Transient RANS and Hybrid RANS/LES
Transcription
Transient RANS and Hybrid RANS/LES
Title 1/4 Transient RANS and Hybrid RANS/LES by K. Hanjalić Hanjali Department of Multi-scale Physics, Delft University of Technology Delft, The Netherlands CONTENT: ¾ Limitations of LES and needs for combined (hybrid) LES/RANS ¾ Hybrid RANS/LES (HRL) ¾ Rationale and a priori tests; zonal and seamless coupling; interface issues ¾ One-, two- and multi-equation RANS models ¾ Examples of application of HRL in attached and separated flows ¾ Plane channel at high Re ¾ Hill flow ¾ Transient -RANS based VLES ¾ Rationale, justification, validation and limitations ¾ Application to thermal R-B convection at extreme Ra’s (“ultra-turbulent” regime) ¾ Examples of practical relevance: ¾ Diurnal dynamics over a mezzo-scale town valley ¾ Diurnal wind over Arctic ice sheet 1 Motivation: Predictions of complex wallwall-bounded turbulent flows and heat transfer at very high Reynolds and Rayleigh numbers ¾ The mainstay of the contemporary industrial CFD are the RANS turbulence closures: affordable, economical,…but: ¾ too much empiricism, lack of universality, difficulties in predicting complex unsteady and nonequilibrium flows, .. ¾ LES: less empirical, captures better the turbulence physics, considered as the future industrial standard,…but: ¾ expensive and time consuming, especially for high Re and Ra number wall-bounded flows in complex geometries: Time- or ensemble-averaged (RANS) or filtered (LES) momentum and energy Equations: ⎞ D Ui 1∂ P ∂ ⎛⎜ ∂ U i ν = Fi − + − τ ij ⎟⎟ ⎜ Dt ρ ∂xi ∂x j ⎝ ∂x j ⎠ DT Dt = q ρc p + ∂ ∂x j ⎛ν ∂T ⎞ ⎜ − τθ i ⎟ ⎜ σ T ∂x j ⎟ ⎝ ⎠ Common practice in RANS approach: Linear Eddy Viscosity/Diffusivity models: 2 3 ⎛ ∂U i τ ij = uiu j = kδ ij − ν t ⎜⎜ τθ i = θ ui = − ⎝ ∂x j + ν t ∂T σ Tt ∂xi ∂U j ⎞ ⎟ ∂xi ⎟⎠ 2 Grids issues for LES and RANS for wallwall-bounded turbulent flows ¾ LES of wall-bounded flows require high resolution grid in all directions for resolving near-wall processes (∆x+ O(50), ∆y+ O(1), ∆z+ O(20)) ¾ For resolving viscous near-wall boundary layer: No of grid cells ∝ Reτ1.8 as compared to Reτ0.4 for outer layer (Chapman, 1979). ¾ For R-B conv.: ∆/H ≈ O(Pr2/NuRa)1/4 ⇒ Total No of grid cells ∝ Ra! ¾ In contrast, for near-wall RANS N∝ ln Reτ , for R-B N∝ Ra1/3 ¾ Hence, for high Reynolds and Raleigh numbers LES still too expensive ¾ Options for very high Re and Ra numbers: • Hybrid LES/RANS (Balaras, Davidson, Spalart, Hamba, Piomelli,… • RANS-based VLES Wall functions versus nearnear-wall RANS (Hybrid) ¾ Wall functions ¾ No universal character ¾ Standard log-law adequate in simple wall-attached flows ¾ Inadequate for separated flows ¾ Hybrid LES-RANS (HRL) strategies (including DES) ¾ Substantial part of turbulence is modelled by RANS ¾ Significantly smaller number of cells (large aspect ratio) ¾ Criteria for location the RANS-LES interface: ¾ Decided by user or ¾ Controlled by cell dimensions –comparison between length scale and typical mesh size (critical in some separating flows) 3 Key questions regarding HRL ¾ Where to locate the interface? ¾ Which matching conditions are to be used at the