Turbulence modelling. Sørensen, Niels N. Publication date: Link back to DTU Orbit

Transcription

1 Downloaded from orbit.dtu.dk on: Dec 19, 2017 Turbulence modelling Sørensen, Niels N. Publication date: 2010 Link back to DTU Orbit Citation (APA): Sørensen, N. N. (2010). Turbulence modelling. Paper presented at 1st Wind Turbine Computational Aerodynamics Lecture Series, United Kingdom. General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. Users may download and print one copy of any publication from the public portal for the purpose of private study or research. You may not further distribute the material or use it for any profit-making activity or commercial gain You may freely distribute the URL identifying the publication in the public portal If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

2 Turbulence modelling Glasgow Niels N. Sørensen Wind Energy Division RISØ-DTU,

3 Outline 1 RANS turbulence modeling 2 Algebraic turbulence models 3 Transport equation based turbulence models 4 LES and DES modeling 5 Laminar turbulent transition 6 The end 2 of 43

4 The Nature of turbulence Irregularity Turbulence is irregular or random. Diffusivity Turbulent flow causes rapid mixing, increases heat transfer and flow resistance. These are the most important aspect of turbulence from a engineering point of view. Three-dimensional vorticity fluctuations (rotational) Turbulence is rotational, and vorticity dynamics plays an important role. Energy is transferred from large to small scale by the interaction of vortices. Dissipation Turbulent flow are always dissipative. Viscous shear stresses perform deformation work which increases the internal energy of the fluid at the expenses of kinetic energy of turbulence. Continuum Even though they are small the smallest scale of turbulence are ordinary far lager than any molecular length scale Flow feature Turbulence is a feature of the flow not of the fluid. 3 of 43

5 How Does Turbulence Look The Onset of Two-Dimensional Grid Generated Turbulence in Flowing Soap Films Maarten A. Rutgers, Xiao-lun Wu, and Walter I. Goldberg 4 of 43

6 Modeling Turbulent Flows, DNS Direct Numerical Simulation (DNS) of channel flow All Scales of the fluid motion spatial and temporal are resolved by the computations. The largest DNS to date is Re τ = uτ H/2 ν N 3 DNS Re 2.25 τ Re H Re τ N 3 DNS Timesteps of 43

7 Modeling Turbulent Flows, LES Large Eddy Simulation (LES) of channel flow Only the large spatial and temporal scales are resolved by the computations uτ H/2 Re τ =. ν ( ) 0.4 N 3 LES Re 2.25 Reτ 0.25 τ Re H Re τ N 3 DNS N 3 LES Timesteps of 43

8 Modeling Turbulent Flows, RANS Reynolds Averaged Navier-Stokes (RANS) of channel flow The resolve the equations are time averaged and do not resolve the eddies Re τ = uτ H/2 ν Re H Re τ N 3 DNS N 3 LES N 3 RANS A hybrid model can be developed using LES/RANS 7 of 43

9 RANS turbulence modeling Navier-Stokes equations Incompressible Navier-Stokes equation The flow equations and additional equations have the following form: Continuity equation: Momentum equations: t (ρu i)+ (ρ(u i U j )) Auxiliary equations: (ρu j ) = 0 t (ρφ)+ (ρ(u j φ)) [ ( Ui µ + U )] j + P = S v, x i x i [ µ φ ] = S φ x i 8 of 43

10 RANS turbulence modeling Reynolds Averaged Navier-Stokes Reynolds averaging of the Navier-Stokes equation, splitting the velocities in the mean and the fluctuating component u i ( r, t) = U i ( r)+u 1 t+t ( r, t), where U i ( r) = lim u i ( r, t)dt T T t Inserting the Reynolds decomposed velocity in the Navier-Stokes and continuity equations Perform time averaging of the equations. The equations are in principle time independent, or steady state. 1 t+t U i ( r)) = lim U i ( r)dt = U i ( r) T T t 1 t+t [ u i (r, t) = lim ui ( r, t) U i ( r) ] dt = U i ( r) U i ( r) = 0 T T t 9 of 43

