28.05.2013 Views

Use of CFD in Civil Engineering (PDF) - SOFiSTiK AG

Use of CFD in Civil Engineering (PDF) - SOFiSTiK AG

Use of CFD in Civil Engineering (PDF) - SOFiSTiK AG

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Use</strong> <strong>of</strong> Computational Fluid Dynamics <strong>in</strong> <strong>Civil</strong> Eng<strong>in</strong>eer<strong>in</strong>g<br />

Pr<strong>of</strong>. Dr.-Ing. Casimir Katz, <strong>SOFiSTiK</strong> <strong>AG</strong>, Oberschleißheim<br />

ir. Henk Krüs, Cyclone Fluid Dynamics BV, Waalre<br />

Zusammenfassung:<br />

Der E<strong>in</strong>satz von <strong>CFD</strong> im Bauwesen stellt nach wie vor e<strong>in</strong>e Nischenanwendung dar. Die Gründe<br />

dafür werden beleuchtet und aufgezeigt, warum sich das jetzt ändern wird.<br />

Summary:<br />

The use <strong>of</strong> <strong>CFD</strong> <strong>in</strong> the civil eng<strong>in</strong>eer<strong>in</strong>g community is still a rare event. The reasons for that will be<br />

discussed and it will be po<strong>in</strong>ted out why that will change now.<br />

1 INTRODUCTION<br />

1.1 Fluid dynamics and their applications<br />

<strong>CFD</strong> is the acronym for „Computational Fluid Dynamics“. Fluids <strong>in</strong> civil eng<strong>in</strong>eer<strong>in</strong>g are mostly<br />

air and water and the questions to be answered are the forces <strong>in</strong>duced by fluid motion and the<br />

transport <strong>of</strong> heat or particles with<strong>in</strong> the fluid. Typical questions are<br />

W<strong>in</strong>d load<strong>in</strong>g on bluff bodies<br />

W<strong>in</strong>d load<strong>in</strong>g on mov<strong>in</strong>g bodies, especially bridges<br />

W<strong>in</strong>d comfort and nuisance to cyclists and pedestrians<br />

W<strong>in</strong>d energy<br />

Heat<strong>in</strong>g and Ventilation (HVAC)<br />

Fire safety eng<strong>in</strong>eer<strong>in</strong>g<br />

Wave load<strong>in</strong>gs<br />

The difference to classical static or dynamic analysis is given by a different mathematical treatment<br />

and thus different mathematical tools [1]. While structural mechanics use the Lagrangian approach<br />

based on displacements, fluid mechanics prefer the Eulerian approach based on the velocity <strong>of</strong> the<br />

fluid. Structural mechanics use the F<strong>in</strong>ite Element Method, most fluid mechanics use the F<strong>in</strong>ite<br />

Volume Method. The basic pr<strong>in</strong>ciples <strong>of</strong> equilibrium and conservation <strong>of</strong> mass, energy and<br />

momentum are common to both methods.<br />

V13 - 1


1.2 S<strong>of</strong>tware for <strong>CFD</strong><br />

S<strong>of</strong>tware for <strong>CFD</strong> has been developed for a long time now, the most well known products today are<br />

the big three FLUENT (now Ansys), STAR-CD/STAR-CCM (CD-Adapco) and CFX (now Ansys),<br />

but there are many more like the <strong>CFD</strong>DRC (now ESI-Group), <strong>CFD</strong>esign (FEM by Blue Ridge<br />

Numerics, now Autodesk) and FDS (by NIST). There are also different techniques available like<br />

vortex particle and Lattice-Boltzmann methods.<br />

The first impression for a structural eng<strong>in</strong>eer is, that everyth<strong>in</strong>g is very complex, that there are<br />

hundreds <strong>of</strong> features and parameters and that it will cost a fortune to start <strong>in</strong>to this field. Over the<br />

years the mantra <strong>of</strong> a technique much too complex for the common eng<strong>in</strong>eer and high licence costs<br />

has established significant barriers. The same happened to structural FE-S<strong>of</strong>tware <strong>in</strong> the late<br />

seventies. <strong>SOFiSTiK</strong> had success because we anticipated the wide spread <strong>of</strong> that technique on<br />

personal computers <strong>in</strong> 1980.<br />

But the <strong>CFD</strong>-market is chang<strong>in</strong>g now, cheaper versions <strong>of</strong> <strong>CFD</strong>-s<strong>of</strong>tware enter the market, and the<br />

open source s<strong>of</strong>tware OpenFOAM has ga<strong>in</strong>ed wide acceptance especially <strong>in</strong> the academic world.<br />