interface? ¾ How does a RANS model react to unsteadiness (“receptivity”)? ¾ Will the dynamics be rightly returned? ¾ What is the impact of RANS layer on the LES region? ¾ Will the modelled contribution correctly compensate the reduction in the resolved contribution? ¾ Which models are suitable? DES approach of Spalart et al. • One-equation transport model (Spalart-Allmaras, 1993, Nikitin et al. 2000) used as RANS model in the near-wall region, and as an ssg model for LES in the outer region 2 2 ⎛ ∂ν ⎞ ⎤ ⎛ ν ⎞ Dν 1 ∂ ⎡⎢ ∂ν i ⎥, ν +ν = cb1 Sν − cw1 f w ⎜ i ⎟ + + cb 2 ⎜ ⎜ ∂x ⎟⎟ ⎥ Dt σ x x ∂ ∂ d ⎢ j j j ⎝ ⎠ ⎝ ⎠ ⎦ ⎣ ( ν ν t = ν fυ1 , χ = , ν fυ1 = g = r + cw 2 ( r 6 − r ) , χ3 χ 3 + cυ31 r= ) χ fυ 2 = 1 − , 1 + χ fυ1 , ν 2 Si di κ 2 , 1 ⎡ 1 + cw6 3 ⎤ 6 fw = g ⎢ , 6 ⎥ ⎣ g + cw3 ⎦ ν Si = S + f 2 υ2 κ 2 di • Switching from RANS to LES: d = min(d , CDES ∆ ), where ∆ = max(∆ x, ∆ y, ∆ z ) and d is the distance from the nearest wall 4 DES: RANS/LES interface location in a channel flow Grid: 96x64x64 Grid: 64x64x32 DES: RANS/LES interface location in a channel flow Grid: 64x64x32 5 A priori test of the response of a RANS model to external LES perturbations ¾ Prescribed RANS layer by reference to the distance to the wall, or separate RANS and SGS models (ideally same type of model) ¾ A-priori test – overlap of the two domains ¾ A-posteriori test – two separate domains A priori study ¾ Rationale: • Identifying / quantifying the response of the RANS layer to LES ¾ Methodology • LES provides information to RANS • RANS does not provide information to LES • LES is solved down to the wall ¾ Case Description • Periodic channel flow • Reb = 10935 – DNS of Moser, Kim and Mansour (1999) • Computational domain: 2πh x 2h x πh • Grid: 96 x 64 x 64 with • Interface location: • SGS model: Smagorinsky 6 A-priori test results Instantaneous streamwise velocity profiles for the a-priori RANS and equivalent LES; Time history for the velocity U (y+ = 30) and Uτ for a-priori RANS and equivalent LES LES RANS A-priori test of Wall Function approach for LES Wall-normal variations of the correlations for the a-priori RANS and LES, (Temmerman, Leschziner & Hanjalic, 2002) 7 A-priori test results Time-averaged kinetic energy and eddy viscosity for the reference DNS, the a-priori RANS and the equivalent LES A posteriori study: coupled RANS and LES ¾ RANS and LES regions are solved using the same solver ¾ Coupling strategy: • Switch from RANS to LES at an imposed location and blending of RANS and LES viscosity on the LES-RANS interface; LES→RANS data transfer RAN S LES ° ° ° ° • • • • • Switch from RANS to LES controlled by wall distance d and cell side ∆ Prescribed y+int 8 Case Description: Fully developed channel flow • Reb = 10935 (Reτ=590) (DNS by Moser, Kim and Mansour (1999) • Reb = 40000 (Reτ=2000) (Experiments by Wei & Willmarth, 1998 • Computational domain: 2πh x 2h x πh Hybrid RANS/LES and Coarse LES • Grid: Nx × Ny × Nz= 64 x 32 x 64 with y+(1) = 0.5 • Interface locations: y+int=65 (Reτ=590); y+int=135 (Reτ=2000); Fine-resolved LES • Reτ=590 : Nx × Ny × Nz= 64 x 64 x 128 and 96x64x64, y+(1) = 0.5 • Reτ=2000: Nx × Ny × Nz= 512x128x128, y+(1) = 0.75 Channel Flow – Results, Reτ=2000 (L. Temmerman) Time-averaged velocity and shear stress profiles for the LES computations. 