> foamCleanPolyMesh >> blockMesh J >> snappyllexMesh -overwrite 3) View final mesh (see above) Copyright © 2015-2016 CFD Direct Ltd CFD DirectsnappyHexMesh: mesh refinement | Tri-surface manipulation 58/102 Task: mesh vehicle geometry for external aerodynamics, max 600k cells m Create a new estateMesh case from inflowOutflow template case >> run >> ep -r $FOAM_ETC/templates/inflow0utflow estateMesh m Copy the estate. obj geometry >> cp $FOAM_RUI/resources/geometry/estate.obj . Find its location, orientation and size ™ Create CAD. obj: front axle at « = 0, vertical z-axis, metre dimensions 2 (0,0, 1) >> surfaceTransformPoints -translate '(1445 32 0)' -rotate '((0 0 1) (0 -1 0))! <+-scale ‘(0.001 0,001 0.001)! estate.obj estateMesh/constant/triSurface/CAD.obj SurgaceTwonsgurmRends - hranelehe *C (Ws 726)' ~ roll Pda Yau 1(G0 oe)! ~Kale CFD Direct Copyright © 2015-2016 CFD Direct LtdsnappyHexMesh: mesh refinement | External patches 59/102 = Set the external patches (inlet, outlet, etc.) in background mesh @ Use alternative blockMeshDict .extPatches file in template case >> ep system/blockMeshDict.extPatches systen/blockMeshDict ground: type wall back, front, top: type symmetry a InblockMeshDict, set backgroundMesh according to diagram (si. 35) m Change patch names, e.g. left —> inlet, etc. Set patch types, see diagram = Run blockMesh CFD Direct Copyright © 2015-2016 CFD Direct LtdsnappyHexMesh: mesh refinement | Initial mesh refinement 60/102 = In snappyHexMeshDict, delete region sub-dictionaries (in CAD. obj geometry and refinementSurfaces) Initial cell length = 2/3 m; level 5 refinement — (2/3)/2° = 21 mm oy) = Run snappyHexMesh ‘ Jed tro Zerbiea (aoe and mort) frommelannhe pe dtbe puroy (5 5) teal -£ log \ gree “Time. Copyright © 2015-2016 CFD Direct LtdsnappyHexMesh: mesh refinement | Review of the initial mesh 61/102 Refinement box A ——— Refinement interfaces separated by 3 cells Y Vehicle surface quite well resolved = Insufficient refinement in the wake; refinement along ground? underbody? = Final Step: add 1 layer of cells to vehicle = ...for more uniform near wall cell size/shape Copyright © 2015-2016 CFD Direct Ltd CFD DirectsnappyHexMesh: mesh refinement | Additional refinement 62/102 = In your own time: add refinement boxes (copy/paste boxA) Box min max level _ Region boxA (3-10) (511.6) 4 Wake boxB (3-10) (811.6) 3 Wake boxC (3-10) (1511.6) 2 Wake boxD (-8 -10) (5 1 0.16) 4 Upstream/underbody m Increase nCel1sBetweenLevels from 3 (default) to 4, 5, ...to extend refinement Upstream/underbody Cell tayer: refinement (cut, purple) ond Copyright © 2015-2026 CFD Direct Ltd CFD DirectObjective data | Probes 63/102 = Aim: to set up estate case to output objective data . Monitored during a simulation = Create an estate case by cloning estateMesh and ; fle dds setting Uinlet to (40 0 0) in0/U Alee grrs ™ Add pressure probes at select locations m Copy packaged probes function object into system directory >> cp $FOAM_ETC/caseDicts/postProcessing/probee/probes systen = Set the fields and probeLocations in probes —\ ™ Include probes from the controlDict file | Foam CloneCure See Sone ln gent wu oo Caer \w Como chs \uor wo FAY Wu softens Hema gee WN gee HY a perdi (92 4 “cael © S a 0): 25 yaw aubrd yer Unies (40 00); 22 ype ves \ 4 (us Suthetweo- due Um f Viale’ 4 copyright © 2015-2016.cRD Direct cd ! CFD DirectObjective data | Force coefficients 64/102 ™ Packaged function objects for force calculation are listed by >> 1s SFOAM_ETC/caseDicts/postProcessing/forces Files ending . cfg contain configuration settings 1 Other files are function