You are on page 1of 5

LES of Laminar Separation Bubble Flows

J. A. Domaradzki
Department of Aerospace Engineering
University of Southern California
Los Angeles, CA 90089-1191
The primary goal of the proposed research is to extend previous ndings regarding laminar separation
bubble (LSB) ow over a at plate (CTR Summer Program 2012) to a more realistic geometrical conguration of a NACA-0012 airfoil. A compelling issue raised by the previous results is the role of the numerical
dissipation in LES of such ows; its quantication in coarse mesh LES constitutes the second goal of this
proposal. Overall, the proposed research should determine necessary requirements for both the numerical
method and the spatial resolution to obtain reliable and accurate LES results for LSB ows.

Physical problem

Reynolds numbers based on wing/blade chord for ows in rotating machinery, for unmanned aerial vehicles (UAV), micro air vehicles (MAV), wind turbines, and propellers are typically less than 2 106 and
often only on the order of 104 to 105 . By comparison, civilian aircraft are characterized by the Reynolds
numbers ranging from a few million to 80 106 for the Boeing 747 at cruising velocity. Recent experimental investigations of low Reynolds number aerodynamics reveal that low to moderate Reynolds number
ows over airfoils and turbine blades are often dominated by the eects of ow separation and turbulent
reattachment [17, 15, 25]. Such phenomena greatly inuence the aerodynamic forces the airfoil or blade
is subjected to. They change the lift and drag characteristics and thus ight stability of UAVs, aect
wind turbine eciency, and cause unsteadiness in turbine ows, which is a determining factor in high cycle
fatigue (HCF) of turbo-machinery components.
LSB ows are qualitatively well understood thanks to numerous experimental investigations, e.g., [13,
16, 3, 14, 21, 4, 27, 5, 6, 17, 15, 25] as well as from direct numerical simulations (DNS) results [20, 24, 1,
22, 18, 19]. The attached laminar boundary layer developing on a wing or blade is subjected to an adverse
pressure gradient due to the airfoils curvature, which causes it to separate. There is an eectively stagnant
ow region immediately behind the separation point, the so-called dead air region, followed by a reverse
ow vortex. The interface between the separated ow moving away from the wing and the recirculating
ow in the vicinity of the wing results in a shear layer with an inectional mean velocity prole. This
shear layer experiences Kelvin-Helmholtz instabilities that develop into turbulence after rst generating
characteristic spanwise vortices. Further downstream, the separated turbulent ow reattaches, thereby
closing o the LSB, and gradually evolves into the classical turbulent boundary layer. The separation
bubbles shape and size changes in time due to vortex shedding, making the problem inherently unsteady.

Numerical diculties

Numerical prediction tools for LSB ows are needed in order to produce more ecient airfoil or blade
designs, to create control schemes to reduce separation eects, and to better predict HCF. The Reynoldsaveraged Navier-Stokes (RANS) approach was shown to be inadequate for such ows by Spalart [24].
Reliable and accurate simulation results for LSB ows are currently dicult to obtain unless DNS or
large eddy simulation (LES) with very high, DNS-like resolution is used. Jones et al. [18] used over 170
million grid points in their DNS for a relatively simple 3-D airfoil conguration. Yang and Voke [26]
reported LES results with the dynamic Smagorinsky model that were in good agreement with experiments
for boundary-layer separation and transition caused by surface curvature at Re = 3, 450. Yet even for
this relatively low Reynolds number, the critical requirements to obtaining experimental agreement was a
high numerical resolution (472 72 64 mesh points) and a high order numerical method. Such stringent
requirements are rarely satised by commercial codes. Similarly, Eisenbach and Friedrich [12] performed
1

LES of ow separation on an airfoil at high angle of attack at Re = 105 using between 50 and 100 million
mesh points. Performing DNS or LES at such high resolutions require substantial computational resources,
long wall-clock runs and long analysis times. If a number of congurations and angles of attack are to
be quickly investigated in the context of airfoil or blade shape optimization, using such high resolutions
becomes prohibitive.
During the CTR Summer Program 2012, Cadieux et al. studied a LSB ow over a at plate using DNS,
under-resolved DNS, and LES with dierent SGS models [8]. Their results demonstrated the feasibility of
very coarse LES for LSB ows (around 1-3% of DNS resolution) for this simple geometrical conguration.
Later work also identied numerical dissipation as playing a signicant role in LES at such low resolutions,
even when nominally high order schemes were used [7]. Another rare example of a low resolution LES of a
LSB ow is given by Almutairi et al. [2]. However, their LES method is a simple application of an explicit
lter. Separately, we have also performed extensive LES simulations of a ow over a NACA-0012 airfoil
with an Immersed Boundary (IB) code INCA developed in Adams group at the Technical University
of Munich [9]. We were able to match the DNS benchmark of Jones [18] using a near-DNS resolution,
but at coarse LES resolutions the accuracy of the Immersed Boundary method at rigid boundaries was
not sucient to obtain agreement for the separation and reattachment points. These results collectively
point to two issues: (1) the extension to realistic geometrical congurations is non-trivial and (2) coarse
resolution LES can be heavily inuenced by the numerical method and associated numerical dissipation.
Addressing these two issues is of paramount importance to promote the use of very coarse LES as a design
tool for industrial applications.

