Sie sind auf Seite 1von 9

Journal of Wind Engineering and Industrial Aerodynamics, 46 & 47 (1993) 145-153 ELsevier

145

Appropriate boundary conditions for computational wind engineering models using the k-E turbulence model
P.J. RICHARDS Senior Lecturer, Department of Mechanical Engineering, University of Auckland, Private Bag 92019, Auckland, New Zealand R.P. HOXEY Research Engineer, Silsoe Research Institute, Wrest Park, Silsoe, Bedford, MK45 4HS, England Abstract In computational models of wind engineering problems within the atmospheric surface layer the approach flow should normally be modelled as a homogeneous flow. Velocity and turbulence profiles associated with the k-e turbulence model are proposed which produce homogeneous conditions. These equations are discussed in the light of full-scale measurements at Silsoe using sonic anemometers. It is suggested that the model constants k = 0.42, C~t = 0.013 and oe = 3.22 more closely match the data obtained. It is also shown that the cospectrum for the Reynolds stress exhibits a characteristic frequency no ~ u,/z which is consistent with the suggested profile equations. 1. INTRODUCTION In 1958 Jensen [1] showed that in wind-tunnel testing it is just as important to correctly model the wind as it is to correctly model the building. This lesson must surely carry over into the relatively new field of computational wind engineering. For many wind engineering applications only the lower 200m or less of the atmospheric boundary layer is of interest. As a result it is this atmospheric surface layer that will be considered in this paper. Ludwig and Sundaram [2] set down the basic conditions for simulating the atmospheric surface layer as follows: "... regardless of the manner of its generation, any flow that is fully aerodynamically rough, horizontally homogeneous, and relatively free from any pressure gradients, constitutes a suitable model for the atmospheric surface layer." For computational modelling purposes "full aerodynamic roughness" implies that the shear stresses should be dominated by the Reynolds stresses. With the k-E turbulence model the Reynolds stress are modelled by using an effective turbulent viscosity ~tt such that [

_ u,w, is rep,ac

by

q,

z* ~w

(1)

where U and W are the mean streamwise and vertical velocities u' and w' are the corresponding velocity fluations, and gt is related to the turbulent kinetic energy k -- (u '2 + V,2 + w'2)/2

0167-6105t93/$06.00 1993 - Elsevier Science Publishers B.V. All fights reserved.

146 and its rate of dissipation e by ~tt = pC~t k2/e (2)

where Ctx is a model constant to be discussed later. Full aerodynamic roughness therefore occurs when ~t >> I-tlwhich is normally the case. Horizontal homogeneity is much harder to achieve. Such homogeneity, which can only exist in regions remote from any obstructions, implies that the streamwise gradients of all variables should be zero. Richards and Younis [3] in their comment on the computational modelling work of Matthews [4] point out that by using empirical equations (power law for velocity etc.) for the inflow boundary conditions he had created a situation which resulted in the approach flow changing rapidly in the inlet region of the solution domain. In particular the flow near the surface accelerated considerably before being retarded by the influence of the building. In order to avoid such problems it is essential that the inlet velocity and turbulence prof'des, the ground shear stress and the turbulence model should be in equilibrium. 2. A HOMOGENEOUS k-e MODEL FOR THE ATMOSPHERIC SURFACE LAYER In steady incompressible 2-dimensional flow modelling of the atmospheric surface layer with the k-e turbulence model the existence of homogeneous flow has the following implications: (i) The vertical velocity is zero (ii) The pressure is constant (iii) The shear stress is constant ie ([-tl + Bt) ~ U z = % = 9u, 2 is the surface shear stress is the air density is the friction velocity. (3)

where Xo O and u,

(iv) The turbulent kinetic energy k and its dissipation rate e satisfy their respective conservation equations which reduce to 3--~-/~tt ~k~ -pc = 0 Oz~ Ok Oz ] + I'ttGk
~ [['l't ~E~ E E2 anO~z-z~k-~e-e~---zj + C 11.l.tGk~--C2P-- ~- = 0

(4)

