Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

JOURNAL OF AIRCRAFT

Vol. 56, No. 2, March–April 2019

Control of Transonic Buffet by Shock Control Bumps on


Wing-Body Configuration

R. Mayer,∗ T. Lutz,† and E. Krämer‡


University of Stuttgart, 70569 Stuttgart, Germany
and
J. Dandois§
ONERA, 92190 Meudon, France
DOI: 10.2514/1.C034969
The aerodynamic shock buffet phenomenon limits the flight envelope of typical transonic aircraft at high Mach
number or incidence. For future aircraft design, its correct prediction by numerical methods and control strategies
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

are of great interest. In the present paper, simplified industrial applicable design guidelines for shock control bumps
are summarized and assessed based on numerical (unsteady) Reynolds-averaged Navier–Stokes simulations of a
wing-body configuration. The design guidelines cover both design objectives, wave drag reduction at dash flight, and
buffet delay/alleviation. Three different shock control bumps have been designed: a two-dimensional bump for wave
drag reduction at dash flight, a two-dimensional bump for delay of buffet onset, and a three-dimensional bump array
with the intention of combining both design objectives. The two-dimensional bumps mark the maximum potential for
each design objective with a drag reduction of 9 drag counts at dash flight and a delay of buffet onset by 4 lift counts.
Even though the three-dimensional bump array combines both design objectives, it does not perform as well as
expected due to massive crest-flow separation and a (so far) minor role of the vortical wake of three-dimensional
bumps in delaying buffet onset.

Nomenclature T st = stagnation temperature, K


bbump = bump width, %c t = time, s
bspacing = spanwise spacing between two adjacent three- xs = chordwise shock position, %c
dimensional bumps, %c y1 = first cell height
c = airfoil chord, m α = angle of attack, deg
cD = drag coefficient Δxsc = chordwise distance between shock position in
cf;x = skin friction coefficient in streamwise direction baseline flow and start of bump crest plateau, %c
cL = lift coefficient Δx2D−3D = shift of bump position for transformation from two-
c^ L = amplitude of lift coefficient variation dimensional to three- dimensional bump, %c
cP = pressure coefficient η = span, %
H12 = boundary-layer shape factor κ = Specific heat ratio
hc = crest height, %c Λ = wing sweep, deg
khc = scaling factor for crest height transformation from Λs = wing sweep at shock position, deg
two-dimensional to three- dimensional bump σ = shock deflection angle, deg
koverlap = factor for overlapping of shock structures of adjacent τ = skew angle for three-dimensional bump, deg
three-dimensional bumps Ψ = rotation angle for three-dimensional bump, deg
L∕D = lift-to-drag ratio ωx = streamwise vorticity, 1/s
lbump = bump length, %c
lLE = bump length upstream of shock, %c
M1 = preshock Mach number I. Introduction
M∞ = freestream Mach number
0
prms
pst
=
=
root mean square of pressure fluctuations, Pa
stagnation pressure, Pa
T HE flight envelope of an aircraft frames all safe operating
conditions. The cruise condition of an airliner is typically set at
high transonic speeds. With increasing loading or flight speed,
PSD = power spectral analysis transonic shock buffet limits the flight envelope. This aerodynamic
q0 = dynamic pressure phenomenon stems from a strong shock-wave/boundary-layer inter-
Re = Reynolds number action on the suction side of the wing and leads to flow instability,
St = Strouhal number, f ⋅ cMAC ∕U0  which can excite structural vibrations and thereby compromise the
structural integrity of the aircraft. Consequently, a margin of 30% on
Received 15 February 2018; revision received 10 August 2018; accepted
the lift coefficient at cruise conditions must be respected by design
for publication 17 December 2018; published online 29 January 2019. standards [1]. In consideration of reducing this safety margin for
Copyright © 2019 by R. Mayer, T. Lutz, E. Krämer, and J. Dandois. Published future aircraft design, current research is focused on the reliable
by the American Institute of Aeronautics and Astronautics, Inc., with prediction of transonic wing buffet with computational fluid
permission. All requests for copying and permission to reprint should be dynamics (CFD) methods. Brunet and Deck [2] performed Reynolds-
submitted to CCC at www.copyright.com; employ the eISSN 1533-3868 averaged Navier–Stokes (RANS) and detached-eddy simulations
to initiate your request. See also AIAA Rights and Permissions www.aiaa. (DESs) on a wing-body configuration at buffet condition and
org/randp. found good agreement with corresponding experiments carried out
*Ph.D. Student, Institute of Aerodynamics and Gas Dynamics, by Caruana et al. [3]. With a zonal approach, they limited the
Pfaffenwaldring 21.

Senior Researcher, Institute of Aerodynamics and Gas Dynamics, scale-resolving methods to regions of separated flow. While they
Pfaffenwaldring 21. successfully captured the massively separated and unstable regions

Professor, Institute of Aerodynamics and Gas Dynamics, Pfaffenwaldring 21. on the outboard portion of the wing with DES simulations, RANS
§
Research Engineer, Applied Aerodynamics Department, 8 rue des methods underestimated the extent of separation, and hence the
Vertugadins. authors suggest that accurate resolution of turbulent scales may be a
556
MAYER ET AL. 557

necessity to model correctly the three-dimensional (3D) buffet are capable of both wave drag reduction at dash flight and delay of
instability. In contrast, Sartor and Timme [4–6] demonstrated by a buffet onset. However, because of the typically large variation in
direct comparison of unsteady RANS (URANS) and scale-resolving shock strength and position between a representative dash condition
delayed-DES simulations that both approaches are capable of and buffet onset of modern airliners, such 2D bumps have to be
describing the main features of transonic wing buffet, like large-scale adaptive in height and position in order to combine both features. The
flow separation and a broadband pressure spectrum on the wing intention of 3D bumps is to overcome this need for an adaptive flow
surface. A comparison with corresponding experimental data by control device. Being positioned for optimum drag reduction at dash
Lawson et al. [7] indicates, however, that high-fidelity approaches flight, it is intended to use the streamwise vortex formation emerging
improve the quality of results. Assessing the effects of sweep angle from the bumps (see, e.g., Refs. [23,25–27]) to suppress incipient
and span on the buffet instability mechanism, Iovnovich and Raveh flow separation aft of the bump and thereby delay buffet onset.
[8] conducted a similar numerical study and also successfully A common approach for studying different bump shapes would be to
predicted transonic buffet on swept wings with URANS simulations. use numerical optimization schemes. However, for a wing-body
Based on these results, they derived a working model for the configuration, this approach comes with high computational costs. In
mechanisms of transonic wing buffet. Sartor and Timme [4] highlighted consequence, industrial applicable guidelines for a preliminary bump
the strong effect of numerical methods, especially turbulence modeling, design have been derived based on airfoil studies. It is the intention
on the prediction of this flow phenomenon. of the present manuscript to present them and demonstrate their
In consideration of the mentioned studies and due to the large applicability.
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

