Turbulence model selector for Simcenter STAR-CCM+
Answer questions about your flow and get a starting turbulence model, wall treatment and time description for Simcenter STAR-CCM+, with the checks that go with it.
The selector could not start. The article below explains its rules, or ask us at support@volupe.com.
Step 1 of 6 · Your case
What are you simulating?
Which turbulence model should I use?
There is no single best turbulence model, so the selector starts from your flow and the result you need. As a starting point, k-omega SST (k–ω SST) suits external flows and flows with separation, rotation or a free surface, and realizable k-epsilon Two-Layer suits most internal flows. Spalart–Allmaras suits attached boundary layers on wings and aircraft, a Reynolds stress model suits swirl-dominated flows such as cyclones, and a transition model is needed when the laminar-to-turbulent transition of a wall boundary layer changes your result.
The decision map and the table below list the rules in the order the selector checks them, and the first rule that matches your answers sets the starting model. The order is Volupe’s engineering judgement. The model names, options and wall treatments come from the Simcenter STAR-CCM+ 2606 User Guide [1], and every result also lists what is still open and what to check in the solution.
The same rules as a table, with every comparison
| # | If your answers say | Starting model | Compared with |
|---|---|---|---|
| 1 | The flow is laminar | No turbulence model | – |
| 2 | The flow regime is not known yet | Provisional: realizable k–ε Two-Layer for internal flow, k–ω SST for external, rotating and marine flow | The other of the two |
| 3 | Resolved fluctuations are needed and a scale-resolving run is affordable | With transition that is not tripped: wall-resolved WALE LES. Wind around buildings: WALE LES with synthetic or precursor inflow. Separation, rotation, a free surface or a vehicle: k–ω SST IDDES. Otherwise: WALE LES | k–ω SST IDDES + Gamma Transition, and k–ω SST + Gamma Transition (URANS) for a forced boundary layer, with transition. k–ω SST DDES, k–ω SST + SRH and Spalart–Allmaras DES, with IDDES. k–ω SST + SRH, otherwise |
| 4 | Transition matters | In a wall boundary layer under an imposed flow, not tripped: k–ω SST + Gamma Transition. Tripped at a known place: the rules below decide, with Turbulence Suppression upstream of the trip. Driven by buoyancy: standard k–ε Low-Re. In a pipe or the bulk flow, or not known: provisional realizable k–ε Two-Layer or k–ω SST, as in rule 2 | Gamma-ReTheta and SA Gamma (airfoils) for the first case. Realizable k–ε Two-Layer and wall-resolved WALE LES for buoyancy-driven transition |
| 5 | Wind around buildings | Equilibrium, constant-density profile: standard k–ε with the Atmospheric Boundary Layer model. Stratified, not in equilibrium or compressible: provisional realizable k–ε Two-Layer | Realizable k–ε Two-Layer and k–ω SST |
| 6 | Buoyancy dominates | Realizable k–ε Two-Layer, Two-Layer Type Buoyancy Driven (Xu) | Standard k–ε Low-Re, EB k–ε |
| 7 | A swirl-dominated core, or a cyclone or separator | Elliptic Blending RST | k–ω SST |
| 8 | Rotor, blade or bend curvature | k–ω SST, with the Curvature Correction Option to assess | Elliptic Blending RST, and Lag EB k–ε for external flow |
| 9 | A jet | Free jet, plume or shear layer: realizable k–ε Two-Layer (realizable k–ε on a wall-function mesh). Impinging jet, or a jet whose wall heat transfer is a result you need: k–ω SST | k–ω SST and Standard (Wilcox) k–ω for a free jet. EB k–ε and realizable k–ε Two-Layer for an impinging jet |
| 10 | Separation, rotating parts, a free surface, or a result that depends on separation onset (outside airfoils) | k–ω SST | Realizable k–ε Two-Layer. Spalart–Allmaras for airfoils and, after the VOF Guidelines, for hulls. k–ω SST IDDES with a scale-resolving budget |
| 11 | Internal flow where heat transfer is a result you need | Wall Y+ below 1 on every wall: EB k–ε. Any other near-wall mesh: realizable k–ε Two-Layer | Realizable k–ε Two-Layer and k–ω SST, or EB k–ε and k–ω SST |
| 12 | Airfoil or aircraft | Result sensitive to separation onset or stall: k–ω SST. Otherwise: Spalart–Allmaras | Spalart–Allmaras, or k–ω SST |
| 13 | Any other internal flow | Realizable k–ε Two-Layer | EB k–ε, k–ω SST |
| 14 | Everything else: external, rotating and marine flow | k–ω SST | Lag EB k–ε |
Transition has its own branch. If the boundary layer on a wall goes from laminar to turbulent by itself, and that changes drag, lift or heat transfer, the selector adds Gamma Transition to k–ω SST. If you trip the boundary layer at a known place, you already know where the transition happens, so the selector treats the flow as fully turbulent and points to Turbulence Suppression for the laminar part upstream of the trip. When buoyancy drives the transition, as on a heated wall, the selector gives standard k–ε Low-Re, which the documentation recommends for natural convection. Low-Re k–ε models are not transition models, though: the documentation says the transition behaviour they can show is a mathematical side effect, not a designed feature, so check the transition location against data, or resolve it with wall-resolved LES. Our post on transition modelling in Simcenter STAR-CCM+ goes through Gamma, Gamma-ReTheta and Turbulence Suppression.
The selector does not switch on fixed limits such as Mach 0.3 or a Richardson number, because whether they matter depends on the quantity you need. Mach 0.3 is the rule of thumb for when density changes reach about 5%, not a limit in the solver. Our dimensionless numbers calculator gives the usual limits, so use it to answer the flow-features step.