However the structural beg<strong>in</strong>ner is still overwhelmed by a wide range <strong>of</strong> features. What is needed is<br />

a robust, easy to use entry po<strong>in</strong>t for this journey.<br />

<strong>SOFiSTiK</strong> has ga<strong>in</strong>ed some experience with an academic multiphysics s<strong>of</strong>tware PHYSICA and is<br />

now support<strong>in</strong>g DOLFYN, an open source <strong>CFD</strong>-Solver used <strong>in</strong> practice <strong>in</strong> many eng<strong>in</strong>eer<strong>in</strong>g fields,<br />

which has been fully <strong>in</strong>tegrated <strong>in</strong> the <strong>SOFiSTiK</strong> environment.<br />

2 BASICS<br />

2.1 Fluid dynamics and their solvers<br />

Two important metrics are the Mach number (v/c) def<strong>in</strong><strong>in</strong>g the ratio <strong>of</strong> the fluid velocity v to the<br />

speed <strong>of</strong> the sound c and the Reynolds number derived from the fluid velocity u, the k<strong>in</strong>ematic<br />

viscosity n, and a characteristic dimension <strong>of</strong> the structure (like the diameter <strong>of</strong> a cyl<strong>in</strong>der or the<br />

height <strong>of</strong> a bridge deck):<br />

d u<br />

Re <br />

<br />

k<strong>in</strong>ematic viscosity<br />

d characteristicdimension<br />

Air<br />

2<br />

6 m<br />

1510 sec<br />

; Water<br />

2<br />

6<br />

m<br />

1. 3110 sec<br />

V13 - 2


Fluid dynamics cover a broad range <strong>of</strong> fields, some <strong>of</strong> them are:<br />

Potential Flow (Re = ∞ )<br />

Creep<strong>in</strong>g Flow (hardly any flow)<br />

Lam<strong>in</strong>ar <strong>in</strong>compressible viscous flow<br />

(Navier Stokes)<br />

Turbulent Flow (Re large)<br />

Compressible Flow (Ma > 0.3)<br />

Supersonic Flow (Ma > 1.0)<br />

Thermal Effects<br />

Combustion & chemical reaction<br />

Free Surfaces<br />

Non-Newtonian fluids<br />

Radiation<br />

In civil eng<strong>in</strong>eer<strong>in</strong>g applications the subsonic <strong>in</strong>compressible turbulent flow is the most common<br />

phenomenon. Compressible flow is needed for shocks and high temperature effects. The major<br />

rema<strong>in</strong><strong>in</strong>g problem is that a direct solution <strong>of</strong> the Navier-Stokes-equations is only possible for<br />

Reynolds-numbers up to approximately 20 000, while practical examples are <strong>in</strong> the range <strong>of</strong> several<br />

millions. Thus turbulence has to be modelled by RANS (Reynolds-Average-Navier-Stokes) models<br />

allow<strong>in</strong>g for very large Reynolds numbers but do not model all effects, that’s why LES (Large Eddy<br />

Simulation) has ga<strong>in</strong>ed some popularity, but requires still high computational effort.<br />