(5)

where

Gk = \-3-f-z]

and o k, (Ye' C 1 and C2 are model constants to be discussed later. All of these may be satisfied by using
U = -~-ln l , ' - - ~ o ]

u. (Z+Zo

(6)

147

U,

2 (7)

3 I1, (8) and e - - K(z + z o) where K is von Karman's constant (-- 0.4) and Zo is surface roughness length. Equations (6)-(8) automatically satisfy equation (4) but they only satisfy equation (5) if

(z + Zo)2 or (~ -

Oe
K2

K2

(9)

(C 2 - Cl)

(10)

In early work on the k-e model Launder and Spalding [5] suggested CI = 1.44, C2 = 1.92 and C~t = 0.09 which with K = 0.4 equation (10) would give ae = 1.11. In the CFD code PHOENICS [6] the above values for C1, C2 and C~t are used in conjunction with K = 0.435 and (~e = 1.314 which also satisfy equation (10). 3. ATMOSPHERIC SURFACE LAYER MEASUREMENTS AT SILSOE Wind velocity measurements at the Silsoe Research Institute in September/October of 1988 using direction pitot tubes up to a height of 25m and tethered kites at heights up to 100m showed that the mean streamwise velocity profile was, as shown in Figure 1, well fitted to a log-law with a gound roughness length of 10ram [7]. 1000 100 10 N 1 .1 .01 0.0 Figure 1. , 0.2 , 0.4

0.6 Uz/U10 Velocity profile at Silsoe in 1988

0.8

1.0

1.2

1.4

The average mean wind speed at a height of 10m was 10.3 m/s. This implies that the friction velocity was u. -- 0.6 rn/s and hence at z = 10m Ixt ~-pKu.z = 2.94 Ns/m 2 whereas l.tl -:--1.71 x 10 -5 Ns/m 2

148 Even at z--_ Zo, I.tt = 2.9 x 10 -3 so t~t]l.tI = 172. More detailed information was obtained on the same site in December 1991 when two ultrasonic, 3-component anemometers were used. These anemometers, manufactured by Gill Instruments Ltd, are marketed as solent ultrasonic anemometers. The instrument has a path length of 150mm and produces digital wind speed information at a frequency of 20.8 Hz. In this study one of the sonic anemometers was mounted at a reference height of 10m, while the other anemometer was positioned for 3-hour periods at 1.01 m and 0.115m. The 3-hour runs were processed as six 26.25 - minute records, each consisting of 32,768 data points. The results presented are for averages of 6 x 26.25 minute non overlapping records. Spectral analysis of this data has been carded out in order to provide the three spectral density functions Suu(n), Svv(n) and Sww(n). In addition the cross-spectral density function Suw(n) has been computed from the fourier transforms of the streamwise velocity U(n) and vertical velocity W(n) as Suw(n) = Cuw(n)-iQuw(n) = U(n).W*(n) (11)

The real part of this, Cuw(n) the coincident spectral density function or cospectrum, is the only part that contributes to the Reynolds stress u'w' and hence Cuw(n) may be viewed as the spectral density function of the Reynolds stress. Figure 2 shows the typical shapes for these spectral density functions. - Cuw(n) has been plotted since its value is generally negative. Integration of the cospectrum gave good agreement with estimates for u'w' obtained by averaging the product of the instantaneous components. 2.5 2.0 r..)

1.5 1.0 0.5

.........
.........

nSuu/(u*.u*~

nsvv/(u*.u*~

nSww/(u*.u*~

0.0 -=,'~.r'-~.'".'7".".~'., .001 Figure 2.

........

.-.'7.

.-.~

--~-

......

.01 n (Hz)

.1

10

Typical spectral density functions at z=10m

From Figure 2 it can be seen that the measurement of u'w' at a height of 10m is not affected by cut off. The von Karman constant was therefore computed from the average of 7 hours of recordings where the mean value of 1 u, (- u'w' )2 - 0.0607 + 0.0017 (12)

UIO

UIO

and from the mean velocity data the profile has a ground roughness length Zo = O.OlOm by a least squares fit. Hence

149
u, 1

KU10 - ln(10/Zo + 1.0) = 0.1447 and K = 0.42+0.01.