Starting points by application
These are the starting points the selector gives for common applications, with the answers each row assumes. Open a row in the selector to change an answer and see what moves.
| Application | Answers assumed | Starting model | Wall treatment | Compared with |
|---|---|---|---|---|
| Pipe or duct, pressure loss | Fully turbulent, steady RANS, mixed near-wall mesh | Realizable k–ε Two-Layer | Two-Layer All y+ | EB k–ε, k–ω SST |
| Diffuser or valve | Separation, steady RANS | k–ω SST | All y+ | Realizable k–ε Two-Layer |
| Electronics cooling by natural convection | Buoyancy dominates, resolved near-wall mesh | Realizable k–ε Two-Layer, Buoyancy Driven (Xu) | Two-Layer All y+ | Standard k–ε Low-Re with Yap, EB k–ε |
| Data centre, heat transfer as a result | Wall-function mesh | Realizable k–ε Two-Layer | Two-Layer All y+ | EB k–ε (the selector starts from it only with Wall Y+ below 1), k–ω SST |
| Room ventilation, mean age of air | Mixed convection, heat and mean age of air as results, wall-function mesh | Realizable k–ε Two-Layer | Two-Layer All y+ | EB k–ε, k–ω SST |
| Cyclone separator | Swirl-dominated core | Elliptic Blending RST | All y+ | k–ω SST with the Curvature Correction Option |
| Centrifugal pump, head and efficiency | Rotor curvature, rotating reference frame, steady RANS | k–ω SST, Curvature Correction Option to assess | All y+ | Elliptic Blending RST |
| Airfoil, fully turbulent | Attached flow, not close to stall | Spalart–Allmaras | All y+ | k–ω SST |
| Airfoil with transition | Not tripped, Wall Y+ below 1 on every wall | k–ω SST + Gamma Transition | Low y+ | k–ω SST + Gamma-ReTheta, Spalart–Allmaras + SA Gamma |
| Vehicle drag | Separated wake, steady RANS | k–ω SST | All y+ | Realizable k–ε Two-Layer, k–ω SST IDDES if scale-resolving is affordable |
| Ship hull | Free surface, steady RANS | k–ω SST | All y+ | Realizable k–ε Two-Layer, Spalart–Allmaras |
| Wind around buildings | Equilibrium atmospheric profile, z₀ known | Standard k–ε with the Atmospheric Boundary Layer model | High y+ | Realizable k–ε Two-Layer, k–ω SST |
| Impinging jet cooling | Heat transfer at the wall, resolved near-wall mesh | k–ω SST | All y+ | EB k–ε, realizable k–ε Two-Layer |
| Vehicle side mirror, broadband noise | Resolved fluctuations, scale-resolving budget | k–ω SST IDDES | All y+ | k–ω SST DDES, k–ω SST + SRH, Spalart–Allmaras DES |
A centrifugal pump where you need the head and the efficiency shows how the steps work. Under Your case you choose Rotating equipment and then Pump or liquid machine, so rotation is already included. The flow is fully turbulent, the result is the mean flow or performance and the budget is steady RANS. Under flow features you tick Rotor, blade or bend curvature, and for What must the rotation represent? you choose The mean effect of rotation, which the result shows as a rotating reference frame. Rule 8 then gives k–ω SST with All y+ Wall Treatment and asks you to assess the Curvature Correction Option, which Siemens keeps for strong streamline curvature or rapidly rotating frames.
For the pressure pulsations from the blades passing the volute tongue, add periodic or transient loads to the result, choose Blade passing or changing relative positions for the rotation and set the budget to transient RANS (URANS). The model stays k–ω SST, and the time description becomes Implicit Unsteady. For the time step, the Axial Fans section of the aeroacoustics guidelines in the documentation aims for about 1° of rotation per step, less with more than 24 blades, and at least 15 steps per blade passage. Our post on the turbomachinery workflow adds that it can be a good idea to run the steady case first and use it as the initial guess for the transient run.
Tick Swirl-dominated core instead, as you would for a cyclone, and rule 7 gives Elliptic Blending RST. Keep it for cases where the swirl itself is what you need to predict.
k-epsilon vs k-omega SST
k-epsilon vs k-omega SST (k–ε vs k–ω SST) is mostly a question of the near-wall region. Siemens calls k–ε [4] a good compromise of cost and accuracy for industrial flows with complex recirculation, with or without heat transfer. k–ω models do better in boundary layers under adverse pressure gradients and can be integrated through the viscous sublayer. k–ω SST combines the two, with k–ω near the wall and k–ε behaviour in the free stream [2].
| Aspect | Realizable k–ε Two-Layer | k–ω SST |
|---|---|---|
| Near the wall | Two-layer formulation with Two-Layer All y+ Wall Treatment, which works on fine meshes and is the least inaccurate between y+ 1 and 30 | k–ω integrated through the viscous sublayer, with All y+ Wall Treatment (High y+ and Low y+ also exist) |
| Adverse pressure gradients | Less accurate than k–ω | What SST was built for [2] |
| Free stream | k–ε behaviour | k–ε behaviour away from the wall, because Wilcox’s k–ω is sensitive to free-stream values |
| Inlet variables | k and ε | k and ω |
| Name in Simcenter STAR-CCM+ | Realizable K-Epsilon Two-Layer, the default k–ε model | SST (Menter) K-Omega, the default k–ω model |
| In the selector | Internal flow, buoyancy-dominated flow, free jets | External flow, separation, rotation, free surfaces, impinging jets |
In the selector, realizable k–ε Two-Layer is the starting model for internal flows and buoyancy-dominated flows, and k–ω SST for external flows, separation, rotation and free surfaces. The two families solve for different variables, k and ε or k and ω, so the inlet values differ too, and the turbulence boundary conditions calculator gives both.
k-omega SST
k-omega SST (shear stress transport), also written SST k-omega, is Menter’s model from 1994: k–ω near the wall, k–ε behaviour away from it, and a limit on the eddy viscosity in adverse pressure gradients [2]. Menter, Kuntz and Langtry describe a 2003 revision and ten years of industrial use [3]. Simcenter STAR-CCM+ calls its SST model SST (Menter) K-Omega, and it is the default k–ω model. For most k–ω applications the User Guide recommends it with All y+ Wall Treatment and the default parameters.