2.2 Materialparameters<br />

The selection <strong>of</strong> fluid material parameters is straight forward: There is a density [kg/m³], a dynamic<br />

viscosity [Pa sec], a compressibility [Pa/m²] and some thermal properties.<br />

2.3 Boundary Conditions<br />

Inflow and outflow boundary conditions are <strong>in</strong> general not complex, but there is a major problem<br />

with two aspects. There exists a boundary layer at every wall. At the wall itself there is no flow,<br />

then we have a t<strong>in</strong>y lam<strong>in</strong>ar viscous sublayer, followed by the turbulent boundary layer. The<br />

treatment <strong>of</strong> the wall boundary condition is quite difficult. Though there are possibilities for near<br />

wall models, the common model is a logarithmic wall law describ<strong>in</strong>g the complete wall boundary<br />

behaviour for a cell sufficiently far away from the boundary:<br />

V13 - 3


However the atmospheric boundary layer has a height larger than one kilometre, thus all civil<br />

eng<strong>in</strong>eer<strong>in</strong>g structures are completely encompassed by it. The design codes for w<strong>in</strong>d load<strong>in</strong>gs<br />

describe the layer with a logarithmic or an exponential law.<br />

It is not only the velocity but also the turbulence characteristics like the k<strong>in</strong>etic energy (turbulence<br />

<strong>in</strong>tensity) and the dissipation rate (Integral length scale) which need to be described. The correct<br />

formulation requires not only to model the roughness <strong>of</strong> the ground with the correct value <strong>of</strong> z0, the<br />

driv<strong>in</strong>g force at the top <strong>of</strong> the fluid doma<strong>in</strong>, but also the treatment <strong>of</strong> the analytic solution <strong>of</strong> the<br />

turbulence equations as <strong>in</strong>itial conditions, which are fully implemented <strong>in</strong> DOLFYN:<br />

*<br />

u ABL zz 0<br />

uz ( ) ln<br />

<br />

z0<br />

<br />

* 2<br />

uABL<br />

kz ( ) <br />

C<br />

<br />

u<br />

( z)<br />

<br />

<br />

V13 - 4<br />

* 3<br />

ABL<br />

zz <br />

0


3 MODELLING<br />

3.1 W<strong>in</strong>dtunnel or <strong>CFD</strong> ?<br />

A w<strong>in</strong>d tunnel is just a model <strong>of</strong> the reality, so is any <strong>CFD</strong> model. The w<strong>in</strong>d tunnel needs scal<strong>in</strong>g,<br />

Reynolds number is not the same and there are cases where this does matter. There is a guidel<strong>in</strong>e<br />

from the WTG (W<strong>in</strong>dtechnologische Gesellschaft) describ<strong>in</strong>g <strong>in</strong> detail how to perform reliable<br />

tests.<br />

But for flexible structures, the measurement equipment may change the effects considerably. On the<br />

other side there are known deficiencies <strong>of</strong> numerical analysis, which do not allow tak<strong>in</strong>g all results<br />

for granted. “The purpose <strong>of</strong> comput<strong>in</strong>g is <strong>in</strong>sight, not numbers.” said R. Hamm<strong>in</strong>g <strong>in</strong> 1962. So the<br />

question is not which technique to be used but how to comb<strong>in</strong>e both methods to their best use.<br />

3.2 Reality or design case ?<br />

Is the requested result the mean values <strong>of</strong> the w<strong>in</strong>d load<strong>in</strong>g for a static analysis or the variation <strong>of</strong><br />

forces <strong>in</strong> time either to account for dynamic effects or to get reasonable loads at all. For example a<br />

flat horizontal ro<strong>of</strong> on columns will have a zero pressure as mean value, but the w<strong>in</strong>d load is <strong>of</strong><br />

course not zero!<br />

What is appropriate for the design <strong>of</strong> a build<strong>in</strong>g for w<strong>in</strong>d load<strong>in</strong>g, a solitaire <strong>in</strong> an empty<br />

environment:<br />

V13 - 5


Or the with<strong>in</strong> the true environment:<br />

Most design codes describe w<strong>in</strong>d pr<strong>of</strong>iles based on the roughness <strong>of</strong> the terra<strong>in</strong>, for urban<br />

environment the value <strong>of</strong> z0 may be over 2.0 m, which is not suitable as roughness for a <strong>CFD</strong> wall<br />

boundary condition <strong>in</strong> general, although it is possible for an <strong>in</strong>let w<strong>in</strong>d pr<strong>of</strong>ile. The ma<strong>in</strong> purpose is<br />

