is given to prediction of
(a)

aerofoil pressure distributions, especially indications of shock strength and location, and
trailing-edge pressure,

(b)

separation boundary associated with trailing-edge pressure divergence, and

(c)

aerofoil drag coefficient for conditions near this separation boundary.

In the comparisons presented the separation boundary is well predicted (generally within 0.08 in C L ) . This
is true when the boundary is associated with either shock-induced separation or trailing-edge separation.
The corresponding values of drag coefficient are fairly well predicted. However the code deals less
satisfactorily with certain details and trends of the experimental pressure distributions. In particular the
often observed gradual increase of trailing-edge pressure prior to its rapid fall (as separation becomes more
extensive) is not predicted. It appears that BVGK tends to under-predict the extent (and effects) of
shock-induced separation.

1.

INTRODUCTION
Some of the calculation methods which couple viscous and inviscid flow elements are capable of predicting
aerofoil flows with shock-induced boundary layer separation. Such codes may therefore be used to predict
separation boundaries directly, and also to predict aerofoil aerodynamic characteristics, such as drag and
pitching moment, at conditions of lift coefficient and Mach number close to a separation boundary.
In this Item one such highly developed viscous-coupled method, known as BVGK, is considered with
particular reference to the accuracy of the codes prediction of:
(a)

a separation boundary defined by trailing-edge pressure divergence,

(b)

drag coefficient at conditions close to this boundary.

Also, since much useful information is provided from prediction of pressure distributions, these too are
considered.

Other codes, embodying approximate solutions to the Navier Stokes equations, are also capable of such predictions. These more complex
codes, although potentially more accurate, have yet to be thoroughly assessed.

The degree of separation is defined (as in Reference 18) by C p = 0.05 , see Section 5.
te

Issued April 1992


1

92008
Experimental aerofoil data, corrected to free air conditions, are used to assess the accuracy of BVGK. Flows
examined involve one, or both, of the following:
(a)

separation commencing at or close to the foot of the shock,

(b)

separation near the trailing edge, well downstream of the shock.

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

Although only a small number of aerofoils is considered, the range of flows is relevant to many advanced
supercritical aerofoils. Separation in shock-free flows is not considered.
In Section 3, following a brief outline of the general nature of viscous-coupled calculation schemes, a
summary is given of the important features which should be included in order to achieve at least reasonably
accurate results for transonic separated flows. This is followed by a description of the details of these
features as they are incorporated in the BVGK code. Further relevant details of BVGK are presented in
Section 4. In Section 5 results obtained using BVGK are compared with wind tunnel data corrected to free
air conditions. These comparisons are evaluated and discussed in relation to the prediction of separated
flows or where severe adverse pressure gradients occur. Recommendations are given in Section 6 regarding
the use of BVGK (and comparable viscous-coupled methods) to predict the particular separation boundary
considered and the associated drag.

2.

NOTATION AND DEFINITIONS

2.1

Notation
All quantities of length made non-dimensional are based on the aerofoil chord, c.
(overall) drag coefficient, equal to C D + C D
V
W

CD
CD

CD

viscous drag coefficient


wave drag coefficient

CL

lift coefficient

Cp

pressure coefficient

aerofoil chord

cf

2
equivalent skin-friction coefficient, equal to 2 w / w u iw

shape parameter, equal to /

transformed shape parameter, see Reference 3

H1

mass flow shape parameter, equal to ( )/

IEIF , INPG
INST , IHH1
ILRN , ICURVB

switches in BVGK program which activate corrections for various higher-order


effects, see Section 3

92008
j

number defining computing mesh (Sketch 3.1)

number defining computing mesh (Sketch 3.1)

free-stream Mach number

Reynolds number based on aerofoil chord, u c/


radial coordinate in plane = re i (Sketch 3.1)

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

r
s

streamwise distance along aerofoil or rear dividing streamline; also coordinate


orthogonal to z, increasing in sense of external flow velocity

aerofoil maximum thickness

velocity component in s coordinate direction, time mean velocity component


where flow is turbulent

free-stream speed

non-dimensional chordwise distance measured from leading edge of aerofoil

coordinate normal to aerofoil surface, zero on aerofoil surface

angle of incidence

boundary-layer thickness

equivalent boundary-layer thickness, see Reference 3

C p

boundary-layer displacement thickness


te

variation of Cp , see Section 2.2 (ix)


te

boundary-layer momentum thickness; also polar coordinate in plane = re i


(Sketch 3.1)

quasi-conservative parameter

viscosity

density

complex variable of interior circle plane (Sketch 3.1)

wall shear stress

Suffixes
i

denotes equivalent inviscid flow

denotes separation point

sep

denotes value where trailing-edge pressure divergence criterion is satisfied

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

2.2

92008
sh

denotes shock location

te

denotes trailing edge more specifically it denotes most rearward measuring


location on upper surface (experiment), or nearest mesh point (on upper surface)
to this location (BVGK)

denotes wall, i.e. aerofoil surface, conditions

denotes conditions just upstream of shock

denotes free-stream conditions

Definitions
(i)

Attached flow
A condition of the flow about an aerofoil corresponding to the mean velocity component, u, being
positive everywhere close to (but not on) the aerofoil profile, see Sketch 2.1.