number of intended design evaluations, effort has been undertaken After a brief summary of the wing-body configuration and the
to improve numerical results of URANS simulations of transonic numerical setup, focus is first put on a comparison between
buffet. Goncalves and Houdeville [9] mention an overproduction of experiment and numerical simulation of the baseline flow. Here, the
eddy viscosity for common eddy-viscosity turbulence models, which intention is to verify that the numerical setup is capable of predicting
limits the development of flow unsteadiness, due to a break of the the main buffet characteristics correctly. Section V then presents the
underlying turbulence equilibrium assumption in the presence of design guidelines, and Sec. VI gives an overview of the bump design.
coherent flow structures as typically observed in transonic buffet. Finally, the effect of SCBs on wave drag reduction at dash flight and
Hence, the proposed modification of the strain adaptive formulation of buffet delay/alleviation of the wing-body configuration is analyzed
the Spalart–Allmaras one-equation turbulence model (SA-SALSA) by and discussed.
Zimmermann et al. [10] intends to reduce turbulent viscosity in the
boundary layer around and aft of the shock wave in case of buffet.
Besides the correct prediction of transonic wing buffet with CFD II. AVERT Model
methods, there is a strong interest in control mechanisms for this flow For the present study, the AViation Emission Reduction
phenomenon. The most prominent approach is the application of Technologies (AVERT) model (see Fig. 1a), designed by ONERA
vortex generators (VGs). While numerous studies [11–14] have and extensively tested in the S2Ma wind tunnel, has been used. This
demonstrated their ability to delay and alleviate buffet (sometimes wing-body model is described in detail by Dandois et al. [15] and
even to suppress buffet entirely), classic mechanical VGs typically mainly represents an evolution of the supercritical OAT15A airfoil
increase viscous drag and deteriorate cruise performance. Fluidic test case described by Jacquin et al. [28], as most of the wing is based
VGs [13,15] seem to overcome this problem. A more profound on this airfoil. The AVERT model has a half-wing span of 1.225 m
overview of different buffet control strategies (including suction and and a leading-edge sweep angle of Λ  30 deg. The root chord
blowing) is given by several reviews [16–18]. equals 0.450 m, the wing tip chord is 0.225 m, and the mean
The current paper assesses the potential of two-dimensional (2D) aerodynamic chord (MAC) is 0.3375 m. Transition is fixed at x∕c 
and 3D shock control bumps (SCBs) for buffet delay/alleviation on a 7% on both sides of the wing as well as on the fuselage. The M∞ − α
wing-body configuration. Previous studies [19–24] indicate that plane in Fig. 1b gives an overview of the experimentally determined
SCBs might yield a double benefit. While improving the buffet buffet boundary.
behavior, they can simultaneously be used for wave drag reduction at Oil-flow visualizations at different incidence angles have been
transonic dash flight (which is their original and more prominent field used to analyze flow separation. While the flow is fully attached at
of application). Whereas neither wave drag nor buffet is an issue for α; M∞   2.5 deg; 0.826, there is massive flow separation at
current aircraft, such flow control devices can be considered as α; M∞   3.5 deg; 0.826. Unsteady pressure measurements as
(potential) enablers for alternative wing planform philosophies (e.g., well as measurements with a balance consistently predicted buffet
laminar flow wings and more general, increased taper to reduce wing onset at α ≈ 3.0 deg for the reference Mach number of M∞  0.826
mass). Comparing different bump geometries with RANS and considered here. All presented wind tunnel test data are wall
URANS simulations, the objective of this paper is to verify whether corrected.
trends found for airfoil buffet control by SCBs (see [24]) can be All simulations presented in this study have been carried out at
transferred from an infinite-span, unswept wing to a wing-body M∞  0.826 with a Reynolds number of ReMAC  2.84 ⋅ 106 , a
configuration. Similar to the study by Mayer et al. [24], two different stagnation pressure pst  60; 000 Pa, and a stagnation temperature
approaches for buffet bumps are considered. In general, 2D bumps T st  293 K. It has to be noted that no model deformation due to

0.88

0.86

0.84
buffet
0.82
M

0.80 no buffet

0.78
buffet onset
0.76
-1 0 1 2 3 4 5 6
[º]
a) AVERT wind tunnel model b) Measured buffet boundary
Fig. 1 AVERT model and experimental results (adapted from Dandois et al. [15]).
558 MAYER ET AL.

aerodynamic loads has been taken into account for the CFD aerodynamic performance is rather small (ΔcD ≈ 1 dc; see Fig. 2),
simulations. local flow features around the 3D SCB array (like vortical structures)
are not properly resolved by the default mesh, and hence their effect
III. Numerical Method on, e.g., buffet onset is not correctly represented. In consequence, the
fine grid resolution has been used to study the local flow topology of
All simulations presented have been carried out with the 3D SCBs inside the steady flow regime and their effect on buffet
compressible, block-structured (U)RANS solver FLOWer [29]. With onset. However, because of the high computational costs of time-
a cell-centered approach, the simulations have been carried out up to a resolved flow simulations, the default grid resolution has been used
density-based residual of less than 10−5 or up to at least 10 buffet to study the effect of the 3D bump array on shock oscillations inside
cycles in fully established flow (based on the most dominant the buffet regime at α  3.5 deg. This is clearly a limitation of
frequency of the broadband pressure spectra). The main equations are the presented results but can be justified by the inferior effect of
spatially and temporally discretized with second-order-accurate local flow structures (like the vortical wake of 3D SCBs) on the
central schemes, and for convergence acceleration, multigrid cycles control of shock oscillations inside the buffet regime, as shown by
and residual smoothing are applied. In case of time-resolved Mayer et al. [24].
simulations, an implicit dual-time-stepping scheme with a predefined To apply the bumps to the suction side of the wing, affected
physical time step size of Δt  1.2 ⋅ 10−5 s and 30 to 100 pseudo- boundary layer and near far-field blocks were removed, the
time-steps per physical time step has been applied. Considering underlying wing geometry was modified, and then the removed
dominant frequencies around St ≈ 0.4 (see Sec. IV), this resolves one
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

blocks were rebuilt. A script-based routine thereof ensures a


buffet period by approximately 265 time steps. As the shock-wave/ constantly high mesh quality, almost independent of the bump
boundary-layer interaction is a key aspect of transonic buffet, geometry.
turbulence modeling strongly affects numerical simulations of the
buffet phenomenon [9,30,31]. Because of the large parameter space
of this study and with regard to high computational costs associated IV. Baseline Flow
with higher-order methods, the one-equation turbulence model SA- At the considered reference Mach number of M∞  0.826, the
SALSA [32] (with slightly adopted model constants for improved local supersonic region on the suction side of the wing is terminated
transonic buffet simulations; see [10]) has been used. The applied by a shock. Outboard of about 45% of span, a single shock wave
numerical setup was first tested for several airfoil buffet test cases; see decelerates the flow to subsonic speeds, and the flowfield is quasi-
also [10]. Similar to the experiments, transition is fixed at x∕c  7% 2D. In contrast, a λ-shock system is present on the inboard section
on both sides of the wing as well as on the fuselage of the (see Fig. 3a), and in the vicinity of the fuselage, the flow becomes
AVERT model. highly 3D.
A structured H-C-type, multiblock grid with boundary-layer Figure 3 compares two pressure distributions at η  55 and 72.5%
refinement, providing a first cell height of y1 < 1, has been used for to experimental results for a moderate incidence of α  2.5 deg,
all simulations. Four hundred one points in the chordwise direction around buffet onset at α  3.0 deg and deep inside the buffet regime
around the wing (including 17 points on the blunt trailing edge) and at α  3.5 deg. The overall agreement with the experimental data
the same number of points per MAC in the spanwise direction lead to is reasonably good. The pressure side, the pressure plateau on the
almost quadratic surface mesh cells on the wing. The fuselage is suction side, as well as the shock position are well captured over this
resolved by 129 points in the radial direction and 409 points in the wide range of inflow conditions. Only between the shock wave and
axial direction. The far field extends to 40 aircraft lengths in all the trailing edge are there minor differences between experimental
directions, leading to a total mesh size of this default grid of and numerical results.
approximately 7.3 ⋅ 107 cells. In Fig. 4, the drag and lift polars of both experiment and baseline
Independence of the simulation results from the grid resolution has CFD are plotted. Whereas the numerically determined drag polar
been verified by a comparison with a significantly refined mesh that shows very good agreement with the experimental results, a shift in
consists of approximately 2.1 ⋅ 108 cells. Here, the local grid the lift polar can be observed. The reason for this shift to higher lift
resolution of the wing surface is equal to the grid resolution of the coefficients is mostly due to an offset in the incidence angle between
study of SCBs on an infinite-span, unswept wing presented by Mayer experiments and numerical simulation. Further support of this
et al. [24]. Considering the global aerodynamic coefficients of the reasoning is given by the slightly more aft predicted shock position in
clean AVERT model at α  2.5 deg, changes of 0.017% in the lift comparison to the experimental results in the steady flow regime (see
coefficient ΔcL  1.02 ⋅ 10−4  and 0.024% in the drag coefficient Figs. 3b and 3c). Beyond that, a certain uncertainty of the CFD results
ΔcD  7.5 ⋅ 10−6  indicate grid-independent results of the default (especially at the wing-body intersection) might affect the results.
mesh resolution. Whereas consequently the default grid resolution is However, as the present study focuses on the effect of bumps on
sufficient for flow simulations of the clean wing as well as with 2D buffet (consideration of Δ values), no further investigation of the
SCBs applied, Fig. 2 indicates that this grid resolution is insufficient origin of this shift in lift has been carried out.
for predicting all flow structures around an array of 3D bumps In the wind tunnel measurements, buffet onset has been found at
correctly. Even though the effect of the refined grid on global α ≈ 3.0 deg for M∞  0.826. In the present paper, the Δα criterion is

