Experiments in Fluids 37 (2004) 281–292

DOI 10.1007/s00348-004-0816-0

Roughness effects on turbulent plane wall jets

in an open channel
M. F. Tachie, R. Balachandar, D. J. Bergstrom

281
Abstract This paper reports the effects of surface rough- interaction between wall-bounded shear flow and a free
ness on the mean flow characteristics for a turbulent plane shear flow. The turbulent wall jet has also been routinely
wall jet created in an open channel. The velocity mea- used to model and conduct fundamental research on
surements were obtained using a laser Doppler anemom- diverse fluid processes of technological and environmental
eter over smooth and transitionally rough surfaces. The importance; for example, separation control on aircraft
power law proposed by George et al. (2000) was used to wings, film cooling in gas turbines, and transport and
determine the friction velocity. Both conventional scaling mixing of effluents in lakes and rivers. In these and many
and the momentum–viscosity scaling proposed by Nar- other applications, knowledge of skin friction character-
asimha et al. (1973) were used to analyze the streamwise istics is essential for flow calculation, and also for pre-
evolution of the flow. The results show that surface dicting mass and convective heat transfer rates. Knowledge
roughness increases the skin friction coefficient and the of spread of the jet is also of interest because it is related to
inner layer thickness, but the jet half-width is nearly mixing and dilution phenomena in the flow.
independent of surface roughness. In this study, we are interested in the characteristics of a
turbulent plane wall jet, that is, a jet issuing from a
two-dimensional rectangular slot into a still surrounding.
1 Such a flow configuration and its nomenclature are shown in
Introduction Fig. 1. Here, we adopt Cartesian coordinate system and
A turbulent wall jet is a shear flow directed along a wall, denote the streamwise (or main flow) and wall-normal
where by virtue of the initially supplied momentum, at any directions by x and y, respectively; bis the slot height,
downstream station, the streamwise velocity over some U denotes the streamwise component of the mean veloc-
region within the flow exceeds that in the external stream ity,Uj is the jet exit velocity; Um is the local maximum
(Launder and Rodi 1981). The flow is confined by a solid velocity; ym and y1/2, respectively, denote the wall-normal
wall on one side, while on the other unbounded side, locations where Um and 0.5Um occur. The region y £ ym and
mixing takes place with the surroundings. The flow field y>ym represent, respectively, the inner and outer layers of
close to the wall has characteristics similar to those of a the turbulent wall jet. As noted earlier, the flow character-
boundary layer and the outer region of flow is structurally istics in the inner region are similar to those of a turbulent
similar to a free jet. Because of its unique features, the boundary layer while the outer layer, though somewhat
turbulent wall jet is a prototypical flow for studying the influenced by the wall, is structurally similar to a free jet.
Because of its diverse technological applications, the
turbulent wall jet has been the focus of numerous prior
Received: 30 November 2003 / Accepted: 10 March 2004 studies. The literature existing prior to 1983 was critically
Published online: 15 April 2004