Sketch 2.1 Notation related to boundary layer


(ii)

Separated flow
A condition of the flow about an aerofoil corresponding to u being negative at some points close
to the aerofoil profile, see Sketch 2.2.

(iii)

Separated flow region(s)


The portion(s) of the aerofoil profile, when separated flow occurs, corresponding to locations where
u is negative at some points close to the aerofoil profile, see Sketch 2.2.

(iv)

Line of zero velocity


Locus of points, when a separated flow occurs, (excluding points on the aerofoil profile) where u
is zero, see Sketch 2.2.

92008
(v)

Point of separation (or separation point)

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

A point on the aerofoil profile approached from downstream along a line of zero velocity, see
Sketch 2.2.

Sketch 2.2 Notation related to separation no re-attachment


(vi)

Point of re-attachment (or re-attachment point)


A point on the aerofoil profile approached from upstream along a line of zero velocity, see
Sketch 2.3.

Sketch 2.3 Notation related to separation with re-attachment


(vii)

Observed separated flow region(s)

The extent(s), if any, of the aerofoil profile, corresponding to locations where either the measured wall
shear stress is negative or zero, or some other experimental separated flow criterion is satisfied.
(viii)

Calculated separated flow region(s) (e.g. for a calculation scheme, such as BVGK, employing an
integral boundary-layer method)
The extent(s), if any, of the aerofoil profile, corresponding to locations where the calculated wall
shear stress, w , is negative or zero; or, what is equivalent, where c f is calculated as negative or
zero. Each extent is bounded upstream by a (calculated) point of separation, and is bounded
downstream, where appropriate, by a (calculated) point of re-attachment.

92008
(ix)

Trailing-edge pressure divergence criterion

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

A criterion for expressing a degree of flow separation by means of either the variation of C p te with
C L , at fixed M , or with M at fixed CL . In particular the criterion is satisfied where one or other
of the variations has shown a decrease in C p te of 0.05 from a trend established at lower values of
C L or M , as appropriate (see Sketch 2.4). For brevity this criterion is referred to as the
C p te = 0.05 criterion. The criterion may be applied either to observed or calculated data for an
aerofoil. (There is some subjectivity in defining the trends but this has a very minor effect.)

Sketch 2.4 Trailing-edge pressure divergence separation criterion


(x)

Trailing-edge pressure divergence separation boundary


The locus in the C L , M plane of the points, denoted by C L , M , at which the trailing-edge
sep
sep
pressure divergence criterion is satisfied.

(xi)

BVGK
BVGK denotes both the scheme of equations (and the solution methods) outlined in Section 4, and
the various program codes by which these solution methods are implemented on a computer.

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

92008

3.

VISCOUS-COUPLED METHODS FOR TRANSONIC SEPARATED FLOWS

3.1

General Nature of Viscous-Coupled Methods


In many problems of aircraft aerodynamics the effects of viscosity and turbulence are confined to relatively
thin shear layers in the vicinity of the wetted surfaces and their wakes. Outside these viscous regions
the inviscid flow equations are applicable to a high level of accuracy. Viscous coupled methods (also known
as viscous-inviscid interaction (VII) methods) take advantage of this fact and employ quite different
calculation procedures for the external inviscid flow and the viscous shear layers. These two flow elements
are linked, or coupled (each element affecting the boundary conditions of the other), and an iterative
procedure is used to produce a converged composite solution. This solution, comprising both the external
inviscid flow and boundary-layer/wake elements, then yields the required aerodynamic parameters, such
as surface pressures, shear stresses, and overall forces and moments.
Viscous-coupled methods have reached an advanced state of development, particularly for two-dimensional
flows. There is a variety of methods however, even for two-dimensional flows, and these methods can be
broadly classified in terms of the equations and solution procedure governing each of their three main
components, namely the inviscid flow, the shear layers and the coupling technique. Thus, the inviscid flow
component may be determined either by solving the Euler equations or the equation for the velocity
potential. In the latter case the adopted inviscid flow solution procedure for dealing with any shock waves
may be either fully conservative, non-conservative, or partially conservative. Similarly, the shear
layers may be treated by a differential or an integral approach, with the latter being subject to numerous
variations, the most important of which involve the incorporation of higher-order terms. Finally, the
coupling technique may be direct, inverse, semi inverse, or quasi-simultaneous, with further
variations involving conditions relating to local boundary layer parameters. Added to these various
categories the particular numerical techniques employed are also of relevance, especially for the inviscid
flow element, where the type and refinement of the grid, the details of the shock capturing scheme, and the
iterative solution procedure, are features which have a bearing on the solution obtained. Further details of
these methods and examples of the influence of various features are given in References 2 and 3.