Y Y
X X

Z Z

cD = 0.03066 cD = 0.03054
L / D = 19.96 L / D = 20.04
a) Default mesh resolution b) Refined mesh resolution
Fig. 2 Surface grid resolution of 3D SCBs on suction side of AVERT wing. Aerodynamic coefficients at M∞  0.826 and cL  0.6118.
MAYER ET AL. 559

Y -1.0
Z
X
-0.5
55 %
72.5 %

cP
0

= 55 %
0.5 = 72.5 %
contour: c p

0 0.2 0.4 0.6 0.8 1.0


x/ c
a) Shock structure with cut sections (CFD) b) α = 2.5°

-1.0 -1.0
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

-0.5 -0.5
cP

cP
0 0

= 55 % = 55 %
0.5 = 72.5 % 0.5 = 72.5 %

0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
x/ c x/ c
c) α = 3.0° d) α = 3.5° (time averaged)
Fig. 3 Shock structure of baseline AVERT model with mid- and outboard cP distributions. M∞  0.826. Symbols: experiment; lines: CFD.

0.70 0.70

0.65 0.65
L

L
c

0.60 Experiment 0.60


CFD

0.55 0.55
0.02 0.03 0.04 0.05
2.0 2.5 3.0 3.5
cD [°]
Fig. 4 Drag and lift polar of AVERT model in baseline configuration M∞  0.826 with estimated buffet onset based on Δα criterion.

used to relate buffet onset to the lift-curve slope break; see, e.g., [7]. boundary-layer separation increases in extent. Further increasing the
As indicated in Fig. 4, the linear part of the lift polar is shifted by incidence α  3.5 deg leads to massive flow separation on the
Δα  0.1 deg, and its intersection with the actual lift curve midboard section of the wing. With regard to the experimental results
in the nonlinear regime is defined as indication for buffet onset at α  3.5 deg, Dandois et al. [15] noted that the flow is fully
(including a mandatory margin to the onset of physical buffet). separated aft of the shock. While shock-induced flow separation at
Considering this criterion, buffet onset is predicted at α ≈ 3.0 deg the midboard region 40% ≤ η ≤ 80% is correctly predicted by the
M∞  0.826; cL  0.672 by the numerical simulations. Incipient numerical simulation, there is also shock-induced flow separation at
flow separation, marking the onset of the nonlinear lift regime, is the outboard region.
consistently predicted, and hence there is excellent agreement Time-resolved flow simulations at α; M∞   3.5 deg; 0.826
between experimental and numerical buffet onset. reveal that the outboard region is the main source of unsteady flow
Figure 5 compares experimental oil-flow images with CFD structures in the CFD simulation; see Fig. 6a. While an amplification
predicted (mean) flow separations at different incidence angles. At of pressure fluctuations in the frequency band of 0.2 ≤ St ≤ 0.6 has
α  2.5 deg, the numerical simulations show a local, shock-induced been observed at η  60% in the wind tunnel tests due to transonic
separation of the boundary layer. It has to be noted that the oil-flow wing buffet (see Fig. 6b), no flow unsteadiness is resolved by the
technique used in the wind tunnel measurements only indicates present numerical setup at this spanwise section. However, an
trailing-edge separation as red paint has to be sucked from the analysis of the power spectral density (PSD) at η  85% highlights
pressure side to the suction side of the wing. Consequently, local the amplification of dominant frequencies in the same broadband
shock-induced boundary-layer separation cannot be seen in the spectrum, which is typical for transonic wing buffet [18]. The
experimental results. With increasing incidence, the wind tunnel generally lower energy level indicates that the numerical simulations
experiments indicate incipient trailing-edge separation at the do not predict buffet as strong as in the experiment. In combination
midboard section α  3.0 deg. In contrast, the numerical results with the more outboard origin of flow unsteadiness, it is supposed
do not predict trailing-edge separation yet. Here, the shock-induced that common weaknesses associated with turbulence modeling as
560 MAYER ET AL.

= 2.5°

= 2.5°
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

= 3.0°

= 3.0°

= 3.5°

= 3.5° 100 80 60 40 20
[%]
a) Experiment: oil-flow (Dandois et al. [15]) b) CFD: cf,x < 0
Fig. 5 Flow separation on suction side of baseline wing at M∞  0.826. Contour lines in CFD results represent cP .

p′rms
[−] 500
q0 0 0.01 0.02 0.03
Experiment: x / c = 0.45
400 Experiment: x / c = 0.70
PSD [Pa2 / Hz]

Fig. 6b: Experiment CFD: x / c = 0.45


300 CFD: x / c = 0.70
Fig. 6b:
CFD 200

100

0
0 0.2 0.4 0.6 0.8 1.0
St
100 80 60 40 20
[%]
a) CFD-predicted pressure fluctuations b) Pressure spectra of baseline configuration. Experiment:
p′rms/q0 on AVERT wing. Contour lines: cP η = 60%, CFD: η = 85%
Fig. 6 Characterization of buffet related pressure fluctuations at M∞  0.826 and α  3.5 deg.

part of the RANS approach are the main source of this deviation from V. Design Guidelines
experimental to numerical results. König [33] originally presented guidelines for the (industrial)
However, the mechanism of the transonic wing buffet observed
design of SCBs, which have been further extended by the authors.
here is in excellent agreement with the model proposed by Iovnovich
The design guidelines originate from the desire to avoid excessive
and Raveh [8]. Outboard of η ≈ 70%, buffet cells propagate toward
the tip of the wing, and for a fixed spanwise position, there is an numerical optimization for a preliminary bump design on a realistic
almost periodic shock movement; see Fig. 7. wing and base on the observation that the performance of SCBs
In conclusion, there are several deviations from the experimental mainly depends on the flow turning they exert at the shock foot and
results, but the present numerical setup predicts transonic wing buffet their position in relation to the shock wave [34,35]. In consequence,
reasonably well and forms in consequence a good platform for the numerous 2D bump shapes have been numerically single-point
subsequent study of the effect of shock control bumps on transonic optimized on different airfoils (conventional NACA0012, super-
buffet of a wing-body configuration. critical OAT15A, and laminar flow OALT25 airfoil) and at different
MAYER ET AL. 561