The proposed project

We propose to study LSB ows at moderate Reynolds numbers with a realistic geometrical conguration
using LES, and to quantify the numerical dissipation in such simulations.
The at plate simulations performed during the previous CTR summer program [8] would be extended
to a more more realistic conguration: a ow over a NACA-0012 airfoil at 5 degree angle of attack. A
graduate student at USC, Giacomo Castiglioni, has already performed extensive LES for this conguration
[9]. Detailed DNS results for this specic case were obtained by Jones et al. using a compressible code in
generalized curvilinear coordinates [18, 19]. Tabulated DNS data for the pressure and the friction coecient
have been obtained from the authors and are available for comparison. We propose to use the same CTR
code [23] that was used for the at plate LES in Summer 2012. Giacomo Castiglioni is already familiar
with this code through conversations with Dr. Taraneh Sayadi at the Munich Summer Program in 2013
and contacts with Prof. Leles current students. Simulations with the generalized curvilinear coordinates
CTR code would be useful in disentangling the numerical eects of IB methods from the eects of LES
models. Furthermore, since in the previous summer program the CTR code demonstrated the capability to
simulate accurately a LSB at around 1-3% of DNS resolution we hope that the same resolution reduction
will be possible for the NACA-0012 airfoil conguration (which was not possible for the IB method).
One surprising conclusion from our work is that the numerical dissipation plays a signicant role in LES
of LSB ows [7, 9]. In simulations performed with the IB method, the best agreement with the benchmark
data was obtained using a dissipative WENO scheme without SGS model rather than a non-dissipative
central dierence scheme with a SGS model. In the at plate simulations, under-resolved DNS provided
a better agreement with the benchmark data than LES with several dierent SGS models tested. For
the CTR code [23], a crude estimate of the numerical dissipation was made and it was determined that
it was principally due to the ltering needed to match solutions of the implicit time stepping used in the
vicinity of the wall and the explicit time stepping used in the rest of the domain [7]. Clearly, knowing
the implicit numerical dissipation and how it compares/interacts with the explicit SGS dissipation should
lead to improvements in LES capabilities. Hence, we plan to use a more rigorous method for estimating
the numerical dissipation and apply it to the analysis of both separated ows (the at plate and the
NACA-0012 airfoil).
2

3.1

Technical approach

During the summer program, the NACA-0012 airfoil case would be run using the CTR code developed by
Prof. Leles group [23]. We would use the same approach as developed for the at plate ow: to perform
rst LES and under-resolved DNS with a resolution around 1% of the benchmark DNS resolution of Jones
et al., and keep increasing the resolution until LES results for the separated ow match satisfactorily the
benchmark DNS data [19].
The eects of the numerical dissipation would be assessed in two ways. We believe that performing
runs with only explicit time-stepping, i.e., eliminating any explicit ltering procedures, would ensure a
spectral-like behavior. Such simulations would be expensive but need to be performed only for limited
time intervals, starting from restart les in runs performed with the production code. Having results
obtained from dissipative and non-dissipative runs, both started from the same initial condition, allows to
compute the numerical dissipation by comparing the local energy decay rates for both cases (see [11, 10]).
Another way, which does not depend on having an access to a non-dissipative code, relies on computing
the residual of the energy rate equation for a selected subdomain, i.e., a dierence between dE/dt and
integrated uxes through the subdomain boundaries. This method is currently being developed and has
been tentatively tested on the Taylor-Green vortex ow with promising results.
J.A. Domaradzki would participate as a PI guiding two graduate students from USC, G. Castiglioni
and F. Cadieux. G. Castiglioni would be involved in simulating the NACA-0012 airfoil ow and F. Cadieux
in the analysis of the numerical dissipation, rst for the at plate problem for which data are available
from the previous summer program, and then for the airfoil problem once the data are generated.

Financial requests:

We request funding only to cover housing expenses at Stanford for the duration of the program. Based
on the previous programs housing rates wed like to request a housing allowance of $3,000 for the PI and
$1,000 for each students, i.e., the total of $5,000 for this project.