Springer-Verlag 2004 reviewed by Launder and Rodi (1981, 1983). The velocity
measurements summarized in these review articles were
M. F. Tachie (&) obtained using hot wires and a Pitot tube on a hydrauli-
Department of Mechanical and Industrial Engineering, cally smooth surface. The review articles revealed that an
University of Manitoba, 15 Gillson Street, accurate determination of the skin friction characteristics
Winnipeg, MB R3T 5V6, Canada in a turbulent wall jet remained a major challenge. Directly
E-mail: tachiemf@cc.umanitoba.ca
related to this are questions regarding the universality of
R. Balachandar the inner region of the mean velocity profile and the
Department of Civil and Environmental Engineering, proper scaling law for the overlap region. A number of
University of Windsor,
401 Sunset Ave., Windsor, N9B 3P4, Canada refined velocity measurements have been reported sub-
sequent to these reviews (Wygnanski et al. 1992; Karlsson
D. J. Bergstrom et al. 1993; Abrahamsson et al. 1994). These studies pro-
Department of Mechanical Engineering,
University of Saskatchewan, vided accurate and comprehensive datasets, especially in
57 Campus Drive, Saskatoon, the near-wall region, and also addressed some of the
SK S7N 5A9, Canada scaling issues related to the mean flow.
The authors gratefully acknowledge financial support of the While the velocity measurements reported to date have
Natural Sciences and Engineering Research Council of Canada greatly advanced our understanding of the characteristics
(NSERC). of turbulent wall jet, the vast majority of these measure-
 turbulent boundary layer and a turbulent wall jet is a
power law with coefficients that depend on Reynolds
number. The present authors (Bergstrom et al. 2001)
successfully applied the power law to model the overlap
region of turbulent boundary layers over smooth and
rough surfaces. The power-law formulation was also
applied to model the overlap region of a turbulent wall jet
over a smooth surface (George et al. 2000; Tachie et al.
2002). In inner coordinates, the power law proposed by
George et al. (2000) for the turbulent wall jet is of the form:
282 Fig. 1. Definition sketch of a turbulent wall jet c
U þ ¼ Ci ðyþ þ aþ Þ ð2Þ
ments were obtained on a smooth surface and so our
knowledge of turbulent wall jet on a rough surface is where, Ci and c are power-law constants and the parameter
deficient. The Pitot-tube measurements by Rajaratnam a+ (=)16) represents a shift in origin for measuring y,
(1965) and Sakipov et al. (1975) appear to be the only associated with the growth of the meso-layer region
rough-wall datasets available in the open literature. (30 < y+ < 300). In contrast to the logarithmic law con-
Because many fluid engineering systems involving wall-jet stants (jand B), the power-law constants depend on the
flows have rough surfaces, additional research is necessary local Reynolds number, yþ 1=2 (=y1/2Us/m).
to advance knowledge of roughness effects on momentum The velocity distribution in the vicinity of a smooth
transport and heat transfer characteristics of turbulent wall is of considerable interest in determining Us and also
wall jet. More specifically, knowledge of roughness effects for constructing a composite velocity profile. Using a
on the skin friction characteristics, decay of maximum Taylor series expansion together with the continuity
velocity, and spread rate for the wall jet will lead to better equation and no-slip condition at the wall, the mean
prediction of momentum as well as mass and convective velocity can be expressed by the following relation:
heat transfer rates. Thus, the focus of the present work is U þ ¼ yþ þ c4 yþ4 þ c5 yþ5 þ ::: ð3Þ
to document the effects of surface roughness on the skin
friction characteristics and the spread rate for a turbulent where the coefficients c4 and c5 may vary slightly with
plane wall jet. Reynolds number as determined by George and Castillo
(1997). Recent laser Doppler anemometer (LDA) mea-
2 surements by Eriksson et al. (1998) suggested c4=)0.0003
Review of scaling laws (±0.0001), and George and Castillo (1997) recommended a
value of c5=13.5·10)6. Neither the very near-wall profile
2.1 (Eq. 3) nor the overlap profile (Eq. 2) is valid in the region
The overlap region of the mean velocity profile of y+=15. Following the earlier work by George and Castillo
In near-wall turbulence research, the scaling law for the (1997), George et al. (2000) applied the following inner
overlap region of the mean velocity profile has received profile to describe the viscous region, the buffer region, the
considerable attention because it leads to the determina- meso-layer region and the overlap region of a smooth wall:
tion of the friction velocity Us. Because of the presumed
similarity between the inner layers of the wall jet and U þ ¼ ½yþ þ c4 yþ4 þ c5 yþ5
exp ðdyþ6 Þ


boundary layer, the classical logarithmic law is routinely þCi yþc 1 þ caþ yþ1 þ 12 cðc  1Þaþ2 yþ2 ð4Þ
used to model the overlap region of the mean flow: ð1  exp ðdyþ6 ÞÞ
1 In Eq. 4, d (=8·10)8) is a damping parameter that fixes
U þ ¼ ln yþ þ B ð1Þ
j the transition from the viscous wall region to the overlap
+
In Eq. 1, U =U/Us, y =yUs/m, Us is the friction velocity region at y =15. The exponential function in Eq. 4 enables
+ +

