.

ReughnessElement Geometry Required

for Wind Tunnel: Simulations of the
I. S. GARTSHORE
Atmo$.phericW~nd
Associate Professor. Using a data' correlation for the wall stress associated with very rough boundaries
and a semi-empirical calculation method, the shape of boundary layers in exact
·K. r•. DE CROOS
,
equilibrium with the roughness beneath them is calculated. A wide range of rouphness
geometries (two- and three-dimensional elements) is included by the use of eq1jivalent
ReSearch Assistant ..
sur/aces of equal drag per unit area. Results can be summarized ~n a sing/4 figure
Dept. of Mechanical Engineering,
The University of British Columbia,
which relates the shape factor of the boundary layer (its exponent if it has a p~er law
"'''''. VancQ.\lver, B. C., Canada 'l velocity profile) .to the height of the roughness elements and their spacing. Mew data
for one turbulent boundn,ry layer developing over a long fetch of uniform roughness is
presented, Wall shear stress, measured directly from a drag plate is combi,ned with
boundary layer integral properties to show ihat the sMar stress correlation a}lopted is
reasOttably accurate and lha,t the boundary layer is close to'equilibrium aft$- passing
over a sireamwise roughness fetch equal to about 350 times the roughiuds element
height. An example is given of the way in which roughness geometryma~ be chosen
from calculated equilibrium results, for one particular boundary layer thif:kne.ss and
a shape useful for sir.tUlating strong atmospheriC winds in a wind tunnel.

1 IntroductiO:l
, lations is often mentioned and is usually implied, although the
term equilibrium is seldom defined precisely. What is usually.
The neutral atmospheric boundary la~' can be simulated in a
meant by equilibrium is that, boundary layer characteristics such
wind tunnel 9Y creating a thick turbulent boundary layer on
as velocity profile shape, nondJmensional spectra, etc" are not
one ~prface of the tunnel (usually the floor) so that measurements
changing significantly in the streamwise direction. A more
can"be made in the turbulent region witli IllDdels of appropriate
exact form of equilibrium is defined by the term "se1f-preserva-
scale. Upstream of the model, which may inchlde scaled sections
of an urban area for example, the thick turbulent boundary la)ler
tion" used by Townsend. [4J. Self preservattorl describes a tur-
bulent shear flow whose turbulence is in exact; dynamic equilib-
is developed by combinations of spires, wedges or trips together
ritlm so that the mean distributions of the turbulence, nondi-
with roughness elements di~tributed on the wall. Various tech-'"
mensionalizea by a single velocity and lengthscale,.do not change
niques have been used (see discussions by Davenport and.
at all in the streamwise direction. Such exacj:. equilibrium is
Isyumov [1],1 Standen [2J, and Counihan [3]) but in each case the .
strictly possible for rough wall boundary layers in zero pressure
wall roughness geometry used waS apparently chosen by trial
gradient only when certain rather artificial! conditions are sat-
.and error and ~ot because.it produced It predicfable wall shear
isfied: either the roughness elements are.high compared to their
, ,stress or velocity distribution. This work pr0vides Il. relatic;1 be-
spacing (which must be regular),p'roducing,J.he •"d-type" rough-
tween roughness geometry and velocity profile shape whi~h should
ness d~cribed by Perry, et al. [51 or else !the ~ouli\hness height
.allow a more rational choice of the. roughness tu be made {or any
and spacing must vary directly with strel"llwise p,istance so that
,\iesired profile or 'effective wall shear stress. .
each remains a constant fra.ction or the bOundary layer height,
. ~ .' '. The equilibrium of bOundary layers used for atmospheric .simu-
as described by Rotta [6). i -
Some anaiytical results implied by exabt self-preservation are
-.-.-. prese~ted in Section \0£ this paper. !. "
, ·1.Numbers in braeketc design ..~ References at end of paPe~. • r%
A restricted region of' a boundary layer with zero pressure
. ',.\ Contributed by the Fluid!, Bngineering· Dh-laion and p,,,,,,,nted at the Winter gre.dirnt and constant roughnesS height and spacing is in.approxi•
'·"Annual Meeting, New York, N. Y., DeC\l!lIber 5, £976 of Tllm AMERI'::"N
J~<>~lEiY OF ,MECIIANICAL ENGlNE>lR8. M~nllscript received at ~\S~I E Head-
mate. equilibrium provided the start of thl'l rQughness or. any
quarters August It,~976. P",per No. 76-WA/FE-18. change in roug:~ne3S is far from the region considered. Mellor
)'" "
';480 (SEPTEM'BER 1977 '. '(raitsactions
,: ,. .'
of the AS.ME
. . '
 nnd (;ihson [71 hn\"c concluded thnt t\lrb\llent, zero-pressure- tern of square hars normal to the flow and it is necessary to geu- ,
gr,, (licnt hotlltd"r~"I\I:'('rs on smooth walls closely ('onforln to the cralizc these results to three-dimensional roughhess elements of
f'('quiremcnts for eX:lct rq\lilihritllll o\'er limited strenmwise nrious shapes. This is done here by defining an effective spacing
length,", and it seems reru;onahle to tlssume that the same is t1\le between t~o-dimensional bars which produces surface drag per
for ro\lgh-wall boundary layers', :\ disc\lssion of this point is in- unit area (or average shear stress) equal to that of the more gen-
duded in Section (i of this paper, eral roughness pattern. rJ
In ordcr (0 !'stimate the rO\lghncs" which is required to pr~uce The drag of a unit width of a single two-dimensional square
('crtain ' houndary layer charnelcristics, evcn for approximate bar (subscript b) is:
equilibrium, it is'nece"ssary (0 formulate'sgmc relationship for the
\\::111 shear stres~ in terms of other hn\lIH.lnry layer propert.ies, The Db = CD. 21 piflk
-
strellll1wise development of the flow <:1I1t then he predicted by
:\nv one cif a varict \' of well known semi-empirical methods , where It is t he height of the bar and (j the average velocity ap-
TI;ese two steps fire 'o utlined in t he next two sections, follow!'d proaching it.. The drag per \\nit area is t.hen D'b = Db/ A. where
h~' a presentation of the ex,,!'t eq\lilibrium \es\llts implied In- the A, i ~ the spacing between the bars in the flow direction.
simple met.hod adopted. Some new experimental e\"idencc '1Il Similarly, the lwemge drag of a unit area of general three-di-
rO\lgh wall bO\lndary la.\·er;'; is th en presented and the approach mellsionnl elements each of similar shape and each 'having a drag
to eCJ\lilibrium is discllssed, Finally a procedure is suggested for
coefficient C IJR hn.c;ed on its frontal area, is
estimating the roughness required to produce a desired boundary
lay~r shape needed for atmospheric simulation, and an example
is given of this procedure.