Three options come up in the selector’s result. The Curvature Correction Option is off by default and meant for strong streamline curvature or rapidly rotating frames, and it is best switched on once the solution is stable. Realizable k–ε and Spalart–Allmaras have it too. The Compressibility Correction accounts for dilatation dissipation where compressibility dominates.
The third is the Realizability Option. For k–ω its default is the Durbin Scale Limiter, which limits the over-prediction of turbulence at stagnation points [21]. The standard and realizable k–ε models offer the same option without a printed default, and EB and Lag EB k–ε do not have it. With VOF Waves you can set it to Vorticity Limiter, which damps the excess eddy viscosity that two-equation models produce beneath waves [22]. Florian Vesting’s post on enhancing turbulence models in Simcenter STAR-CCM+ shows both limiters at work.
SST was built for boundary layers in adverse pressure gradients, where it does better than standard k–ε [2], and that is why the selector uses it when separation matters. Note that separation from a smooth surface is still a weak point of RANS models: on NASA’s wall-mounted hump case they tend to predict too long a separation bubble [31]. Where the separation point decides your result, the User Guide advises a low-y+ mesh, and a second model or measurements tell you how much to trust it. The selector asks about this where it can change the answer: for an airfoil or aircraft it then starts with k–ω SST and keeps Spalart–Allmaras as the comparison, and for other flows it treats the dependence on separation onset like separation, which also gives k–ω SST.
Realizable k-epsilon and the Two-Layer variants
Realizable k-epsilon (Shih et al. [5]) is generally at least as good as standard k–ε and typically better, according to the User Guide. Realizable K-Epsilon Two-Layer is the default k–ε model in Simcenter STAR-CCM+, and Siemens calls it a reasonable choice if you are unsure: close to the high-Re model on a coarse mesh, and close to a low-Re model on a mesh that resolves the viscous sublayer.
The two-layer models give the most freedom in the near-wall mesh. They work on fine meshes and are the least inaccurate in the buffer region between y+ 1 and 30, where other models struggle. The Two-Layer Type can be Shear Driven (Wolfstein) [6], Shear Driven (Norris-Reynolds) or Buoyancy Driven (Xu), the last one for flows where buoyancy forces dominate [7].
For a free jet the selector starts with realizable k–ε. That is our choice, not a Simcenter STAR-CCM+ recommendation: Shih et al. tested the realizable model on free shear flows, round jets among them [5]. The User Guide notes that almost all two-equation models predict a round jet spreading faster than a plane jet, the opposite of what experiments show, and offers the Vortex Stretching Modification of Standard (Wilcox) K-Omega for it. That model is the selector’s second comparison for a free jet, and if the spreading rate matters, compare it with data. An impinging jet is different, because k–ε models over-predict the turbulence near the stagnation point [23]. The selector therefore starts an impinging jet, and a jet whose wall heat transfer is a result you need, with k–ω SST.
Elliptic blending k-epsilon (EB and Lag EB)
EB K-Epsilon (elliptic blending, Billard and Laurence [9]) models the anisotropy near the wall and is valid both in the low-Re boundary layer and in the high-Re bulk flow. The User Guide calls it well suited to internal flows, heat transfer and cases where near-wall modelling matters, and the theory guide credits it with better near-wall accuracy than the realizable model and better stability than SST.
The model runs with Low y+ or All y+ Wall Treatment, but its advantage sits in the near-wall cells, so it shows on a mesh that resolves them. Giving EB k–ε for internal heat transfer only when you plan Wall Y+ below 1 is therefore our own rule, and on a wall-function mesh the selector starts from realizable k–ε Two-Layer, with EB k–ε as the comparison. Lag EB K-Epsilon (Lardeau and Billard [10]) adds a stress–strain lag, a concept from Revell et al., which makes it more sensitive to streamline curvature and rotation, and the User Guide calls it well suited to external aerodynamics. The selector uses it as the comparison for the flows that end on the last rule.
Spalart–Allmaras: when one equation is enough
Spalart-Allmaras (SA) solves one transport equation for a modified eddy viscosity, ν̃ [12]. According to the Simcenter STAR-CCM+ User Guide it is good when the boundary layers are largely attached and any separation is mild, with wings, fuselages and other aerospace external flows as the typical examples. It is not suited to flows dominated by free shear layers, to complex recirculation, especially with heat transfer, or to natural convection.
Standard Spalart-Allmaras is the only RANS variant in Simcenter STAR-CCM+ 2606 (there is no high-Reynolds-number version), and All y+ Wall Treatment is the recommended wall treatment for most, if not all, cases. The selector gives it for an airfoil or aircraft with a fully turbulent, attached boundary layer. If you tell it that the result is sensitive to separation onset or stall, for example the maximum lift, it starts with k–ω SST and keeps Spalart–Allmaras as the comparison. For transition, SA Gamma Transition is the Spalart–Allmaras counterpart of Gamma, with terms for crossflow and roughness-induced transition.
For detached eddy simulation, auto-select in Simcenter STAR-CCM+ picks Spalart-Allmaras Detached Eddy as the recommended DES option. The selector lists it as a comparison beside k–ω SST IDDES, which it uses for separated and vehicle flows when you need resolved fluctuations and can afford them.
