Unsteady_free-surface_wave_induced_separation_cohe

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 28

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228561375

Unsteady free-surface wave induced separation: coherent vortical structures


and instabilities

Article · January 2005

CITATIONS READS

4 400

3 authors:

Mani Kandasa Tao Xing

37 PUBLICATIONS 725 CITATIONS


University of Idaho
104 PUBLICATIONS 2,204 CITATIONS
SEE PROFILE
SEE PROFILE

Frederick Stern
University of Iowa
336 PUBLICATIONS 11,662 CITATIONS

SEE PROFILE

All content following this page was uploaded by Mani Kandasa on 04 June 2014.

The user has requested enhancement of the downloaded file.


UNSTEADY FREE-SURFACE WAVE INDUCED SEPARATION: COHERENT
VORTICAL STRUCTURES AND INSTABILITIES

Manivannan Kandasamy, Tao Xing, Frederick Stern

ABSTRACT

Free-surface wave-induced separation caused by interactions of free-surface waves and


wall boundary layers has relevance to ship and platform hydrodynamics with regard to resistance
and propulsion, stability, and signatures. This paper investigates the physics behind the different
coherent vortical structures and associated instability mechanisms using URANS with
complimentary EFD. The geometry used is a surface piercing NACA0024 foil, which has no
separation in the deep, and significant separation near the free surface solely due to the wave
induced piezometric pressure gradient for high Froude number (Fr= 0.37, and 0.55) turbulent
flows. Laminar simulations were also conducted for Fr=0.37, and for 2D case. Shear layer and
Karman like instabilities were identified for both Fr. Parametric studies using normalized Strouhal
numbers for shear layer (Stθ scaled by momentum thickness at separation θ) and Karman
instabilities (Sth scaled by half wake thickness h) uncovered Fr and free-surface effects on the
instabilities and the separation pattern. Sth for all cases simulated lie in lower range (0.065-0.069),
compared to Sth for flows without free surface (0.07-0.09). Stθ decreases with adverse pressure
gradient at separation such that Stθ (laminar, Fr=0.37) > Stθ (turbulent, Fr=0.37) > Stθ (turbulent,
Fr=0.55). In addition studies identified presence of ‘flapping’ instability associated with pulsation
of the separation bubble for Fr=0.37. A theoretical model constructed based on simple harmonic
motion gives a good prediction of the flapping frequency. The skin friction topology is also
analyzed for Fr=0.37. The overall results, which are verified and validated, provide credible flow
physics on the nature of free surface wave induced separation.

Submitted to Journal of Fluids and Structures; 23 July 2005: revision submitted 9 April, 2006
IIHR—Hydroscience & Engineering, The University of Iowa, Iowa City, IA 52242; Corresponding
author: Frederick-stern@uiowa.edu; 319-335-5215.
INTRODUCTION

Three-dimensional flow separation is a vast area of study and there are still a lot to
understand. The free surface adds to the complications due to waves and their interaction with
boundary layers, vortices, and turbulence and air water interfacial effects such as bubble
entrainment and surface tension. In addition, adverse piezometric pressure gradients due to steep
waves induce, and/or modify boundary layer separation. Gaining insight into the fluid mechanics of
these areas would be of both fundamental and practical interest, especially regarding applications
in ships and ocean engineering.

Relatively less work has been done in the area of free surface wave induced separation
compared to other three dimensional separations. Chow (1967) first discovered this phenomenon
using a surface-piercing foil designed for insignificant separation for the deep condition.
Subsequently, Stern et al. (1989) made similar observations in their study about the effects of
waves on the boundary layer of a surface-piercing flat plate, with superposed Stokes wave,
generated by attached upstream-submerged horizontal hydrofoil. These two studies showed that
separation occurred at high Fr and large wave steepness, initiated just beyond the wave trough, and
extended to the surface piercing foil and plate trailing edges, respectively. On the free surface, the
separation region was wedge shaped, broken, and turbulent. Laminar Navier-Stokes computational
fluid dynamics (CFD) by Stern et al. (1993) using Stern et al. (1989) geometry showed the
topology of the separation for the first time: a saddle point of separation on the plate and a
clockwise outward spiral node on the free surface (starboard-side recirculation region) with vortical
flow transporting fluid to the free surface. Topological analysis of steady RANS simulations by
Zhang and Stern (1996) for Fr=0.37 on a NACA24 foil using Baldwin-Lomax model agreed
qualitatively with Stern et al. (1993), but with more critical points associated with secondary
separations. Steady RANS by Kandasamy (2001) using same geometry as Zhang and Stern (1996)
at Fr=0.37 with k-ω turbulence model showed better comparison of the mean elevation and foil
surface pressures with complementary experimental fluid dynamics (EFD) by Metcalf (2001).
Separation topology was exactly same as Stern et al. (1993). Pogozelski et al. (1997) is the only
experimental work that analyzes the volume flow field for such flows. Unfortunately, analysis was
limited to Fr=0.25 at which there was no flow reversal, whereas current and previous CFD analysis
focus on regimes with flow reversal at higher Fr. Pogozelski et al. (1997) point out that flow
reversal is evident for Fr>0.30 but measurement was not made due to number of entrained bubbles
making velocity measurement difficult. They constructed a sketch of the flow structure based on
video images and PIV measurements at one spatial orientation for Fr=0.25. Sketches depict counter
rotating vortex pairs occurring close to free surface, which begin with shoulder wave breaking.
Most recently, Metcalf et al. (2006) analyzed the unsteady nature of the separation at high Fr
(=0.37) experimentally using frequency spectra of the free-surface elevation and foil surface
pressures. Certain dominant periodic modes were identified, but without flow-field measurements,
the physics behind these frequencies were unresolved. Current study was initiated to resolve the
physics behind these instabilities using 3rd order unsteady RANS (URANS) with free-surface
tracking and complementary EFD.

Unsteady 3D separation and the instability mechanisms governing the unsteadiness have
been thoroughly investigated only for a few basic geometries such as flat plate, cylinders, spheres,
blunt cylinders, and backward facing steps. Studies have uncovered that for all flows the initial
instability is the shear layer instability, which scales with momentum thickness of shear layer at
separation (Stθ=fSLθ/US, US is shear layer velocity at separation), that causes the separated shear
layer to roll up and these vortices either merge and form larger vortices or amalgamate into already
existing large-scale vortices. The instability of the large-scale vortices for non-reattaching flows is

2
caused by Karman instability, which scales with distance between separated shear layers
(Sth=fh/US) that causes the vortices to shed asymmetrically. Reattaching separation exhibits similar
shedding, but different from Karman shedding as the vortex interaction is with the mirror vortex
unlike the staggered vortical arrangement for Karman vortices. Sigurdson (1995) showed that in
spite of this difference, the symmetric shedding scales similar to the Karman shedding for flow past
blunt cylinders. Henceforth, we shall use the term ‘Karman-like shedding’ for such type of
symmetric vortex shedding. Sth for all large-scale Karman and Karman-like vortex shedding
without free-surface has been found to lie in the range 0.07-0.09. Effects of free surface on the
Karman instability was studied by Kawamura et al. (2002) and Lin and Li (2003) using LES for a
surface piercing circular and square cylinder, respectively. It was shown that free surface reduces
strength and frequency of Karman shedding, but values for Sth were not computed. For reattaching
separation, there exist another instability associated with the periodic enlargement and shrinkage of
the separation bubble termed ‘flapping’ instability with a standing wave type nature (Kiya et al.,
1985). The general agreement is that the flapping frequency scales with the mean reattachment
length XR (StR= f XR /U∞). Kandasamy (2005) gives a comprehensive review of the above-
mentioned instabilities in unsteady 3D flows. Comprehensive investigation of all the three above-
mentioned instabilities have been carried out only for a handful of cases namely backward facing
steps (Kiya et al., 1985) and blunt cylinders (Sigurdson, 1995), and never for free-surface wave
induced separation.

