Professional Documents
Culture Documents
The Effect of Convex Wall Curvature On The Structure of The Turbulent Boundary Layer
The Effect of Convex Wall Curvature On The Structure of The Turbulent Boundary Layer
The manuscript was received on 2 January 2009 and was accepted after revision for publication on 9 April 2009.
DOI: 10.1243/09544062JMES1496
Abstract: Effects of convex wall curvature on turbulent boundary layer flow are studied in this
article using a numerical method. Since the non-linear k−ε model often used in engineering
applications cannot satisfy the distribution and wall-limiting behaviour of the Reynolds stress
components, an improved low Reynolds number k−ε turbulence model has been employed to
model turbulences in this study. Based on numerical solutions, turbulent intensity, turbulent
shear stress, and mean velocity are calculated. The results show that the turbulent intensities and
turbulent shear stresses are decreased on convex walls compared with flat plates under similar
conditions. The numerical results also show that for the boundary layer on convex surfaces, the
stabilizing effects lead to less turbulent momentum exchange between fluid particles. The rate
of integral parameters of the boundary layer such as momentum thickness and displacement
thickness is reduced on convex curvature compared to their values on the flat plate. To validate
the numerical method, the numerical results have been compared with previous measured values
and good agreement has been obtained.
Keywords: boundary layer equations, convex curvature, low Reynolds number turbulence
model, shear stress, turbulence intensities
1 INTRODUCTION study. Castro and Bradshaw [2], Muck et al. [4], and
Hoffmann et al. [5] have argued that at least a turbu-
Flow over curved surfaces in the plane of the mean lence closure involving the solution of modelled forms
shear stress has many engineering applications. Exam- of Reynolds stress equations is necessary to account
ples are boundary layers on aircraft wings and bodies, for the over proportional response of the turbulence
turbo machinery nozzle section, aerofoil wakes, and structure to streamline curvature. Irwin and Smith [6],
other free flows deflected by obstacles or pressure Gibson et al. [7], and Rodi and Scheuerer [8] have
forces. The extra rates of strain produced by curvature shown that models of this closure level produce the
are known to have an impact on turbulent structure. effects of a prolonged influence of curvature with very
Accordingly, even for a small ratio of shear layer thick- good quantitative agreement.
ness to curvature radius, significant changes in the As mentioned by So and Mellor [9], the stability of
flow parameters have been observed. The experiments a turbulent flow influenced by centrifugal forces was
by Smits et al. [1] and Castro and Bradshaw [2] showed first studied by Prandtel who extended the analogy
a non-monotonic recovery of the flow from curva- between buoyancy and curvature effect to turbulent
ture, which, on the other hand, was not observed boundary layer flow.
by Gillis and Johnston [3] in their boundary layer Measurement in fully developed turbulent flow in
curved channels by Eskinazi and Yeh [10] and Ellis
and Joubert [11] also indicated stabilization on the
convex side channel and destabilization on the con-
∗ Corresponding author: Faculty of Mechanical Engineering, cave side. However crossover effects near the channel
Shahrood University of Technology, Daneshgah Avenue, Shahrood, centreline prevent the effects from being isolated. In
Semnan, Iran. addition, longitudinal vortices were not observed in
email: hariri.saeed@gmail.com these experiments.
The response of a well-developed turbulent bound- poses a formidable challenge. In their work various
ary layer to suddenly applied convex surface curvature turbulence models, such as k−ε realizable, k−ω, and
was investigated by Muck et al. [4]. They used con- shear stress transport, were used.
ditional sampling techniques so that the turbulent The problem of steady laminar mixed convection
and non-turbulent regions of the flow could be clearly boundary layer flow of an incompressible viscous
distinguished. They concluded that mild convex cur- fluid along vertical moving thin needles with variable
vature, with a radius of curvature of the order of 100 heat flux for both assisting and opposing flow cases
times the boundary layer thickness, tends to atten- was theoretically considered by Ahmad et al. [18].
uate the pre-existing turbulence, apparently without They transformed the governing boundary layer equa-
producing large changes in the statistical average tions first into non-dimensional forms. The curvature
eddy shape. Also, curved shear layers may exhibit effects were incorporated into their analysis, whereas
near-equilibrium behaviour if the ratio of shear layer the pressure variation in the axial direction had been
thickness to streamline curvature is independent of neglected. These equations were then transformed
downstream distance. into similarity equations using the similarity variables,
A general review of curvature effects also was given which were solved numerically using an implicit finite-
by Gillis and Johnston [3]. Several other researchers difference scheme known as the Keller-box method.