and m is the kinematic viscosity of the fluid. The loga- us to smoothly transition from the wall to the overlap
rithmic law constantsj (=0.41) and B (=5.0) are assumed region.
to be universal. However, review of the logarithmic law The skin friction relation for the wall jet is given by:
constants obtained in previous wall-jet experiments Us Co  þ c
showed that only a few investigators (Karlsson et al. 1993; U ¼ C Dy1=2 ð5Þ
m i
Abrahamsson et al. 1994) obtained values that are con-
sistent with the universal values adopted in boundary layer where Co and D are parameters chosen as 1.3 and 1.0,
analysis. For example, Kruka and Eskinazi (1964) obtained respectively, based on refined LDA data by Karlsson et al.
j=0.48 and B=11.4. The data reported by Wygnanski et al. (1993). Additional constraints for the power-law parame-
(1992) provided support for a universal j, but the additive ter were obtained by George and Castillo (1997) using the
constant B varied from 5.5 to 9.5 depending on the exit near-asymptotic method, and the modified constraints in
Reynolds number. terms of wall-jet variables are given by:
George and Castillo (1997) and George et al. (2000) aA
applied the asymptotic invariance principle to show that c ¼ c1 þ ð6Þ
ðln Dyþ Þ1þa
the proper functions describing the overlap region of a 1=2
 2 3  
Um m y1=2 Mo c
Co Co1 6 ð1 þ aÞA 7 ¼C ð12Þ
¼ exp 4 a 5 ð7Þ Mo m2
Ci Ci1 ln Dyþ 1=2
where Cand c are constants that may also depend on initial
conditions. Using prior LDA data, the following values
Based on the work of George and Castillo (1997) and
were recommended: C=1.85 and c=)0.528. Following
LDA data of Karlsson et al. (1993), the following values
George et al. (2000), Eqs. 10, 11 and 12 will be referred to
were recommended: a=0.46, A=2.90, c¥=0.0362,
as momentum–viscosity scaling in this study, while Eqs. 8
Co¥/Ci¥=0.023.
and 9 will be referred to as conventional scaling laws. It
should be noted that the A and c used in Eqs. 10 and 12,
2.2 respectively, are not the same as those used in the power- 283
Streamwise development of the mean flow law formulation for the overlap region (Eqs. 2, 4, 5, 6 and 7
The streamwise evolution of the mean flow has tradition- in Sect. 2.1).
ally been scaled using the slot height (b) and the exit
velocity (Uj). According to the review article by Launder 3
and Rodi (1981), the spread rate for y1/2 and the decay of Experimental set-up and measurement procedure
Um are, respectively, given by the following relations: The experiments were performed in an open channel
flume. Fig. 2 shows an overview of the experimental set-
dy1=2 up. The flume was 10 m long, 0.8 m wide and 0.6 m deep.
¼ 0:073  0:002 ð8Þ The sidewalls of the flume were made of transparent
dx
tempered glass to facilitate optical access. The thickness t
 2 hxi
Um and height b of the slot were 6 mm and 10 mm, respec-
¼ K1 þ K2 ð9Þ tively. The side walls of the wall-jet facility were 6 mm, and
Uj b
the ratio of the slot width w to the slot height was
where K1 and K2 are presumed to be constants. Whenever w/b=78.8. The nozzle had a contraction ratio of 9 to 1 and
the above scaling laws (Eqs. 8 and 9) are applied, the was designed to avoid separation. The wall-jet facility was
velocity decay and spread rates showed a strong depen- fitted on to the bottom of an open channel flume. The inlet
dence on Reynolds number and initial conditions. For of the nozzle was placed 3 m downstream of the channel
example, the review article by Launder and Rodi (1981) contraction, and the slot exit was preceded by straw
showed that the spread rate varied from 0.056 to 0.085. packing to reduce any large-scale turbulence in the
Recent measurements (Wygnanski et al. 1992; Abra- approaching flow. The rough surface was created using
hamsson et al. 1994) also reported spread rates that indi- sand grains of 1.2 mm nominal diameter. The sand grains
cate a distinct dependence on the exit Reynolds number. were attached on to the surface using a double-sided tape.
Recent studies on turbulent boundary layers (Castillo and The sand grains were distributed to achieve the most dense
Johansson 2002) also showed a clear dependence of the arrangement. The height of the water downstream the exit
outer flow on the initial conditions. They further showed was approximately 350–400 mm above the floor of the test
that the Reynolds number dependence observed in the facility.
overlap region of a turbulent boundary layer was mainly The velocity measurements were made using a single-
due to changes in the upstream conditions. component fiber-optics LDA system. The LDA system was
Narasimha et al. (1973) suggested that scaling of the powered by a 300 mW ion–argon laser (Dantec Dynamics
relevant distances by the characteristic dimension of the Inc., Mahwaw, NJ, USA). The optical elements included a
slot and exit velocity is erroneous. Instead, they proposed Bragg cell, a 1.96 beam expansion unit and a 500 mm
the use of kinematic viscosity (m) and exit momentum focusing lens. The probe volume was 0.12·0.12·1.4 mm3.
(Mo) to scale the streamwise evolution of the flow. More In order to obtain data close to the wall, the LDA probe
recent parametric analysis by Wygnanski et al. (1992) and was tilted slightly (at an angle 2) towards the wall. Our
wall-jet similarity theory proposed by George et al. (2000) previous experimental results (Tachie et al. 2003) show
provided further support for Narasimha’s suggestion. that for tilt angle less than 3, tilting has a negligible effect
Using m and Mo, the decay of Um and spread rate for y1/2 on the streamwise mean velocity. Typical sample size at a
are given by the following power laws: measuring point (away from the very near-wall region)
 a varied from 5,000 to 10,000. Measurements were obtained
Um m xMo
¼ A m2 ð10Þ at the jet exit (x/b=0) and at several distances extending to
Mo x/b=130 downstream of the exit. The turbulence intensity
  in the central region of the jet at the exit plane varied from
y1=2 Mo xMo b 3 to 5%.
¼B 2 ð11Þ A total of 10 sets of measurements were made; five sets
m2 m
on a smooth surface, and five sets on a rough surface. The
The power-law constants A, B, a and b may depend on important test parameters are summarized in Table 1.
initial conditions such as source Reynolds number and Here S and R, respectively, denote measurements on a
velocity profile at the exit plane of the jet. Similarity smooth surface and a rough surface. In Table 1,Uj denotes
analysis by George et al. (2000) gave the following power- the centerline velocity at the exit plane of the jet, Uo is the
law relationship between Um and y1/2: bulk exit velocity determined from mass flow rate
284