3.2

Important Features for Transonic Separated Flows


In the course of development of viscous-coupled methods, it has been found that methods which are capable
of giving quite accurate results for transonic aerofoil flows at low and moderate lift coefficient may either
fail or otherwise be quite inadequate for predicting flows with separation. Three main features are necessary
to remedy this situation. These are as follows.
(a)
(b)
(c)

The inviscid flow solution must be capable of accurately predicting flows with strong embedded
shock waves, i.e. where the Mach number just upstream of the shock may be as large as 1.4.
Since the shear layers at and near separation are not thin, carefully evaluated higher-order
terms need to be incorporated into the viscous flow element.
The coupling technique has to be one that can deal with both attached and separated boundary
layer regions. One such technique is the semi-inverse method; another is termed the
quasi-simultaneous coupling technique, see Reference 2.

In most viscous-coupled methods the inviscid outer flow is extended (extrapolated) to the surface (and dividing stream surface) and is
then termed the equivalent inviscid flow (EIF).

92008

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

One viscous-coupled method which includes these features is known as BVGK and it is this particular
method that is assessed in this Item. The BVGK method was devised at RAE, and is a development of an
earlier viscous-coupled method known as VGK, see References 1 and 5. It employs Carters scheme
(Reference 4) for the semi-inverse coupling technique. It is capable of predicting both attached flows and
flows with small extents of separation. Various versions of BVGK have now been generated but they all
incorporate the three main features noted above, and the results obtained from them are largely
indistinguishable. However, the later versions include various numerical refinements and are capable of
yielding smoother results for flows with significant separation.
The performance of a calculation method, in particular its accuracy of prediction, is dependent on the
principal features embodied in the method and also, to a lesser degree, on numerous details of
implementation. Provided a viscous-coupled method embodies the three main features listed above, its
performance may be expected to be largely comparable with that of BVGK. However, details can be
important and for this reason the specific forms of the three features as they are incorporated into BVGK
are presented in Section 3.3.
3.3

Features of BVGK

3.3.1

Inviscid flow
The inviscid flow element of the BVGK method is based on the solution of the velocity potential equation.
This gives advantages of speed and simplicity as compared with methods based on solution of the Euler
equations, but there is a disadvantage for flows with shock waves since flows calculated using the velocity
potential do not satisfy the Rankine-Hugoniot shock relations, nor are the flow fields downstream of a shock
accurately determined. Errors in the potential solution generally become more significant as shock wave
strength increases. However, a modification to the solution procedure due to Lock and Collyer
(Reference 6), which is incorporated into BVGK, results in the computed shock jumps being close to the
correct values. This modification utilises the observation that numerical solutions of the potential equation
cast in two different forms, namely the fully-conservative (FC) form, and the non-conservative (NC)
form, yield differing shock jumps and shock locations, and that these differing solutions have errors of
opposite sign. By introducing a mix of these two basic solution schemes, a so-called
partially-conservative (PC) scheme results, and, properly applied, this yields flow fields which are nearly
correct. The precise mix of the two basic schemes is governed by the quasi-conservative parameter ,
where ranges between 0 and 1.0, with = 1.0 corresponding to the fully-conservative scheme. Lock
and Collyer (Reference 7) present evidence supporting the use of the empirical rule:
= 1 xsh .
However, computed flows, with viscous effects included, are found to be fairly insensitive to changes
in . The results presented in Section 5 generally correspond to = 0.25 and are appropriate to flows
with shocks located aft of about mid-chord.
Solution of the inviscid flow equations in BVGK is performed on an O type mesh, see Sketch 3.1, utilising
successive line over-relaxation (SLOR). Two mesh densities are normally employed, namely 80 15 and
160 30, with the former being used in the initial stage of a calculation.

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

92008

Sketch 3.1 Representation of physical and computing plane meshes used in BVGK


3.3.2

92008
Viscous flow

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

BVGK uses an integral approach to calculate the shear layers. Thwaites method (Reference 8), extended
to allow for compressibility (Reference 9), is used for the laminar boundary-layer portions. The turbulent
boundary-layer portions and wake are computed by the lag-entrainment method (Reference 10). Transition
locations on the upper and lower surfaces need to be specified by the user. Displacement effects of the
shear layers are represented by means of wall transpiration (see Reference 2). The lag-entrainment method
incorporated in BVGK is enhanced by including various higher-order effects. These higher-order effects
are listed below as items (1) to (5) and details are given in Reference 3. The parameter names of the integers
or switches that control each effect in the BVGK code are shown in brackets.
(1)

Equivalent inviscid flow (IEIF)

Various equations in Easts (Reference 11) higher-order form of the lag-entrainment boundary layer method
are based on wall (i.e. aerofoil surface) conditions of the equivalent inviscid flow, EIF, (Reference 12).
These replace the corresponding external-flow conditions, inferred from wall static pressures in the real
viscous flow (RVF), which use the boundary-layer approximation that the pressure is constant normal to
the aerofoil surface or rear dividing streamline.
(2)

Normal pressure gradient (INPG) and normal stress terms (INST)

Allowance is made in the streamwise momentum equation for the effects of normal pressure gradients and
Reynolds normal stresses.
(3)

Revised shape parameter relationship (IHH1)

