Learning Forum

[Allen P Varghese]

I am studying the effect of change in gravity on a tank that is filled up to 1/3rd of its volume with liquid N2. The tank has baffles and I am trying to study the sloshing effects. These are the settings I have enabled,nnMesh:nNumber of cells: 1,226,345nPolyhexacore mesh generated within Fluent watertight geometrynThe mesh is kept as coarse as possible for initial simulatioFluent Setup:nMultiphasenVolume of FluidnExplicit forumulationnImplicit body force enablednSharp interface modellingn2 Eulerian phasesnPhase1: n2gasnPhase2: n2liquidnNo surface tension between themnViscous: LaminarnNo inlet or outlet BCnCell register region defined for region to represent n2liquidnInitialized the whole domain with n2gas = 1, patched defined region with n2liquid = 1nMethodsnScheme: PISOnGradient: Least Squares Cell BasednPressure: Presto!nMomentum: Second Order UpwindnVolume Fraction: Geo-ReconstructnNon-Iterative Time Advancement has been enablednWarped Face Gradient Correction has been enablednControlsnDensity: 0.5nPressure: 0.5nBody Force: 0.5nMomentum: 0.5nSolver SettingsnType: AdaptivenMethod: Multiphase SpecificnDuration specification method: Total timenTotal time: 12secondsnGlobal Courant Number: 2nInitial time step size: 1e-05nFixed time step, reporting interval, time step update interval, profile update interval: 1nnI am running the simulation parallelly on 30 processors of Intel Xeon 4216 CPU @ 2.10 Ghz & 1 GPU NVIDIA GeForce RTX 2080Ti. 64GB 2400MHz RAM and 42GB GPU Memory.nThe simulation is taking a very long time (around 0.25 seconds of flow time has been simulated in 24 hours, 1%). Is there something I am doing wrong? Is there some way I can improve my timings.n

[YasserSelima]

GeForce is not tested with Ansys Fluent. Try running without it. Even if tested, I am not sure about the feasibility of 1 GPU per 30 cores. It might be slowing down the simulations.nNon-iterative is definitely faster, but it might not catch the transition details. nIf you have high acceleration of the tank, you need to enable phase change and take surface tension into consideration. n

[Allen P Varghese]

Thank you for your reply. I will try your suggestions.n

[Rob]

Look at the new Hybrid NITA scheme in 2021R1 and also review dynamic adaption. You may find you've got a fairly fine mesh everywhere which isn't needed. n

[Amine Ben Hadj Ali]

Do not rely on GPU here: it won't add any value. Just invoke the Hybrid NITA schemes which are now accessible from Graphical User Interface. Also you might think about using smart Mesh Adaption + Multiphase Time Constraint in 2021R1.n