1.2 A. Extraction of Baseline Flow Properties


The maximum preshock Mach number M1;n represents a reasonable
indicator for the shock strength and is an easy-to-obtain quantity from
1.0
flow simulations. Considering an isentropic flowfield for simplicity,
cf , x < 0 M1;n can be approximated from the pressure distribution according to
Eq. (1) with the minimum pressure level cp;min (extracted parallel to the
0.8
inflow direction and referenced to inflow conditions) as it is
encountered just upstream of the shock. Independence of the design
t · 102 [s]

shock wave

guidelines from wing sweep is ensured by considering the cosine rule,


0.6
as discussed by Obert [1]. Because the shock region is most critical for
the bump’s performance, the local sweep at the shock position Λs is
considered. The shock position is also extracted from the pressure
0.4
distribution at the location of δcp ∕δxmax in the shock region,

cf , x > 0
M1;n 
0.2
s
    
2 κ −1 2 κ 2 1−κ∕κ
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

⋅ 1 ⋅M∞ ⋅cos Λs ⋅ 1cp;min ⋅ ⋅M∞


2
−1
0 κ −1 2 2
20 60 80 40 100 (1)
x [%clocal ]
Fig. 7 Time-dependent flow separation topology on AVERT wing B. Design of 2D SCB
suction side at η  76% inside the buffet regime at α  3.5 deg
M∞  0.826. Figure 8 clearly highlights that with increasing M1;n the optimal
bump height almost linearly increases and the optimal bump position
moves farther aft on the airfoil. To enhance the bump’s robustness,
inflow conditions for the minimum drag coefficient at constant lift i.e., its performance at bump off-design condition over a range of
with the optimization environment described by Mayer et al. [36]; see varying inflow conditions, König [33] suggests a reduction of crest
Fig. 8. The numerical setup for the bump optimizations is presented in height to about 60–90% (gray shaded area) of the optimal height
[24]. The NACA0012 airfoil has been analyzed in fully turbulent value at the design point in order to get an overall well-balanced result
flow. For the OAT15A airfoil, transition has been set in accordance to between bump design point performance and robustness. Figure 8b
the experiments by Jacquin et al. [28] (fixed at x∕c  7% on both relates the optimum bump position (reference point: start of crest
sides of the airfoil). For the laminar flow OALT25 airfoil, transition plateau of wedge SCB) to the shock position at bump design
on the pressure side has been fixed at x∕c  7% (see experiments by condition on the clean airfoil/wing.
Brion et al. [37]). In this experimental study by Brion et al. [37], it has Again, the effect of wing sweep has to be considered to obtain the
been observed that boundary-layer transition on the suction side is final bump design parameters (referring to the local chord length in
triggered by the shock wave. In consequence, it has been assumed the inflow direction).
that with SCB applied the pressure rise occurring at the bump’s ramp
triggers transition. Hence, and for enhanced numerical stability, hc  hc;n ⋅ cos Λs
transition has been fixed on the suction side of the OALT25 airfoil at
the start of the bump shape. The underlying wedge bump geometry is Δxsc  Δxsc;n (2)
described by Mayer et al. [24], and a symmetric bump (ramp length
equals tail length) has been selected. In consideration of integration In a recent publication [24], the authors outlined how to modify
issues of bumps on a real wing, the bump length has been fixed to such bumps for improved buffet behavior. Concerning the two most
25%c (which is a common value for bump studies [38]). In the important bump parameters, crest height and bump position, it was
following, an overview of the guidelines is given. The design process found that the crest height tunes the control strength of buffet bumps
involves the following steps: analogous to performance SCBs. However, for the optimum crest
1) Extract the relevant flow properties from the baseline flow. position, two distinctive locations have been found. Referring to the
2) Design a 2D SCB. most downstream shock position for varying incidence on the clean
3) Transfer the 2D SCB into a 3D SCB array (optional). airfoil/wing xs;max : , it was found that for delayed buffet onset
4) Perform minor adaption to the local flowfield. (based on the Δα criterion) the crest of the bump should be 7–8%c

1.0
h c,n = 3.984 · M 1, n − 4.608 10 x sc,n = 47.197 · M 1, n − 57.293
R = 0.934 R = 0.875
crest offset xsc, n [%cn ]
crest height hc, n [%cn ]

0.8 8 baseline
shock

0.6 6
xsc
4
0.4
2
0.2
0

0 -2
1.25 1.30 1.35 1.40 1.15 1.20
1.15 1.20 1.25 1.30 1.35 1.40
M1, n M1, n
a) Bump crest height b) Bump crest offset to clean shock
Fig. 8 Bump design guidelines for prediction of bump height and position. ▪: NACA0012; ▴: OAT15A; •: OALT25; numerical optimization of bump
position and crest height for 2D wedge SCB with lbump  25%c.
562 MAYER ET AL.

farther downstream. For alleviation of lift oscillations inside the bump width-to-spacing ratio. It has to be noted that the underlying
buffet regime, there are two equally optimal positions that involve data basis is limited and this study is constrained to the OAT15A
different flow control mechanisms. These are 2–3%c upstream and airfoil so that the derived scaling factors should be considered as an
10–12%c downstream of xs;max: . For further details, see [24]. initial guess for a subsequent adaption of 3D bumps to the local
flowfield. However, the trends are clear with Pearson correlation
C. Transfer of 2D SCB into 3D SCB Array coefficients of R  0.984 (Fig. 10a) and R  0.985 (Fig. 10b).
Whereas the previously shown trends were derived for 2D SCBs, To adapt 3D bumps to the local flow on a swept wing, Jones et al.
[40] suggested aligning the bump to the local flow condition just
König [33] also presented guidelines for transforming a 2D SCB into
upstream of the shock wave by rotation. As part of the present study,
a 3D SCB. For a given bump length (here, 20–30%c), Bruce and
different modifications of 3D SCBs (skew and rotation) have been
Colliss [38] suggest a length-to-width ratio of 3  1. To determine
analyzed on the AVERT model. The center of rotation and skew is the
the spacing of the 3D bump, König [33] considered the shock
start of the crest plateau at the centerline of the bump. Whereas the
structure of a semicone and demonstrated that the angle of the
suggested rotation improves the flow structure, a minor misalignment
spanwise decaying shock wave σ is (for typical bump ramp angles of
around the crest of the bump has been observed due to a skewed
<10 deg) almost independent of the bump ramp angle (see also [39])
alignment of flow onto the ramp with the shock front over span; see
and solely depends on the preshock Mach number M1;n , as shown in
Fig. 11. In consequence, it is proposed to additionally skew the bump
Fig. 9a. With the idea that the oblique shock structures of two
in such a way that the crest of the rotated bump is parallel to the local
adjacent 3D SCBs (see Fig. 9b) should overlap to some extent (e.g.,
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

shock front. However, this adaption has only minor effects on the
50% → koverlap  0.5), König derived Eq. (3) for approximation of
overall bump performance.
the spacing:

bspacing  D. Minor Adaption to Local Flowfield


  It is important to note that the design procedure outlined previously
bbump ∕2  lLE tan σ bbump ∕2  lLE tan σ leads to a preliminary SCB shape; i.e., it represents a reasonable
1 − koverlap  ⋅ 
1  tan Λs tan σ 1 − tan Λs tan σ starting point for assessing the benefit of SCBs on the airfoil/wing
(3) in consideration, but it does not proclaim to be optimum. In
consequence, it is essential to analyze the local flow structures and
(if necessary) to carry out minor design modifications for an optimum
Because of the spanwise decaying shock structure, 3D SCBs performance. Particular focus has to be put on the bump position and
require an increased crest height to exert the same flow turning and its crest height.
hence to achieve the same drag reduction as 2D bumps. Furthermore,
it has been observed that the optimum position of 3D bumps
is slightly more upstream compared to their 2D counterparts
(see, e.g., [24]). Based on the numerical study presented in [24], the VI. Final Bump Design According to Design Guidelines
effect of bump spacing on the optimum position and crest height has The presented design guidelines have been used to define the
been analyzed, and 3D SCBs have been numerically optimized for preliminary design of three different bump shapes; see Table 1 and
maximum drag reduction at constant lift for varying bspacing Fig. 12. Carrying on with the discussion in Sec. I on the adaptivity of
(bbump  const:) on the wing section of the OAT15A airfoil 2D bumps, two different 2D bumps are considered. While the
described in [24]. Figure 10 presents the derived scaling factor for the so-called 2D-P (P: performance) bump is designed for wave drag
crest height and the necessary shift of the bump dependent on the reduction at dash flight, the 2D-B (B: buffet) bump is a buffet bump,
[º]

60
bump shock structure flow
shock deflection angle

55
s bbump

50 lLE
3D SCB
z1
z2 swe
45 pt s
z1 3DSCB hoc
1.15 1.20 1.25 1.30 1.35 1.40 k overlap = z2 k
M1, n
a) Shock deflection angle σ for bump ramp angles b) Geometrical relations of shock structure
< 10º
Fig. 9 Relations for spacing of 3D SCBs.