(e.g. So and Mellor [9, 12], Ramaprian and Shivaprasad The solutions were obtained for a blunt-nosed needle.
[13]) have reported measurements of the Reynolds Numerical calculations were carried out for various
stresses in convex-curved boundary layers. Ramaprian values of the dimensionless parameters of the prob-
and Shivaprasad [13] measured triple products, but lem, which included the mixed convection parameter
used non-linearized constant current hot wires so that λ, the Prandtl number Pr, and the parameter rep-
the results are not reliable. Furthermore, at the time resenting the needle size. It was shown from the
that the work was started, the only experiment that numerical results that the skin friction coefficient,
included triple product measurements (so that all the the wall temperature, and the velocity and temper-
terms in the Reynolds-stress transport equation could ature profiles were significantly influenced by these
be deduced) was the work of Castro and Bradshaw parameters.
[2] on a stably curved mixing layer with a very short Watson and Otto [19] considered the stability of high
region of curvature. Gibson et al. [7] have measured Reynolds number flow past a heated curved wall. Their
velocity and temperature-fluctuation statistics up to results indicated that the influence of both buoyancy
third order in rig. and curvature, with the appropriate sense, can render
A significant contribution to the study of the effect a flow unstable to longitudinal vortices. However, con-
of curvature was made by Bradshaw [14, 15]. In these versely, each mechanism can make a flow more stable,
studies, he described an analogy between stream- as with a stable stratification or a convex curvature.
line curvature and buoyancy and proposed a relation This is partially due to their influence on the basic flow
between mixing length and parameters called the Cur- and also due to additional terms in the stability equa-
vature Richardson number. He also observed that the tions. They observed that the presence of buoyancy in
curvature parameter (ratio of shear layer thickness to combination with an appropriate local wall gradient
the radius of curvature, δ/R) 1/300 was enough to alter can actually increase the wall shear, and these effects
the length scale by as much as 10 per cent. can lead to super velocities and the promotion of a wall
It seemed appropriate to extend the work of jet. This led to the interesting discovery that the flow
Meroney and Bradshaw [16] to include triple-product can be unstable for both concave and convex curva-
measurement and conditional sampling in order to tures. Furthermore, it is possible to observe sustained
provide comprehensive documentation of a mildly vortex growth in stably stratified boundary layers over
curved boundary layer in which the stabilizing effects convexly curved walls. The evolution of the modes was
of curvature significantly attenuate the turbulence, considered in both the linear and non-linear regimes.
but are not strong enough to annihilate the shear stress For the prediction of the boundary layer flow near
in the outer part of the layer as found in some of the convex surface on rotating systems, Kikuyama
the strongly curved experiments. The measurements et al. [20] investigated the effects of system rotation,
were not aimed at the improvement of any particu- radii of the wall, and the axial Reynolds number on
lar calculation method, or even a class of calculation the boundary layer near the convex side, making use
methods, but their paper includes a discussion of the of the curved channel on the rotating table. From the
implications for turbulence modelling in general. measurement results of the velocity distribution, they
A numerical study of turbulent flow over the knew that the Reynolds number based on momentum
S-shaped hydrofoil at 0◦ angle of attack has been thickness is an appropriate parameter to predict the
reported by Kumar and Chatterjee [17]. There, the flow boundary layer transition.