2 , A Wall.StressRelationship for Rough Walls where ( AR), and (AR), an~ spacings between element centers in
the laternl and t.he longitudinal directions respectively and Ap is
The flow around individual roughness elements [It high Iteyn-
one element's fronfal area. Equating D'Rto D\ for equal (j:
old& numbers i3 not Y"et predictable in detail from the equations
of motion. Previous empirical work has related the wall shear
stress to some }ntegral boundary layer thickness, usually,:i~t
displacement thickness (0'), and this empirical approach.. ,is.
followed here. :' .... This can be rewritten as:
A rather extensive review of t.he results obtained using squ!J.re,
two-dimensional bar roughness elements has been made by (3)
Dvorak [8] wh'o ,related the bar height (k) and the bar spacing
(A,) to the effective wal[ shear st.ress (To) and displacement thick-
where A. = (AN), (AR)! is ~he effective plan area associat,ed
ness 0', using the form proposed by CltIllser [91, a.~ follows: , ,with each element. The ratio (A./ k) is nilw the equivalent s~ng
,,"o-dimensional bars of height k whi~~-we.11 stress
~ = kIn {~':k'}
+ A-C. (1) equal to the general roughness whose height is also k.
The drai c~fficients (CD). and (CD)R vary with varying 'up-
stream boundary layers, but we assume that the ratio of one to
where: U =
U,
r
{~r
pU,2
the other will remain constant for identical upstream conditionJ
. " and identical heights of bar and roughness element. We further
't'

W~'\
J "
free ,!'tream velocity
assu1lle that ~e mean characteristics of the turbulent boundary
layer, and its turbulent properties as well, are dependent only
A constant = 4.8 on the"valueof the effective wall shear stress, so that two rough-
ne.ss patterns which produce identica.l.loca.l. wail stress will produce
K constant = 0.41 identical boundary layers. These assumptions remain to be
and C is a constant depending on A, and k as follows: tested in detail hut they produce plausible and tentatively useful
relationshi~s.