URANS has been used to simulate these instabilities for some basic geometry with mixed
results. URANS turbulence models under-predict the level of the turbulent stresses in the detached
shear layer emanating from the separation line (Johnson et al. 1994), and hence sometimes predict
incorrect flow recovery downstream of reattachment, i.e., separation bubble vs. fully stalled regime
(Menter et al. 2002). In addition, the dissipative nature of URANS causes amplitude defect for
strength of the coherent structures and sometimes completely fail to capture the instabilities.
Constantinescu et al. (2001) did a comparative study of numerical simulations of the sub-critical
flow over a sphere using URANS, LES, and DES. URANS predictions was obtained using two-
layer k–e (Chen and Patel, 1998), k–ω (Wilcox, 1988), v2–f (Durbin, 1991), and Spalart-Allamaras
(Spalart and Allamaras, 1994). Simulations were at a Reynolds number of 104. URANS predictions
of the streamwise drag, agreed reasonably with measurements. URANS models also adequately
predicted the pressure and skin friction coefficients along the sphere. URANS predicted the value
of the main shedding frequency accurately, but not its amplitude, and failed to capture the
formation of shear layer vortices. Nevertheless, researchers have shown the ability of URANS to
capture the shear layer instabilities for other geometries. Khorammi et al. (2001) used URANS
with k-ω turbulence model for their simulations for the acoustic analysis of slat free-shear layer and
were able to capture the formation and the subsequent development of the free shear layer
instabilities that were responsible for the radiated noise. Comparison with experimental
measurements showed good correspondence for the frequency spectrum although the amplitude of
the computed noise was lesser than measured. In the follow up paper (Khorammi et al., 2002) the
amplitude defect which was attributed to the excessive dissipation effects was shown to be
corrected by switching off the turbulence production term in the slat cove region.

In particular, URANS is appropriate for simulating turbulent flows when there exists
dominant periodic modes with a spectral gap separating them from the random oscillations, thus
separating the resolved time-scales and modeled scales. It is difficult to get a reliable estimation of
the existence of the spectral gap. In addition, in some cases even if initially two well separated
inertial ranges exist with a spectral gap between them, it is swamped by small scale motions
produced through cascades from large scale turbulence (Poequet et al., 1983). Fortunately, for this
particular case, frequency spectrum of foil surface pressure close to the free surface in the separated

3
region in EFD (Metcalf, 2006) shows a spectral gap separating the inertial ranges of the low
frequency organized oscillations from the high frequency random motions. Metcalf et al., (2006)
were able to extract the dominant modes corresponding to the organized motions using triple
decomposition of the spectra for both the foil surface pressure and free-surface elevation. The free-
surface elevation spectra show the same dominant modes for organized unsteady structures as the
foil surface pressures, but the spectral gap disappears as it is swamped by small-scale motions due
to splashing (Fig 1). Although spectral segregation is absent at the free surface, the existence of the
same dominant modes at the free surface and foil surface suggest that the small scale motions
produced by free-surface splashing and associated turbulent eddies do not significantly affect the
underlying flow pattern but rather mask the underlying coherent structures. As we shall see, current
URANS proves effective in capturing the dominant modes for the organized unsteady structures,
but with amplitude defect due to rapid dissipation of the rolled up vortices. All three instabilities
expected, namely the shear layer instability, Karman-like large-scale vortex shedding, and the
flapping instability are captured for Fr=0.37. The solutions are validated for mean free-surface
elevation and foil surface pressure using Metcalf et al. (2006). RMS and frequency values are also
compared against Metcalf et al. (2006).

Current work is part of a complementary EFD/URANS/DES to investigate flow physics of


such a complicated flow using both surface tracking and level set approaches. The tracking
approach automatically clusters grid points near the deforming free surface and hence is easily able
to resolve the shear layer instability initiating at the free surface. Level set approach required a fine
grid throughout the entire dynamic range of the deforming free surface thus making it too
expensive computationally. The finest affordable grid for level set approach with URANS showed
better comparison of the mean flow, but failed to capture the shear layer instability. However, DES
simulations with level-set capture the shear layer instability, but same as EFD the clutter caused by
the numerous small-scale vortices and turbulent structures make analysis of the coherent vortical
structures almost impossible. URANS with surface tracking successfully isolates the coherent
structures, thus making analysis possible. Thus, each approach has its own merits and the approach
presented in this paper is best suited for the analysis of the coherent structures in the separated
region. Analysis of free-surface splashing, wave breaking and associated turbulent eddies are
beyond the scope of this paper and certain aspects of these issues will be dealt with in forthcoming
paper using DES with level set methods.

COMPUTATIONAL METHOD

The CFD code CFDSHIP-IOWA is a general-purpose URANS research code (Wilson et


al., 2006). Governing equations for Cartesian coordinates, the continuity, and momentum equations
in non-dimensional tensor form are
∂U i
=0 (1)
∂xi

∂U i ∂U i ∂pˆ 1 ∂ 2U i ∂
+U j =− + − uiu j (2)
∂t ∂x j ∂xi Re ∂x j ∂x j ∂x j

4
⎛ p − p∞ ⎞
Where pˆ = ⎜ + z Fr 2 ⎟ and Ui = (U, V, W) are the Reynolds-averaged velocity
⎝ ρU 0
2

components

Temporal discretization uses second order finite differences. Spatial discretization uses
both second and third order finite differences. The pressure implicit split operator (PISO) algorithm
solves the incompressible Navier Stokes equations. Fourth-order artificial dissipation implicitly
added by taking a linear combination of full- and half-cell operators (Sotiropoulos and Abdallah,
1992) overcomes the pressure velocity-decoupling problem caused by the collocated grid. The
overall method is fully implicit and a line-ADI scheme with a pentadiagonal solver and under-
relaxation solves the algebraic equations. Integrating the normal and tangential stresses over all no-
slip boundaries give the fluid forces and moments that act on the system. The turbulence model
used is a blended k-ω/k-e model. The k-ω model has proven to be robust, applicable to complex
geometries and flows, and accurate.