A revised relationship, suitable for flows with trailing-edge separation, is used to link the shear layer
parameters H and H 1 .
(4)

Low Reynolds number corrections (ILRN)

Corrections are made to allow for low Reynolds number effects in the equations yielding the skin friction
coefficient and the equilibrium locus.
(5)

Curvature effects on turbulence structure (ICURVB)

Effects attributable to the influence of longitudinal mean-flow curvature on turbulence are included in the
lag equation.
3.3.3

Coupling
A converged solution to a viscous-inviscid interaction problem is achieved when the boundary conditions
for the viscous and inviscid flow elements are simultaneously satisfied (within a specified error) on the
matching surface. The form of coupling has to be such that an iterative scheme involving successive
computations of viscous and inviscid flow elements does indeed converge, ideally rapidly. Coupling forms
which had been found suitable for attached boundary layers were found to fail as separation was approached,
reflecting a change in the physical nature of the interaction between the boundary layer and the outer flow.
Various schemes were then devised which resulted in satisfactory coupling for both attached and separated
boundary layers and one of these, termed semi-inverse, is utilised in BVGK. In this scheme, due to Carter
(Reference 4), an inverse solution of the boundary-layer equations (i.e. one which employs a specified

distribution of displacement thickness, ) is coupled with a direct solution of the inviscid flow equations.

The method includes a procedure for determining the location of laminar separation. If this is predicted to occur upstream of the specified
transition location, then the method utilises the predicted location of laminar separation as the transition location.

10


4.

92008
THE BVGK COMPUTER PROGRAM

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

There is no single reference source for BVGK; however Reference 3 deals in some detail with the
development of BVGK from the earlier code, VGK. Details of VGK are given in References 1, 5 and 13.
As with VGK, BVGK consists of two parts; the first deals with the mapping of the region outside the
aerofoil onto the interior of a circle, using the method of Reference 14, and the second calculates the flow
solution. The mapping data are retained for use in subsequent calculations; similarly, the flow results may
be stored (in a dump file) to allow subsequent calculations to continue to an improved level of convergence
or to be used as initial conditions for a calculation using the same aerofoil geometry but having different
values of specified Mach number, incidence, lift coefficient or other parameters.
4.1

Computer Requirements
The computer program is in standard FORTRAN 77 and the flow solution is normally divided into
coarse-grid (80 15) flow calculations and fine-grid (160 30) flow calculations.
On the CRAY 2 computer BVGK requires ~ 250K of core store. About 2 seconds of CPU (Central Processor
Unit) time are taken for the mapping calculations. The flow calculations take about one second and four
seconds for 100 coarse-grid, or fine-grid iterations, respectively. A typical flow calculation requires 200
iterations with the coarse grid, followed by 500 iterations with the fine grid. The number of iterations
required to reach satisfactory convergence depends on the particular case, with higher numbers being
necessary for flows with boundary layer separation. Typically, 800 iterations for each case with significant
separation, and a series of perhaps five cases with successively increasing C L , with M fixed, would be
needed to achieve data for a flow with separation extending for several percent of chord.

4.2

Program Input
Program input requires parameters that specify the aerofoil geometry, flow conditions (including Mach
number, Reynolds number, and transition locations, and either the incidence or lift coefficient), iteration
procedure and desired convergence level. In addition, input includes a number of switch parameters (such
as those controlling the higher-order terms).
An input parameter also determines whether the flow calculation is to be continued from a previously
produced dump file, or is to be started from the standard program initial conditions.
It should be noted that a series of flows may be computed using a single input. This is done by use of a
continuation parameter which determines whether a calculation run terminates or whether data for a further
flow calculation are to be read from the input file.
When it is desired to calculate flows with significant separation it is necessary to perform such a series of
runs with successively higher values of lift coefficient, commencing with a lift coefficient corresponding
to a flow with an attached boundary layer or one with at most a very limited separation. Failure to do this
could lead either to the calculation diverging or (depending on the values of the relaxation parameters in
the numerical solution scheme) could result in inadequate convergence. Choices for the lift coefficient
increments, the relaxation parameter values and the numbers of iterations are usually made in the light of
a number of trials and/or experience.

4.3

Program Output
Results output from the flow calculation are given in three printer files and a dump file for use if a
continuation calculation is carried out. The printer files comprise, respectively, a record of the flow

11

92008
iterations, the main output file of the final aerofoil parameters, including those relating to the boundary
layer and wake, and the final pressure coefficient distribution in a form suitable for use with a graph-plotting
program.

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

Flow calculations cease either when the convergence limit is achieved or the specified maximum number
of iterations has been completed. The results corresponding to the final iteration contained in the main
output file are as follows.

(i)

A record of the geometry mapping, or else a comment that the flow calculation was continued from
a dump.

(ii)

A list of the input parameters.

(iii)

A note of the number of iterations computed and the final values of two convergence parameters.

(iv)

Boundary-layer parameters corresponding to the grid nodes on the aerofoil surface and dividing
streamline (wake).

(v)

Mapping and pressure parameters corresponding to the same aerofoil surface and wake locations.

