Professional Documents
Culture Documents
a r t i c l e in fo abstract
Article history: Numerical simulation of turbulent ows past typical truss structures that are common in bridges and
Received 9 May 2008 towers has been conducted. Large-eddy simulation (LES) method is used that has been veried in
Received in revised form computing ows past simple bodies at moderate Reynolds numbers. Truss structures consist of thin
5 October 2009
members that are too small to be accurately resolved by numerical grids that can be handled by most
Accepted 8 October 2009
computer systems. It has been found that when the individual members are long angular bars, the
Available online 10 November 2009
overall ow can be reproduced fairly well by approximating the members with rectangular cylinders
Keywords: whose cross sections are resolved by more than 2 2 computational cells and by taking computational
LES regions extending at least four structure heights downstream and about three heights across the
Truss structure
vertical and spanwise directions. The drag coefcient and the vortex shedding characteristics along with
Drag coefcient
the complex wake structures associated with the truss structures are reproduced reasonably well.
Strouhal number
Rectangular cylinders & 2009 Elsevier Ltd. All rights reserved.
Wake
0167-6105/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jweia.2009.10.007
ARTICLE IN PRESS
134 A. Nakayama et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 133144
Table 1
Congurations for simulation.
Case Conguration
3. Calculation method
Table 2 Table 3
Elements of the calculation method. Summary of test cases with individual members.
Grid Variable interval rectangular grid Case Orientation Aspect Reynolds Number of Total number
(deg) ratio number cells in cross of grid points
Variable arrangement Staggered mesh b/d based on d section (Nx Ny Nz)
Pressure coupling HSMAC Red =Ud/v (Mz Mx)
Convective terms 3rd-order upwind difference
Time advancing 2nd-order AdamsBashforth M00102 0.0 1.0 25,000 22 128 32 96
Subgrid model Standard Smagorinsky with Cs =0.13 M00104 0.0 1.0 50,000 44 128 32 96
Wall model No model M00106 0.0 1.0 75,000 66 128 32 96
M00152 0.0 1.5 25,000 23 128 32 96
M00154 0.0 1.5 50,000 46 128 32 96
M00202 0.0 2.0 25,000 24 128 32 96
M00204 0.0 2.0 50,000 48 128 32 96
M45142 45.0 1.4 35,000 22 128 32 96
M45143 45.0 1.4 44,000 33 128 32 96
M60153 60.0 1.5 37,500 22 128 32 96
M60202 60.0 2.0 50,000 33 128 32 96
Fig. 5. Instantaneous wake structures of cylindrical members: (a) M00102, (b) M00104, (c) M00152, (d) M00154, (e) M45142 and (f) M45202.
Table 4 Table 5
Results of calculation, mean drag coefcient and Strouhal number. Calculation cases of plane truss structures.
Case Orientation Calculated results Experimental values Cases Filling Main Rey- Number Total number Block-
(deg) ratio j frame nolds of cells of grid points age
CD Std CD Std mem- number in main (Nx Ny Nz) ratio
ber b/d based on member (%)
M00102 0.0 2.21 0.083 2.07 0.3 0.13 7 0.01 D ReD = cross
M00104 0.0 1.78 0.143 2.07 0.3 0.13 7 0.01 UD/v section
M00106 0.0 1.75 0.146 2.07 0.3 0.13 7 0.01 (Mz Mx)
M00152 0.0 1.23 0.094 1.77 0.1 0.11 7 0.01
M00154 0.0 1.58 0.128 1.77 0.1 0.11 7 0.01 A2d3 0.48 1.0 4.0 105 33 128 32 96 14.6
M00202 0.0 1.25 0.094 1.57 0.2 0.087 0.01 or B2d2 0.18 1.0 4.0 105 22 128 32 96 5.5
0.12 7 0.01 B2d3 0.34 1.0 4.0 105 33 128 32 96 10.4
M00204 0.0 1.43 0.128 1.57 0.2 C2d3 0.39 1.0 4.0 105 33 128 32 96 11.9
M45142 45.0 1.46 0.080 1.77 0.1 0.11 7 0.01
M45143 45.0 1.61 0.099 1.77 0.1 0.11 7 0.01
M60153 60.0 1.53 0.086 1.57 0.2 0.11 7 0.01
M60202 60.0 1.34 0.054 1.57 0.2 0.087 0.01 or
0.12 7 0.01 The calculation results of the drag coefcient for these plane
trusses are shown in Fig. 7. Since the plane structure is more like
porous screens, the results are compared with that of the
regions away from the joints. The Reynolds numbers based on the empirical design criteria given by the AIJ (2004) for porous
height D of the structure are 4.0 105 which corresponds to the structures. It is seen that the present results are close to the AIJ
Reynolds number obtained at typical wind tunnel tests. In terms criteria and the trend of the effects of the lling ratio is consistent
of the width d of the individual member is about 2.5 104 for with the AIJ line. The drag coefcient of case C2d is largest and
which validation calculations of individual members are done. appears to be due to the effects of the large joint areas.