objects for copying to system directory 1 Copy the forceCoef fsIncompressible file to system 1m #include this file in functions in controlDict Name of patch over which forces are integrated |U.0| for coefficient calculation Tres for moment coefficient Crm Aves for lift/drag coefficients C\, C: Estimate of frontal area } lift and drag directions } moment pitch axis and centre of rotation — midpoint between wheel axles [ogk /oorgsoeo] ere /cedrchs /seh Conan! Fen edt a (2 Andie pavrtune ica eee eae her FAME Ym catcos de am poeko ayers ~ pre el pot Prercmmads k Pam ve te Célelen ey wuvehl ne coai tie ane polar callow cous chr oee foo nope. CFD Direct Copyright © 2015-2016 CFD Direct LtdObjective data | Monitoring data 65/102 = Run simpleFoam m= foamMonitor: displays data graphically; for usage type >> foamMonitor -h m= Execute it on a time-data file, e.g. probe data >> foamMonitor postProcessing/probes/0/p & m Data refreshes every 10s; stops refreshing if data unchanged for 60s = Monitor drag coefficient, setting y—range from 0.3 to 0.4 >> foamMonitor -y "[0.3:0.4]" postProcessing/forceCoeffs/0/forceCoefts.dat & copay vendedo lea Visyelrahen, rr a a a Tie As bok foparceom30/ere Jcare Dict Joat Receosng/ ¥ de age & puede enhbev a prubes ey corpiguacer de, r poremeéhey Inclow ef archive eo el cembel cuhd Diet pile . FUL 29 ; frobtg yy ge auade Gr el curhel Dich che ve Ge x Callen Al yerlear- e poe Copyright © 2015-2016 CFD Direct Ltd CFD DirectObjective data | Residuals. 66/102 Residuals are printed to terminal / log file . but residuals function object writes to file for monitoring Set up residuals to write residuals for U, p and k . . ™ Copy packaged residuals function object into system directory . = #include residuals in functions in the controlDict = Monitor residuals, setting logarithmic y—scale (-1 or -logscale option) >> foamMonitor -1 postProcessing/residuals/0/residuals.dat & os oor oie 10 200 aso a0 35000 Coplar Yenduls en conbdlOict ge yrolle prow stop At wellenow 5 CFD Direct Copyright © 2015-2016 CFD Direct LtdObjective data | Other objective function objects 67/102 f m In $FOAM_ETC/caseDicts/postProcessing = flowRatePatch: flow rate between 2 patches owRateSurface: volume flow rate between 2 tri-surfaces vo. minMax directory: several function objects to write min/max values m .. especially fieldMinMax; useful to check boundedness, e.g. a, T', k, 6, w staticPressure: converts kinematic pressure to pressure (force/area) totalPressure: calculates total pressure pr = Pret + p + p|U|?/2 u pressureDifferencePatch: difference in mean pressure on 2 patches oh tan wenite Corba hone ste} yunbbne } ( Copyright © 2015-2016 CFD Direct Ltd CFD DirectGraphs | Quick graphs 68/102 1m Task: plot pressure downstream of tallest building in windAroundBuildings 1m ParaView has graphs plotting = Load case into ParaView, t = 400 = Extract building patches = Filter + Data Analysis mu... Plot Over Line = Limited format control 204 Pressure 10} Pp im*/s*] 0 20 40 60 80 100 120 140 Height z [m] CFD Direct Copyright © 2015-2016 CFD Direct LtdGraphs | Graphs with the sample utility 69/102 m sample utility: samples data from a sampleDict configuration file = Copy the template sampleDict file to system >> op SFOAM_ETC/caseDicts/postProcessing/graphe/sampleDict system = Edit the file; set same sample line as before = Run sample >> sample m Files in postProcessing/sets/ 10 qo -10 ‘a-20 -30 Samples 0 40 80 120 U and p z{m) @ Use gnuplot to plot result at f = 400 >> gnuplot gnuplot> plot "postProcessing/sets/400/graph_p.xy" with lines Copyright © 2015-2016 CFD Direct