(vi)

Coefficients of lift, drag and pitching moment, including drag coefficient components (viscous
and wave) as well as the total value.

(vii)

A record of the Mach number, incidence and Reynolds number.

The viscous drag coefficient is calculated from the wake momentum thickness far downstream, using
C D = 2
V

(see, for example, Reference 15).


The wave drag coefficient, C D , is calculated using the MACHCONT procedure (Reference 16), which detects each shock and
W
ascertains its strength at various grid lines (contours), and then evaluates the overall wave drag corresponding to each shock by integration.

12

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

5.

92008
ASSESSMENT OF BVGK USING COMPARISONS WITH WIND TUNNEL DATA
In Reference 3, Ashill et al report on detailed comparisons between results computed using BVGK and
experimental data obtained from wind tunnel tests. The experimental data include corrections to allow
for wind tunnel interference effects. However it should be noted that such corrections can not be perfect,
and that noticeable interference effects may remain, particularly for flows with large supercritical regions.
In Reference 3, surface pressures, overall forces and (in some cases) pitching moments are compared for
eight particular flows involving four aerofoils . Free-stream Mach numbers range from 0.598 to 0.753. The
flows include an attached, subcritical case ( M = 0.598 ) , flows with weak shock waves and fully attached
boundary layers, and flows with strong shocks having various extents of boundary-layer separation. Other,
less detailed, comparisons are presented for overall forces for many further flows. Calculations carried out
both with and without each of the various higher-order viscous terms reveal that these terms have a large
influence for flows close to separation, or with some separation present, and very much smaller effects
for the flows far removed from separation. Including all the higher-order terms considerably improves the
accuracy of the calculation for flows near separation or with separation present.
However, in the case of one higher-order term, the turbulence structure curvature effect, controlled by the
switch ICURVB, Ashill et al conclude that it is better omitted when there is significant separation (i.e. a
separated flow region of significant extent) at the rear of the upper surface. Figures 2 and 3 give two typical
comparisons of calculated and observed pressure distributions for attached and separated flows respectively.
Figure 4 gives a further comparison and illustrates the effect of the ICURVB switch on pressure distributions
for a flow having a separated flow region of several percent chord in extent. Since separated flows are of
principal interest in this Item, in all subsequent comparisons the results from BVGK correspond to all the
higher-order terms being included, except that controlled by ICURVB. Further comparisons of experimental
aerofoil data and BVGK calculations are given by Ashill in Reference 19. These also demonstrate the effects
of changing the ICURVB switch.

5.1

Assessment of Prediction of Separation Boundary


In Figure 5 observed and computed variations with lift coefficient, C L , of the pressure coefficient at the
trailing edge, C p , are shown for the RAE 5230 aerofoil for R 6 10 6 and M 0.74 . Also shown
te
are variations of x s , and CD with C L . Points corresponding to the trailing-edge pressure divergence
criterion, ( C p = 0.05 ) are marked with a cross.
te

It is seen in Figure 5 that the values of C L where the crosses occur are very nearly the same on the observed
and calculated Cp te C L variations. Despite this good prediction, the two C p te C L curves do differ
significantly over the range of C L values considered. As shown in the upper part of Figure 5 some
trailing-edge separation is predicted to occur for this aerofoil even at relatively moderate C L values, and
such separation might influence the form of the Cp te C L variations. It is seen that BVGK fails to predict
the observed slight trailing-edge pressure recovery prior to the rapid fall as separation progresses. (It should
be noted that the relations linking c f and the boundary-layer shape parameter in BVGK are based on

The aerofoil designations are RAE 5225, RAE 5230, RAE 5236 and RAE 2822. All are advanced rear-loaded supercritical aerofoils, the

first three having t/c = 14 per cent, with RAE 2822 having t/c = 12 per cent. The profiles are shown in Figure 1, which includes one further
aerofoil profile also considered in this Item.

Close to separation implies that if either M or C L were increased slightly the flow would be separated (i.e. some separation would
be present). Similarly, far removed from separation, implies that relatively large increases in M or C L could be effected before
separation occurred.

Lift at given incidence, pitching moment, and surface pressures are all better estimated by omitting the curvature term, but for some flows
there may be a slight worsening in the estimation of drag coefficient. Ashill et al. indicate that, since the correction term controlled by
ICURVB utilises the displacement surface curvature to represent the longitudinal flow curvature through the full depth of the boundary
layer, it is perhaps the least justified of the correction terms. This is especially so for regions where the flow has separated.

The values of M for the series of wind tunnel tests presented in this (and other) figures increase slightly with C L ; this increase is a
consequence of the wind tunnel blockage corrections varying with C L . Care has been taken to match the M values used in the
calculations with those of the corresponding experiments.

13

92008
observations of attached flows, and are not reliable near separation. This is one reason why the determination
of separation extent in BVGK is unlikely to be very accurate.)