to def<strong>in</strong>e loads just based on the velocity distribution, not to perform a fluid analysis. This can be<br />

clearly seen by the fact that the velocity does not vanish at the ground.<br />

For a w<strong>in</strong>d tunnel test normally the whole environment is modelled. So to compare <strong>CFD</strong> and w<strong>in</strong>d<br />

tunnel, the model size <strong>of</strong> the w<strong>in</strong>d tunnel (<strong>in</strong>clud<strong>in</strong>g the true geometry <strong>of</strong> the equipment) may be<br />

analyzed but the analysis <strong>of</strong> the true scale <strong>of</strong> the natural model <strong>in</strong>clud<strong>in</strong>g the environment should be<br />

more adequate.<br />

And is our <strong>in</strong>terest <strong>in</strong> a model match<strong>in</strong>g reality as close as possible or can we agree on a model on<br />

the safe side?<br />

V13 - 6


4 EXAMPLE BUILDING<br />

Gerhardt [3] has reported considerable deviations between a <strong>CFD</strong>-Analysis and measurements for a<br />

hull <strong>of</strong> a build<strong>in</strong>g (Re = 4·10 7 ), given with a picture as follows:<br />

As all other data was miss<strong>in</strong>g, personal enquiries yielded a pr<strong>of</strong>ile exponent <strong>of</strong> 0.25 (w<strong>in</strong>d tunnel)<br />

and 0.27 (analysis), a rather sufficiently large air volume <strong>of</strong> 2000 x 2000 x 500 m and some<br />

<strong>in</strong>homogenities along the build<strong>in</strong>g and other neighboured build<strong>in</strong>gs <strong>in</strong> the w<strong>in</strong>d tunnel test. Some<br />

tests showed however that the effect <strong>of</strong> those build<strong>in</strong>gs should be neglectable allow<strong>in</strong>g to<br />

concentrate on a 2D section.<br />

Thus the section was digitized from the picture, scaled to the real dimensions <strong>of</strong> 55 m width and 25<br />

m height and an air volume with 500 m distance to the boundaries was used. The mesh density on<br />

the hull was selected with 0.25 cm (approx. 30 y+) yield<strong>in</strong>g a total <strong>of</strong> 12500 cells.<br />

V13 - 7


The first outcome <strong>of</strong> the analysis showed, that the selection <strong>of</strong> the turbulence model had a<br />

significant effect on the pressure distribution, the standard k- model created the distribution with<br />

the large peaks (left picture) while the RNG-model showed better results (right picture):<br />

A deeper look at the <strong>in</strong>flow parameters however showed some deficiencies. Input velocity and<br />

turbulence <strong>in</strong>tensities are well known, but the Eurocode specifies a value for the <strong>in</strong>tegral length<br />

scale <strong>of</strong> more than 200 m <strong>in</strong> a height <strong>of</strong> 25 m, which is considerably larger than the common <strong>CFD</strong><br />

value <strong>of</strong> 0.4*H = 10 m. This contradiction can be overcome if the anisotropic structure <strong>of</strong> the<br />

natural w<strong>in</strong>d is considered. Further the used mesh does not allow for large roughness values. After<br />

correct<strong>in</strong>g those parameters, the follow<strong>in</strong>g pressures have been obta<strong>in</strong>ed:<br />

0.839<br />

0.487<br />

0.103<br />

0.963<br />

1.13<br />

1.28<br />

-0.344<br />

-0.390<br />

-0.383<br />

-0.368<br />

-0.359<br />

1.36<br />

-0.361<br />

-0.362<br />

1.36<br />

1.25<br />

1.14<br />

1.03<br />

0.895<br />

0.821<br />

0.773<br />

0.740<br />

0.689<br />

0.712<br />

0.672<br />

However the w<strong>in</strong>d tunnel test was based on a very dense environment consist<strong>in</strong>g not only <strong>of</strong> other<br />

build<strong>in</strong>gs but also <strong>of</strong> the roughness elements. So any further comparison <strong>of</strong> the <strong>CFD</strong> analysis and<br />