Fig. 2 a. A schematic of top view and side

view of the open channel flume; b
schematic of the wall-jet facility

measurement using an electronic weighing tank, m is the )2, )3, )4 and )5 in Table 1 indicate decreasing Reynolds
kinematic viscosity, h is the height of water level above the number, Reo.
floor of the wall-jet facility, Reo is the source Reynolds The mean velocity profiles at the slot and several
number based on exit conditions (Uo and b), Mo=Uo2b, streamwise locations were reported in Tachie (2001). It
and h is the boundary layer momentum thickness evalu- was observed that the exit profiles obtained in this study
ated from the exit velocity profile. The maximum differ- are not as ‘‘full’’ as reported in some previous studies, for
ence between the bulk velocity determined from the mass example, Karlsson et al. (1993) and Wygnanski et al.
flow rate measurement Uo and the corresponding values (1992). Furthermore, the profiles obtained at x/b ‡ 80 are
obtained by integrating the exit velocity profile was less affected by persistent flow reversal and high background
than 4%. For a given surface condition, the notations )1, turbulence level. Thus, in most part of the discussion that
Table 1. Summary of exit
conditions Test Uj(m/s) Uo(m/s) m·106(m2/s) h (mm) Reo Mo·103(m3/s2) h (mm)

S-1 1.39 1.20 0.96 395 12,500 1.44 0.43

S-2 1.34 1.15 0.97 400 11,900 1.32 0.44
S-3 1.05 0.87 0.96 380 9,100 0.76 0.45
S-4 1.02 0.86 0.98 375 8,800 0.74 0.44
S-5 0.76 0.59 0.96 355 6,100 0.35 0.46
R-1 1.39 1.19 0.96 400 12,400 1.42 0.42
R-2 1.30 1.17 0.97 390 12,100 1.37 0.45
R-3 1.20 1.00 0.97 378 10,300 1.00 0.44
R-4 0.99 0.83 0.98 376 8,500 0.69 0.45
R-5 0.72 0.58 0.98 350 5,900 0.34 0.46
 285

Fig. 3. Variation of maximum velocity

with streamwise distance: a conventional
scaling; b momentum–viscosity scaling.
Error bars represent measurement uncer-
tainties. Lines in (b) are Eq. 10 with
appropriate values of A and a

follows, only data in the range 20 £ x/b £ 80 are examined. clarity, only selected datasets are plotted. The datasets
At these streamwise locations, the flow exhibits good two- plotted in Fig. 3a (and subsequent figures) are chosen to
dimensionality, and the collapse of the profiles is reason- include the lowest, intermediate, and maximum source
able (Tachie 2001). Reynolds numbers. Here bs=(b)2h), and h is the boundary
layer momentum thickness determined using the mean
4 velocity profile at jet exit. Values ofh are summarized in
Results and discussion Table 1. The choice of bs(rather than b) as the relevant
length scale is expected to minimize any effect of the shape
4.1 of exit velocity profile on streamwise development of the
Streamwise development of the mean flow flow (Ramaprian and Chandrasekhara 1985), for example,
2
The variation of normalized maximum velocity Um =Uj2 when the present data are compared with prior work in
with streamwise distance x/bs are shown in Fig. 3a. For which the exit velocity profile is ‘‘full’’. The error bars
286