9 HRL Modelling Practice: One-eqn. RANS (L.Temmerman, 2001) L. ¾ RANS model: one-equation transport model for turbulence energy (Wolfshtein, 1969); ¾ SGS model: One-equation transport model for SGS energy (Yoshizawa and Horiuti, 1985) ¾ Assumption: RANS and LES grids are identical at the interface; ¾ Target: • Viscosity: ν tRANS = ν tLES ,int , int • Velocity: RANS U int = U int LES • Modelled energy: RANS . kmod, int = k mod,int LES A two-layer hybrid scheme: Matching criteria ¾ Matching criteria: continuity of total eddy viscosity at the interface ν SGS + ν res LES = ν t +ν ν res RANS res LES (ui'u j' − uk' uk' δ ij / 3) S ij = S ij S ij with overbar denoting filtered, and <> some local smoothing. ¾ Resolved stresses continuous across the interface ⇒ ν SGS =νt Cµ at the interface: One-eqn model: ν t = Cµ lµ k Cµ = ν SGS 0.5 lµ k RANS ,int 0. 5 k-ε model: ν t = Cµ f µ f µ ( k / ε )ν SGS 2 Cµ = ( f ( k / ε )) µ 2 k2 ε 2 10 Adjustment of Cµ Variant 1 ⎡1 − exp ( − y ∆ ) ⎤⎦ Cµ = 0.09 + ( Cµ ,int − 0.09 ) ⎣ ⎡⎣1 − exp ( − yint ∆ int ) ⎤⎦ Variant 2 ⎧ y+ + ⎪ C µ = 0.09 27 for y ≤ 27 ⎪ ⎨ ⎡C − 0.09 ⎦⎤ ⎣⎡1 − exp( − ( y − y ( y + = 34)) / ∆ ) ⎦⎤ ⎪ C = 0.09 + ⎣ µ ,int for y + > 27 + ⎪ µ ⎣⎡1 − exp( − yint − y ( y = 34) / ∆ int ) ⎤⎦ ⎩ Channel Flow – Results, Reτ=2000 One-equation RANS, (L. Temmerman) Time-averaged velocity profiles for the hybrid RANS-LES computations. 11 HRL: Channel Flow, Reτ=2000, One-equation RANS, (L. Temmerman) Interface issues in Two-equation RANS model ¾ Interface B.C. for 2-eqn RANS, kint and εint - options for k: c: Scale similarity a: k = k res int LES kint = k SGS = 0.5( Û i − U i )2 b: Isotropic spectrum distrib. kint = k SGS = 3 Cκ −2 / 3 ⎛ vSGS ⎞ π ⎜ ⎟ 2 CS8 / 3 ⎝ ∆ ⎠ where Û i - test-filtered velocity U i - filtered velocity 2 ¾ Interface B.C. for 2-eqn RANS, kint and εint - options for εint : ε int = k3/ 2 2.5 yn Or, from least-square error between the total viscosity on both sides of interface. 12 Channel Flow – Results, Reτ=2000 Two-equation RANS (M. Hadziabdic) Streamwise vorticity, Reτ=590 fine-resolved LES (96x64x64) coarse LES (64x64x32) ∆ z+ hybrid RANS/LES (64x64x32) RANS/LES interface 13 Zonal k-v2-f RANS/LES: Equations (Hadziabdic & Hanjalic 2003) • RANS model: k − ε −υ 2 − f ∂ ∂k ∂U j k = + ∂x j ∂x j ∂t ⎛ LRANS ⎝ LLES α = max⎜⎜1, ⎞ ⎟⎟ , ⎠ ⎛ ⎞ ⎜ (ν + ν t ) ∂k ⎟ + P − α ⋅ ε ⎜ ∂x j ⎟⎠ ⎝ LRANS = Cl α ≤1 RANS region 1 < α ≤ 1.5 Buffer region α >1 LES region 1.5 ktot ε , LLES = 0.8 ⋅ (∆X ⋅ ∆Y ⋅ ∆Z )1 / 3 ktot = k res + k mod • LES model: Dynamic Smagorinsky model Hybrid RANS (k-v2-f) / LES (dynamic): Velocity profiles 14 Hybrid RANS (k-v2-f) / LES (dynamic): Shear stress and kinetic energy Reτ=2000 Reτ=20000 Hill Flow – Case Description ¾ Periodic channel flow with constriction at both ends ¾ Re number based on channel height and bulk velocity is 21560 ¾ Data from highly resolved LES computations (5 x 106 nodes) by Temmerman et al (2003) ¾ Domain size: 9h × 3.036h × 4.5h (h=hill height) ¾ Grid details: • Discretisation for HRL 112 × 64 × 56 + • Near-wall resolution: yc (1) ≈ 1 • Spanwise and streamwise resolution: ∆x = ∆z 15 Hill Flow - Results (x/h)sep. = 0.22 (x/h)reat. = 4.72 (x/h)sep. = 0.21 (x/h)reat. = 5.30 (x/h)sep. = 0.23 (x/h)reat. = 4.64 (x/h)sep. = 0.23 (x/h)reat. = 5.76 Averaged streamlines for the reference simulation, LES, DES and RANS-LES cases. Hill Flow - Results Streamwise velocity profiles at x/h = 2.0. 