Ltd CFD DirectGraphs | Automated data processing 70/102 m While running an application, e.g. simpleFoam, extra data can be processed @ ... through function objects, specified under functions in controlDict m Requires knowledge to configure m ... but pre-packaged tools are in $FOAM_ETC/caseDicts/postProcessing = For example, singleGraph processes graph data m To use, copy the file to system >> cp $FOAM_ETC/caseDicts/postProcessing/graphs/singleGraph system = Set the start and end of the sampling line in singleGraph Copyright © 2015-2016 CFD Direct Ltd CFD DirectGraphs | Graph data with a function object 71/102 @ Include singleGraph in functions in controlDict = Function objects execute while a solver is running = = to test, run for 1 solution step and output data; in controlDict 1e end of the solution step = Run simpleFoam m Graph data written to postProcessing/graph/401 m Same graph as with sample; uses same sampling library Copyright © 2015-2016 CFD Direct LtdGraphs | Customising graph data 72/102 ™ graph function object file: #include "sampleDict.cfg" Data interpolation method for sampling ‘Output format, e.g. csv, vtk Uniform distribution of sampling locations ‘Number of of sampling locations ‘Output sampling location data: distance: distance from start point x ordinate +s Y Z) Co-ordinates = Settings can be modified or (more conveniently) over-ridden m...e.g. change to write (, y, z) co-ordinates; edit singleGraph file 8 keyword entry over-rides icate in sampLeDict .cfg ler of entries matters ™ Run simpleFoam again; check data in files in postProcessing/graph/402 Copyright © 2015-2016 CFD Direct Ltd CFD DirectTurbulence: introduction | Encountering turbulence 73/102 = Copy the axisynmetricJet case; naming it axisymmetricJet-LU = Generate the mesh a In constant /turbulenceProperties, set RASModel to laminar = Run simpleFoam; and display velocity field = Results at step 100; subsequently blows up Numerical instability caused by fluctuations inlet Large/medium eddies: wer tay Copyright © 2015-2016 CFD Direct LtdTurbulence: introduction | Turbulent scales and mixing 74/102 = Copy the axisynmetricJet-LU case; naming it axisymmetricJet-linear u In system/fvSchemes, set div(phi,U) to Gauss linear 1... reducing numerical diffusion; maintaining fluctuations ™ Run simpleFoam; at step 1000... Disordered fluctuations 2 Low mixing — not realistic Copyright © 2015-2016 CFD Direct Ltd CFD DirectTurbulence: introduction | Nature of turbulence 75/102 @ Perturbations grow chaotically instead of being damped by viscous forces = Fluctuations in time, disorder in space, wide range of scales ™ Characterised by Reynolds number = inertia/viscous forces UL Res == v a ... Disa characteristic length = 10° < Re < 10*: laminar—turbulent transition Me log(Z) Large eddies, size | o weer Ee gt Lye ele oe Ddent? oe y oe Inertial subrange spa (wee 3 dake Smallest scales ~ Re" dissipate energy log(wave number, k) Nonna af CONS Copyright © 2015-2016 CFD Direct Ltd CFD DirectTurbulence: introduction | Need for turbulence modelling 76/102 m Smallest scales (Kolmogorov microscales) for jet ~ 10° m = Cost: impractical to mesh down to these scales = Stability: energy not dissipated in small eddies ->‘blows up’ = U = random; statistical properties, e.g. mean U — non-random = Cost: solving/averaging U over time is expensive corccer la vel pronedio = = Solve for U, e.g. steady, inc ib 9 ', incompressible fluctuations affect — ain a mean flow V+ (UU) — V-(vVU) = -Vp-V-(U'U) Arvest cl JO cgeego ane eco ~ How do we treat this? Eqn. for UU" has terms in UUU! Eqn. for WU has terms in WUE = no closure wuwa fevminn — clayre pablo colar cre te nega sqonpme €o. menhrer Copyright © 2025-2016 CFD Direct Ltd CFD DirectTurbulence: introduction | Turbulence madell 77/102 fabelan? bee! in Sete exewey) m Eddy viscosity 14: UU" = (2/3k—ve(VO+ VU") se arenes o> leviarnes de le volecele! prumeclio = Turbulent kinetic energy: k = U’?