2.5 0
crest position shift [%c]
hc scaling factor khc

2.0 -0.5

1.5 -1.0

1.0 -1.5
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
bbump / bspacing bbump / bspacing
a) Crest height b) Bump position
Fig. 10 Adaption of bump parameters for transfer of 2D SCB to 3D SCB.
MAYER ET AL. 563

Z Z Z Z
X X X X

inflow

Z
ck
sho

baseline skew ( ) rotation ( ) skew + rotation X


Fig. 11 Adaption of 3D SCBs to local flow by rotation and skew. Local flow features of 3D SCB at η ≈ 60%. Contour indicates pressure coefficient cP .

which has been designed for maximum delay of buffet onset. Hence, 0.08
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

these two bumps outline the full potential of 2D bumps if a sole


design objective is considered. In contrast, the intention of 3D bumps

buffet
is to overcome the necessity for an adaptive flow control device, and 0.06
consequently, only one 3D bump design, which was designed for
wave drag reduction at dash flight and buffet onset delay by the 0.04

cD
vortical wake of the 3D bump array, has been considered.
For all bump designs, the bump length lbump has been set to 25%c.
Furthermore, for improved robustness, the crest height has been 0.02
Experiment
scaled by 0.8. Figure 13a shows the crest height over span, resulting
CFD
from the preliminary design guidelines. Because of the shock
0
structure with the λ shock on the inboard section, bumps have only 0 1 2 3 4 5
been applied between approximately 30% ≤ η ≤ 90% of span. The [º]
position of the bumps over span is not explicitly shown here but can
Fig. 12 Drag rise of baseline configuration M∞  0.826 with bump
be seen in Fig. 14. design conditions.
The 3D SCB array has been evolved from the 2D-P bump
according to the guidelines presented in Sec. V.C. For this design, a
length-to-width-ratio of 3 has been used. The resulting bump width
agrees very well with that of the preliminary bump study on an a more disturbed boundary layer between the bumps, and this can
infinite-span, unswept wing [24]. In detail, the midsection of the 3D lead to higher viscous drag [26,41].
bump has been set to 6%c, and the overall bump width (including the Comparison of the 2D-P bump with the 2D-B bump highlights the
side flanks) has been set to 10%c. According to Eq. (3), the spacing effect of the bump design condition and design objective. While the
has been estimated to approximately 23%cMAC (koverlap  0.5, former has been designed for drag reduction at dash flight, the latter
khc ≈ 2.0, and Δx2D−3D ≈ −1.05%c). The final 3D bump array has been designed for delay of buffet onset. Because of the more
consists of nine bumps. With a rotation of ψ  10 deg (orientation downstream position of the 2D-B bump, its potential for drag
of supersonic flow onto the ramp) around the bump reference point reduction is shifted toward higher lift coefficients. At dash flight, the
and a skew of τ  15 deg, especially the bumps on the midboard 2D-B increases drag by 1.7 dc. This performance deficit stems from a
section are very well aligned with the local flow. It has to be noted misalignment of the 2D-B bump with the shock wave at dash flight.
that, due to the demonstrative character of the SCB design presented Figure 16 shows the surface pressure coefficient in the shock region
here, no further adaption has been carried out. along the cross-section shown in Fig. 13b. In addition, the bump
shape cross-sections are included in the figure. The 2D-P bump and
the 3D bump array reduce the shock strength by splitting the shock
VII. Analysis of Bump Designs into a λ-shock system. In contrast, because of the more downstream
A. Wave Drag Reduction at Dash Flight bump position, the 2D-B bump decelerates the flow locally to
Both the 2D-P bump and the 3D bump array have been primarily subsonic speeds (local cP is larger than the critical pressure
designed for drag reduction at dash flight. Figure 15 shows the coefficient cP , which has been calculated as discussed by Lock [42]),
achieved drag savings over the considered cL range. At the dash flight which significantly increases drag.
bump design condition cL  0.6118, the 2D-P bump reduces drag The different cL ranges for which the 2D bumps are effective point
by 9 drag counts (dc; 10−4 cD ) and the 3D bump array achieves a drag out the typically observed necessity for an adaptive control device to
reduction of 6.0 dc. Both bump designs reduce drag over a wide range combine both wave drag reduction at dash and buffet alleviation.
of cL around their design point. As the 2D bump represents a specific However, this also strongly depends on the cL range between dash
design of a 3D bump (zero spacing), it can be assumed that the flight and buffet onset. In fact, it depends on the extent of shock
performance deficit of the 3D bump array can be reduced by a movement as shock misalignment is the main reason for the
reduction of bump spacing, transforming the 3D bump array toward a detrimental effect at the bump off-design condition. Hence, shock
2D bump design. However, placing the 3D bumps too close results in fixing mechanisms like variable camber in combination with a static
2D bump might be also a promising approach, as discussed by, e.g.,
Werner [43].

Table 1 Bump design conditions B. Delay and Alleviation of Transonic Buffet


Bump M∞ ReMAC α, deg cL cD To delay buffet, the effect of all three bumps on buffet onset
2D-P 0.826 2.84 ⋅ 10 6
2.5 0.6118 0.03126 (Δα criterion) has been analyzed. Figure 17 shows the corresponding
2D-B 0.826 2.84 ⋅ 106 2.9a 0.6601 0.03708 lift polars. It is not surprising that the 2D-B bump performs
3D 0.826 2.84 ⋅ 106 2.5 0.6118 0.03126 particularly well (by shifting buffet onset by ΔcL  4.07 lift counts
(lc; 10−2 cL [6.06%]) as it has been solely designed for delayed buffet
a
Point of shock reversal/most downstream shock position. onset based on the findings of Mayer et al. [24]. Even though not
564 MAYER ET AL.

1.5 2D-P 1.5


cross-section in Fig. 13b 2D-B
3D array

hc [%clocal ]

hc [%clocal ]
1.0 1.0

0.5 0.5

0 0
90 80 70 60 50 40 30 50 60 70 80
[%] x [%clocal ]
a) Crest height hc over span b) Streamwise cross-section of midboard
bumps
Fig. 13 Final design of 2D-P, 2D-B, and 3D bump array. Linear scaling of crest height of 2D bumps to zero at boundaries.

tip root -40


Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

20 2D-P
baseline -30 2D-B
3D SCB array
40 buffet design point
-20
c [%]

cD [ dc]
60 dash design point
cf , x < 0
-10
80
0
20
2D-P 10
40
20
c [%]

60 0.56 0.58 0.60


0.62 0.64 0.66 0.68
cL
80 Fig. 15 Drag savings by bumps (M∞  0.826).