11 RANS turbulence modeling The Reynolds Averaged Navier-Stokes The flow equations and additional equations have the following form: Continuity equation: Momentum equations: (ρu j ) = 0 t (ρu i)+ (ρ ( [ ( ) U i U j + u i u j Ui ) µ + U )] j + P = S v, x i x i Auxiliary equations: t (ρφ)+ (ρ(u j φ+u jφ )) [ µ φ ] = S φ x i 10 of 43

12 RANS turbulence modeling Reynolds Stresses Performing the Reynolds Averaging Process, new terms has arisen, namely the Reynolds stress tensor: τ ij = ρu i u j This brings us at the turbulent closure problem, the fact that we have more unknowns than equations Three velocities+pressure+six Reynolds-stresses Three momentum-equations+continuity equation To close the problem, we need additional equations to model the Reynolds-stresses 11 of 43

13 RANS turbulence modeling The Reynolds Averaged Momentum-equations The Reynold Stresses originates from the convective terms t (ρu i)+ (ρ ( [ ( ) U i U j + u i u j Ui ) µ + U )] j + P = S v, x i x i The Reynold Stresses are often treated together with the diffusive terms t (ρu i)+ (ρ(u i U j )) [ ( Ui µ + U ) ] j u i u j + P = S v, x i x i 12 of 43

14 RANS turbulence modeling Reynold-stress equations Performing the following operation on the Navier-Stokes equation, equations for the Reynold-stresses can be derived: u i NS(u i)+u j NS(u i) = 0 τ ij τ ij t +U k x k U i = τ jk +2µ u u i j +u p i + u p j + [ ν τ ] ij +ρu i x k x k x k x k x u j u k k The procedure introduces new unknowns (22 new unknowns) ρu i u j u k 10 unknowns u i 2µ u i x k u j x k p + u j p 6 unknowns 6 unknowns 13 of 43

15 RANS turbulence modeling Boussinesq Eddy Viscosity Approximation Eddy-viscosity models Compute the Reynolds-stresses from explicit expressions of the mean strain rate and a eddy-viscosity, the Boussinesq eddy-viscosity approximation τ ij = ρu i u j = 2µ ts ij 2 3 ρkδ ij,, where S ij = 1 2 ( Ui + U ) j, and 2ρk = τ ii = ρu i x u j. i The k term is a normal stress and is typically treated together with the pressure term 14 of 43

16 RANS turbulence modeling The Reynolds Averaged Navier-Stokes The flow equations and additional equations have the following form: Continuity equation: Momentum equations: (ρu j ) = 0 t (ρu i)+ (ρu i U j ) [ ( Ui (µ µ t) + U )] j + ˆP = S v, x i x i Auxiliary equations: t (ρφ)+ (ρu j φ) [( µ+ µt σ φ ) ] φ = S φ x i 15 of 43

17 RANS turbulence modeling RANS turbulence models Algebraic turbulence models Prandtl Mixing Length Modeling Cebeci-Smith Modeling Baldwin-Lomax Model One equations turbulence models Spalart-Allmaras Baldwin-Barth Two equation turbulence models k ǫ model k ω model k τ models Reynolds stress models 16 of 43

18 Algebraic turbulence models Algebraic Turbulence Model Prandtls mixing length hypothesis is based on an analogy with momentum transport on a molecular level Molecular transport t xy = µ du dy, where µ = 1 2 ρv thl fmp Turbulent transport τ xy = µ t du dy, where µt = 1 2 ρv mixl mix y U(y) and, v mix = c 1 l mix du dy, and l mix = c 2 y 17 of 43

19 Algebraic turbulence models Prandtl Mixing Length Model The mixing length model closes the equations system Turbulent transport τ xy = µ t du dy, where µt = 1 2 ρv mixl mix and, v mix = c 1 l mix du dy, and l mix = c 2 y The proportionality constant for the mixing velocity c 1 and for the mixing length c 2 needs to be specified The equation for the turbulent eddy viscosity is a part of the flow solutions, as it depends on the mean flow gradient turbulence is not a fluid property but a property of the flow 18 of 43