C = - 5.95 [ O.4S In ~ - 1]
We must now det~mine the ratio of (CD). to (CD)R measured
(2)
with identical element heights and identical upstream boundary
IlIyers. If the upstream boundary la:yeds considerably smaller
This con elation is valid for (A,/ k) > .':i and for (U,It / v) > 70, than. the height k, the drag co-efficient,of a surfaoo:-mounted
both satisfied in most atmospheric simulations of high winds .. , square plate normal to the flow is about 1.15, and that of s. two-
Few natural rough boundaries can be al?proximated by p. pat- dimensional surface-mounted plate infinitely wide nonnal to the

-------Nomenclature&---------------------------------------------------~~­
Ap frontal 'lrea of one roughness ele- K von Karman'si' constant; taken to Z distance normal to wall • \
ment ' be equal to 0.41 ('" Ii r.ominal bounde,ry layer thlckn.~
Ap plan a; <!a assOcillted. with one roughness element height · related too· and n by equation
roughness element fl = shape fa.ctor related: to H by equa- (11)
A ,;.. 4.8, constant associated with tion (13); equal to the~El,Xionent 0· displacement ~hickness
equation (1) of a ,power law veloqliy. trofiHl e momentum thlckness
C runction of roughness geometry, if equation (7) is val~tt ~ A. or
effective streamwise spacing two
equa tior: (2) . mean velocity at height Z~ dimensional bars; see also ~ua­
H EO o· /e, shape fietor for boundary , free stream velocity (constant) tions (3) and (4)
layer .. [To/pPI'l shear velocity p - density of fluid (constant)
H, shape factor related te H by equa- x = streamwise distance, measured To ~ average shear strElS& on roughened
tion (6) from start of roughneSs wall

Journa1 of Fluids Engineering SEPTEMBER 1971/481

. ~
 I
flow is about 1.20. A cube in the same circumstances has a drag properties UT) fJ and 8*. Equations (1) and (2), relating ([!r/Ut)
co-efficient of about 1.17 and a long square bar normal to the to 0' for given k a nd h,.are shown graphically in Fig. 1. The dif-
flow, four times as wide as it is high, has a value of 1.2.. These ferential equ ations (5) and (8) must be solved numerically for
results, all taken fr om the compilation of referen ce [lOJ, suggest. particular cases of interest, unless the boundary layer is in exact
that the ratio (Cnh/(Cn)n is close to unity for some shapes of eq uilibrium. In this case, the...,Q.ifferential equations become
roughness element. In. particular, this approximatioll appears arithmetic as shown in the ne~ection.
valid for parallelepiped-shaped elements having one fa ce normal
to the flow, depth less than or eqllal to height and width greater 4 Results for Exact Equilibrium
than about 0.4 times the heigh t [sce reference 110\\. For other
shapes of roughness element,,.rlrag r,oe fficients must be estimated If the velocd}' profile shape is not chllnging at all, i.e. 8' / fJ
or rneasured for use in equation (3). and Ht are constant, the equations of the previolis section adopt
Since the rectangular elements described in the previous para- even simpler form s. Iqs useful to demonstrate this, since zero-
graph imply a drag co-efficient ratio of approximately one, equ a- pressure-gradient. bou,n,dllry layers developing over.a constant
tion (3) for these elements becomes: roughness geometry may conform approximately. to this equilib-
rium condition.
X. ~ Ap
(4) If (8*10) = constant, then HI = constant from equation (6).
k Ap Equation (5) then shows .that dfJ/dx is a constant, so that the
boundary layer grows linla"rly. Equation (8) shows that UTIU l
Equations (3) or (4) allow the shear stress correlation of Dvorak = .constant in anyone such equilibriutn case, which is strictly
to be used for a wide variety of two- and three-dimensional possible from equations (1) and (2) only when X.l t and 8* Ik
roughness shapes. remain· constant, asalready mentioned. By combining equations
(8) and (5) to elimjnate fJ, we may relate H to UT IU I as follows:
3 A Calculation Method
Having used integral properties in the shear stress correlation
[~J = FI(H)IHI = function of (H) (9)