For a ship hull, the VOF Guidelines suggest Spalart–Allmaras when there is little or no separation, with k–ω as the second choice. The selector starts a hull with k–ω SST and lists Spalart–Allmaras as the comparison, so if your hull has little separation, run both.
Reynolds stress model for swirl and curvature
A Reynolds stress model (RSM, called RST in Simcenter STAR-CCM+) solves for the six Reynolds stresses and the dissipation, seven equations in all, instead of using an eddy viscosity. It therefore accounts for anisotropy, streamline curvature, swirl and high strain rates by construction, at the price of more memory, more time and more iterations.
Siemens recommends RST where the turbulence is strongly anisotropic, with the swirling flow in a cyclone separator as its example. Of the four pressure-strain models it prefers Elliptic Blending, because it handles all-y+ meshes. Elliptic Blending goes back to Manceau and Hanjalić [11].
In the selector, only a swirl-dominated core, as in a cyclone, a swirl burner or a vortex chamber, goes to Elliptic Blending RST, with k–ω SST and the Curvature Correction Option as the comparison. In a cyclone the swirl is the result, and eddy-viscosity models fail to predict the vortex where a Reynolds stress model gets close [24]. The curvature of a rotor, a blade or a bend goes to k–ω SST with the Curvature Correction Option to assess, and RST becomes the comparison. In a pump or compressor passage the result is usually the head, the pressure ratio or the efficiency, and the template of the Turbomachinery Import Workflow in Simcenter STAR-CCM+ also starts from a k-omega model.
RANS, DES or LES?
RANS (Reynolds-averaged Navier–Stokes) models all the eddies and gives you the mean flow. LES (large eddy simulation) resolves the large, energy-carrying eddies and models only the smallest ones. DES (detached eddy simulation) is a hybrid that keeps RANS in the attached boundary layers and resolves the eddies where the flow has separated. Most simulations rely on RANS, and the User Guide advises scale-resolving models only when you are confident that the grid can be afforded and the cost is justified.
The selector moves to a scale-resolving model only when two answers agree: the result needs resolved fluctuations, such as broadband noise or fluctuating peak loads, and the budget allows a scale-resolving run. A separated wake or a rotor still gets a RANS model unless you ask for resolved fluctuations. Our computational fluid dynamics explainer shows what RANS, LES and DNS resolve.
DES vs LES
DES vs LES: LES resolves the large eddies everywhere, also in the boundary layers, where the eddies are small and the number of cells grows fast with the Reynolds number. A wall-resolved LES of a high-Reynolds-number boundary layer needs far more cells than a wall-modelled one [25]. DES keeps a RANS layer on the walls and resolves the eddies only where the flow has separated, and for massively separated flows at high Reynolds number it is more capable than either URANS or LES [18].
Simcenter STAR-CCM+ has three DES models: SST (Menter) K-Omega Detached Eddy, Spalart-Allmaras Detached Eddy and EB K-Epsilon Detached Eddy. For the SST and Spalart-Allmaras models the Formulation Option is DDES or IDDES, and the EB model uses DDES. DDES delays the switch to LES inside attached boundary layers [15], and IDDES makes the RANS layer thinner and gives DES some of the abilities of a wall-modelled LES [16, 17]. When you need resolved fluctuations and can afford them, the selector starts from k–ω SST IDDES for separated, rotating, free-surface and vehicle flows.
For LES, WALE Subgrid Scale is the default subgrid model [19], and auto-select pairs it with All y+ Wall Treatment. The LES Guidelines call WALE cheaper, more thoroughly validated and better near walls than the Smagorinsky models. Since Simcenter STAR-CCM+ 2022.1 the wall shear stress of a wall-modelled LES is taken from the second cell off the wall by default (Reference Location: Off-Wall Cell), which reduces the log-layer mismatch. For the convection scheme, Siemens recommends Bounded-Central for LES, and Hybrid-BCD is the default for DES.
Scale Resolving Hybrid (SRH) is the continuous alternative. It sits on top of a k–ε or k–ω model and moves towards LES where the mesh and the time step allow, while a shielding function keeps the near-wall region in RANS mode.
For a scale-resolving run, the LES Guidelines ask for a convective Courant number close to 1 where you resolve the turbulence, second-order time, a mesh with its cut-off in the inertial subrange and a ratio of turbulent to subgrid kinetic energy (k/ksgs) above 5, preferably above 20. They run 2 to 5 flow-through times, the time the flow takes to pass the region you study, before sampling starts.
Where an attached boundary layer enters the resolved region, the Synthetic Eddy Method (a Synthetic Turbulence Specification at the inlet) gives it resolved inflow turbulence. It cannot be combined with motion, so with a moving rotor the selector points to the other method in the User Guide, Anisotropic Linear Forcing in a volume upstream. It lists these checks with every scale-resolving result.
Steady or transient RANS
A steady RANS solution is enough when the mean flow is steady and the result is a mean quantity. The selector switches to Implicit Unsteady when the result needs periodic or transient loads or resolved fluctuations, when you say the flow behaviour is unsteady, when a rotor must move relative to a stator, and for sloshing, waves and transient mixing. Mean age of air can use a steady scalar equation on a steady flow field.
You only tell the selector what the project can afford, and the result you need and the flow set the time description. If the result needs a transient run and the budget is steady RANS, the result shows this as an open requirement, so you can either find the budget or narrow the target. If you can afford a transient run that the result does not need, the selector stays steady and tells you so.
URANS follows unsteadiness that is tonal, such as blade passing or regular vortex shedding. For broadband noise the aeroacoustics guidance in the User Guide advises against URANS and points to DES or LES.