A free surface tracking method with nonlinear kinematic and approximate dynamic free
surface boundary conditions simulates the free surface. With the free surface tracking approach, the
computational domain covers the water region only and the grid is dynamically conformed to the
ship hull and free surface location. The KFSBC used updates the grid every time step to compute
the free surface location and requires that the free surface is a material surface. The DFSBC
provide boundary conditions for velocity and pressure and require that stresses are continuous at
the free surface. The KFSBC computes the evolution of the free surface, while the DFSBC
provides boundary conditions for velocity and pressure. Considering the KFSBC, the following
condition satisfies the requirement that the wave elevation ζ  be a stream surface
D (ζ − z )
=0 (3)
Dt
Expanding equation (3) gives a 2D hyperbolic PDE for ζ
∂ζ ∂ζ ∂ζ
+U +V −W = 0 (4)
∂t ∂x ∂y
At the intersection of free- and no-slip surfaces (i.e., the contact line), equation (4)
becomes singular when the contact line is in motion but the fluid velocity is zero due to the viscous
no-slip boundary condition. To overcome the problem a small near-wall region is “blanked out”
when solving equation (4) and the solution in this region gets linearly extrapolated from the interior
of the domain.

The DFSBC requires that the normal and tangential stresses are continuous across the free
surface.

τ ij n j = τ ij* n j (5)
Where nj is the unit normal vector to the free surface and τij and τij* are the fluid and external-
stress tensors, respectively. The following approximations obtain free surface boundary conditions
from equation (5): (i) the external stress and surface tension assumed zero so that τ ij n j = 0 ; and
(ii) the gradients of the free surface and normal velocity in the tangential directions assumed small
(i.e. ∂ζ / ∂x = 0 , ∂ζ / ∂y = 0 , ∂W / ∂x = 0 , and ∂W / ∂y = 0 ).

5
The applicability and quality of the code CFDSHIP-IOWA for complex free-surface flows
and other ship hydrodynamic applications has been well tested and established (e.g. Simonsen and
Stern, 2005; Kim et al., 2006; Wilson et al., 2006). The status of ship hydrodynamics CFD was
assessed at the recent Gothenburg 2000 Workshop (Larsson et al., 2003) and levels of verification
and validation and overall code performance demonstrated that CFDSHIP-IOWA was among the
best codes for naval combatant test case.

Geometry and Flow Conditions and complementary EFD data

Fig. 1 shows a photo of the foil at Fr=0.37 and Re=1.52 × 106 in the tow tank. The tow tank
dimensions are w/c=2.5 and h/t =2 (where w=width of tow tank, h=depth of tow tank, c= chord
length, t=foil draft). For simplicity of the simulations, instead of modeling the exact tow tank
conditions, we model the foil with extended foil draft (2 × t) where the foil reaches all way down to
the bottom boundary with no flow beneath foil (Fig. 2). The extended foil draft prevents the
restricted water effects due to the bottom from affecting the free surface solutions. The simulations
also ignore the tow tank walls. In addition, in the tow tank there is also flow beneath the foil along
its flat bottom, which forms a separation bubble near the leading edge affects the pressure
distribution near the foil bottom and local flow. Previous CFD and EFD established relatively small
dependence of the free surface separation pattern to both the wall, and foil bottom. Current study
consists of both laminar and turbulent simulations. Laminar cases simulated are 2D and Fr=0.37 for
two Re (=1500 and 2500). Turbulent case simulated are Fr=0.37 and 0.55 with respective Re= (
1.52, and 2) × 106. Validation data (Metcalf et al., 2006) include free-surface mean elevation and
foil surface mean pressure for both turbulence cases, and RMS and frequency spectra for Fr=0.37.

Computational Domain and Grids

The C-grid used (Fig 2) has a far-field boundary that replaces the tow tank-wall boundary.
The C-grid is a body-fitted grid whose far-field boundary is a semicircle of radius R encircling the
foil with the trailing edge as its focal point. A rectangular grid of length Le and half-width R
extends from the trailing edge to the exit. The free surface and the foil surface have grid clustered
close to them. Y+ < 1 for all grids. The grids were generated using commercial code GRIDGEN.
Domain size was finalized based on results from domain convergence studies. R/c= 5, 7, 10 and
Le/c = 5.5, 6.6, 8.1, respectively for the 3 domains in the study. The solutions show convergence
for both the mean and the FFT. For the mean, the convergence ratio is 0.6, and the percent change
for the largest two grids is 0.15 %. The dominant frequency for the small, medium, large domains
are 1.39, 1.41, 1.4 Hz respectively, showing an oscillation of 0.7 %. The convergence ratio for the
magnitude of the dominant frequency (1.4 Hz) is 0.55, and the percent change for the largest two
grids is 10%. The largest domain was chosen for the simulations.

Boundary Conditions

The boundary condition on each boundary is as follows. On the foil surface (no slip), (U,
V, W) = ∂P/∂n =0 (where n is normal to the body). On the exit plane, axial diffusion and pressure
gradient are assumed negligible, i.e., ∂2(U, V, W)/ ∂x2 = ∂P/∂x=0. On the bottom deep boundary, an
impenetrable slip condition is used, i.e. ∂ (U, V, P)/ ∂z=W=0. On the outer boundary the far-field
boundary condition, U=U∞, V=W=P= 0. On the free surface ∂ (U, V, P)/ ∂z=W=0 and p = Z /Fr2.
The conditions on the turbulent quantities are as follows: on the foil surface, k=0, ω= 60 / (Re ×
0.075 × ∆y2), eddy viscosity vt = 0; on the exit plane, on the bottom deep boundary, on the outer
boundary and on the free surface ∂k/∂n=∂w/∂n=∂vt/∂n=0. These boundary conditions along with
the surface tracking method used are inadequate to accurately model the turbulent/splashing free

6
surface, which require a level-set approach and modified boundary conditions (Brocchini, 2002).
However, since the scope of the present paper is to resolve just the coherent structures, the
approach remains valid as the turbulent/splashing free surface does not significantly affect the
organized coherent structures, but rather just mask it at the free surface as discussed previously.

Analysis Methods

The subjects of analysis include free surface wave effects, foil surface pressure effects,
unsteady separation pattern, instability mechanisms, skin friction topology, and vortex and
turbulence structures. The methods used for vortex detection in the separated region are velocity
vector eigen-modes vortex detection (a Tecplot feature) and the Q- criterion (Hunt et al., 1995).
Studying the time evolution of the vortices and the frequencies associated shed light into the
mechanisms of these vortical structures occurring within the separation region. Parametric studies
on RMS, frequency, vortex strength and total head show the effects of Re, Fr and the free surface
on these structures and the instabilities associated with them. Normalized Strouhal numbers serve
to correlate the instability frequencies with those identified for other fundamental geometries.
Correlating these structures and frequencies with those already identified for other 3-D separations
and factoring in the influence and effects of the free surface completes the over all picture.
Topological analysis of the separation pattern at key time steps shows how these instabilities affect
the topology and how the flow topology permits behavior such as vortex merging and breaking.