the w<strong>in</strong>d tunnel becomes useless.<br />

V13 - 8<br />

0.650<br />

0.627<br />

0.613<br />

0.626<br />

0.646<br />

0.637<br />

0.574<br />

0.497<br />

0.422<br />

0.398<br />

0.391<br />

0.389<br />

0.386<br />

0.389<br />

0.391<br />

0.391<br />

0.388<br />

0.386<br />

0.386<br />

0.382<br />

0.384


Let’s use complex environments for the <strong>CFD</strong> as well:<br />

V13 - 9


V13 - 10


5 EXAMPLE BRIDGE SECTIONS<br />

5.1 Millau Bridge<br />

A section <strong>of</strong> the Millau Bridge has been analyzed with <strong>CFD</strong> and compared to measurements <strong>in</strong> a<br />

w<strong>in</strong>d tunnel.<br />

Despite the good agreement <strong>of</strong> the measurements with the analysis, some questions rema<strong>in</strong> also <strong>in</strong><br />

this case. The turbulence parameters <strong>in</strong> the w<strong>in</strong>d tunnel have not been fully specified. And nobody<br />

really knew the w<strong>in</strong>d conditions at the site <strong>of</strong> the bridge. There is a deep valley and fundamental<br />

<strong>CFD</strong> analysis had been undertaken to get the w<strong>in</strong>d conditions <strong>in</strong> the nature. But to perform a<br />

dynamic analysis <strong>of</strong> a mov<strong>in</strong>g bridge <strong>in</strong> the w<strong>in</strong>d field, the drag coefficients may be used.<br />

V13 - 11


5.2 Bridge <strong>in</strong> the w<strong>in</strong>d tunnel<br />

The second example is the analysis <strong>of</strong> a bridge section <strong>in</strong> a w<strong>in</strong>d tunnel [4]. Here the <strong>CFD</strong> model<br />

matches exactly the geometry <strong>of</strong> the w<strong>in</strong>d tunnel. The <strong>in</strong>flow parameters are well known (I=3 %,<br />

L=0.03 m)<br />

The lift and moment coefficients are quite closely matched, but for the horizontal drag, the<br />

measurements deviate and the simulation is closer to the pressure measurements.<br />

V13 - 12


When analyz<strong>in</strong>g this deck the results deviate to a certa<strong>in</strong> extend. First the flow field is considerably<br />

different for the standard k- and the RNG variant:<br />

The pressure distribution is very sensitive to the mesh def<strong>in</strong>ition, but the pr<strong>in</strong>cipal distribution is:<br />