20
2D-B
40
-1.0 baseline
c [%]

60 2D-P
-0.8 2D-B
80 3D SCB array

-0.6
20
cP

3D array (default grid) c *P


40 -0.4
c [%]

60 SCB cross-sections;
-0.2 see Fig. 13
80
0
0.4
0.6 0.8
20
3D array (fine grid) x/c
Fig. 16 cP distribution (shock region) of midboard cross-section at dash
40
design point (M∞  0.826 and cL  0.6118).
c [%]

60

80 degradation for cL < cL;design but also a rather robust bump behavior
for cL > cL;design.
It has to be noted that the positive impact of the 2D-P bump on
90 80 60 50 70 40 30
buffet observed here strongly depends on the considered dash flight
[%] condition, which is set rather close to buffet onset in the presented
Fig. 14 Distribution of skin friction cf;x on wing suction side at buffet study. The 3D bump array has been designed to combine both design
onset of baseline configuration (M∞  0.826 and cL  0.672). objectives: wave drag reduction at dash flight and buffet control. But
with a delay of buffet onset by ΔcL  0.55 lc 0.82% , it is not as
effective as the 2D-P bump. A comparison with the simulation results
specifically designed for buffet control, the 2D-P bump also delays of the default grid resolution reveals that the grid resolution does not
buffet by ΔcL  1.44 lc 2.14% . A combination of a significant affect the predicted buffet onset; see Table 2. While it is a general
increase in shock strength and a rather moderate shock movement observation that 2D SCBs are slightly more effective in reducing
(compared to the preliminary 2D studies discussed by Mayer et al. wave drag compared to 3D SCBs [20,26,27,35], this trend seems also
[24]) still leads to a shock weakening by the 2D-P bump at the buffet to be true for the delay of buffet onset and is an indication that the
design point, which can also be seen in Fig. 15. This is in line with intention to delay buffet onset by using 3D SCBs as smart vortex
literature [35,38], which shows a severe bump performance generators has not yet been successful.
MAYER ET AL. 565

0.72 tail, which is most prominent on the downwind tail flank of the single
3D SCBs due to the spanwise crossflow velocity component on the
0.70 swept wing. This separation topology highlights an important issue
for delaying buffet by using 3D SCBs as smart vortex generators. For
cL the AVERT model considered here, shock-induced boundary-layer
0.68 separation mainly causes the nonlinearity of the lift polar and
consequently dominates buffet onset (Δα criterion). However, the
vortical wake of 3D SCBs can only be used for boundary-layer
0.66 control aft of the bump shape. This implies that such an approach
might only be successful if either the shock wave moves significantly
cL

Baseline aft of the 3D SCB array when approaching buffet onset condition or if
0.64 2D-P
2D-B
flow separation close to the trailing edge dominates the separation
3D (default) topology. Consequently, the limited shock movement between dash
0.62 flight and buffet onset as well as the fact that the (comparably) larger
3D SCBs cause more stress to the boundary layer explain the inferior
n
rio

buffet control ability of the 3D SCB array (even in comparison to the


rite

0.60
-c

2D-P bump). A detailed discussion on the vortical structures


Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

observed in the wake of the 3D SCB array is given in Sec. VII.C.


0.58 Deep inside the buffet regime at α  3.5 deg M∞  0.826,
additional URANS simulations have been carried out to assess the
2.5 3.0 3.5
effect of bumps on buffet related pressure fluctuations. At this flow
[º] condition, experimental data indicate buffet in terms of a broadband
Fig. 17 Effect of bumps on buffet onset based on Δα criterion pressure spectrum; see Fig. 6b.
M∞  0.826. In line with the delay of buffet onset, all three bumps suppressed
any (numerically predicted) flow unsteadiness, and no pressure
fluctuations were observed. However, as transonic buffet on a swept
The nonlinearity of the lift polar at high values of cL is due to wing is a highly challenging phenomenon for numerical simulations,
incipient flow separation. Figure 14 compares the surface skin wind tunnel measurements are necessary for more profound
friction cf;x at the buffet onset lift coefficient of the baseline conclusions.
configuration cL  0.672. It has to be noted that the wing geometry
has been transformed into a rectangular shape for better present-
ability. The baseline flow is characterized by severe shock-induced C. Flow Topology of 3D Bump Array
flow separation for η > 40%. However, the flow reattaches at the rear Wind tunnel tests on isolated 3D bumps (see, e.g., [23]) as well as
part of the wing. At the inboard section η < 40%, the shock-wave/ corresponding flow simulations of 3D bumps on infinite-span,
boundary-layer interaction is less severe due to the λ-shock system. unswept wings (see, e.g., [24]) indicate the presence of a pronounced
With the 2D-P bump applied, flow separation is limited to the crest streamwise vortex system generated by the bump. However, the
of the bump. However, the shock print in Fig. 14 suggests that the detailed character of the vortical wake (e.g., the rotation direction of
bump is positioned slightly too far upstream to completely suppress the primary vortex pair) is highly sensitive to flow phenomena like
flow separation. Furthermore, the inboard end of the bump is not flow separation around the crest/tail of the bump and the superposed
optimally designed. Here, the shock wave naturally splits into a crossflow velocity in case of swept wing flows.
λ-shock system for η < 40%. As the bump shape follows the stronger To study the effect of wing sweep on the vortical wake of a 3D
shock leg, it causes severe crest-flow separation inside the λ-shock SCB, the numerical setup for 3D SCBs on infinite-span, unswept
system. This effect is slightly alleviated by the blending of the crest wings, described by Mayer et al. [24], has been used. The 3D SCB
height to zero between 32% ≤ η ≤ 40% (see Fig. 13a), but considered here has been designed for maximum drag reduction at
performance could be further improved by shortening the bump’s cL  0.8 M∞  0.73 on the OAT15A airfoil. In a parametric
spanwise extent at the inboard end. The 2D-B bump efficiently sweep, the sweep angle Λ has been varied between 0 and 25 deg. With
suppresses the shock-induced boundary-layer separation over most the wing sweep relations as discussed by Obert [1], the aerodynamic
parts of span. Only in the tip region, the shock wave causes a characteristics normal to the leading edge of the infinite-span, swept
separation bubble. In consequence, this bump design could be further wing have been kept constant. Figure 18 shows the relative effect of
improved by extending the bump shape toward the wing tip. the 3D SCB on the flowfield; i.e., the flow solution of the clean airfoil
However, the almost complete suppression of flow separation clearly has been subtracted from the flow solution with the applied SCB at
highlights the beneficial effect of the 2D-B bump on buffet onset. The the same flow condition and sweep angle. It can be clearly seen that
array of 3D bumps leads to a completely different flow topology. the vortex system is generated almost independently of the sweep
Here, the shock strength is reduced in the vicinity of the bumps. angle but it gets distorted with increasing crossflow. However,
However, because of the greater crest height of 3D SCBs, it can be considering the total streamwise vorticity ωx (instead of its relative
argued that they operate closer to their limit to achieve drag savings change Δωx due to the bump shape), shear effects cover the (rather
similar to those of 2D SCBs. And, here, it can be clearly seen that the weak) counterrotating vortex system for Λ > 5 deg, and a single
increased crest height causes local boundary-layer separation on the vortex dominates the wake of the 3D SCB.
As the bump in Fig. 18 was analyzed at moderate dash flight
conditions, there was no flow separation present around the bump
shape. In contrast, the 3D bump array caused severe crest-flow
Table 2 Delay of buffet onset by separation on the AVERT model at buffet onset flow condition of the
bumps (Δα criterion)
baseline configuration. So, while in general similar flow structures as
Bump ΔcL , lc ΔcL , % described with reference to Fig. 18 can be observed on the AVERT
model, the crest-flow separation sheds further vortices into the
2D-P 1.44 2.14
2D-B 4.07 6.06
flowfield. In fact, these vortices dominated the boundary-layer health
3D (default)a 0.53 0.79 aft of the 3D bump array, as shown in Fig. 19. So, even though the
3D (fine)a 0.55 0.82 counterrotating vortex system, which strengthened the boundary
layer aft of the bump along its centerline, was still present, it was
a
Default and fine grid resolution shown in Fig. 2. outbalanced by the detrimental effect of crest-flow separation.
566 MAYER ET AL.

blanked
Δω x [s −1]
-500 -300 -100 100 300 500

Y Y
X X
Z Λ = 0° Z Λ = 15°
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

Y Y
X X
Z Λ = 5° Z Λ = 20°

Y Y
X X
Z Λ = 10° Z Λ = 25°
Fig. 18 Effect of wing sweep on vortical wake of a 3D SCB at cL  0.8 M∞  0.73 on the OAT15A airfoil. Streamwise vorticity Δωx 
ωx jwith bump − ωx jwithout bump shown in slices between 0.72 ≤ x∕c ≤ 0.98. Surface streamlines show local flow direction.