VERIFICATION AND VALIDATION

Verification and validation used Stern et al. (2001) methodology with updated correction
factors by Wilson et al. (2004). Grid studies were conducted for Fr=0.37. Fr=0.55 results were
unsteady only on the fine grid as the coarser grids had insufficient resolution in the wake; hence
quantitative verification on the three grids was not performed. Table 1 and 2 summarize the grids
used for the grid studies, and the different cases conducted, respectively. Fr=0.37, being the main
case of interest undergoes 3 separate grid studies (case 1 using URANS2 with refinement ratio (rK)
= 1.189, and case 2 and 3 using URANS3 with rK=1.189 and 1.414, respectively). Verification was
performed for mean CTX, and dominant higher end frequency (fDOM) caused by shear layer
instability. Validation of mean CTX was not done due to unavailability of data. Verification and
validation of the point variables (free-surface elevation, foil surface pressure) was performed for
Fr=0.37 using case3 grid study. Grid densities were determined considering initial understanding of
flow physics and computer resources, prior to detailed analysis of results. Final URANS3 solutions
using a non-systematic finer grid (solution grid) ensured sufficient grid resolution to support
conclusions for Fr=0.37.

Iterative and Statistical Convergence

Parametric studies on sub iterations (free-surface momentum coupling) for each time step
ensured iterative solution convergence at each time step. Results showed difference about 5% and
< 1% for CTX by changing free surface/momentum coupling outer iterations from 3 to 4, and 4 to 5,
respectively. Simulations then used 4 outer iterations. Statistical convergence of the running
average on the time histories of CTX established statistically stationary unsteady solutions. The
criterion for statistical convergence is that the fluctuations of the running mean drop down to less
than 1 % of the mean value.

7
Time Step Studies

Time step studies were conducted on grid3 (case2 fine grid) for Fr=0.37 CTX with ∆t
=0.00707, 0.01 and 0.014 using URANS3. The results show oscillatory convergence for CTX with
the uncertainty (UT = 0.3% mean CTX) in par with iterative uncertainties (UI), being order of
magnitude lesser than the case2 grid uncertainties (UG). UT~UI << UG, so that simulation numerical
uncertainty USN = √(UG2+ UT2+ UI2 ) ~ UG.

Verification and Discussion of Integral Variables

Table 2 tabulates results from grid studies for mean CTX and fDOM. For Fr=0.37, URANS2
(case1) with rK = 1.189, observed order of accuracy (pG=3.7) is greater then the estimated order of
accuracy (pGest=2) with moderate USN (=6.9%) for CTX, and observed pG =1.47 for fDOM with high
uncertainty (=44%). For URANS3 with rK=1.189 (case2), observed pG for both CTX and fDOM is
lower than pGest with moderate and high uncertainties for CTX and fDOM, respectively. For case3 with
rk=1.414, observed pG is close to pGest (=3) for both CTX and fDOM (pG=2.9 and 3.24, respectively).
The uncertainties are much lower compared to cases 1 and 2 (USN= 4% and 9.3% for CTX and fDOM,
respectively). Cases 1 and 2 with the coarse grid show high variability for pG indicating that the
asymptotic range is not reached. Fig 3a shows convergence of the CTX running mean for case3
along with the solution grid. The mean values for CTX on their respective fine grids for URANS2
(case1) and URANS3 (case3) are 0.01284 and 0.01335, respectively and the difference is 3.8%,
which lies within the uncertainty levels of both cases. The mean CTX for the solution grid is 0.013,
which lies within the uncertainly interval of case3 fine grid. Fig 3b shows the FFT for the solution
grid. FFT shows three distinct peaks corresponding to the shear layer instability (2 Hz), symmetric
Karman-like vortex shedding (0.74 Hz) and flapping instability (0.32 Hz).

Verification, Validation and Discussion of Point Variables

For Fr=0.37, detailed verification and validation was performed using case3 grid study.
Validation studies were performed for wave profile and elevation, and uncertainties are got over
the entire free surface, thus locating regions where improvement is warranted. Reader is referred to
Kandasamy (2005) for details. CFD shows monotonic convergence, and the solution grid
accurately predicts the toe of the bow wave, where separation initiates, at x/c=0.4 (Fig 4a).
However, EFD profile shows sharper recovery of wave elevation after the toe.

CFD over predicts wave elevation magnitudes (Fig 4b). The uncertainties are highest at the
toe and the Kelvin wave crest (USN = 9 % dynamic range). In addition, the comparison error
(E=Data-Simulation) is highest at both these regions (25% of data) and the percent by which
validation not achieved is 17% and 6% at the toe and the Kelvin wave crest, respectively.
Tabulated in table 3 and 4 are the profile and domain averaged verification and validation results,
respectively. Validation is almost achieved in a global sense; percent by which global validation
fails is 0.44% and 1.08% for profile and domain averaged elevations, respectively. Comparison of
the solution grid with the case3 fine grid showed domain averaged elevation difference < 1% with
maximum difference at the toe and Kelvin wave crest (8%) where USN is highest. Since the overall
pattern is very much similar and the differences in elevation lie close to uncertainty levels, all
further analysis for Fr=0.37 will be based on the solution grid.

8
Fig 5a compares free-surface elevation RMS values for CFD and EFD. Maximum RMS
(=0.01) occurs at the toe and along the separated shear layer. The dominant frequency content
contours in Fig. 5b have a cut off frequency of 0.2 Hz (as the time used for the FFT was 10s) with
RMS<0.003 blanked out as they lie within noise levels. URANS shows three distinct regions in the
separation region: The high frequency 2 Hz region within the separation region (Shear layer
instability dominates), the intermediate frequency 0.74 Hz right after reattachment (Karman-like
instability dominates), and the low frequency 0.32 Hz region around the Kelvin wave crest
(flapping instability dominates). All three are also evident in the EFD, but not quite as distinct.
CFD RMS values compares better when the shear layer first separates with about 20% amplitude
defect, but the defect increases downstream as URANS rapidly dissipates the rolled up vortices.
The free surface splashing not modeled in CFD also contributes to the amplitude defect, but to a
lesser extent as EFD showed that the organized structures contribute most to the free-surface RMS.