Whichever you run, monitor the quantities you report and not only the residuals, as Christoffer Johansson explains in our post on residuals in Simcenter STAR-CCM+.
Wall treatment and the near-wall mesh
The wall treatment decides how the model meets the wall, and it sets the near-wall mesh you need. Simcenter STAR-CCM+ has four. Low y+ Wall Treatment resolves the viscous sublayer, High y+ Wall Treatment puts the first cell in the log layer with a wall function, and All y+ and Two-Layer All y+ Wall Treatment blend between the two. Siemens advises an all-y+ treatment whenever the model allows it, especially when a pure low- or high-y+ mesh is hard to get.
| Wall treatment | Wall Y+ to aim for | Models in the selector |
|---|---|---|
| Low y+ Wall Treatment | Below 1 (a few faces up to 5, if they stay in the viscous sublayer) | EB k–ε, Lag EB k–ε, standard k–ε Low-Re and k–ω SST + Gamma when y+ below 1 holds on every wall, and wall-resolved LES with WALE |
| High y+ Wall Treatment | Above 30 | Realizable k–ε and standard k–ε |
| All y+ Wall Treatment | Below 1 or above 30, with few cells between 1 and 30 | k–ω SST, Spalart–Allmaras, Elliptic Blending RST, DES and LES, and the low-Re k–ε models on mixed meshes |
| Two-Layer All y+ Wall Treatment | Below 1 or above 30, with few cells between 1 and 30 | Realizable k–ε Two-Layer |
| Transition models | Between 0.1 and 1 over the whole of every wall where transition can occur, with fine streamwise spacing near transition | Gamma and Gamma-ReTheta with k–ω SST |
Note that you only get Wall Y+ from a solution. There is no automatic way to build a mesh for a chosen wall treatment, so run an exploratory case and plot Wall Y+. The y+ calculator gives a first cell height and the prism layers to start from, and the selector’s result links to it with the target filled in: 1 for a resolved wall and 30 for wall functions, the calculator’s own presets. For wall functions the y+ calculator works with 30 to 300.
For the transition models, the Transition Guidelines ask for Wall Y+ between 0.1 and 1 over the whole of every wall where transition can occur, not only near the transition point, and fine streamwise spacing near the transition region to capture the onset. In our transition post Christoffer Johansson also asks for y+ below 1 over the entire surface, and adds that prism layers should never grow by more than 20% (stretch factor 1.2) per layer.
Surface roughness enters as the Roughness Height, an equivalent sand-grain roughness ks, not the arithmetic roughness Ra. Siemens advises the Rough Displaced Origin specification over Rough where both exist, because it does not limit the roughness height by the wall distance. With Low y+ Wall Treatment, rough walls are available only for the k–ω models, so with roughness the selector moves EB k–ε, Lag EB k–ε and standard k–ε Low-Re to All y+. The DES models and SRH are not in the roughness table of the User Guide, so the result asks you to confirm the roughness method for them.
For wind around buildings the selector asks for the terrain roughness length z0 instead. The Atmospheric Boundary Layer model takes it as Roughness Length (default 0.01 m), and the ground is a rough wall with wall-function constants of its own. In the documented setup the ground uses Rough with E 8.75 and C 0.25, a Roughness Height of about 35 z0 (0.35 m for z0 = 0.01 m), RplusRough of 90 or more and the Roughness Limiter off. The first cell centre must then sit at least 10 z0 off the ground, with y+ of 30 or more. Before you add the buildings, check in an empty domain that the inlet profiles hold all the way to the outlet [20, 27].
Turbulence models in Simcenter STAR-CCM+
These are the Simcenter STAR-CCM+ 2606 models the selector returns, with the names you find in the model selection, the wall treatments each one allows and what the selector uses it for. If you come from another code, note that we found no GEKO, BSL k–ω or v2f model in the 2606 model lists, and that RNG k–ε is there only for internal combustion engines, with the In-Cylinder model.
| Model | Name in Simcenter STAR-CCM+ | Wall treatment | The selector uses it for | Reference |
|---|---|---|---|---|
| k–ω SST | SST (Menter) K-Omega | All y+ (also High y+, Low y+) | External flow, separation, rotation, free surfaces, impinging jets | [2, 3] |
| Realizable k–ε Two-Layer | Realizable K-Epsilon Two-Layer | Two-Layer All y+ | Internal flow, buoyancy-dominated flow, free jets | [5, 6, 7] |
| Realizable k–ε | Realizable K-Epsilon | High y+ | Free jets on a wall-function mesh | [5] |
| Standard k–ε | Standard K-Epsilon | High y+ | The atmospheric boundary layer, with the Atmospheric Boundary Layer model | [4, 20] |
| Standard k–ε Low-Re | Standard K-Epsilon Low-Re | Low y+, All y+ | Buoyancy-driven flow, with the Yap Correction to assess, and buoyancy-driven transition, where the transition location needs a check against data | [8] |
| EB k–ε | EB K-Epsilon | Low y+, All y+ | Internal heat transfer on a resolved mesh | [9] |
| Lag EB k–ε | Lag EB K-Epsilon | Low y+, All y+ | Comparison for external flows | [10] |
| Standard k–ω (Wilcox) | Standard (Wilcox) K-Omega | All y+ (also High y+, Low y+) | Comparison for free jets, with its Free-Shear and Vortex Stretching modifications | – |
| Elliptic Blending RST | Elliptic Blending | All y+ | Swirl-dominated cores | [11] |
| Spalart–Allmaras | Standard Spalart-Allmaras | All y+ (also Low y+) | Attached aerodynamic boundary layers | [12] |
| Gamma, Gamma-ReTheta | Gamma Transition or Gamma-ReTheta Transition, with SST (Menter) K-Omega | Low y+, All y+ | Wall boundary-layer transition | [13, 14] |
| SA Gamma | SA Gamma Transition, with Standard Spalart-Allmaras | All y+ | Transition on airfoils | – |
| Turbulence Suppression | Turbulence Suppression (Optional Models, with a RANS model) | Turbulence Suppression Option on the walls | The laminar region upstream of a trip | – |