of the previous section, it is now sensible to adopt an integral

calculation method to predict the stream wise development of where FI and H t are related to)!'ln equation (6). The shape
the boundary layer. . factor H can be relllted to n WhiCh can in turn be interpreted as
As pointed out by Dvorak [8], an integral atLx;lliary equation the exponent of the power law profile if equation (7) is adopted.
which is based on the entrainment of fluid through the outer Notice that n can be regarded as a shape factor even if equation
edge of the turbulent region is preferable to those based on, in- (7) is not an accurate description of the velocity profile however.
tegral properties of shear stress or kinetic energy through the The shear stress ratio (UTIU I ) occuring in (9) can be replaced
layer. The latter methods ·may be particularly sensitive to the by suitable functions of 15' /k and A,/k as shown in Fig. 1 or equa-
roughness geometry through the uncertain conditions very close tions (1) and (2) so that (9) can be written as:
to the individual roughness elements. For this reason, we follow
Dvorak in using Head's empirical correlations. Head's method n = H - 1 =.fn [UT]
-2~ U
I
is described by Dvorak and in reference [11]. Applied to zero-
pressure-gradient boundary layers, it becomes:
(10)
(5)
Since,
where FI is a known empirical function of the ratio (8* If}), fJ is
~[n: 1 ]
15*
the usual momentum thickness and HI is a shape factor related (11)
k
empirically to (8* I fJ) [see equation (6)]. For the range of (8* IfJ)
expected in atmospheric simulations, F I (/)* / 0) can be represented,
following Dvorak, as:

Fl ( 8;) t
= exp -3.512 - 0.617 In (HI - 3)]

where: HI = 3.3 + exp [0.4667 - 2.722 In (H - 0.6798)] (6)

.8-
and H=-
8
It is not necessary to assume a form for the velocity profile
to use this method, but if the simple power law is used for other
purposes, it can be related to H in the usual way: if

~ " [rJ (7)

the exponent n is equal to (H - 1)/2.

All in all integral methods, the momentum integral equation: 2 5 10 20 50

(8)

is also used. Thus, equation (1) describing the shear stress, and Flg.l Wall shear stress expected for various roughness \leometrles.
Dvorak's correlation (8) __ - - - - ' present dau 0: O'uughlln
equations (5) and (8) describing the development, can be used from Woodlnll, et al. [14], >',/k - 64-<:;): Antonla rnd Luxton, [12].
to predict the streamwise variation of the three boundary layer >',/k - 4, 8.

482/ SEPTEMBER 1977 Transactions of the ASME

 ~ 1
can be used to define a nominal boundary layer thickness 0, equal 38.1 mm thick and with 203 mm chord placed across the entire
to the' netual bOllndllry Inyer thickness if a power profile exists, t \lnncl so thnt it created a gradual two-dimensional ramp and
equation (10) can be written 1lS: then a backward facing step 3S.1 mm high.
Wall shear stress was measured directly from a drag plate, a
lOA, ] (12) large isolated section of roughened floor fllStened to an accurate
11 = fll Lk' k wind tunnel balance. The drag force on this section could then be
measured directly and the average force per unit area deduced.
IlS plotted in Fig.:2'. Notice thnt. this rclat.ionship is valid only for
This drn.!I: plate WIlS 0.91 m wide and approximately 2 m long,
equilibrium situations, sincc t.hc arithmetic relationship of equa- so that reasonably l/lorge drag values were obtained. Drag read-
tion (9) is applicable only for exact equilibrium. ings wcre found to be insensitive to the size of the small gap
The ratio [V,o' /(VTo)] is related to the defect i1lw, IlS pointed around the drag plate but were sensitive to tunnel pressure
out by Clauser [9] (see also reference [11 D. He found t.hat bot.h gradient. The latter was set as carefully as possible to zero'for
rough and smooth wall boundary layers have the same value of each test. Drag readings were found to be proportional to the
[Vlo' /(VTo)] in zero-pressure-gradient, implying a truly universal square of the free stream velocity, as expected, indicating that
defect law. The present equilibrium calculations do not give a no Reynolds number effects were present. The drag plate meas-
precisely constant value of this parameter over the entire range urements and the boundary layer parameters 0, 0·, 0 and n
of A,/k and 11 con;;idercd. Calcuillted values lie between 3.S0 deduced from the measured velocify profiles are discussed here,
and 3.99, compared to Clauser's value of 3.60 for boundary since they are directly comparable to the correlations of previous
layers developing over constant roughness. These values will be sections. Measured values of o·
and 0 and values of nand 0
dip cussed in Section 6. deduced from them are given in Table 1.