Fig. 4. Variation of maximum velocity

with half-width over smooth and rough
surfaces. Lines are Eq. 12 with appropriate
values of C and c

shown in Fig. 3a represent measurement uncertainties at rough-wall data are comparable to measurement uncer-
95% confidence level. The present smooth-wall data at tainties. The best fit of Eq. 12 to the present data sets and a
Reo=9,100 are in good agreement with previous LDA data fit with the constants recommended by George et al. (2000)
(Reo=10,000) in the developed region, i.e., x/bs>22. The based on previous LDA data are also shown. The values of
figure also demonstrates that, at similar source Reynolds C in both fits are identical. The difference between the
number, especially in the region of flow development power-law exponentsc is approximately 1%. The near
2
(x/bs £ 22), values of Um =Uj2 are lower for the rough constancy of the power-law constants, i.e., C and c, may
surface than for the smooth surface. At source Reynolds suggest a universality of Eq. 12.
2
number Re12,500, for example, the values of Um =Uj2 at The variation of the inner layer thickness ym with xis
x/bs=11 are 0.78 and 0.65, respectively, for smooth and examined in Fig. 5. Accurate determination of ym is gen-
rough surfaces. Corresponding values at x/bs=22 are 0.53 erally difficult. To ensure that ym is accurately determined,
and 0.39 for smooth and rough surfaces, respectively. This the probe volume was moved in the y direction at incre-
difference is attributed to enhanced mixing owing to ments of 1.0 and 1.5 mm for the smooth- and rough-wall
surface roughness. data, respectively, in the neighborhood of ym. Figure 5
Figure 3b shows the variation of Um with x using the demonstrates that enhanced mixing caused by surface
momentum–viscosity scaling, i.e., Eq. 10. Only data in the roughness significantly increases ym. Specifically, the local
range 20 £ x/b £ 80 are plotted in this figure. As explained value of ym is about 40 and 25% higher for the rough-wall
earlier, the flow is approximately self-preserving and two- data at x/b=30 and x/b=70, respectively.
dimensional at these downstream locations (20 £ x/b £ 80). The variation of the jet half-width y1/2 with x is shown
Figure 3b shows that the momentum–viscosity scaling in Fig. 6. Conventional scaling is used in Fig. 6a. The
removes the Reynolds number dependency noted in Fig. 3a. spread rates vary from 0.085 for the largest source
This scaling also collapses the smooth- and rough-wall data Reynolds number to 0.090 for the lowest Reynolds num-
reasonably well. The best fits (in a least square sense) of ber. That is, the present datasets show a Reynolds number
Eq. 10 to the present datasets and those reported by dependence as observed in a number of prior studies. The
Wygnanski et al. (1992) are also shown in Fig. 3b. The values spread rates obtained in this study are higher than the
of the power-law exponent a for both fits are within 4% of average value of 0.073 recommended by Launder and Rodi
each other. However, the multiplicative constant A is (1981). This may be due to upstream conditions. Unlike
significantly different. These differences may be due to the distinct dependence of ym on surface roughness
different upstream conditions. (Fig. 5), Fig. 6a shows that spread rate for y1/2 is nearly
A plot of Um versus y1/2 is shown in Fig. 4. The data do independent of surface condition. That is, for the transi-
not indicate any systematic Reynolds number effect. tional rough regime examined in this work, roughness
Furthermore, the differences between the smooth- and effects on the mean flow are confined to the inner region
 287

Fig. 5. Inner layer thickness for smooth

and rough surfaces

(y £ ym). The ratio of ymto y1/2 is in the range 0.16 £ ym/y1/2 example, the rough-wall profiles are less ‘‘full’’ compared
£ 0.18 for the smooth-wall data and 0.20 £ ym/y1/2 £ 0.25 with the smooth data close to the wall as evidenced in
for the rough-wall data. Figure 6b reveals that the profiles turbulent boundary layer measurements (Tachie et al.
are nearly independent of source Reynolds number and 2000). As already noted in Fig. 5, it is seen that ym (i.e.,
surface roughness when the momentum–viscosity scaling location of U/Um=1) for the rough wall is farther away
is applied. The best fits of Eq. 11 to the present data and from the wall compared with the smooth wall.
those reported by Wygnanski et al. (1992) are also in good
agreement in spite of differences in the source Reynolds 4.2.2
numbers and the exit velocity profiles. The power-law Mean profiles in inner coordinates
exponents are within 2% of each other. The fact that the Figure 8a shows typical mean velocity profiles obtained on
value of b is less than 1 in both studies is an indication of a a smooth surface in inner variables (U+, y+). The values of
non-linear spread rate. Us for the smooth-wall dataset were obtained from the
velocity gradient at the wall, or by fitting a fifth order
4.2 polynomial, that is, Eq. 3 truncated at fifth order, to the
Transverse profiles at various streamwise locations near-wall data. The coefficients obtained in this work were
Consideration is now turned to the mean velocity profiles, c4=)0.00026 and c5=13.6·10)6; these values are in good
U(y), at various streamwise locations. Only selected agreement with those suggested by Eriksson et al. (1998)
velocity profiles obtained at 20 £ x/b £ 80 are discussed. and George and Castillo (1997), and in our recent LDA
At these locations, the mean flow is two-dimensional and measurements in turbulent boundary layers created in an
nearly self-preserving. The profiles obtained in the present open channel (Tachie et al. 2003). The wall-jet data by
work are compared with the refined LDA data by Karlsson Karlsson et al. (1993), low Reynolds number boundary
et al. (1993) at source Reynolds number Reo=10,000. layer data obtained in open channel (Bergstrom et al.
2001) and the logarithmic law (with j=0.41 and B=5.0) are
4.2.1 also shown in Fig. 8a. The region over which the loga-
Mean profiles in outer variable rithmic profile collapses with the wall-jet datasets is lim-
The mean velocity profiles obtained on a smooth surface ited to 25 £ y+ £ 80. For the present datasets, only six or
in the present work and a dataset by Karlsson et al. (1993) less data overlap with the logarithmic profile. Obviously
are shown in Fig. 7a using outer variables (Um, y1/2). The the use of the Clauser plot technique to determine the wall
agreement between the present and prior datasets is shear stress in a turbulent wall jet at low source Reynolds
reasonable. The profiles obtained on smooth and rough numbers may lead to a significant error. The boundary
surfaces are compared in Fig. 7b. The collapse of the layer and wall-jet profiles collapse fairly well in the region
smooth- and rough-wall profiles is good in the outer re- y+ £ 80. Therefore, contrary to prior arguments
gion. As shown in the inset, however, systematic and sig- (Hammond 1982) and experimental results (Wygnanski
nificant roughness effects occur in the inner region. For et al. 1992), Fig. 8a supports the premise that the turbulent
288