20 Algebraic turbulence models Baldwin-Lomax Model { µ t = Ω = 2Ω ij Ω ij, and Ω ij = 1 2 µ inner = ρl 2 Ω if y y cross µ outer = ρkc wpf wake F Kleb (y) if y > y cross. y cross = min(y) where µ outer = µ inner ( Ui ( udiff 2 F wake = min y maxf max, C wk y max, and u diff = max F max U ) ) j, and l = κy (1 e y+ A x + i ) [, and F Kleb = ( ) ( ) Ui U k min Ui U k ( yckleb y max ) 6 ] 1 ) y max and F max is determined by the maximum of F(y) = y Ω (1 e y+ A + 19 of 43 A + = 26, C cp = 1.6, C Kleb = 0.3, C wk = 0.25, κ = 0.4, K =

21 Algebraic turbulence models Algebraic Models Gives good results for simple flows, flat plate, jets, simple shear layers and airfoils at low AOA Typically the algebraic models are fast and robust Needs to be calibrated for each flow type, they are not very general They are not well suited for computing flow separation Typically they need information about the boundary layer properties and wake location, and are difficult to incorporate in modern flow solvers. 20 of 43

22 Transport equation based turbulence models One and Two Equation Turbulence Models The derivation is again based on the Boussinesq approximation τ xy = µ t du dy, where µt = 1 2 ρv mixl mix The mixing velocity is determined by the turbulent kinetic energy v mix k 1 2, and k = 1 2 u i u j The mixing length is determined from an algebraic relation or using another transport equation l mix k 3 2 ǫ, where ǫ = ν u i x k u i x k 21 of 43

23 Transport equation based turbulence models Second equation Proposer(s) (year) Variable Symbol Kolmogorov (1942) k 1 2 /l ω Saffman (1970) Wilcox et al. (1972) Chou (1945) k 3 2 /l ǫ Davidov (1961) Harlow-Nakayama (1968) Jones-Launder (1972) Rotta (1951) l l Rotta (1968,1971) kl kl Rodi-Spalding (1970) Ng-Spalding (1972) Spalding (1969) k/l 2 W 22 of 43

24 Transport equation based turbulence models The k-equation By taking the trace of the Reynolds Stress equation we get ρ k t +ρu k j = τ ij U i ρǫ+ [ µ k 1 ] 2 ρu i u i u j p u j Using that 1 2 ρu i u i u j p u i = µ t k σ k ρ k t +ρu k j where k = 1 2 u i u i and ǫ = ν u i x k u i x k = τ ij U i assuming u i φ µ t φ ρǫ+ [( µ+ µt σ k we get ) ] k 23 of 43

25 Transport equation based turbulence models Dissipation of turbulent kinetic energy The equation for the dissipation of turbulent kinetic energy is derived by the following operation on the Navier-Stokes equations. ρ ǫ t +ρu ǫ j = 2µ U i NS(u i ) = 0 [ u i u i + u k x k x k x i [ 2ν u i [ ] 2 u i 2 2µ x m x k u k ] u j ǫ+ν p x m u j x m + 2µ u i u i u k x k x m x m ] 24 of 43

26 Transport equation based turbulence models The k ǫ model The standard k ǫ model is only applicable for high Reynolds numbers, and therefore needs wall modelling. Eddy viscosity µ t = ρc µ k 2 Transport equation for turbulent kinetic energy ρ k t +ρu k j = τ ij U i ǫ ρǫ+ [( ) ] µ+ µt k σ k Transport equation for dissipation of turbulent kinetic energy ρ ǫ t +ρu ǫ j ǫ = C ǫ1 k τ U i ǫ 2 ij ρc ǫ2 k + [( ) ] µ+ µt ǫ σ ǫ C ǫ1 = 1.33, C ǫ2 = 1.92, C µ = 0.09, σ k = 1.0, σ ǫ = 1.0, 25 of 43