5 Present Measurements Table 1 Measured boundary layer properties

To compare with the calculated results of Figs. 1 and 2, par- x o· 0 n 0 Ut Uto·
ticularly for three-dimensional roughness elements, measurements k T r k U; ~ U.o
were made in a boundary layer developing over a long fetch: of
140.8 2.17 1.28 0.348 8.38 16.48 4.27
uniformly roughened wall. 26S.S 2.S3 1. 75 0.309 11.93 17.97 4.2'(;
The U.B.C. wind tunnel used for these experiments is an open- 332.S 3.25 2.06 0.289 14.44 18.41 3.14
circuit, blower type and has a test section 24.4 m in length, 2.44 396.S 3.51 2.27 0.273 16.49 18.74 3.99
m wide, and initially 1.5 m in height. The test section roof is 492.S 4.0S 2.67 0.264 19.73 18.98 3.92
adjusted to maintain ambient room pressure throughout the test
section I~ngth, Mean velocity mellSurements were made using
NOTES
a linearized hot wire anemometer.
The floor roughness WIlS created by fllStening thin vertical 1) k = 1 1/2 in. (38.1 mm,Y ,
strips of aluminum to the wall at regular intervals. Each strip 2) The boundary layer thi'cknesses 0* and 0 were obtained by
presented a frontal area of 3S.1 mm high and 19 mm wide to the integration from measured velocity profiles; 0 and 71 were
flow. The strips were placed in lines across the tunnel, the strips then deduced from 0* and e. ,
being 152.4mm apart in each line and the lines being 152.4 mm 3) V T WIlS found from' forCe measprements on the drag plate.
apart using a staggered pattern. 4) Roughness Reynolds numbet VIe/V, was approximately
The resultant value of Ap/ Ap for equation (3) is 32 and, using 3 X 104.
the assumption, that the drag coefficients in this equation are
equal, the value of A./k obtained from equation (4) is also 32. e
The values of o· and were obtained by integration directly,
A small trip was used at the beginning of the roughness fetch from the measured velocity profiles. Values of H ar& equal to
to reduce the importance of any length of smooth surface up- the ratio (0*/8) {see equation (6)] and 71 may be calculated from:
stream of the roughened wall and to reduce the importance of H - 1
the initial roughness elements. (See a discussion of this point by n= - 2 - (13)
Antonia and Luxton, reference (12]). The trip used was a wedge
4
The quantity n is used here simply as R shape factor; related to
H through equation (13). In fact, the velocity distributions were
not particularly well desctibed by a relatio::lShip of the form of
equation (7), so that the more taborious but' precise approach
described above was necessary. The nominal boundary layer
thickness 0 was then found from equation (11). Values of 11 as
calculated were close to the values of 11 which might have been
chosen from the velocity profiles themselves. Aa is well known,
the definition of 0 directly from the measure velocity profiles is
somewhat arbitrary, and has been avoided through the present
n procedure. .
For experimental convenience, measul'(}ffients with different
lengths of roughness fetch were made, not by moving the <Ira«
plate from one place to another, but by adding or subtracting
roughness sections upstream of the fixed point of measurement.
'I\us, the measurements do not relate precisely to a single bound-
ary laY'er, since upstream 0: each fet-::h of roughness (x), there is a
2 variable length of smooth floor equal to t,be tunnel length minus
x. Perhaps for this reason, the measured values of 0 were slighUy
6 but consistently larger than those obtained by use of the two--
k dimensional mometitum equation [equation (8)] together with
FIII.2 Effect of roughness geometry on profile shape. Exact _quill- mellsured values of shear stress. A difference of 11 percent was
br~um conditions, ; present data, 0. found between the two values of 0 at the m8J'lln\lI'l fetch of ..