Fig. 6a, b. Variation of half-width with

streamwise distance. a Conventional scal-
ing; b momentum–viscosity scaling. Lines
in (b) are Eq. 11 with appropriate values of
B and b

wall jet has a universal inner region, which is also identical Castillo (1997) and George et al. (2000) were also used. In
to the inner region of a turbulent boundary layer. The determining the power-law constants (Ci and c), the fits to
success of the logarithmic law to model the open channel the experimental data were restricted to the region
boundary layer data up to the edge of the boundary layer yþ 60:1yþ1=2 . The power-law constants obtained vary only
(y+500) is due to the characteristic low value of the wake weakly because of the narrow range of Reynolds numbers
parameter in open channel flows compared with values obtained in this study. Our values of Ci and c at a given
reported for a canonical boundary layer. yþ
1=2 and corresponding values obtained by George et al.
The composite velocity profile (Eq. 4) was also used to (2000) are reasonably close, the maximum difference being
process the smooth-wall data. Typical results are shown in 3%. The composite profile matches the velocity data at
+
Fig. 8a. The values of c4 and c5 used to fit the fifth order yþ
1=2 ¼ 880 from the wall to y =80, and the data
polynomials to the near-wall data were used. The values of yþ +
1=2 ¼ 1; 930 from the wall to y =150. The ability of the
a+ (=)16) and d (=8·10)8) recommended by George and composite profile to match the experimental data
 289

Fig. 7a, b. Mean velocity profiles in outer

coordinates. a Comparison with previous
data; b Comparison between smooth- and
rough-wall profiles

throughout the viscous sublayer, the buffer region and the (Eq. 5) in Table 2. In computing values of Us/Um from
overlap region implies that more data points were Eq. 5, the values of Ci and c used to fit the composite
employed in the profile matching compared with the use of profile to the experimental data were employed together
the Clauser plot technique. For this reason, we expect the with D=1.00 and Co=1.3 as recommended by George et al.
composite profile to provide a more accurate estimate of (2000). The table shows that the maximum difference
the friction velocity Us than the Clauser plot technique. between the two sets of values is less than 5%. The
To demonstrate the reliability of the power law to agreement noted above is good in view of uncertainties in
estimate Us, values of Us/Um obtained independently by determining y1/2, ym, Um and Us. It should be noted that
fitting to the near-wall data were compared with the cor- the values obtained from the near-wall fit are consistently
responding values predicted from the power-law theory higher than those obtained from the power-law theory. As
290