| k–ω SST IDDES or DDES | SST (Menter) K-Omega Detached Eddy, Formulation Option IDDES or DDES | All y+ (also High y+, Low y+) | Resolved fluctuations in separated, rotating, free-surface or vehicle flow, and with Gamma Transition as a comparison for transition | [15, 16, 17] |
| Spalart–Allmaras DES | Spalart-Allmaras Detached Eddy, Formulation Option IDDES or DDES | Low y+, All y+ | Comparison for k–ω SST IDDES | [18] |
| SRH | Scale Resolving Hybrid Turbulence, on SST (Menter) K-Omega | All y+ | Hybrid comparison for DES and LES | – |
| LES with WALE | WALE Subgrid Scale | All y+, wall-modelled | Resolved fluctuations without large separated regions, for example jets and ducts | [19] |
| Wall-resolved LES with WALE | WALE Subgrid Scale | Low y+ (Wall Y+ below 1) | Resolved fluctuations with transition | [19, 25] |
The selector also uses options of these models: the Two-Layer Type Buoyancy Driven (Xu) for realizable k–ε Two-Layer, the Yap Correction for standard k–ε Low-Re, the Curvature Correction Option and the Compressibility Correction for k–ω SST, and the Atmospheric Boundary Layer optional model for standard k–ε. In the result they are listed in the order of the Physics Model Selection dialog: the flow solver and the time, then Viscous Regime, Turbulence, the model family, the model, the wall treatment and the Optional Models.
Available in 2606 but not used by the selector are Standard K-Epsilon Two-Layer, the Linear and Quadratic Pressure Strain RST models (the quadratic one only on wall-function meshes), EB K-Epsilon Detached Eddy and the Smagorinsky and Dynamic Smagorinsky subgrid models.
Other physics in the same setup
Turbulence, motion, heat transfer and transport are separate choices in the same simulation, and the result adds rows for them. Gravity and a density model for buoyancy, and Volume of Fluid for a free surface, come from the flow-features step. Fluid and solid energy for conjugate heat transfer (with solid regions), a passive scalar or multi-component species, porous regions or baffles, cavitation and radiation come from the optional last step. In our tests that step never changed the turbulence model, which is why it is optional.
Some combinations come with conditions. Shock waves call for the coupled flow solver, and so does natural convection at high Rayleigh numbers, where the coupled solver copes better, according to the documentation. The Segregated Flow solver handles mildly compressible flow and natural convection at low Rayleigh numbers, but it does not suit shocks, high Mach numbers or high Rayleigh numbers. With LES or DES and shocks, check that the coupled solver is available with your model, because the User Guide is not consistent on this point. For a free surface, the VOF Guidelines run most cases transient even when you want a steady answer.
Spherical Harmonics (P1) radiation cannot handle refraction, specular reflection or overset meshes, and k-Distribution does not work with VPMC. A passive scalar must not change the material properties. Reacting flows need a reaction model and a turbulence–chemistry interaction model on top of RANS or LES.
For dispersed multiphase flow, with Lagrangian particles or Eulerian phases, the selector only adds a pointer. The turbulence options there are a choice of their own, such as mixture or phasic turbulence in an Eulerian model, and our post on turbulence for multiphase simulations walks through them.
What the selector does and does not do
The selector asks about your case in six steps: the case, the flow regime, the result and the budget, the flow features, the walls and the mesh, and an optional last step for the rest of the setup. From the second step it shows the model it would give so far, and from the fourth you can view the result straight away, with the open answers marked as not decided yet. The six examples under the first question each open a complete result.
The result gives the starting model with its wall treatment and time description, the answers that decided it and what would change it, and the near-wall target with a link to the y+ calculator. Below that come the model set in the order of the Simcenter STAR-CCM+ model selection, what is still open, what to check in the solution and the models to compare with. Copy report gives it all as plain text with a link back, for a project log or a support case. If you change an answer, the answers that no longer apply are cleared and the others are kept, and you can undo the change.
The selector gives a starting point, and it does not validate a setup. Check mesh sensitivity, time-step sensitivity where it applies and the statistical convergence of every quantity you report, and compare with measurements or benchmark data before you rely on the result. The rules are Volupe’s, and the names, options and section titles come from the Simcenter STAR-CCM+ 2606 User Guide. The result cites it by section title, so look the sections up in the documentation of your release and check the names and options there if you run another release.
If you are unsure about your case, feel free to contact us at support@volupe.com, or join one of our Simcenter STAR-CCM+ training courses.
References and changelog
References
[1] Siemens Digital Industries Software, Simcenter STAR-CCM+ 2606 User Guide, 2026: Turbulence; Deciding on a RANS Turbulence Model; Deciding on a K-Epsilon Model; Deciding on a K-Omega Model; Deciding on a Spalart-Allmaras Model; Reynolds Stress Transport (RST) Turbulence; Detached Eddy Simulation (DES); Selecting LES and a Subgrid Scale Turbulence Model; LES Guidelines; Deciding on a Wall Treatment; Setting up a Wall Treatment; Wall Treatment Models Reference; Transition Guidelines; Setting Up Atmospheric Boundary Layers. Cited without a link, because the documentation needs a Siemens login.