(13) (14) energy

Also apparent from Figure 2 is the relatively high ratio of turbulent kinetic

(k = 0.5(Ou 2 + Ov2 + Ow2)) to the Reynolds stress "(u'w' = - u,2). With the results plotted as nS(n)/u, 2 versus log(n) curves the areas under the curves are indicative of the size of the variances (02) and the area under the cospectrum is one unit of area since
-oo

SOnCuwdln(n) = u'w'

= -u 2 .

(15)

These ratios are compared with the turbulent boundary layer on a flat plot data of Klebanoff [8] and other wind engineering data in Table 1. Table 1. Surface layer turbulence level Source Klebanoff [8] (Y/8 = 0.1) Panofsky & Dutton [9] Hagen et al. [10] ESDU [11] (z = 10m, zo = 0.01) Silsoe Ou2/U,2 3.60 5.71 7.6 9.65 Ov2/U,2 2.01 3.68 4.62 6.11 Ow2/U,2 1.09 1.56 2.3 1.73 k/u, 2 3.35 5.48 6.2 7.26 8.75

It is clear from Table 1 that all the atmospheric surface layer measurements show k/u, 2 > 3.33 which is the value given by equation (7) with C~t = 0.09. In their computational modelling work Hagenet al. recognised this observation and reduced C~t to 0.026. However, they also used C1 = 1.44, C2 = 1.92, oe = 1.22 and K = 0.4 which do not satisfy equation (10). Beljaars et al. [12] similarly used Panofsky and Dutton's [9] flat terrain data to justify reducing Cla to 0.032 but also argued that oe should be adjusted in order to satisfy equation (10). The possible cause for the higher ratios of k/u, 2 in the atmospheric surface layer may lie in the significant low frequency contribution to Ou2 and Ov2 which contribute little or nothing to the Reynolds stress. The use of more responsive instrumentation may also be partially responsible for higher values in recent measurements. For CFD modelling of the surface layer at Silsoe it appears that appropriate constants are: C~t = 0.013, K = 0.42 and oe = 3.22. Although the Silsoe data brings into question some of the constants used in the k-e model it does lend some support to the form of equations (7) and (8). From these it may be argued that we would expect the Reynolds stress to exhibit a characteristic frequency (say no) for which dimensional analysis suggests
U~

n o o~ ~ ie

no ~ K(z + Zo)

(16)

As a result Figure 3 shows the simultaneously recorded cospectra at heights of 1.01m and 10m plotted as -nCuw/U, 2 versus log (nKz/u,). The resulting curves are very similar with the central frequency

150

no--

0.4 u, K~-

u, z

(17)

The cospectrum appears to be normally distributed about no and may be reasonably fitted by -nCuw(n) 1 2 - ~7 U, with an = 1.7.
0.5. . . . . . .

( exp

(ln(n)-ln(no)) 2 ) 20~n

(18)

1-01 10 Eq18

tlIv 0
I
....

10-~
Figure 3.
10 -~

i
10 -~ I0 o I0' 102

nkz/u*
Normalised Cospectra The primary imlications of Figure 3 are that it shows that the significant frequencies which contribute to the Reynolds stress scale inversely with height. Also frequencies an order of magnitude larger or smaller than the central frequency are equally less significant. An alternative interpretation of the normalised frequency nKz/u, is that this is approximately equivalent to the ratio of frequency to the velocity gradient since with a logarithmic boundary layer n o~ nK(z + Zo) u. (19)

151 4. APPROPRIATE BOUNDARY CONDITIONS In computational models of wind engineering problems the boundary conditions used should be capable of producing a honogeneous boundary layer flow in the absence of the object of interest, usually a building. To do this the boundaries should be located sufficiently remote from the object so that they have negligible effect in the region of interest. Further for flow through a rectangular domain with the inlet through one face the inlet boundary should provide velocity, k and e profiles u, / z + Zo/ -K - In \ ~ j
U, 2
k

U = =

(20)

a/r~ ~
,tt *

(21)

U, 2

= K(z + Zo)

(22)

where the friction velocity is usually calculated from a specified velocity Uh at a reference height h as
U,

K Uh z,

/
/

(23)