27 Transport equation based turbulence models The k ǫ model, low Reynolds number version ρ k t +ρu j k = τ ij U i k 2 µ t = ρc µf µ ǫ ρǫ+ [( ) ] ( ) µ+ µt k k µ σ k Transport equation for dissipation of turbulent kinetic energy ρ ǫ t +ρu ǫ j ǫ = C ǫ1 f 1 k τ U i ǫ 2 ij ρc ǫ2 f 2 k + E + [( ) ] µ+ µt ǫ σ ǫ where ǫ = ǫ ǫ 0 and ǫ 0 is the value of ǫ at the wall and boundary conditions k = ǫ 0 = of 43

28 Transport equation based turbulence models The k ω model The k ω model of Menter is a blend of the original k ω model of Wilcox, the k ǫ model and the Bradshaw et al. model. Eddy viscosity: µ t = ρa 1 k max(a 1 ω; F 2 Ω) The model is tuned through the F 2 function to switch to the Bradshaw assumption in adverse pressure gradients µ t = τ Ω = ρ a 1k Ω, a 1 = of 43

29 Transport equation based turbulence models The k ω model Eddy viscosity µ t = ρa 1 k max(a 1 ω; F 2 Ω) Transport equation for turbulent kinetic energy ρ k t +ρu k j = τ ij U i ρβ kω + [( ) ] µ+ µt k σ k Transport equation for the specific dissipation rate ρ ω t +ρu ω j = γ U i τ ij βρω ρ(1 F 1 ) ν t σ ω2 ω k ω + [( µ+ µt σ ω ) ] ω 28 of 43

30 Transport equation based turbulence models Blending Function s F 1 and F 2 ( ) ( ) F 1 = tanh arg1 4 and F 2 = tanh arg2 4 where arg 1 = min [ max ( ) k 0.09ωy ; 500µ ; y 2 ω 4ρk CD kω y 2 σ ω2 ( ) k arg 2 = max ωy ; 500µ 2, y 2 ω ], and CD kω represents the cross-diffusion term ( CD kω = max 2ρ 1 ) k ω ; σ ω2 ω. 29 of 43

31 Transport equation based turbulence models Constants for the k ω SST model Constants for the inner region (k ω) β 1 β γ 1 σ k1 σ ω1 a Constants for the outer region (k ǫ) β 2 β γ 2 σ k2 σ ω2 a The actual constant is obtained by blending the inner and outer constants α = α inner F 1 +α outer(1 F 1 ) 30 of 43

32 Transport equation based turbulence models Wall boundary conditions for the k ω model The k ω model is valid all the way to the wall, and do not need low Re modifications The boundary conditions are very simple to apply. At the wall we have k = 0, and ω = 10 6ν β 1 y 2 The velocity boundary conditions is a simple no-slip condition. The model is robust in the low Re version, and only demands y of 43

33 Transport equation based turbulence models Farfield boundary conditions for the k ω model Typically, the inflow turbulence intensity is known k = IU 2 For many aerodynamic applications, like airfoil computations, inflow is often nearly laminar in the farfield ν t << ν This can be approximated by ν t = ν This leads to ω = C µ k ν t = C µ k ν For wall bounded cases, the inflow value of ω can often be computed using the mixing length hypothesis assuming a velocity profile. 32 of 43

34 LES and DES modeling Large Eddy Simulation Filtering of the Navier-Stokes equation, splitting the velocities in the resolvable-scale or filtered velocity and the sub-grid scale (SGS) velocity u i = u i + u u i ( r, t) = G( r ξ; )u i ( ξ, t)d 3 ξ with G( r ξ; )d 3 ξ = 1 A typical filter could be the volume-averaged box filter { G( r ξ; ) 1/ 3, r = i ξ i 0, otherwise 33 of 43

35 LES and DES modeling LES filtering of the Navier-Stokes equations As with the Reynolds Averaging procedure, the filtering generates new terms from the convective terms where u i u j = u i u i + L ij + C ij + R ij L ij = u i u j u i u j C ij = u i u j + u j u j R ij = u i u j Filtering differs from standard averaging in one important respect u i u i 34 of 43

36 LES and DES modeling frame The Leonard stresses (L) are of the same order as the typical truncation error in a second-order accurate scheme The cross-term stress tensor (C) are typically modeled together with the Reynolds stresses Q ij = R ij + C ij The first model for the sub-grid scale stresses (SGS) was the model by Smagorinsky (1963) based on the gradient diffusion approximation τ ij = 2µ ts ij, where S ij = 1 2 ( ui + u ) j x i µ t = ρc 2 s 2 S ij S ij, and C s = [0.10 : 0.24] 35 of 43