takes place over concave and convex surfaces and is The process of laminar-turbulent boundary tran-
accompanied by the favourable and adverse pressure sition on a concave wall of 1 m in the radius
gradients and flow separation. Modelling such a flow of curvature was studied experimentally under a
free-stream velocity of 3 m/s by Yamada and Ito [21]. pictures with the light sheet oriented in a longitudi-
Flow visualization was achieved using smoke tech- nal plane as well as along two-cross-stream planes at
niques and flow measurement by an X -type probe. +45◦ and at −45◦ to the flow direction. Their experi-
Helen et al. [22] studied the effect of including ments were conducted under conditions of nearly zero
wall and streamline curvature terms in swept-wing pressure gradients along the wall. Their results were
boundary-layer stability calculations. They cast lin- also obtained in a flat-wall region following the curved
ear disturbance equations on a fixed, body-intrinsic, wall, as the flow began to recover from the effects of
curvilinear coordinate system. They observed that streamline curvature. Both sets of results were com-
convex-wall curvature has a stabilizing effect, while pared with those obtained from a fully developed
streamline curvature is destabilizing if the curvature flat-wall boundary layer. Their study clearly showed
exceeds a critical value. that convex curvature strongly attenuates most of the
Brighi and Hoernel [23] investigated the simila- organized large-scale structures in the flow. These
rity solutions of a plane mixed convection boundary structures are not restored to their earlier state quickly
layer flow near a semi-vertical plate, with a pre- after the removal of streamline curvature. Their visua-
scribed power law function of the distance from the lization results supported those obtained from earlier
leading edge for the temperature that was embed- measurements of spectra, intermittency, and length
ded in a porous medium. They showed the existence scales.
and uniqueness of convex and concave solutions for An experimental study of developing turbulent flow
positive values of the power law exponent. through a rectangular-sectioned bend of decreasing
The results of an experimental study of developing cross-sectional area was reported by Launder and
turbulent boundary layers on the concave and convex Loizou [27]. In their investigation over the first 90◦ of
walls of a 90◦ bend of square cross-section using hot- arc, the duct aspect ratio changes from 1:1 to 3:1 due to
wire anemometry were presented by Mokhtarzadeh- the decrease in width of the plane end walls (the width
Dehghan and Yuan [24]. Results are presented for of the curved wall remaining uniform). They selected
the conventional mean turbulence quantities and two duct Reynolds numbers for study, 8.2 × 104 and
the period of bursting. They found that turbulence 6 × 104 , which gave rise to mean values of the acceler-
enhancement on the concave wall and the opposite ation parameter K̄ of 1.5 × 10−6 and 2.1 × 10−6 . They
effect on the convex wall are most pronounced in measured streamwise mean velocity profiles and tur-
the outer layer, and the effect on the normal inten- bulence intensities at eight stations through the bend
sity is greater than that on the streamwise intensity. with particular emphasis on the near wall velocity pro-
But on the convex wall, they observed that suppres- files on the plane of symmetry of the duct. Their results
sion of turbulence shear stress to insignificant levels showed a marked thickening of the viscous sublayer
occurs for normal distances from the wall greater and damping of the turbulence intensity due to accel-
than y/δ ≈ 0.4. Their results of conditional sampling eration, with the effect being particularly strong on the
showed the opposite effects of the concave and convex convex wall where the streamline curvature also acts to
walls on the bursting frequency, and that none of the inhibit turbulence. They confirmed that at the higher
scaling laws results in the collapse of this parameter. level of acceleration, the boundary layer on this sur-
Receptivity of a shear layer separating at a con- face reverts essentially to laminar, with the turbulence
vex corner in the wall was examined experimen- level being reduced practically to background levels. In
tally by Asai and Kaneko [25]. Acoustic forcing was contrast, on the concave wall it appears that a greater
applied to excite instability waves. In their study par- degree of laminarization occurs at the lower value
ticular attention was focused on the comparison of of acceleration. They suggested that this unexpected
receptivity coefficient between a sharp-edged corner result may be due to differences in the secondary flow
and a rounded corner whose radius of curvature is pattern between the two test cases.