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

Similar comparisons of BVGK results and experimental data for the same aerofoil (RAE 5230) and Mach
number (0.74), but at the higher Reynolds number of 17.7 10 6, are shown in Figure 6. As for the lower
Reynolds number case, the prediction of the CL value at the point where the trailing-edge pressure
divergence criterion is satisfied is good. Comparing Figures 5 and 6, it is seen that at the higher Reynolds
number, extents of the predicted trailing-edge separation are lower. Also, the observed slight recovery in
trailing-edge pressure (prior to the rapid fall as separation progresses) is virtually absent. Comparison of
the observed and predicted separation boundaries for the RAE 5230 aerofoil, corresponding to the higher
Reynolds number, is shown in Figure 7 for five Mach numbers in the range 0.72 to 0.75. It is seen that the
predictions are very good throughout this Mach number range. Only trailing-edge separation is predicted
for these cases.
Further comparisons of BVGK predictions with wind tunnel experiments are given in Figures 8 and 9 for
the RAE 5234 aerofoil at R 20 10 6 with M 0.70 and M 0.75. It is noted that some minor
extrapolation is necessary for both the experimental and calculated results in order to determine points
corresponding to the separation boundary at M 0.70 . At this lower Mach number the flows are attached
at all but the highest C L values (see upper portions of Figure 8). Even so, there is a difference of about
0.05 between the observed and predicted C p te values for the majority of the range of CL and, again, a
failure of BVGK to predict the slight pressure rise (observed in this case for C L values between 0.7 and
1.0) prior to the rapid decrease which occurs as separation progresses. For both Mach number values, BVGK
overestimates the value of C L on the separation boundary at M 0.70 by about 0.05, and at M 0.75
by about 0.08. At the higher Mach number the observed variation of C pte with CL (Figure 9) is particularly
abrupt at C L 0.8, but no similar rapid variation is predicted by BVGK. The RAE 5234 aerofoil was
designed to yield pressure distributions over the rear upper surface which (for attached flows) appear
(approximately) as two straight line segments, as shown in the example pressure distribution of Figure 10,
with the adverse pressure gradient being significantly higher in the more aft segment. It is known that in
such a situation a separation region extending downstream from the shock grows very rapidly, with
increasing C L , as soon as it reaches the point where the pressure gradient rises a situation corresponding
to flow breakdown (References 20, 21, 22). Using Reference 22, together with data from BVGK, it is
estimated that flow breakdown does indeed occur in this case at a value of C L very close to 0.8 and it thus
seems that this is the cause of the observed abrupt change of Cp te with C L at this point. The reason that
BVGK does not yield a similar abrupt variation of C p te with C L is probably due to the code significantly
under-predicting the separation induced by the shock.
It may be noted that, for conditions on the separation boundary at M 0.75 , BVGK indicates a
trailing-edge separation of some significant extent but no shock-induced separation, whilst at M 0.70 ,
separations both at the shock and near the trailing edge are predicted.
Other comparisons of computed and observed separation boundaries are given in Reference 17. For free
stream Mach numbers of 0.7 and above, the comparisons shown in Figures 6 to 9 and 11 are typical.

Comparisons at the lower Mach number of 0.6, which are presented in Reference 17 and an example of which is shown in Figure 11, are
less well established, due to the paucity of adequately converged BVGK calculations. Comparisons presented in Reference 17 at an even
lower value of M , namely 0.5, are unreliable for a different reason, i.e. the flows contain shocks close to the forced transition location
on the upper surface with the probable result that the shock waves interact with the boundary layer whilst the latter is in a transitional,
rather than a fully turbulent, state.

14


5.2

92008
Assessment of Prediction of Drag at Conditions Near the Separation Boundary
In Figures 5, 6, 8 and 9, data are given permitting a comparison to be made of predicted and measured drag
coefficients at values of C L close to the separation boundary.

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

In Figure 5 it is seen that the calculated CD values for the RAE 5230 aerofoil at M 0.74 and R 6 10 6
are good predictions at both ends of the C L range, but they are over-predictions (approaching 10 percent)
at intermediate C L values. A very similar result is seen in Figure 6. It may be significant that it is in the
range of these intermediate CL values that C p te is increasing with C L a feature (as noted in Section 5.1)
which is not reproduced by the calculations. It appears that at least for some cases where there is a shock
wave of moderate strength (i.e. M 1 < 1.3 ) such as not to induce separation to occur near the shock, (see,
for example, Figures 3 and 4) the actual upper surface boundary-layer development results in rather lower

values of and at the trailing edge than would have occurred had the same overall upper surface pressure
rise taken place gradually (i.e. without a shock). This rather surprising beneficial effect of a shock does
not seem to be reproduced by the calculation scheme as evidenced by comparisons of measured and
calculated C pte . In these cases the relatively high calculated values of overall drag coefficient are also partly
the result of an over-prediction of wave drag . At higher C L values, separation occurs, and viscous drag
increases rapidly. By under-predicting separation, BVGK may then significantly under-predict the viscous
drag component, and thereby under-predict the overall drag coefficient.
A somewhat similar behaviour in C p te is observed for the RAE 5234 aerofoil in Figure 8 for a small range
of CL values around 0.85. The discrepancies between calculated and experimental C D values are generally
smaller in this case, and the measured trailing-edge pressures are lower than those predicted throughout
the C L range. At the highest Mach number for the RAE 5234 aerofoil (Figure 9) there is an under-prediction
of C D for most of the C L range shown. In this case, where the observed C p te values fall rapidly with
increasing CL near the (observed) separation boundary, there is a corresponding rapid rise in drag due
undoubtedly to a rapid rise in the viscous component. As noted, BVGK tends to under-predict such effects
due to its under-prediction of shock-induced separation.