/2 = Transport equation for k V+ (Uk) — V+ (DienVk) = G(VU)— ec ce penbw ve peede predectn —SSaS Te NS Cemrchamendte le velecdd advection afsion production —dlssipation m= What about ¢? Empirical equation, with C,, Cz tuned for shear flows... Eb whee UK excemneachs Ve(Ue) V+ (DeenVe) = ACF ~ C27 % = wo exole we valdect& Y= CyR/e Diet = Ver =U+Vi Dee =U/Ge+Y Clee Le gure lepevle, wero co (a gale euyivicr Aerie wares de sete m Copy the axisymmetricJet case m In constant/turbulenceProperties, set RASModel to KEpsilon = The k-e equations above will be solved ‘ ™ Run simpleFoam; view a diffuse, mean U field in ParaView Copyright © 2015-2016 CFD Direct Ltd CFD DirectTurbulence modelling essentials | k — w SST turbulence model 78/102 m= SnappyHexMesh:Intro module: generated basic cyclone mesh = Task: simulate steady-state flow in the cyclone, inlet speed 10 m/s m inflowQutf low case: setup with kOmegaSST turbulence model m k-w SST (Shear-Stress Transport): Industrial standard (default) model = ... blend of k-w at wall with k-e in freestream m Turbulent kinetic energy k lore 1 (grr Te k=5U =3(@ +U; +02) m Specific dissipation w w le ~ Ok = = Turbulent viscosity 1% _k or Up 210 meh ves 3 lo? 0% CFD Direct Copyright © 2015-2016 CFD Direct LtdC Turbulence modelling essentials | Initialisation of turbulent fields m €, W — estimate using turbulent length scale 3/2 L= oyES (C,, = 0.09) Fully-developed pipe flow: L = 0.07x Dp ... Hydraulic diameter D, = 4xArea/(Wetted Perimeter) Wall bounded flows: L = 0.4x Dy; Dp = boundary layer height w= Onl VE/L a k —estimate k = 1.5(UI)*; I = turbulent intensity 1 Fully-developed pipe flow: 1% < I < 10% =... =0.16Re™*; Re = |U|D,/v ™ External flows, e.g. wind tunnel: typically, 0.05% SS 1% ™ Lower estimate of k is safer; w > k, so 4 is sensible Copyright © 2015-2016 CFD Direct Ltd 79/102Turbulence modelling essentials | Running the cyclone case 80/102 & Dp = 4x0.5x0.2/1.4 = 0.285 = TL = 0.070.285 ~ 0.02 Inlet flow speed = 10 m Re = 10X0.02/1.5e-5 ~ 13,000 m1 =0.16 ReM ws 5% (hk ww = 009° Vk/L 56 @ Run simpleFoam to step 1000 ‘Streamtubes: centre (0, 0, 0), radius 0.18, 3 points 9 |U| 20 O21 debate weleves ecgulen haan ater ce dino selo ey Vas entrads - Copyright © 2015-2016 CFD Direct LtdTurbulence modelling essentials | Wall boundary layers 81/102 m k, omega and nut include special boundary conditions at walls c ‘= Why?...Why not k = 0? => % =0...What about w ~+ 00? _— Friction velocity; Tw = uz U,/ut 1 = nev") +. 25 viscous sub-layer 20 —— 15 10 log-law layer <>! ( 1 10 1000 ury/v ( = Small near-wall cell (ys 1) for accurate wall shear stress Tw OU; /Oy) ( m= Requires very large mesh — impractical for most CFD users sro hey cenlicues de furl gam cards ge pueden My Sefudes er bo wnedlos CFD Direct \— Copyright © 2015-2016 CFD Direct LtdTurbulence modelling essentials | Wall functions 82/102 m nutkWallFunction: sets 4 = f(y*) log-law for better 7, prediction ® ... allowing centre heights of near-wall cells of y+> 30 m Sets % =O foryt<11 = nutUSpaldingWal1Function: continuous 1 function to wall (y+ = 0) G/e © 1.5-2.0 f k 10 20 30 40 y7 = Large variation in k, ¢, w near wall = Modelled cell-average values more accurate than simulation for y+ > 1 m omegal/al1 Function: calculates near-wall cell w ™ epsilonWallFunction: calculates near-wall cell © wo Salo . eeptkel Tnetany zeroGradient: condition for k, g, R " duched gud Ts pe € CFD Direct Copyright © 2015-2016 CFD Direct LtdTurbulence modelling essentials | Further work and summary 83/102 = Test running the cyclone case with alternative BCs on wall: Field Boundary condition K fixedValuc, 0 omega zeroGradient nut calculated m The case should run successfully =... but wall shear Ty too low m Test on external aerodynamics case and note lower drag = Summary: to reduce computational cost ™ ... generate mesh with near-wall cell centres y+ > 30 m= ...and apply the following boundary conditions at walls Field Boundary condition nut nutkWallFunction omega —omegalal1 Function epsilon epsilonWallFunction k,q/R __kqRWal] Function CFD Direct Copyright © 2015-2016 CFD Direct LtdDiscretisation: introduction | Discretisation: overview 84/102 = Differential equations, continuous fields > matrix equations, discrete values m For generic field Q (scalar, vector ...), standard transport equation: cee esas) + Mellel | rate of advection diffusion sources ae : density a T°: diffusivity, e.g. viscosity j1, thermal conductivity k = S(Q): sources, e.g. pressure gradient —Vp, gravity pg, surface tension .. Copyright © 2015-2016 CFD Direct Ltd CFD DirectDiscretisation: introduction | Case for testing discretisation 85/102 To illustrate numerical issues, create a new estateN0 case by cloning estate Copy patchAverage packaged function object to measure average p at inlet >> cp $FOAM_ETC/caseDicts/postProcessing/faceSource/patchAverage systen/ In patchAverage, set patch to inlet m #tinclude this file in functions in controlDict ™ Make a backup of the 0 directory >> cp -r 0 0.org Copyright © 2015-2016 CFD Direct Ltd CFD DirectDiscretisation: introduction | fvSchemes overview 86/102 fvSchemes: organised by type of derivative... aU Gp + V+ (UU)-V+ VU) = —Vp oe a OL fatect adveion aston pressure Eiinge recent Requires a default scheme for every type of derivative Other entries over-ride the default Specify no default for divSchenes; includes a) advection b) other (diffusive) terms Copyright © 2015-2016 CFD Direct Ltd CFD DirectDiscretisation: introduction | Potential flow 87/102 ® potentialFoam calculates potential flow (w = 0), solving (in effect) u ...V?@ = 0: Laplace equation for potential = ...U = V®: calculate velocity field U = With -writep option, solves additionally .Vpt V+ (F+V+(UU)) = ... where flow direction tensor F = UU/|UU| = 2 diffusion equations = good demonstration of diffusion discretisation Bernoulli equation to evaluate p m= Run potentialFoam, redirecting output to file >> potentialFoam -vritep -withFunctionObjects ™ Need to add an entry to each of fvSolution and fvSchemes (p1em)" ” get ce (ach, CFD Direct Copyright @ 2015-2016 CFD Direct LtdDiscretisation: introduction | Running potentialFoam 88/102 a * t Residuals increase z a z between solution steps * =>matrix equation is changing Solving for Phi, Initial residual = 1,fFinal residual = 010686, No Iterations 5 Solving for Phi, Initial residual = (0,00, Final residual = 0.00076, No Iterations 1 Solving for Phi, Initial residual = 0.000909, Final residual = 7.83e-05, No Iterations 3 Solving for Phi, Initial residual = 0.00036, Final residual = 2.43e-05, No Iterations 2 Solving for Phi, Initial residual = 5.63e-05, Final residual = 4.42¢-06, No Iterations 3 Solving for Phi, Initial residual = 1.60-05, Final residual = 1.32e-06, No Iterations 2 Solving for Phi, Initial residual = 2.61e-06, Final residual = 6.5¢-07, No