20 times the boundary-layer displacement thickness. The numerical study was conducted by Kim and
Their results clearly showed that the receptivity coef- Patel [28–30] to study the fate of artificially introduced
ficient is quite sensitive to the sharpness of the corner vortex pairs in an otherwise two-dimensional (2D)
and that the amplitude of the excited instability wave turbulent boundary layer, with emphasis on the influ-
is twice larger for the sharp corner compared with ence of streamwise surface curvature. In their study
the rounded corner. Also they found that the same Reynolds-averaged Navier–Stokes equations for three-
dependency of the receptivity coefficient on the cor- dimensional turbulent flows, together with a two-layer
ner sharpness is verified by analysing the shear layer turbulence model to resolve the near-wall flow, were
response to wind-tunnel background noise without solved by a numerical method for a configuration
acoustic forcing. in which some measurements were available from
The large-scale structure of the turbulent boundary previous experiments. Longitudinal vortex pairs, with
layer along a convex wall was studied by Chiwanga and common flow between them down or up, were intro-
Ramaprian [26] by using smoke and a laser light sheet duced in a flat-plate boundary layer, and in a boundary
for flow visualization. They obtained flow visualization layer subsequently subjected to concave or convex
wall curvatures to elucidate the effects of streamwise In the above equations, up is the potential velocity and
surface curvature and attendant pressure gradients in ue (x, y) is the external velocity of the flow as it varies
a curved rectangular duct. They found that circula- on the x, y direction on curved surfaces. kr (x) and h1
tion around the primary vortex, introduced in the flow, are curvature parameters, which are defined as
decreases on flat and convex surfaces but increases on
1
concave surface. Also their results showed that surface kr (x) = + (6)
curvature has a rather dramatic effect not only on the R(x)
development of the primary vortices, but also on the h1 = 1 + k r y
growth of secondary vortices.
R(x) is the radius of the curved surface.
The present work expands considerably upon those
The turbulence term in equation (1) can be defined
papers, providing details about numerical method for
as [30]
solving turbulent boundary layer equations. Based on
the numerical method, the effects of curved walls ∂u kr u
−u v = νt
− (7)
especially convex curvature on the turbulent bound- ∂y h1
ary layer are studied. The turbulent computations
have been carried out by an improved form of k−ε where νt is the eddy viscosity and will be calculated by
low Reynolds turbulence model. Also the turbulence the turbulence model, which will be described in the
intensities are calculated by employing a simplified next section.
form of the Launder eddy viscosity. To validate the
numerical method, a comparison has been carried out 2.2 Turbulent model
between numerical results and previously measured
values. The turbulence model employed in this article is an
improved form of low Reynolds k−ε turbulence model.
The model could be applied for whole regions in the
2 NUMERICAL METHOD boundary layer both the inner layer (linear sublayer,
buffer layer, log law region) and the outer layer. The
2.1 Boundary layer equations model is defined as [30, 33]
The 2D, steady-state, incompressible turbulence u ∂k ∂k 1 ∂ ∂k ∂u kr u 2
boundary layer flow equations with coordinate system +v = νt h1 + νt −
h1 ∂x ∂y h1 ∂y ∂y ∂y h1
(x, y) and velocity components u, v, where the x-axis 1/2 2
is taken along the streamwise direction and the y-axis ∂k
− ε − 2ν (8)
in the normal direction, can be written as [31, 32] ∂y
u ∂ε ∂ε 1 ∂ νt ∂ε
x-momentum equation +v = h1
h1 ∂x ∂y h1 ∂y σe ∂y
u ∂u ∂u kr 1 ∂p
+v + uv = − ε2 ε ∂u kr u 2
h1 ∂x ∂y h1 h1 ρ ∂x − f2 cε2 + cε1 νt −
2 k k ∂y h1
∂ u kr ∂u kr2 u ∂u v u v 2 2
+ν + − − − 2k (1) ∂ u
∂ 2y h1 ∂y h12 ∂y
r
h1 + 2ννt (9)
∂y 2
In this article, it is assumed ∂p/∂x ≈ 0. cμ = 0.09, cε1 = 1.44, cε2 = 1.92, σk = 1.0, σε = 1.3
y-momentum equation −3.4 k2
fμ = exp , R t = (10)
(1 + Rt /50)2 νε
kr u2 1 ∂p
= (2) f2 = 1 − 0.3 exp(−Rt2 ) (11)
h1 ρ ∂y
where k is the kinetic energy and ε is the rate of
Continuity equation turbulence dissipation.