The main reasons for the MACHCONT procedure in BVGK resulting in a wave drag over-prediction are (a) any shock is assumed to be
everywhere normal to the flow and (b) any shock is taken to extend to the aerofoil surface, without allowance for speed reduction in the
boundary layer.

15


6.

92008
RECOMMENDATIONS CONCERNING THE USE OF BVGK AND COMPARABLE VISCOUSCOUPLED CODES

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

In Section 5, results presented have shown that BVGK is capable of providing good predictions for transonic
aerofoil flows close to separation, and also where there are limited regions of separation near the shock
and/or near the trailing edge. The particular consequence of separation selected to define a separation
boundary, namely the trailing-edge pressure divergence, is also a feature that is predicted with reasonable
accuracy by BVGK. Separation boundary prediction is generally within 0.08 in C L of the observed value.
Prediction of C D near the separation boundary within a 10 per cent error is achieved generally. A tendency
of BVGK to under-predict shock-induced separation (and its effects) has been noted.
On the basis of these comparisons (and others, of which those presented are typical) it is recommended that
BVGK may be used to predict both the separation boundary defined by C pte = 0.05, and the associated
C D values. The probable errors of such predictions are likely to be similar to the errors noted in the examples
given although, where aerofoil flows markedly different from those examined are involved, a degree of
caution is necessary. Caution is particularly relevant where separation is affected by shock waves far forward
on the chord.
Since the performance of a viscous-coupled code is linked to the principal features it embodies, similar
recommendations may be made for any viscous-coupled code that includes the determining features of
BVGK listed in Section 3. Where possible, however, any such viscous-coupled code should first be
evaluated against experimental data, by making comparisons similar to those in Section 5, since the details
of the calculation scheme might have a significant influence on the computed results.

7.

REFERENCES

1.

COLLYER, M.R.

An extension to the method of Garabedian and Korn for the calculation


of transonic flow past an aerofoil to include the effects of a boundary
layer and wake. ARC R & M 3828, 1978.

2.

LOCK, R.C.
WILLIAMS, B.R.

Viscous-inviscid interactions in external


Aerospace Sci. Vol. 24, pp. 51-171, 1987.

3.

ASHILL, P.R.
WOOD, R.F.
WEEKS, D.J.

An improved semi-inverse version of the viscous Garabedian and Korn


method (VGK). RAE TR 87002, January 1987.

4.

CARTER, J.E.

A new boundary-layer inviscid iteration technique for separated flow.


AIAA Paper 79-1450, 1979.

5.

ESDU

Methods for estimating the pressure distribution on a two-dimensional


aerofoil in viscous transonic flow. Engineering Sciences Data Unit
Transonic Data Memor. 81019, March 1982.

6.

LOCK, R.C.
COLLYER, M.R.

A note on the prediction of viscous effects on aerofoils with particular


reference to transonic flows with shock waves. RAE tech. Memor.
Aero. 1735, October 1977.

7.

LOCK, R.C.
COLLYER, M.R.

An extension to the method of Garabedian and Korn. In Numerical


methods for the computation of transonic flows with shock waves.
Fredr. Vieweg und Sohn, Braunschweig, 1980.

16

aerodynamics.

Prog.

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

92008
8.

THWAITES, B.

Approximate calculation of the laminar boundary layer. Aero. Quart.,


Vol. 1, pp. 245-280,1949.

9.

STEWARTSON, K.

Correlated compressible and incompressible boundary layers. Proc. R.


Soc. Lond. Series A, Vol. 200, pp. 84-100, 1949.

10.

GREEN, J.E.
WEEKS, D.J.
BROOMAN, J.W.F.

Prediction of turbulent boundary-layers and wakes in compressible flow


by a lag-entrainment method. ARC R & M 3791, 1973.

11.

EAST, L.F.

A representation of second-order boundary layer effects in the


momentum integral equation and in viscous-inviscid interactions. RAE
tech. Rep. 81002, 1981.

12.

LOCK, R.C.

Prediction of the drag of wings at subsonic speeds by viscous/inviscid


interaction techniques. AGARD Rep. R-723 Aircraft drag prediction
and reduction, 1985.

13.

COLLYER, M.R.

A users guide to a program for computing viscous transonic flows past


aerofoils (Garabedian and Korn method). RAE tech. Memor. Aero.
1693, September 1976.

14.

CATHERALL, D.
FOSTER, D.N.
SELLS, C.C.L.

Two-dimensional incompressible flow past a lifting aerofoil. RAE tech.


Rep. 69118, June 1969.

15.

YOUNG, A.D.
SQUIRE, H.B.