Journal of Fluids Engineering SEPTEMBER 197'1 /483

 e
roughness \fhen measured values of at· the short est. ronghncss URcfuJ" in predicting rough-wall boundary layer characteristics
fetch were l\sed ns the starting condition for the int egr:ttion nf and that It hOlllHhry b ye r developing over constant wall rough-
equation (8). ness can approach eq uilibrilltll after n rather long (x/ k>' 350 in
The experimental l\llcertainty of the data is estimated IL'i fol- this c:L,e) retch of roughness. The fetch required will certainly
lows: U, / U I , ± 3 percent; o'/k, ± 2 percent; el k ± 2 perce nt; \'n.ry with t.he roughn ess used and the initial conditions imposed,
n (bnsed on uncertainties in o' and e), ± 4 pcrecn\.; o/ k (b:L'cd bllt is not always us large u..q suggested by Counihan [13] (x/t;
on uncertainties in o· and n), ± 6 percent. Uncertainties in x = 1,000) on other grounds.
and J; are negligible.
,.
6 Discussion of Experimental Results 7 Estimating Roughness for Use in Atmospheric
Simulations .
The measured values of o·/k, elk, Ut/UT and n are plotted in
Figs. 1 and 2. The roughness required to generate a velocity prolue of a
The calculated curves of Fi~. 1 do not lISSume an equilibrium given shape with a boundary layer of a given depth can be
condition, but merely represent the shear stress correlation of estimated from the equilibrium results described in Section 4.
Dvorak [8]. The measured yalues in this figl!re agree with the Great accuracy will not be obtained, for the degree of equilibrium
trends of the data and indicate that the effective value of A./k for reached in any situation depends on the upstream history of the
the present data is about 30, which is cloRe to the value expected layer and in particular the fetch used and the trip used (iI any).
from the geometrical arrangement ns calculated from equation First estimates are possible however, as shown.in the following
(4) (A./k = 32). Thus, we have some confidence in the shear pamgraphs. .
stress correlation of Dvorak and in the present extension of it For example, if a boundary layer is required for simule.tions of
to three-dimensional roughness elements. a suburban wind at a scale of 1 :400, then a boundary layer of
A comparison with the data of two other experiments is made about 0.9 m thick is required with a shape given by n = 0.28. If
in Fig. 1: that of O'Laughlin, reported by Wooding, et a!. [14], the defect law integral parameter Ulo·/(U) is about 4 for a
and that of Antonia and Luxton [12]. The data of O'Laughlin boundary layer developing over uniform roughness, as found in
was obtained in an aii' duct 0.11 m by 0.46 m in cross-section and the tests described in S€ction 5, then, from equ",tion (11), 0·/0
7.31 m long. Cubes 4.8 mm on a side covered one of the 0.46 m is about 0.22 and Ut/U, is therefore about 18. Also, the nominal
walls of the duct and sh~ar streSJl was mensured directly using a value of o· is about 0.2 m.
force plate. The closest cube spacing used by O'Laughlin pro- Appropriate combinations of roughness, height and density
vided a ratio of A./k, in present notation, of 64 and one result, may now be chosen from Fig. 1, since 0* and Ut/UT are known.
inferred from reference 14, is plotted in Fig. 1. Agreement is If 2.5 ~m roughness elements are used, then 0* /t = 8.0 and for
not good in this case, perhaps because of the rather small duct U, / U I = 1/ 18 = 0.055, the effective value of X./k is about 14
size used by 0' Laughlin. from Fig. 1. This means that roughness elements with square
Al1tonill. and Luxton studied a boundary layer developing over frontal projections 2.5 mm square would [from equation (4)] have
square bars placed normal to the flow such that ';<.. / 1; = 4. For to be placed on approximately 95 mm centers to provide the
large x/k, (x/k ;;> 250) H became constant and then decreased correct roughness on the floor. Alternatively, if square elements