Iterations 1 ‘continuity exror = 0.000141 Interpolated U error = 0.000158 Convergence fairly slow Galculasing Euler pressure field Solving for p, Initial residual = j/FIWB1 residual = 0.0584, No Iterations 2 Solving for p, Initial residual = (0.118, Final residual = 0.00673, No Iterations 2 Solving for p, Initial residual = 0.0223, Final reeidual = 0.0018, Wo Iterations 2 Solving for p, Initial residual = 0.00541, Final residual = 0.000228, No Iterations 3 Solving for p, Initial residual = 0.00i41, Final residual = 9.78¢-05, No Tteratins 2 Solving for p, Initial residual = 0.000394, Final residual = 1.79e-05, No Iterations 3 Solving for p, Initial residual = 0.000127, Final residual = 9.26e-06, No Iterations 2 Solving for p, Initial residual = 4.4ie-05, Final residual = 1.75e-06, No Iterations 4 Solving for p, Initial residual = 2.47905, Final residual = 1.51e-06, No Iterations 2 Solving for p, Initial residual = 1.2e-05, Final residual = 8.760-07, Mo Iterations 2 Solving for p, Initial residual = 7.37606, Final rosidual = 5.¥5e-0/, lo Iterations 2 CFD Direct Copyright © 2015-2016 CFD Direct Ltd1999999007093 Discretisation: introduction | potentialFoam solution 89/102 m Load case in ParaView, orientate in y-normal plane m Slice: origin y = +0.2, y-normal plane box; color by cell pressure -olor by cell pressure m...check crinkle slice . m Extract Block — C, = Produces diffusive pressure and velocity fields = Irrotational flow = no separation/vortices = => matching pressure up/downstream; Ap = -7.3 (* 0) from inlet to outlet = => d’Alembert’s paradox: zero lift/drag Copyright © 2015-2016 CFD Direct LtdDiscretisation: diffusion | Diffusive transport 90/102. 1 Finite volume discretisation begins with integral over volume V = Demonstrate using pressure p field > V « (T'Vp) fy (LVp) dv = [essere ss LNs (Vp) = LPs (p/dn), sum for all faces Gauss’s theorem Discretisation = Discretisation uses Gauss theorem m Face I's: usually by linear interpolation of adjacent cell values @ Key issue: (Op/On) — discretisation of gradient of p, normal to face, snGrad. Copyright © 2015-2016 CFD Direct Ltd CFD Direct7 c ( ¢ ¢ Discretisation: diffusion | Surface normal gradient 91/102 face f implicit explicit = Example face intersecting cell A (centre, left) and B (centre, right) @ Non-orthogonal: Ono # 0 = Gradient of p in direction A-+B uses ps and pa; => matrix coefficients @ To produce normal (n) gradient, transerse gradient (Op/Ok); can be calculated = ...and added as an explicit source CFD Direct Copyright © 2015-2016 CFD Direct LtdDiscretisatio iffusion | Non-orthogonal correction 92/102 m potentialFoam was run previously with corrected normal gradient @ .../.e. included (Op/Ok), source term => gradient points in normal (n) direction m™ Change to uncorrected in fvSchemes = Run potentialFoam reading/writing data to time directory 1 >> ep -r O.org 1 >> potentialFoam -writep > log.uncorrected & = Convergence is rapid because source does not change Solving for Phi, Initial residual Solving for Phi, Initial residual Solving for Phi, Initial residual Solving for Phi, Initial residual Solving for Phi, Initial residual ="3.73e-05, No Iterations 4 = 2.66e-06, lio Iterations 4 Copyright © 2015-2016 CFD Direct LtdDiscretisation: diffusion | Non-orthogonal correction (cont/d) 93/102 131709 } - = Error in p accumulates across faces with high non-orthogonality ( 1 Pressure change is Ap = -338 from inlet to outlet C ™ uncorrected scheme highly inaccurate in regions of non-orthogonal faces te ( = ... but more convergent/stable due to zero source ¢ c 7 ( re C c r re \— Copyright © 2015-2016 CFD Direct LtdDiscretisation: diffusion | Limited correction 94/102 = Non-orthogonal correction can be limited @ |(0/Ok)s| < some % of orthogonal part (pa — py...) u ...where limiting coefficient ¢ = 0.33 means 50% of orthogonal part m ...and ¢ = 0.5: 100% of orthogonal part wee % (pe — pa)/|d| ——> [#: :| b| = (}] ak * Incorrect gradient , direction reduces accuracy K limit magnitude FF source (@p/any~e °F COPIPK improves stability = Add limited in fvSchenes Copyright © 2015-2016 CFD Direct Ltd CFD DirectDiscretisation: diffusion | Limited correction (cont/d) 95/102 = Run potentialFoam reading/writing data to time directory 2 >> ep -r O.org 2 >> potentialFoam -writep > log.limited & m Log file shows slower convergence with limiting = ... but improved stability — residuals decrease more smoothly a Ap =-101 from inlet to outlet — significantly lower than uncorrected = Try limited, p = 0.5 > Ap =-20.8 m Recommendation for (Op/On)s: corrected scheme @ but Limited when unstable — 6), > 70°-80° CEOOCP RE ANNA G0G92I1791790399399090039 C Copyright © 2015-2016 CFD Direct LtdDiscretisation: advection | Advective transport 96/102 Advection is transport of property, e.g. energy, by bulk fluid motion = Demonstrate using turbulent kinetic energy k field > V + (pUk) [von av [8-008 ~ D800 ky = So de hr v 7 ‘te—_ ? mass Mux dr cls knaus ao m Discretisation uses Gauss theorem CLdumebric (lux = ge: calculated and stored as separate field phi m ...often needed in multiple terms, e.g. V+ (pUU), V+ (pUk), V+ (Ue) m Key issue: ky — interpolation to faces of advected field (at cell centres) (54> wa CEO wequeveg le lenmrnd Copyright © 2015-2016 CFD Direct Ltd CFD DirectCCCP OCR T FA 4790004N29N9997I9N9339I99000000 Discretisation: advection | Setting advection schemes 97/102 = Advection schemes are set in divSchemes sub-dictionary of fvSchemes file m Change div(phi,U) to upwind; revert laplacianSchemes to corrected 1 Run potentlalFoam again advection of momentum advection of turbulence = Ap = -23.5 from inlet to outlet — significant error compared to linearUpwind = Note: advection scheme inaccuracy gives lower Ap than for diffusion schemes ™ ...since Laplacian terms in both ® and p equations = errors compounded Copyright © 2015-2016 CFD Direct Ltd CFD DirectDiscretisation: advection | Navier-Stokes simulation 98/102 ™ Clone a new case estateAdvection from the estateNU case @ Reset the momentum advection to linearUpwind m Run simpleFoam to t = 200 = developed wake vortices = Monitor drag coefficient Cy; oscillates around Cg = 0.34 Copyright © 2015-2016 CFD Direct Ltd CFD DirectDiscretisation: advection | Accuracy and upwind = Switch momentum advection to upwind = Wake shrinks rapidly = Drag coefficient increases to Cy ~ 0.46; increase of 30%! 99/102 Copyright © 2015-2016 CFD Direct LtdDiscretisation: advection | Interpolation of advected fields 100/102 = Plot field Q vs distance « between cell and face centres Q Flow direction, calculated 1 gradient 1 cellU face f cell D © = upwind uses upwind (U) cell value only linear, linearUpwind use downwind (D) cell value also ™ Limited schemes evaluate Q; = Quy + 8(r)(Qi — Qu) m...where r = 2(VaQ)u/(VaQ);—1 u LimitedLinear, vanLeer: popular limited schemes Copyright © 2015-2016 CFD Direct Ltd CFD DirectDiscretisation: advection | Advection: accuracy and boundedness 101/102 m Copy advect:ion2D example of scalarTransportFoam to run directory m Advects 0 < T < 2 by velocity at angle (30°); test schemes for div(phi ,T) Unbounded (TP >2and Moderately T <0) unbounded linear upon Sclhe Tine _Hmnear_ linearUpwing S