Within the surface layer the flow is partially maintained through interaction with the flow at higher levels. Hence at the top of the domain a constant shear stress 'c = pu, 2 should be applied in the streamwise direction. At the ground boundary a retarding shear stess will exist, but in order to allow this to adapt to changes caused by a building or other obstruction this should be calculated on a local basis. If a velocity Ug is the near ground velocity at a height Zg then a local friction velocity can be deduced from
U,g

KU 8
- -

Zo

and a retarding shear stress 3, = pu** 2 (25)

applied. If this stress is assumed to apply across a near ground cell of height 2Zg and plan area A then the rate of doing work on this cell is given by
u*g 3A K Ink

~sU(2g)A "~ P

12zs +z.)
z~

(26)

which is equivalent to the rate of transfer of energy from the mean flow to turbulent kinetic energy.

152 Also from equation (16)


E~

kno
U,

i c e ,~ k ( z + Zo) or to be consistent with equations (7) and (8) k U,g ' ' K(z + Zo)

(27)

e =

(28)

which when integrated over the near ground cell gives a nett dissipation rate paf-~K A k u * g l n / 2 z 8 ) + o z z (29)

Peoel~

It is suggested that equations (26) and (29) should be used in the conservation equation for k in the near ground cell, in which case generation will equal dissipation when

k =

U, 2 ~u

(30)

The near ground dissipation rate at the cell centre should be specified as ~-U-u k u, e = g K(z 8 + Zo) (31)

These boundary conditions if properly applied will lead to an equilibrium boundary layer with no pressure gradient in regions remote from any obstructions. 5. CONCLUSIONS In order to adequately model the atmospheric surface layer the boundary conditions, turbulence model and associated constants must be consistent with each other. A suitable set of boundary conditions for use with the k-E turbulence model has been proposed. Full scale measurement at Silsoe have led to an evaluation of yon Karman's constant K = 0.42 + 0.01. These measurements also suggest C~t = 0.013, however such a change would also require oe to be increased to 3.22 in order to be consistent. This set of constant has not been rigorously tested, but is presented to highlight the differences.

153 6. REFERENCES M. Jensen, The model law for phenomena in the natural wind, Ingenioren, 2 (1958) 121. G.R. Ludwig and T.R. Sundaram, On the laboratory simulation of small-scale atmospheric turbulence, CAL Report No VC-2740-S-1, 1969. 3 P.J. Richards and B.A. Younis, Comments on "Prediction of the Wind-generated Pressure Distribution around Buildings" by E.H. Mathews, J. Wind Wng. Ind. Aerodyn., 34 (1990) 107-110. 4 E.H. Mathews, Prediction of the wind-generated pressure distribution around buildings, J. Wind Eng. Ind. Aerodyn., 25 (1987) 219-228. 5 B.E. Launder and D.B. Spalding, The numerical computation of turbulent flows, Comput. Methods Appl. Mech. Eng., 3 (1974) 269-289. 6 PHOENICSReference Manual TR200, CHAM, London, 1991. 7 P.J. Richards and R.P. Hoxey, Computational and wind-tunnel modelling of mean wind loads on the Silsoe Structure Building, Proc. 8th Int. Conf. on Wind Eng., University of Western Ontario, Canada, July 1991. 8 P.S. Klebanoff, Characteristics of turbulence in a boundary layer with zero pressure gradient, NACA Rep. 1247 (1955). 9 H.A. Panofsky and J.A. Dutton, Atmospheric Turbulence, Wiley-interscience, 1984. 10 L.J. Hagen, E.L. Skidmore, P.L. Miller and J.E. Kipp, Simulation of effect of wind barriers on airflow, Trans ASAE, 24 (1981) 1002-1008. 11 ESDU 85020, Characteristics of atmospheric turbulence near the ground, ESDU, 1985. 12 A.C.M. Beljaars, J.L. Walmsley and P.A. Taylor, A mixed spectral finite-difference model for neutrally stratified boundary-layer flow over roughness changes and topography, Boundary Layer Meteorology 38 (1987) 273-303. 1 2

Das könnte Ihnen auch gefallen