[2] F. R. Menter, “Two-equation eddy-viscosity turbulence models for engineering applications”, AIAA Journal 32(8), 1598–1605, 1994.
[3] F. R. Menter, M. Kuntz and R. Langtry, “Ten years of industrial experience with the SST turbulence model”, Turbulence, Heat and Mass Transfer 4, Begell House, 625–632, 2003.
[4] B. E. Launder and D. B. Spalding, “The numerical computation of turbulent flows”, Computer Methods in Applied Mechanics and Engineering 3(2), 269–289, 1974.
[5] T.-H. Shih, W. W. Liou, A. Shabbir, Z. Yang and J. Zhu, “A new k–ε eddy viscosity model for high Reynolds number turbulent flows”, Computers & Fluids 24(3), 227–238, 1995.
[6] M. Wolfshtein, “The velocity and temperature distribution in one-dimensional flow with turbulence augmentation and pressure gradient”, International Journal of Heat and Mass Transfer 12(3), 301–318, 1969.
[7] W. Xu, Q. Chen and F. T. M. Nieuwstadt, “A new turbulence model for near-wall natural convection”, International Journal of Heat and Mass Transfer 41(21), 3161–3176, 1998.
[8] C. R. Yap, Turbulent heat and momentum transfer in recirculating and impinging flows, PhD thesis, University of Manchester, 1987.
[9] F. Billard and D. Laurence, “A robust k–ε–v2/k elliptic blending turbulence model applied to near-wall, separated and buoyant flows”, International Journal of Heat and Fluid Flow 33(1), 45–58, 2012.
[10] S. Lardeau and F. Billard, “Development of an elliptic-blending lag model for industrial applications”, AIAA Paper 2016-1600, 54th AIAA Aerospace Sciences Meeting, 2016.
[11] R. Manceau and K. Hanjalić, “Elliptic blending model: a new near-wall Reynolds-stress turbulence closure”, Physics of Fluids 14(2), 744–754, 2002.
[12] P. R. Spalart and S. R. Allmaras, “A one-equation turbulence model for aerodynamic flows”, AIAA Paper 92-0439, 1992.
[13] R. B. Langtry and F. R. Menter, “Correlation-based transition modeling for unstructured parallelized computational fluid dynamics codes”, AIAA Journal 47(12), 2894–2906, 2009.
[14] F. R. Menter, P. E. Smirnov, T. Liu and R. Avancha, “A one-equation local correlation-based transition model”, Flow, Turbulence and Combustion 95, 583–619, 2015.
[15] P. R. Spalart, S. Deck, M. L. Shur, K. D. Squires, M. Kh. Strelets and A. Travin, “A new version of detached-eddy simulation, resistant to ambiguous grid densities”, Theoretical and Computational Fluid Dynamics 20(3), 181–195, 2006.
[16] M. L. Shur, P. R. Spalart, M. Kh. Strelets and A. K. Travin, “A hybrid RANS-LES approach with delayed-DES and wall-modelled LES capabilities”, International Journal of Heat and Fluid Flow 29(6), 1638–1649, 2008.
[17] M. S. Gritskevich, A. V. Garbaruk, J. Schütze and F. R. Menter, “Development of DDES and IDDES formulations for the k–ω shear stress transport model”, Flow, Turbulence and Combustion 88, 431–449, 2012.
[18] P. R. Spalart, “Detached-eddy simulation”, Annual Review of Fluid Mechanics 41, 181–202, 2009.
[19] F. Nicoud and F. Ducros, “Subgrid-scale stress modelling based on the square of the velocity gradient tensor”, Flow, Turbulence and Combustion 62, 183–200, 1999.
[20] P. J. Richards and R. P. Hoxey, “Appropriate boundary conditions for computational wind engineering models using the k–ε turbulence model”, Journal of Wind Engineering and Industrial Aerodynamics 46–47, 145–153, 1993.
[21] P. A. Durbin, “On the k–ε stagnation point anomaly”, International Journal of Heat and Fluid Flow 17(1), 89–90, 1996.
[22] B. E. Larsen and D. R. Fuhrman, “On the over-production of turbulence beneath surface waves in Reynolds-averaged Navier–Stokes models”, Journal of Fluid Mechanics 853, 419–460, 2018.
[23] T. J. Craft, L. J. W. Graham and B. E. Launder, “Impinging jet studies for turbulence model assessment, II: an examination of the performance of four turbulence models”, International Journal of Heat and Mass Transfer 36(10), 2685–2697, 1993.
[24] A. J. Hoekstra, J. J. Derksen and H. E. A. Van den Akker, “An experimental and numerical study of turbulent swirling flow in gas cyclones”, Chemical Engineering Science 54(13–14), 2055–2065, 1999.
[25] H. Choi and P. Moin, “Grid-point requirements for large eddy simulation: Chapman’s estimates revisited”, Physics of Fluids 24, 011702, 2012.
[26] S. B. Pope, Turbulent Flows, Cambridge University Press, 2000, chapters 6 and 7.
[27] B. Blocken, T. Stathopoulos and J. Carmeliet, “CFD simulation of the atmospheric boundary layer: wall function problems”, Atmospheric Environment 41(2), 238–252, 2007.
[28] D. B. Spalding, “A single formula for the ‘law of the wall’”, Journal of Applied Mechanics 28(3), 455–458, 1961.
[29] R. A. W. M. Henkes, F. F. van der Vlugt and C. J. Hoogendoorn, “Natural-convection flow in a square cavity calculated with low-Reynolds-number turbulence models”, International Journal of Heat and Mass Transfer 34(2), 377–388, 1991.
[30] C. G. Speziale, “On nonlinear K–l and K–ε models of turbulence”, Journal of Fluid Mechanics 178, 459–475, 1987.