Point comparisons in the three distinct regions show a good match with the EFD (Fig 6a)
for the frequencies, but with amplitude defect. At the shear layer instability dominant region
(x=0.75, y=0.14) a clear peak occurs at 2 Hz for CFD and 2.2 Hz for EFD. At the Karman-like
shedding region (x=0.90, y=0.104), a clear peak occurs at 0.74 Hz for CFD and 0.75 Hz for EFD.
Unlike CFD, the EFD peak is of lesser magnitude than the 2 Hz peak. This deters the 0.75 Hz
frequency from showing up more in the dominant FFT contours (Fig 5b). At flapping dominant
region (x=0.90, y= 0.32) both CFD and EFD show peak at 0.32 Hz. Fig 6b shows CFD/EFD point
comparisons of pressure oscillation and FFT at three locations corresponding to the different
instabilities. Similar to the free-surface comparison the frequencies are captured quite well, but
with an amplitude defect. Fig 7 gives a comparison of the foil surface mean pressure, and Fig 8
compares foil surface RMS values. Note that EFD pressure taps on the foil surface were sparse and
hence data is not available near the toe. RMS values in Fig 8 show the path of the shear layer
vortices. Similar to the free-surface RMS, the rapid dissipation of the vortices results in a smaller
unsteady region in CFD.

Fr=0.55 solutions exhibit higher wave steepness (Fig 9a) and the toe shifts towards the
trailing edge (x/c=0.72). The far field wave elevation contours show good qualitative match with
the EFD (Fig 9). Near-field EFD measurements are not available. Maximum RMS=0.04 occurs at
toe (Fig 10a) which is four times higher than Fr=0.37. Similar to Fr=0.37, the contour of dominant
frequency (Fig 10b) shows three distinct frequency regions: shear layer instability region (0.9 Hz),
Karman instability region (0.246), and interference region with twice Karman frequency due to
response to shedding from both sides (0.492). Fig 11 qualitatively compares the foil surface
pressure with EFD.

MEAN SEPARATION PATTERN

Fr=0.37

Fig 12 shows the mean separation pattern. The shear layer separates in the region close to
the free surface at x/L=0.40 and the reattaching separation region forms a wedge shape on the free
surface. The skin friction lines at the free surface and the foil surface illustrate the owl shaped
topological pattern. The mean topological pattern agrees qualitatively with Stern et al. (1993) and
Kandasamy (2001), but with the extra spiral node and associated saddle resulting from the unsteady
shear layer vortices not seen in the previous steady simulations. The volume streamlines show
vortical flow transporting fluid upward toward the free surface. The vortex strength along the
vortex core decreases as it nears the free surface. At the trough, the total head or stagnation

9
pressure (PStag) decreases to 58% of the original level due to dissipation. Pogozelski et al.(1997)
report that near the trough the total head decreases 50%-60% for their case.

Fr=0.55

Fig 13 shows the mean separation pattern for Fr=0.55. Separation is non-reattaching
starting near the trailing edge at x/c=0.72, region of reverse flow extending to x/c=1.3. The vortical
structures extend from the foil trailing edge upward towards the free surface going into the wake.

INSTABILITY ANALYSIS

Laminar Solutions

Laminar flow simulations for cases Re = 1500 and 2500 (2D case and Fr=0.37) revealed
unsteady non-reattaching separation with shear layer and Karman instabilities. For all cases, Stθ
(~0.013) lies in range similar to airfoils (Pauley et. al., 1990). The free surface does not alter the
non-dimensional pressure gradient enough to cause significant changes in Stθ. On the other hand,
the free surface reduces Sth (=0.069), compared to the 2D cases (Sth ~ 0.08). For laminar flows,
assuming UC scales with US the consensus is that the shear layer frequency scales with Re½.
Sigurdsen (1995) proposed the following reasonable approximation, not completely accurate, as
individual calculation of θ is not made.
FSL D
= Constant (6)
U S Re1D/ 2

For the 2D cases the constant is about 0.013 and for Fr=0.37 cases the constant is about 0.025.

Fr=0.37 Turbulent solutions

For Fr=0.37 there are two types of vortices associated with the separation. The first forms
due to the recirculating fluid in the separation bubble at the free surface. The second is a result of
the shear layer instability, which causes the separated shear layer to roll up. There are three
instabilities associated with this case: the shear layer instability, the Karman-like instability for
reattaching flows and flapping instability.

Shear layer instability

The shear layer vortices form at the toe just after separation. Fig. 14 shows the flow pattern
over one time-period corresponding to the shear layer vortex shedding. Vortical flow occurs in
regions of positive Q shown by the blue isosurfaces with imbedded core lines contoured by vortex
strength. The free surface seems to inhibit the shear layer roll up and hence the shear layer vortices
have higher strength in the region below. The shear layer vortex convects downstream; then merges
with the stationary recirculation vortex for a brief time and then breaks down as it further convects
downstream. It is a periodic process of frequency about 2 Hz. The mean flow visualization (Fig 13)
shows the structures similar to what we see during the merger of the two vortices. There is a low-
pressure field created near the vortex core, magnitude of the low-pressure dependent on the
strength of the vortex core. The free-surface elevation has direct relation to the pressure and hence
the free surface depresses along the vortex motion causing shoulder waves with frequency 2 Hz.
The shear layer vortex has a higher strength when it first forms near the toe. This explains why the
RMS of free-surface elevation is the highest near the toe. US=1.45 and θ=0.0028 giving Stθ = 7.7 x

10
10-3. The value is little more than half of that got for the laminar cases. This seems to contradict
findings from studies on mixing layer (Ho and Huerre, 1986), where turbulent Stθ is higher than
laminar Stθ. However, considering the fact that this separation is APG induced and Stθ is dependent
on the APG, our simulations suggest that Stθ varies inversely with the non-dimensional APG,
turbulent cases having higher APG than the laminar cases.

Brown and Lopez (1990) criterion for vortex breakdown in swirling flows states that for
breakdown to occur the helix angle of velocity should exceed that of vorticity along some stream
surface (necessary not sufficient condition). i.e. If, δ=tan-1 (ν/w)- tan-1 (η/ζ), where the ratios (ν/w)
and (η/ζ) are of the azimuthal and axial components of velocity and vorticity respectively, then δ
has to be positive in order for breakdown to occur. Contours of δ along the vortex core (Fig. 15)
show the criterion satisfied with the breakdown occurring at maximum positive δ (0.4) on the core.
Fig. 15 also shows the diverging streamlines and stagnation in the vortex core at the breakdown
point with flow converging from the top and bottom.

Karman-like instability

Fig. 16 illustrates the symmetric type Karman-like shedding process for reattaching flows,
namely the merger of the shear layer vortices to form bigger vortices that then shed. Such type of
shedding is seen in other reattaching flows such as backward facing steps and flow past the leading
edge of a blunt cylinder. The blue isosurfaces denoting positive Q portray the behavior of vortical
structures over one shedding period. Two or three shear layer vortices merge after the reattachment
point and then shed downstream with frequency 0.74 Hz. The normal distance of the separated
shear layer from the foil at the free surface, used as the equivalent for half wake thickness h
(=0.13), gives Sth = 0.066.