with increasing x, from which the authors assumed that an 40 mm high and 40 mm wide are used, a similar procedure sug-
equilibrium condition had been reached. The data quoted here gests that they should be placed on approxillll\tely 175 mm
. is for x/k = 372. For A./ k = 4, it is likely that the bars are so eenters.
close together that mutual sheltering takes place. This is If the local effective wall shear stress determines the mean
anticipated by Dvo.rak's results [8] and by the limitation A./k characteristics of an equilibrium boundary layer and its impor-
> 5 placed on equation (2). We expect, therefore that the noCl- tant turbulent properties as well (as already assumed in Section
dimensional shear stress (U ,/U I ) would be lower than, but close 2), then both of the roughness element geometries d&8cdbed will
to, that cakulated for A. /k = 5. The data from Antonia and produce identical boundary layers once equilibdum is reached.
Luxton's report [12], ploited in Fig. 1, confirms .this expeCtation. The important question of which flow would reach equilibdum
The data from the present experiments plotted in Fig. 2 do more quickly requir.~ further investigation and would de.;>elld on
not all show good agreement with the plotted curves which were the initial conditions used in each case.
calculated using the lISSumption of equilibrium. As is clearly When exact equilibrium is assumed, as in the present example,
seen from the figure, the measured values do approach the ex- predictions can be made directly from Fig. 2. For the required
pected equilibrium, and are quite well given by the appropriate. o and n, combinations of spacing (density) and height can be
equilibrium curve (X./k = 30) for the two smallest values of n selected from the figure and give the same results as those cal-
(i.e., the two largest valu~ of x / I.). We therefore conclude thn.t culated in the previous paragraph.
a roughness fetch of approximately 350 I. is required to reach The possibility of using graded roughness, closer spacing or
equilibrium, at lee.st for the upstream ccnditions present in this greater height at the start, is an obvious possibility. AB an ex-
experiment. tension of this idea, very large initial roughness elements in the
Another indication of the present boundary layer's approach from .'lf spires or wedges have been used to develop the boundary
to the calculated equilibrium condition is apparent in the values layer (reference [2] or [3]) with much smaller but uniform rough-
of the defect" law integral parameter, Ulo· / (U,O), discussed in ness being used downstream. The likelihood of the boundary
Section 4. Values of this ratio, deduced from mensured shel\r layer being close to equilibrium seems somewhat remote in these
and n, are given in Table J. for the present case, and range from cases however, unless u. very long fetch of roughness is used down.-
4.27 at the first data point (x/k ~ 141) to 3.92 at the last data stream of the spires. In Counihan's simulation oi the urban
point (x ~ 493). The equilibrium value of the ratio, as found boundary layer using wedges (reference [3]), the inttigral param-
from Head's method and the lISSumption of equilibrium is 3.99 eter Uto· /(U) at the point where equilibrium is assumed has a
for f.. / k = 30 and n ~ 0.26, showing that the pr~nt boundary value of over 4.50 wherenspresent calculations suggest that a
layer becomes very clo~.to the equilib~ium condition predicted 'value doser to 4.00 would repl'!l8ent equilibrium. More work
for it. needs to be done in this area, both to assess the degree of equilib-
From this limited comparisol1 of anticipated and measured rium present downstream of vadous spires or graded roughness
results, we conclude that Dvorak's correlation of shear stress with elements and to assess the importance of equilibdum for the
integral boundary ls~e~ properties and roughness geometry is testing being done.