[31] NASA Turbulence Modeling Resource, “2D NASA wall-mounted hump separated flow validation case”, tmbwg.github.io/turbmodels/nasahump_val.html.
Review
Reviewed by Christoffer Johansson, M.Sc., Senior Application Specialist at Volupe, in October 2026. The selection rules are Volupe’s. The model names, options and documentation sections come from the Simcenter STAR-CCM+ 2606 User Guide.
Changelog
October 2026, new version of the selector: six steps in a new order, a provisional model from the second step, examples, and a result that names the answers behind the model, a near-wall target with a link to the y+ calculator and models to compare with. There are new or reworked rules for tripped and buoyancy-driven transition, wind around buildings, impinging jets, swirl and cyclones, and turbomachinery, and a provisional model instead of no answer when the flow regime is not known. Saved links from the earlier version still open, with the answers moved to the new steps.
19 October 2026, new page layout: we tested the rules on 20,000 random answer sets for internal consistency, for example that laminar answers never get a turbulence model. This is not a validation of the recommendations.
Next in a CFD set-up
All CFD Handbook tools →y+ = uτy / ν
Calculator
y+ calculator
Size the first cell for the wall treatment the selector suggests, from a resolved sublayer to wall functions.
Open calculator →
k = ³⁄₂ (U I)2
Calculator
Turbulence boundary conditions calculator
Turn turbulence intensity into inlet k, ε, ω and ν̃.
Open calculator →
GCI = Fs|ε| / (rp − 1)
Calculator
Grid Convergence Index (GCI) calculator
Check mesh independence with a three-grid refinement study.
Open calculator →
Questions about turbulence models
Selector reviewed by
Christoffer Johansson
Senior Application Specialist, M.Sc.
Unsure which model fits your case? We reply within one day.
Ask a technical question →
Which turbulence model should I use in Simcenter STAR-CCM+?
Start from the flow regime and the result you need. For fully turbulent flow, k-omega SST is the usual start for external flows and for separation, rotation or a free surface, realizable k–ε Two-Layer for internal flows, Spalart–Allmaras for attached boundary layers on wings, and a Reynolds stress model for swirl-dominated flows such as cyclones. Add Gamma Transition to SST when the transition of a wall boundary layer changes the result, and use DES or LES only when you need resolved fluctuations and can afford them. The selector above goes through these choices in order.
Is k-omega SST the best turbulence model?
k-omega SST is the usual start for external flows and for flows with separation, but no single model is best for every flow. SST was designed for boundary layers in adverse pressure gradients, where it does better than standard k–ε (Menter, 1994), and the Simcenter STAR-CCM+ User Guide recommends it with All y+ Wall Treatment for most k–ω applications. For internal flows with complex recirculation, realizable k–ε Two-Layer is often the better start, and for swirl-dominated flows a Reynolds stress model. Where the separation point decides the result, compare SST with a second model or with measurements.
Which wall treatment should I use: High y+, Low y+ or All y+?
Use All y+ Wall Treatment (Two-Layer All y+ with realizable k–ε Two-Layer) whenever the model allows it, as the Simcenter STAR-CCM+ User Guide advises, especially when a pure low- or high-y+ mesh is hard to get. Aim for a Wall Y+ below 1, or above 30, with few cells in between. Use Low y+ only when the near-wall cells stay in the viscous sublayer on every wall, and High y+ with wall functions, about 30 to 300 as in our y+ calculator, when the wall detail matters less than the cell count. The y+ calculator gives the first cell height for your target.
DES or LES: when do I need a scale-resolving model?
When the result depends on resolved turbulent fluctuations, such as broadband noise or fluctuating peak loads, and you can afford the mesh, the time steps and the sampling. DES keeps RANS in the attached boundary layers and resolves the separated regions, which is why the selector starts separated and vehicle flows with SST IDDES. The selector uses LES for flows without large separated regions, such as jets. A separated wake or a rotor on its own still gets a RANS model.
When does transition modelling matter?
When the laminar-to-turbulent transition of a boundary layer changes drag, lift, separation or heat transfer, for example on an airfoil with laminar flow over the front part. For many applications it does not drive the result and is not worth modelling. Gamma Transition with SST needs no extra model input, no correlations or free-stream edge to define, but it is sensitive to the free-stream turbulence that reaches the leading edge, so set the inlet values with their decay in mind. It also needs a mesh with Wall Y+ between 0.1 and 1. If the boundary layer is tripped at a known place, model it as fully turbulent from the trip. Our post on transition modelling in Simcenter STAR-CCM+ goes through Gamma, Gamma-ReTheta and Turbulence Suppression.
Which turbulence model suits turbomachinery?
For mean performance, such as head, pressure ratio or efficiency, a rotating reference frame with SST is a common start, and the template of the Turbomachinery Import Workflow in Simcenter STAR-CCM+ also starts steady with a k-omega model. Assess the Curvature Correction Option where the curvature or the rotation is strong. Blade-passage loads need the rotor to move relative to the stator and an Implicit Unsteady run. A Reynolds stress model is the comparison when the swirl itself is what you need to predict.
Does the selector replace validation?
No. It gives a starting point. Check mesh and time-step sensitivity, monitor the quantities you report and not only the residuals, and compare with measurements or benchmark data before you rely on the result. The GCI calculator helps with the mesh study.
Talk to an engineer in your region
Every Volupe business developer holds an engineering degree. Find your regional contact below, or send us a message.
Ask us about Turbulence model selector
Use this form for licence quotes, demos, training and technical questions.
What happens next
- A reply within one day.
- A scoping call with one of our engineers.
- A quote, demo or other next step that fits your enquiry.
After purchase: support from Volupe simulation engineers.