Flapping instability

Analysis of the flow field shows periodic increase and decrease of the separation bubble
size with frequency 0.32 Hz. A half-domain medium-grid (187,824 grid points) with symmetry
conditions imposed on the center-plane serves better to study the flapping instability as it cuts off
the asymmetric influences. Fig 17 shows two extremities of the flapping: the low mode (smaller
separation and higher wave amplitude) and the high mode (bigger separation with smaller wave
amplitude). The flapping frequency with the half domain medium grid (0.18 Hz) is lower than that
got from the solution grid. Both domain difference and grid density cause this difference. The
separation-bubble reattachment length for the high mode is 0.7 L and for the low mode is 0.46 L.
The entrainment of fluid from the separation bubble increases with increase in number of the
coherent vortices, which in turn increase with reattachment length. The number of coherent
vortices is quantified by the parameter we term entrainment number (En = LR/ λ SL). En = 2.2 for the
high mode and 1.4 for the low mode. The pressure difference (∆ P) between separating and
reattaching points which is higher for the low mode (∆ p= 0.86) compared to high mode (∆ p=
0.72) dictates the re-injection rate (Fig 18). During the high mode the increased entrainment and
decreased re-injection starts shrinking the separation bubble and at the low mode the decreased
entrainment and increased re injection starts enlarging the bubble, causing the observed pulsation.

The free surface boundary conditions dictate that d (U, V, W)/dz = 0, which implies that at
the crest and the trough of the free shear layer stream lines (Fig 18), where the slope is zero the
convective derivatives in the normal direction vanish. Assuming the free surface at the crest and
trough oscillates only in the vertical direction (ignoring the minute transverse travel); the only
vertical forcing factor is the pressure gradient at the free surface. Approximating this motion as

11
similar to simple harmonic motion of a spring mass system at the crest and the trough gives a
simple predictive model for the flapping instability. In our model, a coefficient in terms of pressure
replaces the spring coefficient giving

If the equilibrium pressure is PO and the pressure at the free surface is pfs. Then the force
balance on the volume of fluid (V) is
∂ 2ζ
Fz= -(pfs–po)dxdy = m a = m ∂ ζ2 = ρ dx dy D 2
2
(7)
∂t ∂t
∂ 2ζ - ( PFS - PO ) − (ζ − ζ O )
= = (8)
∂t 2
ρD Fr 2 ρ D
∂ 2ζ 1
= -k2 ζ where k2 = (9)
∂t 2
Fr ρ D
2

ζ (t)= A Cos ( k t ) + B Sin ( k t ) (10)

k gives the natural angular frequency

k 1
f= = (11)
2π 2 π Fr ρ D

The only parameter required for this model is D, which is half the magnitude of oscillation
at the crest. For the half domain case, D=0.0062, which gives f = 0.17. In spite of such crude
approximation, the result matches surprisingly well with the simulation results i.e. f=0.18 Hz. The
formula is also applicable for the full domain solutions. The solution grid D=0.0016 gives f= 0.34
and the actual observed frequency is 0.32 Hz.

Instability Effects on Flow Topology

The shear layer shedding changes the topology of the separation most. The Karman and
flapping instabilities have no effect on the topology at the foil surface as they occur away from the
foil. The separation for Fr=0.37 has features of both open separation and closed separation.
Separation starts at a saddle point at the toe. The resulting recirculation region on the free surface
will be termed the primary closed separation. The shear layer vortex is of the ‘tornado kind’ or
‘foci’, which is a special kind of open separation. This differs from other open separation because it
alters the flow topology unlike the usual open separation. The primary free surface separation
causing the recirculation vortex (node N1) starts at saddle S1 (Fig 19a). The spiral node N2 (initial
stage of the shear layer vortex) occurs just below this saddle point. The separatrixes LS2 and LA2,
which converge at saddle pt S2, enclose the primary separation. LS2 is the line of separation and
LA2 is the line of attachment. LS2 terminates at the spiral node N2. This is a ‘horn-type’ dividing
surface (Wang, 1998). LA2 terminates at node Na at the reattachment point.LS2 and LA2 form a
wedge shaped pattern. Note that this wedge does not completely close, as LS2 does not reach the
free surface, but rather ends at a node N2, just below the free surface. This topological structure
plays an important role in allowing the merger of the shear layer vortex with the recirculation
vortex, as they are not cut off from each other. As the shear layer vortex (node N2) convects
downstream (Fig 19b), the wedge opens up, and since the topology of the primary separation is
such that the separatrix does not cut off the shear layer vortex, but instead terminates at it, it allows
the body vortex to enter the recirculation region. Thus, the primary separation remains open to the

12
shear layer vortex. The convection of the node N2 causes the saddle S1 attached with this node also
to convect downstream. This saddle S1 has two separatrix LS1 and LA1. LA1 terminates at Na too
and the dividing surface of LA1 serves to cut off N2 from the LA1, LS1 wedge enclosure, as these
separatrixes are not open to N2. As N2 convects down stream, the wedge enclosed by LA1 and
LS1 enlarges cutting off N2 from the primary separation. The phenomenon facilitates the vortex
breakdown. When the vortex breaks, N2 cuts off from the primary separation, now enclosed by
LS1 and La1. On the free surface, another node/saddle pair forms and now LS1 terminates at the
newly formed node (shear layer vortex) to which the LS1, LA1 wedge enclosure is open. The cycle
repeats. The topological rule for our case considering the free surface to be a wall is ΣN-ΣS=0
(Zhang and Stern, 1996). That is for every node there has to be a corresponding saddle. The
topology satisfies this rule at every instance in the cycle.

Fr=0.55

The flow has shear layer instability and asymmetric Karman instability in this case. This
differs from Fr=0.37 which has symmetric Karman shedding. In addition, the flapping instability is
absent. The vortex dynamics of this separation is also different.

Shear layer instability

Fig 20 shows the time evolution of the shear layer vortices along one period. The shear
layer vortices form at the toe and amalgamate into the trailing edge vortex. The distinct 1.1 s period
corresponds to the merger of two shear layer vortices and their subsequent amalgamation into the
trailing edge vortex. Thus, the frequency of shear layer instability is two times that seen in the time
history i.e. f=1.84. θ =0.002 and Us =1.4 giving Stθ = 0.0051. Value is lesser than Fr=0.37 case as
Fr=0.55 has a higher adverse pressure gradient than Fr=0.37 and Stθ varies inversely with non-
dimensional pressure distribution.

Karman instability

The asymmetric type Karman instability (Fig 21) that occurs in the wake has a frequency
0.246 Hz. The half wake thickness (h)=0.4 and Us =1.45 giving Sth = 0.0678. This lies in the same
range as all free surface simulations conducted (0.065-0.069). There is a region in the centerline,
which has twice this frequency because of response to both sides.

CONCLUSIONS

For the first time the unsteady flow pattern of the free-surface wave induced separation has
been elucidated, with detailed verification and validation providing credibility to the solutions.
Solutions were compared with previous CFD/EFD, strengthening understanding and resolving
unanswered questions by comprehensive treatment of the unsteady separation through juxtaposed
topological and structural dynamics analysis tied to previous literature.