Fig. 8a, b. Mean velocity pro-

files in inner coordinates. a
Comparison with previous
data; b comparison between
smooth- and rough-wall pro-
files. The lines are fits to the log
law and composite/power-law
profiles. k+=kUs/m, where k is
the average roughness height of
the sand grains (k=1.2 mm)
and Us is the local friction
velocity; Rem=Umym/m

noted in our previous work (Tachie et al. 2002), the An accurate determination of the friction velocity in
agreement would be better still if a slightly higher value of rough-wall turbulent flows presents a greater challenge
Co¥/Ci¥ was used. For example, if Co¥/Ci¥=0.0237 is used than for a smooth wall. Fig. 8a and Table 2 demonstrate
(instead of 0.023), the maximum difference between the that the power law provides accurate and consistent esti-
two approaches would reduce to 1.5%. In this case, addi- mates of the friction velocity in a turbulent wall jet over a
tional refined measurements are required over a wider smooth surface. The reliability and accuracy of the power
range of Reynolds number to calibrate the power-law law to determine the friction velocity for rough-wall tur-
constants. bulent boundary layers were shown in our previous work
Table 2. Comparison between
values of Us/Um obtained yþ
1=2 Rem Us/Um
from near-wall fit
(Eq. <equationcite>3 Near-wall fit Theory %
</equationcite>) and (Eq. <equationcite>3 (Eq. <equationcite>5 difference
values obtained from </equationcite>) </equationcite>)
power law theory,
Eqs. <equationcite>5 880 2,700 0.0620 0.0599 3.5
</equationcite> and 1,110 3,300 0.0611 0.0583 4.6
<equationcite>6 1,590 5,000 0.0585 0.0564 3.6
</equationcite> with a=0.46, 1,700 5,400 0.0574 0.0556 3.1
A=2.9, c¥=0.0362, Co¥/ 1,930 5,800 0.0569 0.0549 3.5
Ci¥=0.023 2,430 7,670 0.0561 0.0536 4.4
291

(Bergstrom et al. 2001). In previous application of the shown for comparison. Selected values of Cf predicted
power law to the rough-wall boundary layer, the values of from the power-law theory (George et al. 2000) are also
Co and a+ for a smooth surface were adopted; only the shown in Fig. 9. The present smooth-wall values are in
power-law constants Ci and cwere optimized. In this study, good agreement with those predicted from the power-law
the power law was applied to the rough-wall data as well. theory (George et al. 2000), and from the correlations of
Fig. 8b shows typical rough-wall velocity profiles and a Bradshaw and Gee (1960) and Eriksson et al. (1998).
smooth-wall profile. The corresponding fits of Eq. 2 to Hammond’s correlation is also in reasonable agreement
selected rough-wall data are also shown. The power-law for Rem (=Umym/m)‡7,000, but indicates substantially
constants (Ci and c) were chosen to ensure a reasonable fit higher values than the present data and other correlations
of Eq. 2 to the experimental data. It is well known that the at Rem<7,000. Figure 9 also demonstrates that surface
virtual origin (for y) for a rough surface depends on the roughness significantly increases the skin friction com-
specific geometry, distribution and density of the rough- pared with the smooth-wall values. The increase owing to
ness elements. The particular sand-grain roughness used surface roughness varies from 15% at Rem3,000 to about
in this study has been used in our previous rough-wall 30% at Rem7,000. It is also clear from Fig. 9 that Cf de-
boundary layer studies (e.g., Tachie et al. 2000; Bergstrom cays gradually with increasing Rem for the smooth wall,
et al. 2001). Based on these prior studies, we estimated the but nearly independent of Rem for the rough wall.
virtual origin to be 0.3 mm, that is, 25% of mean diameter
of the sand grains. Because of the uncertainty in deter- 5
mining the virtual origin, the fits were limited to y+ > 40. Conclusions
The power law is able to model the rough-wall data in the The conclusions of the present study are summarized as
region 40<y+<250. The relatively wider range noted in this follows:
case compared with the smooth-wall data is consistent
with the relatively thicker inner layer observed in Figs. 5 1. For the Reynolds numbers considered in this work, the
and 7b. The values of the local Reynolds number Rem inner region of the mean velocity profiles for the
(=Umym/m) and the dimensionless roughness height, smooth-wall turbulent wall jet and low Reynolds
k+ (=kUs/m) are given for each dataset. The value of k+ for number turbulent boundary layer collapse reasonably
test R-1 (x/b=33) is higher than for test R-1 (x/b=70) be- well. Therefore, the lack of universality noted in the
cause the friction velocity decays with downstream loca- inner region in some prior wall-jet studies may be due
tion. Figure 8b demonstrates that one effect of surface to inaccurate estimate of the friction velocity, which
roughness on the mean flow is to shift the normalized may be partly due the narrow region over which the
velocity profile downwards and to the right. Since the logarithmic law applies to the smooth-wall turbulent
values of the roughness shift, DU+, are low (less than 5), wall jet.
the rough surfaces are in the transitional rough regime. It 2. The momentum–viscosity scaling removes the source
is evident from Fig. 8b that the roughness shifts depend on Reynolds number effects as well as initial and surface
both the local Reynolds number (Rem=Umym/m) and conditions on the streamwise development of the mean
dimensionless roughness height k+. The lack of systematic flow, at least for the smooth surface and transitionally
trend between roughness shift and the k+ values may be rough regime reported in this study. That is, the max-
due, in part, to the fact that the rough surfaces are in imum velocity, Um, and jet half-width, y1/2, obtained on
transitional rough regime, and the fact that the velocity smooth and rough surfaces, and at various Reynolds
profile U(y) decays with downstream distance. numbers collapse onto a single curve when scaled with
the kinematic viscosity and source momentum.
4.3 3. At similar source Reynolds number and downstream
Skin friction coefficient location, the value of maximum velocity, Um, obtained
The values of the skin friction coefficient,Cf=2(Us/Um)2 on a rough surface is lower than on a smooth surface.
obtained from the power-law constants and Eq. 5 (with Surface roughness increases the inner layer thickness,
D=1.0,Co=1.3) are shown in Fig. 9. The smooth-wall ym, of the wall jet, but the spread rate for the jet half-
Cf-correlations proposed by Bradshaw and Gee (1960), width, y1/2, is nearly independent of surface roughness,
Hammond (1982) and Eriksson et al. (1998) are also at least for the transitionally rough regime. The results
292