ARTICLE IN PRESS
138 A. Nakayama et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 133144
Fig. 6. Computational grids used for computing ow past plane truss structures: (a) A2d3, (b) B2d3 and (c) C2d3.
Fig. 7. Drag coefcient of plane truss structures. Fig. 8. Instantaneous drag coefcient of plane truss structures.
The time variation of the instantaneous drag coefcient is shown though not shown here. This is due to the fact that vortex sheddings
in Fig. 8. There are uctuations but the magnitude is smaller than off members of different sizes and orientations have different
those of the instantaneous drag coefcients of individual members, frequencies and phases and tend to cancel each other out.
ARTICLE IN PRESS
A. Nakayama et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 133144 139
Table 6
Calculation cases of three-dimensional truss structures.
Cases B/D L/D Filling ratio j Main member Reynolds number Total number of grid
thickness b/D based on D ReD = UD/v points (Nx Ny Nz)
Fig. 11. Computational grids used for computing ow past three-dimensional truss structures: (a) A3d2, (b) B3d2 and (c) C3d3.
ratio j given by Scruton and Newberry (1963), Pagon (1958), and members crossing each other. The short-span case A3d2 and
AIJ design standard (2004). It is seen that the calculated values are wide-span case A3d2w show little difference implying the basic
generally in agreement with these empirical relations, although span taken for the other short span cases is sufcient.
the trend with respect to j of cases A3d and C3d are opposite of Time uctuations of the instantaneous drag coefcients for the
the correlations. In case C3d, it may be due to large diagonal three different truss congurations with the member cross section
ARTICLE IN PRESS
A. Nakayama et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 133144 141
resolved by 2 2 cells are shown in Fig. 13. It is seen that the drag
coefcients of these three-dimensional trusses uctuate in time
considerably more compared with the drags on plane truss
structures shown in Fig. 8. This is due to the effects of the
wakes of the upstream members inuencing the downstream
ones which do not exist in plane trusses. In order to see this, time
uctuations of the drag forces on individual members are shown
in Fig. 14. cDm is the contribution to the total drag coefcient from
individual members. The magnitudes of the drag forces on the
downstream members are signicantly smaller than those on the
upstream members but the amplitudes of the uctuations are an Fig. 16. Instantaneous lift force on different members of case A3d2: (a) upwind
order of magnitude larger. It is also seen that the frequencies of members, (b) downwind members and (c) connecting members.
ARTICLE IN PRESS
142 A. Nakayama et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 133144
beating harmonic. The case C3d2 which has the most diagonal
members shows smallest uctuations. Also the dominant
uctuations of A3d2 have larger periods than those of B3d2 and
C3d2. Details of spectral characteristics of uctuating lift
coefcient are discussed later.
Fig. 16 shows the time uctuations of the lift forces on
individual members of case A3d2. It is seen that the downstream
members receive an order of magnitude larger lift forces than the
upstream members, and the connecting members get yet smaller
lift than the upstream ones. Larger uctuations of cLm of
downstream members start only sometime after the calculation
is started. It implies that they are caused by the wakes and
vortices shed by the upstream members that take time to develop
and convect downstream to make an inuence. There are not
appreciable differences between the magnitudes of the lift forces
Fig. 17. Power spectra of uctuating total lift forces.
on horizontal and diagonal members.
The power spectra of the instantaneous lift coefcients are
shown in Fig. 17. There are two distinct peaks in case of A3d2 but
one broad one in case B3d2 and no clear peaks in case of C3d2.
The peak at about fD/U= 0.2 seen in case A3d2 appears the
oscillation due to ow structures of size comparable to D but not
quite low enough for vortex shedding off solid body of size
D. Since d/D= 0.062 for all these cases, this and other peaks in the
other cases are much lower than the Strouhal frequencies of the
individual members rather than their own wakes. This implies
that the uctuations on the lift that mainly come from the
unsteady ow past the downstream members are induced by the
wake of the upstream members. The upstream members do shed
vortices at their characteristic frequencies related to their
thicknesses but the stronger unsteadiness of the lift forces
comes from the undulation of the wakes of the upstream
members on the scale of the whole structure. These will be seen
better in instantaneous ow pictures discussed below.
List of symbols
References Murakami, S., Mochida, A., 1995. On turbulent vortex shedding ow past 2D square
cylinder predicted by CFD. Journal of Wind Engineering and Industrial
Aerodynamics 5455, 191211.
Architectural Institute of Japan, 2004. Recommendations for loads on buildings, Nakaguchi, H., Arai, I., Matsuzaka, M., 1964. Static wind loads on towers of frame
Maruzen, Japan. work structure. Journal of Aeronautical Society of Japan 12 (121), 18.