Order of accuracy reached estimated value, signifying asymptotic range for the finest grid
studies, with CTX USN = 4% for Fr= 0.37. Validation is almost achieved in a global sense; percent
by which global validation fails is 0.44% and 1.08% for profile and domain averaged elevations,
respectively. Validation fails by the highest margin at the toe (17% defect) due to the intense
splashing seen in EFD not modeled by CFD. Point comparisons of elevation show good match for
the frequencies, but with RMS under-predicted.

13
Two separation regimes were simulated; Fr=0.37 has closed reattaching separation with
dominant shear layer instability, weaker symmetric Karman like shedding and flapping instability,
and Fr=0.55 has closed non-reattaching separation with dominant shear layer instability and weaker
Karman shedding. Current simulations support previous findings that the free surface reduces both
strength and frequency of large scale Karman type vortex shedding. All cases simulated in this
study show reduced Sth (0.66 – 0.69). Shear layer vortex formation is also hampered on the free
surface. Stθ for the shear layer vortices varies inversely with the pressure gradient at separation
such that Stθ (laminar, Fr=0.37) > Stθ (turbulent, Fr=0.37) > Stθ (turbulent, Fr=0.55). The mean
separation topology bears close resemblance to Stern et al. (1993) and Kandasamy (2001), the
difference being that the current results have the extra node/saddle combination, missing in the
previous two, making current pattern topologically equivalent to owl shaped pattern of the second
kind. The previous two simulations do not resolve the shear layer vortices that are responsible for
the extra node/saddle combination arising in the mean flow pattern.

To the authors’ knowledge, this work gives the most detailed explanation of unsteady free-
surface separation to date. Current work is part of a complementary EFD/URANS/DES approach
to investigate flow physics of such a complicated flow. Forthcoming paper using URANS and
DES, with both free surface tracking and unsteady single-phase level set method (Carrica et. al.,
2005), will address the pros and cons of each approach. Initial analysis of ongoing DES with free
surface tracking and level-set approaches also show reattaching separation in the mean with
topological equivalence to owl shaped pattern of the second kind. This provides confidence that the
fundamental URANS flow topology downstream of separation in this case is not adversely affected
due to under prediction of turbulent stresses.

ACKNOWLEDGEMENTS

The Office of Naval Research under Grant N00014-01-1-0073, administered by Dr. Patrick
Purtell, sponsored this research.

14
Table 1 Grids used for verification

5 (Non-
Grids 1 2 3 4 systematic final
solution grid)

Size 70,680 114,048 187,824 531,006 1,004,400

Y+ 0.36 0.3 0.25 0.18 0.9

Table 2 Grid study for integral quantities

pG USN
Case Simulation Grids rK
CTX fDOM CTX fDOM
URANS2
1 1,2,3 1.189 3.7 1.47 6.9 44
Fr=0.37
URANS3
2 1,2,3 1.189 1.94 2.4 15 41
Fr=0.37
URANS3
3 1,3,5 1.414 2.9 3.24 4 9.3
Fr=0.37

Table 3 Profile and domain average verification of point variables

RG PG CG UG UGC

Profile 0.535 1.8 0.47 2.26 1.2


Elevation 0.535 1.8 0.47 1.78 0.95

Table 4 Profile and domain average validation of point variables

E% UV % UD % USN %

Profile 5.92 5.48 5 2.26


Elevation 3.72 2.64 1.95 1.78

15
Fig 1 EFD, Surface-piercing NACA24 foil at Fr=0.37

Fig 2 Computational domain, boundary conditions, and fine grid

(a) (b)

Fig 3 Forces: a) Time history, and b) FFT

16
(a)

CFD

EFD
(b)

Fig 4 Fr=0.37: a) wave profile, b) wave elevation

Flapping
CFD CFD
SL vortices
Karman-like
vortices

EFD EFD
(a)
(b)
Fig. 5 Fr=0.37 free surface elevation: a) RMS, b) dominant FFT

17
CFD x= 0.74, y=0.12 (SL) EFD CFD x= 0.69 z= - 0.11 (SL) EFD

CFD x= 0.90, y=0.104 (Karman-like) EFD CFD x= 0.86 z= - 0.14 (Karman-like) EFD

CFD x= 0.90, y=0.32 (flapping) EFD CFD x= 0.79 z=-0.14 (flapping) EFD
a b
Fig 6 Time history and FFT: a) elevation, b) pressure

18
(a) (b)
Fig. 7 Qualitative comparison of foil surface pressure for Fr=0.37: a) EFD, b) CFD

(a) (b)
Fig. 8 Fr=0.37 surface pressures RMS: a) EFD, b) CFD

CFD

EFD

Fig. 9 Fr=0.55 elevation

19
(a)

(b)

Fig. 10 Fr=0.55 free-surface wave elevation CFD: a) RMS, b) FFT

(a) (b)
Fig. 11 Fr=0.35 mean foil surface pressure: a) CFD, b) EFD

20
Fig 12 Fr=0.37 Mean separation pattern

Fig 13 Fr=0.55 Mean separation pattern

21
Fig 14 Shear layer instability –positive Q contours with imbedded vortex cores colored by strength;
Pressure time history at x,z = (0.9, -0.095)

Fig 15 Contour plots of δ at breakdown

22
Fig 16 Karman instability -positive Q contours showing vortex amalgamation and shedding.
Pressure time history at x,z = (0.9, -0.095)

a b
Fig 17 Top view of shear layer flapping: a) high mode, b) low mode

23
Fig 18 pressure plots along shear layer and foil surface at high mode and low mode

t sec

Flow
Free surface

Sa
Na

Foil surface

t + 0.2 sec
Free surface

Sa
Na

Foil surface

Fig 19 Skin friction topology Fr=0.37

24
Merger
Amalgamation

Fig 20 Fr=0.55 Shear layer vortex merger and amalgamation shown over one period; pressure time
history at x, z = (0.86, -0.3)

Fig 21 Fr=0.55 separation pattern; elevation time history at x, y= (1.11, 0.15)