Fig. 9. Variation of skin friction coefficient

Cf with Reynolds number Rem over smooth
and rough surfaces

reported in this paper clearly demonstrate that surface International Symposium on Applications of laser techniques to
roughness enhances the level of skin friction (typically fluid mechanics, 20–23 July, Lisbon, Portugal, paper 1:5.
Kruka V, Eskinazi S (1964) The wall jet in a moving stream. J Fluid
15–30%) compared with the smooth-wall case. Mech 20:555–579
Launder BE, Rodi W (1981) The turbulent wall jet. Prog Aerospace Sci
19:81–128
Launder BE, Rodi W (1983) The turbulent wall jet-measurement and
References modeling. Ann Rev Fluid Mech 15:429–459
Abrahamsson H, Johansson B, Lofdahl L (1994) A plane two- Narasimha R, Narayan KY, Parthasarathy SP (1973) Parametric
dimensional wall jet in a quiescent surrounding. Eur J Mech B analysis of turbulent wall jets in still air. Aeronaut J 77:355
Fluids 13:533–556 Rajaratnam N (1965) Plane turbulent wall jets on rough boundaries.
Bergstrom DJ, Tachie MF, Balachandar R (2001) Application of power Dept Civil Eng, University of Alberta, Canada
laws to low Reynolds number boundary layers on smooth and Ramaprian BR, Chandrasekhara MS (1985) LDA measurements in
rough surfaces. Phys Fluids 13:3277–3284 plane turbulent jets. J Fluids Eng 107:264–271
Bradshaw P, Gee MT (1960) Turbulent wall jet with and without an Sakipov ZB, Kozhakhmetov DB, Zubareva LL (1975) Analysis of
external stream. Aero. Res. Council R & M 3252 turbulent jets flowing over smooth and rough surfaces. Heat
Castillo L, Johansson G (2002) The effects of the upstream conditions Transfer-Soviet Res 7:125–133
on a low Reynolds number turbulent boundary layer with zero Tachie MF (2001) Open channel turbulent boundary layers and wall
pressure gradient. J Turbulence 3 jets on smooth and rough surfaces. PhD thesis, Dept Mech Eng
Eriksson JG, Karlsson RI, Persson J (1998) An experimental study of a University of Saskatchewan, Canada
two-dimensional plane wall jet. Exp Fluids 25:50–60 Tachie MF, Balachandar R, Bergstrom DJ (2002) The inner region of a
George WK, Abrahamsson H, Eriksson J, Karlsson RI, Lofdahl L, turbulent wall jet. Exp Fluids 33:351–354
Wosnik M (2000) A similarity theory for the turbulent plane wall Tachie MF, Balachandar R, Bergstrom DJ (2003) Low Reynolds
jet. J Fluid Mech 425: 368–411 number effects in open-channel turbulent boundary layers. Exp
George WK, Castillo L (1997) Zero pressure gradient turbulent Fluids 34:616–624
boundary layer. Appl Mech Rev 50:689–729 Tachie MF, Bergstrom DJ, Balachandar R (2000) Rough wall turbulent
Hammond GP (1982) Complete velocity profile and ‘‘optimum’’ skin boundary layers in shallow open channel flow. J Fluids Eng
friction formulas for the plane wall-jet. J Fluids Eng 104:59–66 122:533–541
Karlsson RI, Eriksson JG, Persson J (1993) LDV measurements in a Wygnanski I, Katz Y, Horev E (1992) On the applicability of various
plane wall jet in a large enclosure. In: Proceedings of the 6th scaling laws to the turbulent wall jet. J Fluid Mech 234:669–690