The calculation of the profile drag of aerofoils. ARC R & M 1838,


1938.

16.

BILLING, C.M.
BOCCI, A.J.

The MACHCONT method for calculating the wave drag of a


two-dimensional aerofoil. Aircraft Research Association Memor. 272,
1986.

17.

FREESTONE, M.M.

Direct prediction of separation boundaries and associated drag on


aerofoils using two viscous-coupled calculation methods. Ministry of
Defence S and T Memor. 1-90, April 1990.

18.

ESDU

A method of estimating a separation boundary for two-dimensional


aerofoil sections in transonic flow. Engineering Sciences Data Unit
Transonic Data Memor. 81020, with Amendment A, June 1990.

19.

ASHILL, P.R.

Calculation and measurement of transonic flows over aerofoils with


novel rear sections. Proc. Int. Congress of Aerospace Sciences,
88-3.10.2, 1988.

20.

ESDU

Introduction to transonic aerodynamics of aerofoils and wings.


Engineering Sciences Data Unit Transonic Data Memor. 90008, April
1990.

21.

FULKER, J.L.
ASHILL, P.R.

A study of the factors influencing shock-induced separation on swept


wings. RAE tech. Rep. 83088, 1983.

22.

ESDU

A method of estimating a flow breakdown boundary for aerofoils and


swept wings in transonic flow. Engineering Sciences Data Unit
Transonic Data Memor. 91021, October 1991.

17

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.


92008

FIGURE 1 PROFILES OF AEROFOILS CONSIDERED

18

92008

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

FIGURE 2 COMPARISON BETWEEN COMPUTED AND OBSERVED PRESSURE DISTRIBUTIONS


FOR ATTACHED FLOW
RAE 5225, M = 0.598 , C L = 0.433 , R = 20 10 6

19

92008

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

FIGURE 3 COMPARISON BETWEEN COMPUTED AND OBSERVED PRESSURE DISTRIBUTIONS


FOR SEPARATED FLOW
RAE 5225, M = 0.737 , CL = 0.659 , R = 6 10 6

20

92008

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

FIGURE 4 COMPARISON BETWEEN COMPUTED AND OBSERVED PRESSURE DISTRIBUTIONS


RAE 5230, M = 0.736 , C L = 0.706 , R = 6 10 6

21

92008

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

FIGURE 5 COMPARISON OF COMPUTED AND OBSERVED AERODYNAMIC PARAMETERS


RAE 5230, M 0.74 , R 6 10 6

22

92008

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

FIGURE 6 COMPARISON OF COMPUTED AND OBSERVED AERODYNAMIC PARAMETERS


RAE 5230, M 0.74 , R 17.7 10 6

23

92008

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

FIGURE 7 COMPARISON OF COMPUTED AND OBSERVED SEPARATION EFFECTS BOUNDARIES


FOR THE RAE 5230 AEROFOIL AT R 17.7 10 6

24

92008

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

FIGURE 8 COMPARISON OF COMPUTED AND OBSERVED AERODYNAMIC PARAMETERS


RAE 5234, M 0.70 , R 20 10 6

25

92008

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

FIGURE 9 COMPARISON OF COMPUTED AND OBSERVED AERODYNAMIC PARAMETERS


RAE 5234, M 0.75 , R 20 10 6

26

92008

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

FIGURE 10 COMPARISON BETWEEN COMPUTED AND OBSERVED PRESSURE DISTRIBUTIONS


RAE 5234, M = 0.736 , C L = 0.606 , R = 20 10 6

27

92008

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

FIGURE 11 COMPARISON OF COMPUTED AND OBSERVED SEPARATION EFFECTS BOUNDARIES


FOR THE RAE 5234 AEROFOIL AT R 20 10 6

28

92008
THE PREPARATION OF THIS DATA ITEM

ESDU product issue: 2007-02. For current status, contact ESDU. Observe Copyright.

The work on this particular Data Item was monitored and guided by the Transonic Aerodynamics
Committee, which first met in 1960 and now has the following membership:
Chairman
Mr D.R. Stanniland

Aircraft Research Association

Vice-Chairman
Mr D.R. Holt

British Aerospace Defence Ltd, Brough

Members
Mr S.E. Allwright
Mr M.C.P. Firmin
Mr J.L. Fulker
Mr G.R. Hargreaves
Mr M.D. Hodges
Mr A. Pagano
Dr A.J. Peace
Mr G.R. Richards
Mr B. Probert

British Aerospace Regional Aircraft Ltd, Woodford


Independent
DRA Aerospace Division, Bedford
British Aerospace Regional Aircraft Ltd, Woodford
Independent
British Aerospace Airbus Ltd, Bristol
Aircraft Research Association
British Aerospace Defence Ltd, Kingston
British Aerospace Defence Ltd, Preston

The technical work involved in the assessment of the available information and the construction and
subsequent development of the Data Item was undertaken by
Dr M.M. Freestone

Senior Engineer.

The person with overall responsibility for the work in this subject area is Mr D.J. Mitchell.

29

Potrebbero piacerti anche