References
[1] M. Alam and N.D. Sandham. Direct numerical simulation of short laminar separation bubbles with
turbulent reattachment. J. Fluid Mech., 410:128, 2000.
[2] J. H. Almutairi, L. E. Jones, and N. D Sandham. Intermittent bursting of a laminar separation bubble
on an airfoil. AIAA J, 48(2):414426, 2010.
[3] A. E. Alving and H. H. Fernholz. Turbulence measurements around a mild separation bubble and
downstream of reattachment. J. Fluid Mech., 332:297328, 1996.
[4] S. Burgmann, C. B
ucker, and W. Schroder. Scanning PIV measurements of a laminar separation
bubble. Exper. Fluids, 41:319326, 2006.
[5] S. Burgmann, J. Dannemann, and W. Schroder. Time-resolved and volumetric PIV measurements of
a transitional separation bubble on an SD7003 airfoil. Exper. Fluids, 44:609622, 2007.
[6] S. Burgmann and W. Schroder. Investigation of the vortex induced unsteadiness of a separation bubble
via time-resolved and scanning PIV measurements. Exper. Fluids, 45:675691, 2008.
[7] F. Cadieux, J. A. Domaradzki, T. Sayadi, and T. Bose. DNS and LES of laminar separation bubbles
at moderate Reynolds numbers. ASME J. Fluids Eng., in press, 2013.

[8] F. Cadieux, J. A. Domaradzki, T. Sayadi, T. Bose, and F. Duchaine. DNS and LES of separated ows
at moderate Reynolds numbers. In Proceedings of the 2012 Summer Program, pages 7786. Center
for Turbulence Research, 2012.
[9] G. Castiglioni, J.A. Domaradzki, V. Pasquariello, S. Hickel, and M. Grilli. Numerical simulations
of separated ows at moderate reynolds numbers appropriate for turbine blades and unmanned aero
vehicles. Int J Heat Fluid Flow, submitted, 2014.
[10] J. A. Domaradzki and S. Radhakrishnan. Eective eddy viscosities in implicit modeling of decaying
high Reynolds number turbulence with and without rotation. Fluid Dyn. Res., 36:385406, 2005.
[11] J. A. Domaradzki, Z. Xiao, and P. K. Smolarkiewicz. Eective eddy viscosities in implicit large eddy
simulations of turbulent ows. Phys. Fluids, 15(12):38903893, 2003.
[12] S. Eisenbach and R. Friedrich. Large-eddy simulation of ow separation on an airfoil at a high
angle of attack and Re = 105 using Cartesian grids. Theor. Comp. Fluid Dyn., 22:213225, 2008.
10.1007/s00162-007-0072-z.
[13] M. Gaster. On the stability of parallel ows and the behaviour of separation bubbles. PhD thesis,
University of London, 1963.
[14] C. Haggmark. Investigation of disturbances developing in a laminar separation bubble ow. PhD
thesis, Royal Institute of Technology, Sweden, 2000.
[15] R. Hain, C. J. Kahler, and R. Radespiel. Dynamics of laminar separation bubbles at low-Reynoldsnumber aerofoils. J. Fluid Mech., 630:129153, 2009.
[16] H.P. Horton. Laminar separation bubbles in two and three dimensional incompressible ow. PhD
thesis, University of London, 1968.
[17] H. Hu, Z. Yang, and H. Igarashi. Aerodynamic hysteresis of a low-Reynolds-number airfoil. J. Aircraft,
44(6):20832085, 2007.
[18] L. E. Jones, R. D. Sandberg, and N. D. Sandham. Direct numerical simulations of forced and unforced
separation bubbles on an airfoil at incidence. J. Fluid Mech., 602:175207, 2008.
[19] L.E. Jones, R.D. Sandberg, and N.D. Sandham. Stability and receptivity characteristics of a laminar
separation bubble on an aerofoil. J. Fluid Mech., 648:257296, April 2010.
[20] J. C. M. Lin and L. L.Pauley. Low-Reynolds-number separation on an airfoil. AIAA J., 34(8):1570
1577, 1996.
[21] O. Marxen, M. Lang, and S. Wagner. A combined experimental/numerical study of unsteady phenomena in a laminar separation bubble. Flow, Turbul. Combust., 71:133146, 2003.
[22] O. Marxen and U. Rist. Mean ow deformation in a laminar separation bubble: separation and
stability characteristics. J. Fluid Mech., 660:3754, 2010.
[23] S. Nagarajan. Leading edge eects in bypass transition. PhD thesis, Stanford University, June 2004.
[24] P.R. Spalart and M.K. Strelets. Mechanisms of transition and heat transfer in a separation bubble.
J. Fluid Mech., 403:329349, 2000.
[25] G.R. Spedding and J. McArthur. Span eciencies of wings at low Reynolds numbers. J. Aircraft,
47(1):120128, 2010.

[26] Z. Yang and P.R. Voke. Large-eddy simulation of boundary-layer separation and transition at a change
of surface curvature. J. Fluid Mech., 439:305333, 2001.
[27] Serhiy Yarusevych, Pierre E. Sullivan, and John G. Kawall. Coherent structures in an airfoil boundary
layer and wake at low Reynolds numbers. Physics of Fluids, 18(4):044101, 2006.

You might also like