Biggs, J.M., 1954. Wind loads on truss bridges. Transactions of the American Society Nakayama, A., Noda, H., 2000. LES simulation of ow around a bluff body tted
of Civil Engineers 119, 879898. with a splitter plate. Journal of Wind Engineering and Industrial Aerodynamics
Constantinides, Y., Oakley, O.H. Jr., Holmes, S., 2006. Analysis of turbulent ows
7778, 643651.
and VIV of truss spar risers. In: Proceedings of the 25th International
Nakayama, A., Vengadesan, S.N., 2002. On the inuence of numerical schemes and
Conference on Offshore Mechanics and Arctic Engineering, OMAE2006-92674.
subgrid-stress models on large-eddy simulation of turbulent ow past a square
Fadlun, E.A., Verzicco, R., Orlandi, P., Mohd-Yusof, J., 2000. Combined immersed-
cylinder. International Journal of Numerical Methods in Fluids 38 (3), 227253.
boundary nite-difference methods for three-dimensional complex ow
Okajima, A., 1983. Flow around rectangular cylinders of various breath to depth
simulations. Journal of Computational Physics 161, 3560.
ratio. Journal of Wind Engineering 17, 119.
Grinstein, F.F., Margolin, L.G., Rider, W.J. (Eds.), 2007. Implicit Large Eddy
Ower, E., 1948. The wind resistance of lattice girder bridges. Institution of Civil
Simulation: Computing Turbulent Fluid Dynamics. Cambridge University
Engineers, vol. R7556.
Press, Cambridge.
Pagon, W.W., 1958. Wind forces on structures. In: Proceedings of the A.S.C.E.
Hirt, C.W., Cook, J.L., 1972. Calculating three-dimensional ow around structure
Structural, vol. 84(4), p. 1711. Journal of the Structural division.
and over rough terrain. Journal of Computational Physics 10, 324340.
Piomelli, U., Balaras, E., 2002. Wall-layer models for large-eddy simulations.
Ikeda, T., Durbin, P.A., 2002. Direct simulations of a rough wall channel ow. Report
TF-81, Stanford University, Mechanical Engineering. Annual Review of Fluid Mechanics 34, 349374.
Japan Road Association, 1991. Handbook of wind loads on road bridges, Maruzen, Pope, S.B., 2002. Turbulence 2002. Cambridge University Press, Cambridge.
Japan, pp. 5461. Rodi, W., 1997. Comparisons of LES and RANS calculations of the ow around bluff
Kuroda, M., Tamura, T., Suzuki, M., 2007. Applicability of LES to the turbulent wake bodies. Journal of Wind Engineering and Industrial Aerodynamics 69, 5575.
of arectangular cylindercomparison with PIV data. Journal of Wind Rodi, W., 1998. Large eddy simulation of the ow past bluff bodies, State-of-the-art.
Engineering and Industrial Aerodynamics 95, 12421258. JSME International Journal Series B, Fluids and Thermal Engineering 41,
Kuwahara, K., 1999. Rearrangement of Karman vortex street. In: Proceedings of the 361374.
13th Symposium of Numerical Fluid Dynamics, JSFM, Tokyo. pp. 227228. Sagaut, P., 2006. Large Eddy Simulation for Incompressible Flows, third ed Springer,
Launder, B., Sandham, N., 2002. Closure Strategies for Turbulence and Transitional Berlin Heidelberg.
Flows. Cambridge University Press, Cambridge. Scruton, C., Newberry, C.W., 1963. On the estimation of wind loads for building and
Leschziner, M. A., 2007. Statistical modelling and simulation of turbulent separated structural design. Proceedings of the Institution of Civil Engineers 25, 97126.
ows: limits, hard lessons and symbiotic relationships. In: Proceedings of the Selvam, R.P., Tarini, M.J., Larsen, A., 1998. Computer modeling of ow around
JSFM Annual Symposium, Paper AM07-00-000. bridges using LES and FEM. Journal of Wind Engineering and Industrial
Lubcke, H., Schmidt, S., Rung, T., Thiele, F., 2001. Comparison of LES and RANS in Aerodynamics 7778, 643651.
bluff-body ows. Journal of Wind Engineering and Industrial Aerodynamics 89, The Japanese Electrotecnical Committee, 1979. Design standard on structures for
14711485. transmissions, JEC-127-1979, Denkishoin, Japan.
Miyake, Y., Tsujimoto, K., Agata, Y., 2000. A DNS of a turbulent ow in a rough-wall Washizu, K., Ohya, A., Otsuki, Y., Fujii, K., 1978. Aeroelastic instability of
channel using roughness elements model. JSME International Journal 43 (2), rectangular cylinders in a heaving mode. Journal of Sound and Vibration 59
232242. (2), 195210.