∂u ∂ Boundary conditions are
+ (vh1 ) = 0 (3)
∂x ∂y y=0 k=0 ε=0
Fig. 5 Variation of mean velocity by the effect of convex curvature (uref = 12 m/s, δ/R = 0.023)
Fig. 6 Variation of mean velocity by the effect of convex curvature (uref = 33 m/s, δ/R = 0.01)
Fig. 7 Variation of turbulence shear stress by the effect of convex curvature (uref = 12 m/s,
δ/R = 0.023)
Fig. 8 Variation of turbulence shear stress by the effect of convex curvature (uref = 33 m/s,
δ/R = 0.01)
compared with flat plate cases. This means the tan- stress increases downstream, whereas the experimen-
gent of angles, which are created between (Y , U ), is tal data show a slow decrease. It can be seen that the
greater on the convex wall than the flat plate under convex curvature prevents the turbulence shear stress
the same conditions. and turbulent intensity growing. Also the conventional
Reynolds shear stress is shown in Figs 7 and 8. average of Reynolds stress profiles indicates that the
The numerical value of turbulent shear stress on response of turbulence to suddenly applied stabiliz-
the convex surface is compared with corresponding ing streamline curvature is rather rapid, the decrease
numerical values on the flat plate. In all cases the in turbulence intensity implying a preceding increase
shear stress decreases rapidly after the start of the in the turbulence energy dissipation rate, followed by
convex curvature, and thereafter the predicted shear a decrease in dissipation as the flow settles down at the
Fig. 9 Variation of turbulence intensity by the effect of convex curvature (uref = 12 m/s,
δ/R = 0.023)
Fig. 10 Variation of turbulence intensity by the effect of convex curvature (uref = 33 m/s,
δ/R = 0.01)
new, lower, level of intensity. The different behaviour which show the variation of turbulence intensities
of turbulent intensity and of dissipation is not para- and turbulent shear stresses, three regions can be
doxical because the dissipation rate equals the rate of observed. In the first region the turbulent parameters
energy transfer from larger eddies. Then it can be said are increased in the normal direction (y or η). In the
that a reduction in intensity can still be associated with second, a constant region appears, whereas the thick-
an increase in ε. ness of this region for the flat plate is larger than convex
Variations of turbulent intensities are shown in Figs 9 wall. In the third region, turbulent intensity and shear
to 14. Due to the effect of convex curvature the nor- stress are decreased and at the end of this region all
mal turbulent terms are lower when they are compared turbulence intensities take a value equal to the free
with the same condition on the flat plate. In all figures, stream turbulence intensity. The numerical values of
Fig. 11 Variation of turbulence intensity by the effect of convex curvature (uref = 12 m/s,
δ/R = 0.023)
Fig. 12 Variation of turbulence intensity by the effect of convex curvature (uref = 33 m/s,
δ/R = 0.01)
v 2 and w 2 differ noticeably from those of Gibson et al. with distance but the rate of growth on the flat plate
[7]. Their results agree closely with the classical mea- is slightly faster than the convex wall. As expected,
surements of Kelebanoff [39], whereas current results the turbulence in shear layers is highly sensitive to
show smaller difference between v 2 and w 2 , which streamline curvature and the turbulent shear stress
are found in most of the experimental and numerical and intensity are reduced by curvature when the angu-
researches (Figs 11 to 14). lar momentum of the flow increases in the direction of
The variation of momentum thickness and displace- the radius of curvature, and are increased when the
ment thicknesses is shown in Figs 15 and 16. Note angular momentum decreases with radius.
from these figures, by comparing flows on the flat plate As it can be seen from Figs 7 to 14, generally good
and on convex surfaces, that both δ ∗ and θ increase agreement has been obtained between numerical
Fig. 13 Variation of turbulence intensity by the effect of convex curvature (uref = 12 m/s,
δ/R = 0.023)
Fig. 14 Variation of turbulence intensity by the effect of convex curvature (uref = 33 m/s,
δ/R = 0.01)
Fig. 15 Variation of momentum thickness by the effect of convex curvature (uref = 12 m/s,
δ/R = 0.023)
Fig. 16 Variation of displacement thickness by the effect of convex curvature (uref = 12 m/s,
δ/R = 0.023)
results and experimental data. This suggests that the The improved version of low Reynolds number k−ε
turbulence model, which is employed in this study, can turbulent model proposed in this study has been
predict the turbulent flow over curved surface accu- employed to model turbulence. The numerical results
rately. It is also worth mentioning that although low have been compared with available measured values.