16 Some observations ¾ Resolved motion in the URANS region is as strong as the resolved motion in the equivalent LES region. ¾ Hence, in the RANS region, both resolved and modelled contributions to the motion are substantial. ¾ The sum of both contributions is too high, hinting at the need of an ad hoc modification to reduce the total motion. ¾ Channel Flow: Flow Encouraging results. ¾ The response to the parameters change is small. ¾ Response to the location of the interface: in proportion of modelled motion; ¾ Hill Flow: Flow agreement with the reference data reasonable; ¾ Compared to the channel case, Cµ has a similar behaviour. ¾ Difficult to draw definitive conclusions because of the low Re. Concluding Remarks on HRL ¾ New hybrid RANS-LES (HLR) method allowing: • Freedom in locating the interface; • Dynamic adjustment of the RANS model to ensure continuity across the interface. ¾ For identical grids, the HRL results are significantly better than those obtained with LES for the same (coarse) mesh. ¾ Application to a recirculating flow: • Results are non-conclusive due to low Reynolds number; • The hybrid RANS-LES approach overestimates the recirculation zone length. ¾ Fundamental inconsistency in on the LES side next to RANS ¾ (unrealistic streaks structure, insufficient stress); ¾ Needs for further adjustment (smoothing, extra forcing, artificial backscatter, …) irrespective of RANS model 17 Schematics of TRANS –VLES rationale Semi-deterministic Modelling (SDM), (Ha Minh et al.) T-RANS Niche: HighHigh-Ra challenge in thermal RB convection ¾ Nu∝ Ra1/3 for Ra<1012 (Pr O(1)) ¾ Nu∝ Ra1/2 for Ra→∞ ¾ λv /H ∝ Ra-1/7 ¾ λθ/H ∝ Ra-1/3 18 T-RANS EQUATIONS AND SUBSCALE MODELS: ⎞ 1 ∂( P − Pref ) ∂ Ui ∂ Ui ∂ ⎛⎜ ∂ Ui + βgi ( T − Tref ) + Uj = ν − τij ⎟ − ⎟ ρ ∂xi ∂t ∂x j ∂x j ⎜⎝ ∂x j ⎠ ∂T ∂t ∂C ∂t + Uj + Uj ∂T ∂x j = ⎞ ∂ ⎛⎜ ν ∂ T − τ θj ⎟ ⎜ ⎟ ∂x j ⎝ Pr ∂x j ⎠ ⎞ ∂ ⎛⎜ ν ∂ C = − τcj ⎟ ⎟ ∂x j ∂x j ⎜⎝ Sc ∂x j ⎠ ∂C +final closure: 3eqn. model Dk = Dk + Pk + Gk − ε Dt Dε Dt D θ2 Dt = Dε + Pε1 + Pε 2 + G ε − Y = Dθ + Pθ − ε θ assuming weak equilibrium (D / Dt − Diff)ϕui = 0 Subscale ASM/AFM/ACM ⎤ ⎡ ∂T ∂ Ui + ξτθj + ηβ g i θ 2 ⎥ ⎢τij ∂x j ⎥⎦ ⎣⎢ ∂x j τθi = −Cφ k ε τ ci = −C φ ∂ Ui ⎤ k ⎡ ∂C + ξτ cj ⎢ τ ij ⎥ ε ⎣⎢ ∂x j ∂x j ⎥⎦ ⎛ ∂ Ui ∂ Uj τ ij = − ν t ⎜ + ⎜ ∂x j ∂x i ⎝ ⎞ 2 ⎟ + k δ + C k β g θu ij i j ⎟ 3 ε ⎠ Verification: Long-term averaged temperature profiles and heat flux in R-B convection for different Ra numbers; DNS and TRANS (Kenjereš and Hanjalić, 1999-2003) ^^ ~~ Φ Ψ − ΦΨ = Φ Ψ + φψ α ∂T ~ ~ q − T W − θw = w = const. ∂z ρcp Wall scaling with heat-fluxbased buoyancy velocity 19 MON1(z/D=0.5,x/L=0.5, y/L=0.5) MON2(z/D=0.01,x/L=0.5,y/L=0.5) T-RANS LES Time spectra of <U>, <V>, <W> and <T> signals at characteristic