37 LES and DES modeling LES filtering of the Navier-Stokes equations Smagorinsky model t (ρu i)+ (ρu i u j ) = τ ij = ρ [ ( ui µ + u j x i (Q ij 13 ) Q kkδ ij = 2µ ts ij ) τ ij ] P + S v P = p Q kkδ ij Q ij = R ij + C ij µ t = ρ(c s ) 2 S ij S ij, and C s [0.10 : 0.24] 35 of 43

38 LES and DES modeling LES modeling LES models are by nature unsteady LES models are by nature fully three dimensional They resolve the large scales and only model the isotropic small scales The standard SGS model needs damping of the eddy viscosity near solid walls similar to the van Driest damping Resolving the anisotropic eddies in the near wall region where the scale are small may require a very fine grid Pure LES models are to computational demanding for most airfoil and rotor computations LES models can be combined with approximate wall boundary conditions, or even zero, one or two equation models for the region near walls 36 of 43

39 LES and DES modeling Hybrid models The original DES model of Spallart et al. based on the Spallart-Allmaras model The k ω DES model The k ω DDES model The SAS model 37 of 43

40 LES and DES modeling DES and DDES modeling The idea behind the DES modeling is to exchange the turbulent length scale with the grid size when the turbulent length scale become larger than the grid size, and the grid is capable of resolving some of the scales. The turbulent length scale in the k ω model is given by L k ω t = k 3 2 k ǫ = β ω using ǫ = β kω The turbulent length scale in a LES would be L LES t = C Des with = min[ x, y, z] To enforce the L LES t when the grid allows, we will simply scale the dissipation term in the ω equation by the ratio between L k ω t and L LES ( ) β ωk = β L k ω t ωk max F L DES Shield ; 1 t t. 38 of 43

41 LES and DES modeling DES and DDES modeling The idea behind the DES modeling is to exchange the turbulent length scale with the grid size when the turbulent length scale become larger than the grid size, and the grid is capable of resolving some of the scales. The turbulent length scale in the k ω model is given by L k ω t = k 3 2 k ǫ = β ω using ǫ = β kω The turbulent length scale in a LES would be L LES t = C Des with = min[ x, y, z] To enforce the L LES t when the grid allows, we will simply scale the dissipation term in the ω equation by the ratio between L k ω t and L LES β ωk = β ωk max ( ) L k ω t (1 F L DES Shield ); 1 t t. 39 of 43

42 LES and DES modeling DES and DDES modeling Grid Induced Separation, or Modeled Stress Depletion (MSD) One should avoid that the LES part of the model is active within the boundary layer The boundary layer can be shielded by a damping function eg. the F 1 or the F 2 blending functions from the k ω SST model 40 of 43

43 Laminar turbulent transition Laminar turbulent transition Often the flow is not fully turbulent, this was already noticed by Osbourne Reynolds in 1883 The transition process depend on many parameters Reynolds Number Free stream turbulence level Laminar separation bubbles Cross flow Surface Roughness Mass injection Typically approaches for transition modeling e n method (Orr-Sommerfeld eqn.) Empirical Correlations Michel Mayle Abu-Ghannam and Shaw Suzen 41 of 43

44 Laminar turbulent transition The γ Re θ Correlation based transition model The model is based on comparing the local Momentum Thickness Reynolds number with a critical value from empirical expressions Re θ = Re θt The following relation is used to simplify the computations in a general CFD code Re θ = Reν t max The model is based on transport equations, and can easily be implemented in general purpose flow solvers In the present form the model handles natural transition, by-pass transition, and separation induced transition The transition model is coupled to the k ω SST model through the production and destruction terms in the k-equation 42 of 43

45 The end Conclusion I have talked briefly about the basic property of turbulence The boussinesq approximation has been given The RANS, LES and DES methods has been introduced A few popular modelling approaches has been shown 43 of 43