Reynolds k−ε has been employed in this study, the The results show that convex curvature basically
model has been modified by authors. attenuates the existing turbulence structure, perhaps
producing significant preferential changes in struc-
ture parameters like the stress intensity ratio. The
5 CONCLUSIONS effects of stabilizing and destabilizing curvature on
the turbulent shear layer are essentially different
In this article, a numerical method has been developed phenomena. The most obvious evidence is the very
to study turbulent boundary layer on curved surfaces. rapid response of a boundary layer to the application
The boundary layer equations in a curved surface have or removal of stabilizing convex curvature as com-
been discretized using the finite-difference method. pared with the slow reaction to concave (destabilizing)
curvature and also changes in turbulent structure 13 Ramaprian, B. R. and Shivaprasad, B. G. The struc-
parameters on convex surfaces; though interesting in ture of turbulent boundary layer along mildly curved
detail and important enough to need special atten- surfaces. J. Fluid Mech., 1978, 85, 273–303.
tion in turbulence models are generally not very 14 Bradshaw, P. The analogy between streamline curvature
and buoyancy in turbulent shear flow. J. Fluid Mech.,
large.
1969, 36, 177–191.
Regarding the turbulence model used in this arti-
15 Bradshaw, P. Effects of streamline curvature on turbulent
cle, it can be concluded that because the wall-normal flow. AGARD-AG, No. 169, 1973.
Reynolds stress component is important for precise 16 Meroney, R. N. and Bradshaw, P. Turbulent boundary
predictions and the non-linear k−ε model often used layer growth over a longitudinally curved surface. AIAA
in engineering applications cannot satisfy the distri- J., 1974, 13(11), 1448–1453.
bution and wall-limiting behaviour of the wall-normal 17 Kumar, T. M. P. and Chatterjee, D. Numerical study of
Reynolds stress component. Thus, an improved ver- turbulent flow over an S-shaped hydrofoil. Proc. IMechE,
sion of the low-Reynolds-number turbulence model Part C: J. Mechanical Engineering Science, 2008, 222(C9),
was proposed by the curvature parameter to be used 1717–1734. DOI: 10.1243/09544062JMES929.
on curved surfaces so as to reproduce exactly the 18 Ahmad, S., Arifin, N. M., Nazar, R., and Pop, I. Mixed
convection boundary layer flow along vertical moving
wall-limiting behaviour of turbulence quantities.
thin needles with variable heat flux, heat and mass
The numerical results, which have been compared
transfer, 2007 (Springer-Verlag, Berlin, Heidelberg). DOI:
with measured value, show good agreement. This 10.1007/s00231-007-0263-6.
validates the numerical method and especially the 19 Watson, C. E. and Otto, S. R. The stability of a curved,
proposed turbulence model. heated boundary layer: linear and nonlinear problems.
Anziam J.: The Australian & New Zealand Indust. Appl.
© Authors 2009 Math. J., 2005, 46(4), 507–544.
20 Kikuyama, K., Hasegawa, Y., Matsumoto, N., and
Nishikawa, M. Experimental study on boundary layer
REFERENCES stability over a convex surface in rotating system. Nihon
Kikai Gakkai Nenji Taikai Koen Ronbunshu J., Journal
1 Smits, A. J., Young, S. T. B., and Bradshaw, P. The effect code: X0587B, 2001, 2, 209–210.
of short regions of high surface curvature on turbulent 21 Yamada, K. and Ito, A. Process of boundary layer tran-
boundary layers. J. Fluid Mech., 1979, 94, 209–242. sition on a concave wall. JAXA Spec. Publ. J., 2006, 5(12),
2 Castro, I. P. and Bradshaw, P. The turbulence structure of 29–30.
a highly curved mixing layer. J. Fluid Mech., 1976, 73, 265. 22 Helen, L. R., Ray, S. L., and Media, M. Streamline-
3 Gillis, J. C. and Johnston, J. P. Experiments on turbulent curvature effect in three-dimensional boundary
boundary layer over convex wall and its recovery to flat layers. SAE Tech. Papers, document number: 921987,
wall condition. In Turbulent shear flows 2 (Eds L. J. S. 1992.