monitoring points, Ra=109 Mean vertical profiles of temperature for different Ra 20 High-Ra number challenge in thermal RB convection ¾ Nu∝ Ra1/3 for Ra<1012 (Pr O(1)) ¾ Nu∝ Ra1/2 for Ra→∞ Ra1/2 Ra1/3 Ra-1/7 ¾ λv /H ∝ Ra-1/7 Ra-2/9 ¾ λθ/H ∝ Ra-1/3 Ra-1/3 z z CASE (I): weak CASE (II): strong ∆Τ=1 T=T(x,y,z,τ) z/H=2/3 1600m 800m ∆Τ=2 ∆T=4 z/H=1/3 C=C(x,y,z,τ) Residential Industrial INITIAL STRATIFICATIONS y TEMPERATURE CONCENTRATION y T-RANS of pollutant dispersion in a town valley 21 Q0>0 Q0<0 TIME INITIAL STRATIFICATION CASE (I): weak CASE (II): strong Instantaneous trajectories in vertical plane over hilly terrain T-RANS of pollutant dispersion in a town valley Passive pollutant dispersion visualized by concentration isosurface Evolution of the pollutant front (C=0.05 Cmax): strong stratification 22 Diurnal winds over Arctic ice sheet 380 km Solution domain: 380x45x3 km Mesh: 180x40x40 45km 3 km Ra~1010 , Pr~ 1 Assumed near-ground temperature V. van Huijen, S. Kenjeres and K.Hanjalic Some instantaneous streamline patterns over an ice sheet Wind velocity profiles: comparison with measurements 23 Some Conclusions on T-RANS/VLES: ¾ Both RANS and LES will long be in use, each in its niche, but industry cannot count on LES for large-scale problems in the foreseeable future ¾ There will be an increasing research on merging RANS and LES strategies for very high Re and Ra numbers and complex flows ¾ T-RANS based VLES captures well main flow features in flows dominated by large-scale (pseudo)deterministic structures ¾ T-RANS can be used to predict natural convection at very high Ra and in complex domains, which are inaccessible to LES, DNS or other methods. References: 1. 2. 3. 4. 5. 6. 7. Balaras, E., Benocci, C., Piomelli, U., Two-layer approximate boundary conditions for large-eddy simulations, AIAA Journal 34 (1996), 1111-1119. Cabot, W., Moin, P., Approximate wall boundary conditions in the large eddy simulation of high Reynolds number flow, Flow, Turbulence and Combustion, 63 (1999), 269-291. Hanjalic, K., Hadziabdic, M., Temmerman, L. and Leschziner M., Merging RANS and LES strategies: zonal or seamless coupling, Invited lecture DLES V, Munchen Aug. 27-29, 2003 (to appear in R. Friedrich, B. Geurs and O. Metais, (eds) Durect and Large-Eddy Simulations V, Kluwer Acad. Publ. 2004 L.Temmerman, M.A.Leschziner, K.Hanjalic, A priori studies of a near-wall RANS model within a hybrid LES/RANS scheme , 5th Internacional Symposium on Engineering Turbulence Modelling and Measurements, Mallorca, Spain, 16-18 September, 2002 Spalart, P.R., Jou, W-H., Strelets, M., Allmaras, S.R., Comments on the feasibility of LES for wings and on the hybrid RANS/LES approach, in Advances in DNS/LES, 1st AFOSR Int. Conf. On DNS/LES (Greden Press) (1997). Spalart P.R. and Allmaras, S.R., A one-equation turbulence model for aerodynamic flows. AIAA Paper 92-0439. (1992). Temmerman, L., Leschziner, M., Mellen, C. and Froehlich J., Investigation of subgrid-scale models and wallfunction approximations in Large Eddy Simulation of separated flow in a channel with streamwise periodic constrictions, Int. J. Heat Fluid Flow (to appear). 24