20 Δ H 12
1.0
deterioration
0.8
40 0.6
0.4
c [%]

0.2
60 0
-0.2
-0.4
80 -0.6
-0.8
improvement
-1.0
90 80 70 50 60 40 30
[%]
Fig. 19 Effect of 3D bump array on boundary-layer shape factor H12 at buffet onset of baseline configuration (M∞  0.826 and cL  0.672).
ΔH12  H12 j3D bump array − H12 jbaseline .

VIII. Conclusions characteristics were in reasonably good agreement with experimental


The applicability of (industrial) guidelines for the efficient design results and formed the baseline for the subsequent bump study. The
bump design process of three different bump shapes was described in
of shock control bumps for a wing-body configuration has been
detail, and design guidelines were presented. Two different 2D
demonstrated with (U)RANS simulations, and the potential of two- bumps were considered, one specifically designed for wave drag
dimensional (2D) and three-dimensional (3D) shock control bumps reduction at dash flight (2D-P bump) and one designed for the delay
(SCBs) for wave drag reduction, buffet onset delay, and buffet of buffet onset (2D-B bump). Furthermore, a 3D bump array was
alleviation has been assessed. First, the numerical setup was validated designed with the intention of combining both design objectives. In
by a comparison of the clean model with wind tunnel measurements. summary, all three bumps served their purpose, with the 2D bumps
Lift and drag over the polar as well as buffet onset and buffet marking the maximum potential for each design objective with a drag
MAYER ET AL. 567

reduction of 9 dc at dash flight (2D-P bump) and a delay of buffet [13] Molton, P., Dandois, J., Lepage, A., Brunet, V., and Bur, R., “Control of
onset by 4 lc (2D-B bump). Even though the 3D bump array Buffet Phenomenon on a Transonic Swept Wing,” AIAA Journal,
combines both design objectives (drag reduction by 6 dc and delay of Vol. 51, No. 4, 2013, pp. 761–772.
buffet onset by 0.55 lc), it did not perform as well as expected due to doi:10.2514/1.J051000
[14] Kouchi, T., Yamaguchi, S., Koike, S., Nakajima, T., Sato, M., Kanda, H.,
massive crest-flow separation and the (so far) minor role of the and Yanase, S., “Wavelet Analysis of Transonic Buffet on a Two-
vortical wake of 3D bumps in delaying buffet onset. An in-depth Dimensional Airfoil with Vortex Generators,” Experiments in Fluids,
analysis of the relevant flow features of 3D bumps for the delay of Vol. 57, No. 11, 2016, pp. 1–15.
buffet onset on finite-span, swept wings revealed a significant doi:10.1007/s00348-016-2261-2
influence of crossflow on the wake topology of 3D SCBs. For the [15] Dandois, J., Molton, P., Lepage, A., Geeraert, A., Brunet, V., Dor, J.-B.,
swept wing of the AVERT model, local flow separation on the 3D and Coustols, E., “Buffet Characterization and Control for Turbulent
bumps’ tails dominated the character of the vortical wake. In line with Wings,” AerospaceLab: The ONERA Journal, Vol. 6, No. 6, 2013,
the ability to delay buffet onset, all three bumps suppressed any pp. 1–17.
numerically predicted flow unsteadiness inside the buffet regime at [16] Lee, B. H. K., “Self-Sustained Shock Oscillations on Airfoils at
Transonic Speeds,” Progress in Aerospace Sciences, Vol. 37, No. 2,
α  3.5 deg M∞  0.826. 2001, pp. 147–196.
doi:10.1016/S0376-0421(01)00003-3
[17] Raghunathan, S., Early, J. M., Tulita, C., and Benard, E., “Periodic
Acknowledgments Transonic Flow and Control,” The Aeronautical Journal, Vol. 112,
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

The experimental data used in this Paper stem from ONERA No. 1127, 2008, pp. 1–16.
S2MA wind tunnel tests, which were conducted within the FP7 doi:10.1017/S0001924000001949
AVERT European project (contract number AST5-CT-2006- [18] Giannelis, N. F., Vio, G. A., and Levinski, O., “A Review of Recent
Developments in the Understanding of Transonic Shock Buffet,”
030914), funded by the European Commission and project partners
Progress in Aerospace Sciences, Vol. 92, May 2017, pp. 39–84.
(Airbus Operations, Ltd.; Airbus Operations SL; Dassault Aviation; doi:10.1016/j.paerosci.2017.05.004
Alenia Aeronautica; and ONERA). [19] Birkemeyer, J., Rosemann, H., and Stanewsky, E., “Shock Control on a
Swept Wing,” Aerospace Science and Technology, Vol. 4, No. 3, 2000,
pp. 147–156.
References doi:10.1016/S1270-9638(00)00128-0
[1] Obert, E., Ronald, S., Debbie, J. W. L., van den Berg, T., Justin, H. K., [20] Eastwood, J. P., and Jarrett, J. P., “Toward Designing with
and van Tooren, M. J. L., Aerodynamic Design of Transport Aircraft, Ios Three-Dimensional Bumps for Lift/Drag Improvement and Buffet
Press, Amsterdam, The Netherlands, 2009. Alleviation,” AIAA Journal, Vol. 50, No. 12, 2012, pp. 2882–2898.
[2] Brunet, V., and Deck, S., “Zonal-Detached Eddy Simulation of doi:10.2514/1.J051740
Transonic Buffet on a Civil Aircraft Type Configuration,” Notes on [21] Bogdanski, S., Gansel, P., Lutz, T., and Krämer, E., “Impact of 3D Shock
Numerical Fluid Mechanics and Multidisciplinary Design, Vol. 97, Control Bumps on Transonic Buffet,” High Performance Computing in
No. 2003, 2008, pp. 182–191. Science and Engineering 14, Springer International, Cham, Switzerland,
doi:10.1007/978-3-540-77815-8 2015, pp. 447–461.
[3] Caruana, D., Mignosi, A., Corrège, M., Le Pourhiet, A., and Rodde, A. M., doi:10.1007/978-3-319-10810-0_30
“Buffet and Buffeting Control in Transonic Flow,” Aerospace Science and [22] Bogdanski, S., Nübler, K., Lutz, T., and Krämer, E., “Numerical
Technology, Vol. 9, No. 7, 2005, pp. 605–616. Investigation of the Influence of Shock Control Bumps on the Buffet
doi:10.1016/j.ast.2004.12.005 Characteristics of a Transonic Airfoil,” New Results in Numerical and
[4] Sartor, F., and Timme, S., “Reynolds-Averaged Navier-Stokes Experimental Fluid Mechanics IX, edited by A. Dillmann, G. Heller, E.
Simulations of Shock Buffet on Half Wing-Body Configuration,” Krämer, H. P. Kreplin, W. Nitsche, and U. Rist, Vol. 124, Notes on
53rd AIAA Aerospace Sciences Meeting, AIAA, Reston, VA, 2015, Numerical Fluid Mechanics and Multidisciplinary Design, Springer,
pp. 1–11. Cham, 2014, pp. 23–32.
doi:10.2514/1.J055186 doi:10.1007/978-3-319-03158-3
[5] Sartor, F., and Timme, S., “Delayed Detached-Eddy Simulation of [23] Colliss, S. P., Babinsky, H., Nübler, K., and Lutz, T., “Vortical Structures
Shock Buffet on Half Wing-Body Configuration,” AIAA Journal, on Three-Dimensional Shock Control Bumps,” AIAA Journal, Vol. 54,
Vol. 55, No. 4, 2016, pp. 1230–1240. No. 8, 2016, pp. 2338–2350.
doi:10.2514/1.J055186 doi:10.2514/1.J054669
[6] Sartor, F., and Timme, S., “Mach Number Effects on Buffeting Flow on a [24] Mayer, R., Lutz, T., and Krämer, E., “Numerical Study on the Ability of
Half Wing-Body Configuration,” International Journal of Numerical Shock Control Bumps for Buffet Control,” AIAA Journal, Vol. 56, No. 5,
Methods for Heat and Fluid Flow, Vol. 26, No. 7, 2016, pp. 2066–2080. 2018, pp. 1978–1987.
doi:10.1108/HFF-07-2015-0283 doi:10.2514/1.J056737
[7] Lawson, S., Greenwell, D., and Quinn, M. K., “Characterisation of [25] Holden, H., and Babinsky, H., “Shock/Boundary Layer Interaction
Buffet on a Civil Aircraft Wing,” 54th AIAA Aerospace Sciences Control Using 3D Devices,” 41st Aerospace Sciences Meeting and
Meeting, AIAA Paper 2016-1309, 2016. Exhibit, AIAA Paper 2003-447, 2003.
doi:10.2514/6.2016-1309 doi:10.2514/6.2003-447
[8] Iovnovich, M., and Raveh, D. E., “Numerical Study of Shock Buffet on [26] Ogawa, H., Babinsky, H., Pätzold, M., and Lutz, T., “Shock-Wave/
Three-Dimensional Wings,” AIAA Journal, Vol. 53, No. 2, 2015, Boundary-Layer Interaction Control Using Three-Dimensional Bumps
pp. 449–463. for Transonic Wings,” AIAA Journal, Vol. 46, No. 6, 2008,
doi:10.2514/1.J053201 pp. 1442–1452.
[9] Goncalves, E., and Houdeville, R., “Turbulence Model and Numerical doi:10.2514/1.32049
Scheme Assessment for Buffet Computations,” International Journal [27] Wong, W., Qin, N., Sellars, N., Holden, H., and Babinsky, H., “A
for Numerical Methods in Fluids, Vol. 46, No. 11, 2004, pp. 1127–1152. Combined Experimental and Numerical Study of Flow Structures over
doi:10.1002/(ISSN)1097-0363 Three-Dimensional Shock Control Bumps,” Aerospace Science and
[10] Zimmermann, D.-M., Mayer, R., Lutz, T., and Krämer, E., “Impact of Technology, Vol. 12, No. 6, 2008, pp. 436–447.
Model Parameters of SALSA Turbulence Model on Transonic Buffet doi:10.1016/j.ast.2007.10.011
Prediction,” AIAA Journal, Vol. 56, No. 2, 2018, pp. 874–877. [28] Jacquin, L., Molton, P., Deck, S., Maury, B., and Soulevant, D.,
doi:10.2514/1.J056193 “Experimental Study of Shock Oscillation over a Transonic Supercritical
[11] Bur, R., Coponet, D., and Carpels, Y., “Separation Control by Vortex Profile,” AIAA Journal, Vol. 47, No. 9, 2009, pp. 1985–1994.
Generator Devices in a Transonic Channel Flow,” Shock Waves, Vol. 19, doi:10.2514/1.30190
No. 6, 2009, pp. 521–530. [29] Rossow, C.-C., Kroll, N., and Schwamborn, D., “The MEGAFLOW
doi:10.1007/s00193-009-0234-6 Project: Numerical Flow Simulation for Aircraft,” Progress in Industrial
[12] Huang, J. B., Xiao, Z. X., Liu, J., and Fu, S., “Simulation of Shock Wave Mathematics at ECMI 2004, edited by A. Di Bucchianico, R. Mattheij,
Buffet and its Suppression on an OAT15A Supercritical Airfoil by and M. Peletier, Vol. 8, Mathematics in Industry (The European
IDDES,” Science China: Physics, Mechanics and Astronomy, Vol. 55, Consortium for Mathematics in Industry), Springer, Berlin, 2006,
No. 2, 2012, pp. 260–271. pp. 3–33.
doi:10.1007/s11433-011-4601-9 doi:10.1007/3-540-28073-1
568 MAYER ET AL.