484/ SEPTEMBE~ 1977 Transactions of the ASME

Acknowledgments 6 Rotta, J. C., Progrcss in Aero. Sciences, Vo!. 2, Ferri-Per-
gamon Press, 1962.
MilCh of the'work reported here was started while the first 7 I\fellor, G. L., and Gibson, D. 1\1., "Equilibrium Turbu-
author was on sabbaticallea\'e at the National Research Council lent Boundary Layers," Journal of Fluid Mechanics, Vol. 24,
1966, p. 22.5.
of Canada. Tfie cooperation and encouragement of those at the S Dvorak, F. A., "Calculation of Turbulent Boundary
N.n.C. is appreciated, as is the financial support from the Layers on Rough Surf aces in Pressure Gradient," A I A A Journal,
N.R.C. through grant A430S and from the Atm;)spheric En- Vol. 7, 1969, p. 1752.
vironmental Service through grant 810:3. 9 Clauser, F. H., "Turbulent Boundary Layers in Adverse
Pressure Gradients," Journal of the Aeronautical Sciences, Vol. 21,
1954, p. Ql.
10 Engineering Sciences Data Unit; Item No. 70015 "Fluid
References Forces and Moments on Flat Plates," October, 1972 and Item
No. 71016, "Fluid Forces, Pressures and Moments on'Rectangu-
1 Davenport, A. G., and Isyumov, N., The Application of lar Blocks," July, 1973.
the Boundary Layer Wind Tunnel to he Prediction of Wind 11 Gartshore, I. S., "A Relationship Between Roughness
Loading; Seminar, "Wind Effects on Buildings and Struclures;" Geometry and Velocity Profile Shape for Turbulent Boundary
Ottawa, Sept., 1967, Proceedings, Vol. I, p. 201. Layers," N.A.E., LTH-LA-140, 1973, National Research Council
2 Standen, N. M., "A Sp'ire Array for Generating Thick of Canada.
Turbulent Shear Llwers for Natural Wind Simulation in Wind 12 Antonia, R. A., and Luxton, R. E., "The Response of a
TUllnels," NHC (Cailada), NAE LTIt-LA-94, May 1972. Turbulent Boundary Layer to a Step Change in Surface Rough-
3 COllnihan, J., "The Structure and the Wind Tunnel Simu- ness," Journal of Fluid Mechanics, Vol. 48, 1971, p. 721.
lation of Hural and Urban Adiabatic Boundary Layers," Sym- 13 Counihan, J.; "Wind Tunnel Determination of the Rough-
posi1l1n of External Flows, University of Bristol, July 1972. nes~ Length as a Function of the Fetch and the Roughness Den-
4 Townsend, A. A., The Structure of Turbulent Shear Flow, sity of Three-Dimensional Elements," At71WSpheric Environ-
Cambridge University PresQ • 1956. ment, Vol. 5, 1971, p. 637.
!i Perry, A. E., Schofield, W. H., and Joubert, P. N., 14 Wooding, R. A., Bradley, E. F., and l\Iarshall, J. K.,
"Hough Wall Turbulent Boundary Layers," Journal of Fluid Drag Due to Regular Arrays of Roughness Elements of Varying
/If echanics, Vo!. 37, 1969, p. 383. Geometry," Boundary Layer Meteorology, Vol. 5, 1973, p. 285.

CALL FOR PAPERS

SYMPOSIUM ON FLUID TRANSIENTS

AND ACOUSTICS IN THE POWER INDUSTRY
American Society of Mechanical Engineers (ASME)
1978 Winter Annual Meeting

ASME's Fluids Engineering Division is sponsoring a symposium on Fluid 'Trllonsierits and

Acoustics in the Power Industry at the 1978 Winte, Annual Meeting in San Francisco, California,
December 10-15, 1978. Papers are requested in (bpt not limited to) the following areas: com-
pressible and incompressible unsteady flows; thermofluid transients; multi-phase flow transients;
transients in hydraulic machinery; modeling of devices for surge suppression; transients in rup-
turing lines; column separation; passive acollstic diagnosis; acoustics of cavitation; uIt.rasonic
detection and monitoring; plant noise control; component sound transmission; pump noise;
noise generation mechanisms; background noise; pipe noise; and valve noise.

Program Format

One to two days of paper presentations are planned. The symposium will commence with a
session of papers hroadly covering fluid tra;l.~ients fwd acoustics as related to the power industry.
The remaining sessions will be devoted to symposium papers on more specialized topics.

D.adllnes
Those wishing to participate should prepare 1,000 word abstracts for submission by February 1,
1978. Abstracts in the area of fluid transients Hhould be sent to Const.antine Papadakis, Bechtel
Inc., P. O. Box 1000, Ann Arbor, l\fichigan 48106, tel. (313)994-7179; and those in acoustics to
Henry Scarton, Mechanical Engineering Department, Hensselaer Polytechnic Institute, Troy,
New York 12181, tel. (518)270-6334. Selection or papen; will be made by February 28, 1978.
Completed papers should be submitted by May 31, 1978 on author-prepared mats which will be
sent to each author at the time of accept,lnce.

publication
Preprinting of the symposium papers on the AS:\I:E author-prepared mats will not impair
later consideration for publication in Transactions of ASME.

Journal of Fluids Engineering SE PTE M'8 ER 1977 I 485