1.15<br />

1.15<br />

0.722<br />

0.779<br />

0.481<br />

0.606<br />

0.252<br />

0.482<br />

0.375<br />

0.270<br />

-0.448<br />

Location / Value Experiment Simulation DOLFYN DOLFYN<br />

openFOAM k-<br />

RNG<br />

top spt 01 (tap 31) 0,494 0,650 0,723 0,689<br />

top spt 05 (tap 35) -0,413 -1,090 -0,676 -0,662<br />

top spt 06 (tap 36) -0,902 -1,230 -1,307 -1,181<br />

top spt 18 (tap 8) -0,202 -0,270 -0,216 -0,149<br />

top spt 19 (tap 9) -0,264 -0,260 -0,199 -0,135<br />

top spt 21 (tap 11) -0,270 -0,250 -0,133 -0,096<br />

dwn spt 40 (tap 30) 0,496 0,670 0,779 0,799<br />

dwn spt 34 (tap 24) -0,708 -1,110 -0,730 -0,677<br />

dwn spt 33 (tap 23) -1,083 -1,250 -1,180 -1,065<br />

dwn spt 26 (tap 16) -0,295 -0,340 -0,335 -0,261<br />

dwn spt 25 (tap 15) -0,305 -0,300 -0,297 -0,218<br />

dwn spt 22 (tap 12) -0,303 -0,250 -0,133 -0,109<br />

Drag coefficient<br />

(range <strong>of</strong> values)<br />

-1.80<br />

0.160<br />

-1.68<br />

0.0384<br />

-0.586<br />

-0.158<br />

-0.413 -0.391<br />

-0.260<br />

-0.217<br />

-0.192<br />

-0.168<br />

-0.413<br />

-0.490<br />

-0.589<br />

-0.842<br />

-1.06<br />

-1.70<br />

0,088<br />

0,075<br />

-0.152<br />

-0.139<br />

-0.352<br />

-0.315<br />

-0.127<br />

-0.280<br />

-0.116<br />

-0.250<br />

-0.109<br />

-0.235<br />

-0.102<br />

V13 - 13<br />

-0.223<br />

-0.0977<br />

-0.213<br />

-0.0955<br />

-0.208<br />

-0.0948<br />

-0.202<br />

-0.0921<br />

-0.199<br />

-0.0932<br />

-0.204<br />

-0.0958<br />

-0.213<br />

-0.225<br />

-0.100<br />

-0.239<br />

-0.106<br />

-0.272<br />

-0.114<br />

-0.424<br />

0,073 0,093<br />

0,081<br />

-0.123<br />

-0.260<br />

-0.136<br />

-0.202<br />

-0.153<br />

-0.182<br />

-0.169<br />

-0.183<br />

-0.158<br />

-0.290<br />

-0.146<br />

-0.143<br />

-0.138<br />

-0.121<br />

0,082<br />

0,069<br />

-0.113


6 CONCLUSION<br />

The critical def<strong>in</strong>itions for a <strong>CFD</strong> analysis are the selection <strong>of</strong> the mesh and the <strong>in</strong>flow conditions.<br />

Handled with greater success are bluff bodies with sharp edges, slender structures (e.g. airfoils)<br />

need more attention.<br />

So there is no free lunch by just buy<strong>in</strong>g a <strong>CFD</strong> s<strong>of</strong>tware, which is also valid for any other type <strong>of</strong><br />

complex simulation s<strong>of</strong>tware (like structural analysis F<strong>in</strong>ite Element). Each journey starts with the<br />

first step. It’s time to take that first step.<br />

7 DOLFYN<br />

DOLFYN [1] is an open source collocated face based F<strong>in</strong>ite Volume s<strong>of</strong>tware to solve<br />

<strong>in</strong>compressible fluid dynamic problems <strong>in</strong> 3D. The key features are:<br />

Standard k- and RNG turbulence models<br />

Stable numerical procedures<br />

Temperature / scalars / particles <strong>in</strong>cluded<br />

Postprocess<strong>in</strong>g with ParaView / VisIt (VTK-files)<br />

The implementation <strong>in</strong> the <strong>SOFiSTiK</strong> environment gives additional<br />

Mesh generation with SOFIMSHA / SOFIMSHC<br />

Postprocess<strong>in</strong>g with WINGRAF<br />

Full CADINP support <strong>in</strong>clud<strong>in</strong>g formulas for boundary and <strong>in</strong>itial conditions<br />

Easy w<strong>in</strong>d def<strong>in</strong>ition (atmospheric boundary layer) directly or via SOFiLOAD<br />

Direct generation <strong>of</strong> result<strong>in</strong>g w<strong>in</strong>d load<strong>in</strong>g <strong>in</strong> the data base<br />

Possible (planned) extensions are:<br />

Compressible subsonic flow<br />

Free surfaces (VOF)<br />

Conjugate heat transfer (heat transfer with different materials)<br />

8 LITERATURE<br />

[1] www.dolfyn.net<br />

[2] J.H.Ferziger, M.Peric, Numerische Strömungsmechanik, Spr<strong>in</strong>ger, 2008<br />

[3] H.J. Gerhardt, Experimentelle und numerische Verfahren bei der Bauwerks-Bemessung, Der<br />

Prüf<strong>in</strong>genieur Vol. 24, April 2004<br />

[4] A. Sarkic, C. Neuhaus, R. Höffer, Numerical and experimental determ<strong>in</strong>ation <strong>of</strong><br />

aerodynamic forces at a long span bridge girder, Eurodyn 2011, Leuven<br />

V13 - 14

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!