[30] Grossi, F., Braza, M., and Hoarau, Y., “Prediction of Transonic Buffet by [37] Brion, V., Abart, J. C., and Paillart, P., “Experimental Analysis of the
Delayed Detached-Eddy Simulation,” AIAA Journal, Vol. 52, No. 10, Shock Dynamics on a Transonic Laminar Airfoil,” 6th European
2014, pp. 2300–2312. Conference for Aeronautics and Space Sciences [CD-ROM], EUCASS,
doi:10.2514/1.J052873 Rhode-St-Genese, Belgium, 2015, pp. 1–11.
[31] Illi, S., Lutz, T., and Krämer, E., “On the Capability of Unsteady RANS [38] Bruce, P. J. K., and Colliss, S. P., “Review of Research into
to Predict Transonic Buffet,” 3rd Symposium “Simulation of Wing and Shock Control Bumps,” Shock Waves, Vol. 25, No. 5, 2015,
Nacelle Stall,” DFG Research Unit FOR1066, Braunschweig, Germany, pp. 451–471.
2012, pp. 1–13. doi:10.1007/s00193-014-0533-4
[32] Rung, T., Bunge, U., Schatz, M., and Thiele, F., “Restatement of [39] AMES Research Staff, “Equations, Tables, and Charts for Compressible
the Spalart-Allmaras Eddy-Viscosity Model in Strain-Adaptive Formu- Flow,” NACA TR 1135, 1953.
lation,” AIAA Journal, Vol. 41, No. 7, 2003, pp. 1396–1399. [40] Jones, N. R., Eastwood, J. P., and Jarrett, J. P., “Adapting Three-
doi:10.2514/2.2089 Dimensional Shock Control Bumps for Swept Flows,” AIAA Journal,
[33] König, B., “Shock Control Bumps on Transonic Transport Aircraft,” Ph.D. Vol. 55, No. 3, 2017, pp. 861–873.
Thesis, Univ. of Stuttgart, Stuttgart, Germany, 2013. doi:10.2514/1.J055169
[34] Sommerer, A., Lutz, T., and Wagner, S., “Numerical Optimization of [41] Qin, N., Wong, W. S., and Le Moigne, A., “Three-Dimensional Contour
Adaptive Transonic Airfoils with Variable Camber,” Proceedings 22nd Bumps for Transonic Wing Drag Reduction,” Journal of Aerospace
ICAS Congress [CD-ROM], Optimage Ltd., Edinburgh, U.K., 2000. Engineering, Vol. 222, No. 5, 2008, pp. 619–629.
[35] Pätzold, M., Lutz, T., Krämer, E., and Wagner, S., “Numerical doi:10.1243/09544100JAERO333
Optimization of Finite Shock Control Bumps,” 44th AIAA Aerospace [42] Lock, R. C., “An Equivalence Law Relating Three- and Two-
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

Sciences Meeting and Exhibit, AIAA Paper 2006-1054, 2006. Dimensional Pressure Distributions,” Aeronautical Research Council
doi:10.2514/6.2006-1054 TR 3346, Cranfield, U.K., 1964.
[36] Mayer, R., Lutz, T., and Krämer, E., “Toward Numerical Optimization [43] Werner, M., “Application of an Adaptive Shock Control Bump for Drag
of Buffet Alleviating Three-Dimensional Shock Control Bumps,” 6th Reduction on a Variable Camber NLF Wing,” 2018 AIAA Aerospace
European Conference for Aeronautics and Space Sciences [CD-ROM], Sciences Meeting, AIAA Paper 2018-0789, 2018.
EUCASS, Rhode-St-Genese, Belgium, 2015, pp. 1–12. doi:10.2514/6.2018-0789
This article has been cited by:

1. Jack A. Geoghegan, Nicholas F. Giannelis, Gareth A. Vio. 2020. A Numerical Investigation of the Geometric
Parametrisation of Shock Control Bumps for Transonic Shock Oscillation Control. Fluids 5:2, 46. [Crossref]
2. Vilas Shinde, Jack McNamara, Datta Gaitonde. 2020. Control of transitional shock wave boundary layer interaction using
structurally constrained surface morphing. Aerospace Science and Technology 96, 105545. [Crossref]
Downloaded by KOREA AEROSPACE INDUSTRIES LTD on May 21, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.C034969

You might also like