Bradbury et al.), 1980, pp. 116–128 (Springer, New York). 23 Brighi, B. and Hoernel, J. D. On the concave and
4 Muck, K. C., Hoffmann, P. H., and Bradshaw, P. The convex solutions of a mixed convection boundary layer
effect of convex surface curvature on turbulent boundary approximation in a porous medium. Appl. Math. Lett.,
layers. J. Fluid Mech., 1985, 161, 347–369. 2006, 19(1), 69–74.
5 Hoffmann, P. H., Muck, K. C., and Bradshaw, P. The 24 Mokhtarzadeh-Dehghan, M. R. and Yuan, Y. M. Mea-
effect of concave surface curvature on turbulent surements of turbulence quantities and bursting period
boundary layers. J. Fluid Mech., 1985, 161, 371–403. in developing turbulent boundary layers on the concave
6 Irwin, H. P. A. H. and Smith, P. A. Prediction of the effect and convex walls of a 90◦ square bend. Exp. Therm. Fluid
of streamline curvature on turbulence. Phys. Fluid, 1975, Sci., 2002, 27(1), 59–75.
18(6), 624–630. 25 Asai, M. and Kaneko, M. Experimental investigation of
7 Gibson, M. M., Jones, W. P., and Younis, B. A. Calculation receptivity of a shear layer separating at a convex corner.
of turbulent boundary layer on curved surfaces. J. Phys. Fluid Dyn. Res., 2000, 27(2), 117–128.
Fluids, 1981, 24(3), 386–395. 26 Chiwanga, S. C. and Ramaprian, B. R. The effect of
8 Rodi, W. and Scheuerer, G. Calculation of curved shear convex wall curvature on the large-scale structure of the
layers with two equation turbulence models. Phys. turbulent boundary layer. Exp. Therm. Fluid Sci., 1993,
Fluids, 1983, 26(6), 1422–1436. 6(2), 68–176.
9 So, R. M. C. and Mellor, G. L. Experiments on turbulent 27 Launder, B. E. and Loizou, P. A. Laminarization of
boundary layer on a concave wall. Aero. Quart., 1975, 26, three-dimensional accelerating boundary layers in a
25–40. curved rectangular-sectioned duct. Int. J. Heat Fluid
10 Eskinazi, S. and Yeh, H. An investigation of fully devel- Flow, 1992, 13(2), 124–131.
oped turbulent flow in curved channel. J. Aerosp. Sci., 28 Kim, W. J. and Patel, V. C. Influence of streamwise
1956, 23, 23. curvature on longitudinal vortices imbedded in tur-
11 Ellis, L. B. and Joubert, P. N. Turbulent shear flow in a bulent boundary layers. Comput. Fluids, 1994, 23(5),
curved duct. J. Fluid Mech., 1974, 62, 65–84. 647–673.
12 So, R. M. C. and Mellor, G. L. Experiments on convex 29 Patel, V. C. Effects of curvature on turbulent boundary
curvature in turbulent boundary layers. J. Fluid Mech., layer, 1969, R. M. No. 3599 (Ministry of Technology,
1973, 60, 43–62. London).
− f2 cε2
δ boundary layer thickness 2k i+(1/2),j
δ∗ displacement thickness
Di+(1/2),j
= Ai+(1/2),j
εi,j+1 + B̄i+(1/2),j
εi,j + Ci+(1/2),j εi,j−1
ε turbulent dissipation rate
η dimensionless parameter (η = y/δ) + Li+(1/2),j (Ni+(1/2),j )2 + G1i+(1/2),j (Mi+(1/2),j )2
θ momentum thickness
νt kinematic eddy viscosity − G2i+(1/2),j (Ni+(1/2),j ) + G3i+(1/2),j