25
REFERENCES

BROWN, G. L. & LOPEZ, J. M., 1990, “Axisymmetric Vortex Breakdown: Part 2. Physical
Mechanisms”, Vol. 221, 553
CARRICA, P.M., WILSON, R.V., & STERN, F., “An Unsteady Single-Phase Level Set
Method for Viscous Free Surface Flows,” IIHR Report No. 444, April 2005
CHEN, H. C., & PATEL, V. C., 1998, “Near-Wall Turbulence Models for Complex Flows
Including Separation”, AIAA Journal,Vol.26, 641
CHEN, T. AND CHWANG, T., 2001, “Trailing Vortices in a Free Surface Flow”, Physics of
Fluids, Vol.14, No. 2, 827
CHOW, S.K., 1967, “Free Surface Effects on Boundary Layer Separation on Vertical Struts”,
PhD Thesis, The University of Iowa, Iowa City, IA.
CONSTANTINESCU, G.S., CHAPELET, M.C., & SQUIRES, K.D., 2001, “Prediction of
Turbulent Flow Over a Sphere”, AIAA Journal, Vol. 26, 155
DURBIN, P.A., 1991, “Near-Wall Turbulence Closure without Damping Functions”,
Theoretical and Computational Fluid Dynamics, Vol. 3(1), 1.
HO, C.M., & HUERRE, P., 1986, “Perturbed Free Shear Layers”, Annual Rev. Fluid. Mech.
16, Vol. 16, 365
HUNT, J. C. R., WRAY, A. A., & MOIN, P., 1988, “Eddies, Stream, and Convergence Zones
in Turbulent Flows”, Report CTR-S88, Center for Turbulence Research.
JOHNSON, D.A., MENTER, F.R., & RUMSEY, C.L., 1994, “The Status of Turbulence
Modeling for Aerodynamics”, AIAA Paper No. 1994-2226.
KANDASAMY, M., 2001, “RANS Simulation of Free Surface Wave Induced Separation
Around a Surface Piercing NACA-0024 Hydrofoil”, M.S. Thesis, The University of Iowa,
Iowa City, IA.
KANDASAMY, M., 2005, “URANS for Unsteady Free Surface Wave Induced Boundary
Layer Separation”, Ph.D. Thesis, The University of Iowa, Iowa City, IA.
KAWAMURA, T., MAYER, S., GARAPON, A. & SRENSEN, L., 2002, “Large Eddy
Simulation of a Flow Past a Free Surface Piercing Circular Cylinder”, AIAA Journal, Vol. 33,
No. 8, 1398.
KHORAMMI, M.R., SINGER, B.A., & BERKMAN, M.E., 2001, “Time-Accurate Simulations
and Acoustic Analysis of Slat Free-Shear Layer”, 7th AIAA/CEAS Aeroacoustics Conference,
AIAA Paper Number 2001-2155.
KHORAMMI, M.R., SINGER, B.A., & LOCKARD, D., 2002, “Time-Accurate Simulations
and Acoustic Analysis of Slat Free-Shear Layer”, 8th AIAA/CEAS Aeroacoustics Conference,
AIAA Paper Number 2002-2579.
KIM, J., PATERSON, E., & STERN, F., “Verification & Validation and Sub-Visual and
Acoustic Modeling for Ducted Marine Propulsor,” ASME J. Fluids Eng, in press.
KIYA, M., & SASAKI, K., 1985, “Structure of Large Scale Vortices and Unsteady Reverse
Flow in the Reattaching Zone of a Turbulent Separation Bubble”, Journal of Fluid Mechanics,
Vol. 154, 463
LARSSON, L., STERN, F., & BERTRAM, V., “Benchmarking of Computational Fluid
Dynamics for Ship Flows: The Gothenburg 2000 Workshop,” Journal Ship Research, Vol. 47,
No. 1, March 2003, 63
LIN, P., & LI, C.W., 2003, “Wave-Current Interaction with a Vertical Square Cylinder”, Ocean
Engineering, Vol. 30, 855
MENTER, F.R., 1993, “Zonal Two-Equation K-Omega Turbulence Model for Aerodynamic
Flows”, AIAA Paper 1993-2906.

26
MENTER, F.R., KUNTZ, M., Ansys-CFX, 2002, “Adaptation of Eddy-Viscosity Turbulence
Models to Unsteady Separated Flow behind Vehicles”, The Aerodynamics of Heavy Vehicles:
Trucks, Busses and Trains, Asilomar, Ca, Published By Springer
METCALF, B., 2001, “Experimental Investigations on Free Surface Wave-Induced
Separation”, M.S. Thesis, The University of Iowa, Iowa City, IA.
METCALF, B., LONGO, J., GHOSH, S., & STERN,F., 2006, “Unsteady Free-Surface Wave-
Induced Boundary-Layer Separation for a Surface-Piercing NACA 0024 Foil: Towing Tank
Experiments”, J. Fluids and Stuctures, Vol 22
PATERSON, E.G., WILSON, R.V., & STERN, F., 2003, “General-Purpose Parallel Unsteady
RANS Ship Hydrodynamics Code: CFDSHIP_IOWA”, IIHR Report No. 432
PAULEY, L.P., MOIN, P., & REYNOLDS, W.C., 1990, “The Structure of Two Dimensional
Separation”, J. Fluid Mech., Vol. 220, 397
POUQUET, A., FRISCH, U., & CHOLLET, J.P., 1983, “Turbulence with Spectral Gap”,
Physics of Fluids, Vol 24 (6), 877
POGOZELSKI, E.M., KATZ, J., & HUANG, T.T., 1997, “The Flow Structure around a
Surface Piercing Strut”, Physics of Fluids, Vol.5, 1387.
SIGURDSEN, L.W., 1995, “The Structure and Control of Turbulent Reattaching Flow”,
Journal of Fluid Mechanics, Vol. 248, 267
SIMONSEN, C. & STERN, F., “Flow Structure around an Appended Tanker Hull Form in
Simple Maneuvering Conditions,” Computers & Fluids, 34, 2005, 169
SOTIROPOULOS, F., & ABDALLAH, S., 1992, “A Primitive Variable Method for the
Solution of Three Dimensional Incompressible Viscous Flows”, J. Comp. Phy., Vol. 103, 336.
SPALART, P.R., & ALLMARAS, S.R., 1994, “A One-Equation Turbulence Model for
Aerodynamic Flows”, La Rech. Aerospatiale, Vol. 1, 5.
SPALART, P.R, JOU, W. H., STRELETS, M. & ALLMARAS, S.R., 1997, “Comments on the
Feasibility of LES for Wings, and on a Hybrid RANS/LES Approach”, 1st AFOSR Int. Conf.
on DNS/LES, Aug.4-8, 1997, Ruston, LA. In Advances in DNS/LES, C. Liu & Z. Liu Eds.,
Greyden Press, Colombus, OH.
STERN, F., HWANG, W.S., & JAW, S.Y., 1989, “Effects of Waves on the Boundary Layer of
a Surface-Piercing Flat Plate: Experiment and Theory,” Journal of Ship Research, Vol. 33, 63
STERN, F., CHOI, J.E., & HWANG, W.S., 1993, “Effects of Waves on the Wake of a
Surface-Piercing Flat Plate: Experiment and Theory,” Journal of Ship Research, Vol. 33, 102
WILCOX, D.C., 1988, “Reassessment of the Scale-Determining Equation for Advanced
Turbulence Models”, AIAA Journal, Vol. 26, 1299
WILSON, R.V., CARRICA, P.M., & STERN, F., 2006, “Unsteady RANS Method for Ship
Motions with Application to Roll for a Surface Combatant,” Computers & Fluids, Vol. 35,
Issue 5, 501
WILSON, R., SHAO, J., & STERN, F., 2004, “Discussion: Criticisms of the “correction
factor” Verification method [1]”, Tran. ASME, Vol. 126, 704
ZHANG, Z., & STERN, F., "Free Surface Wave-Induced Separation," ASME J. Fluids Eng.,
Vol. 118, September 1996, 546

27

View publication stats

You might also like