4.17.3. Interfaces Treatment

Figure 4.13: Backward Facing Step Flow Using ELES shows a typical example of an ELES scenario. The geometry of the simulation is divided into RANS and LES zones with interfaces between them. The upstream and the downstream zone are covered by RANS and the middle section with the reversed flow is covered by LES.

Figure 4.13: Backward Facing Step Flow Using ELES

Backward Facing Step Flow Using ELES

As the flow passes between the RANS domain and the LES domain, and from the LES domain into the downstream RANS domain, the treatment of the interfaces must be considered, as described in the following sections.

4.17.3.1. RANS-LES Interface

The most critical interface is the interface where the flow leaves the RANS domain and enters the LES region (RANS-LES interface). At this interface, it is necessary to convert modeled turbulence kinetic energy into resolved energy and you must select an appropriate method for this transfer. This can be achieved by selecting one of the two existing methods for generating resolved turbulence at inlets – namely the Vortex Method (VM, see Vortex Method) or the Spectral Synthesizer (SS, see Spectral Synthesizer). Both methods have been augmented so that they can be used at interior interfaces between the RANS and LES zone. The RANS information is obtained from the RANS model used upstream of the interface. For details and recommendations on these methods, see Zonal Modeling and Embedded LES (ELES) and Setting Up the Embedded Large Eddy Simulation (ELES) Model in the User's Guide.

If a DDES, SAS, SDES, or SBES model is run in the entire domain, the interface treatment ensures that the turbulent kinetic energy that is converted from RANS to LES is subtracted from the RANS model. Thereby avoiding double-accounting of turbulence.

4.17.3.2. LES-RANS Interface

There is also an ambiguity concerning how to revert back from the LES to the RANS zone (LES-RANS interface). The first option is to freeze the background RANS model during the ELES simulation in the LES zone. This assumes that the ELES is started from a reasonably converged RANS simulation (this is recommended in any case). The momentum equations in the LES zone are not affected by the freeze as they are provided by the eddy-viscosity from the LES model. At the ‘downstream’ LES-RANS interface, the RANS model is then simply reactivated, using the frozen RANS solution inside the LES zone as an ‘inlet’ condition for the downstream RANS zone. This works well, as long as there is no significant change in mean flow topology between the full RANS and the ELES run, or if the flow downstream of the LES-RANS interface is not relevant to the goals of the simulation. It should be stressed, however, that for this option, any non-converged RANS starting solution will affect the flow for all times downstream of the LES-RANS interface. The option of freezing the turbulence variables of the underlying RANS model inside the LES domain is employed with all standard RANS models.

A second option would be to run the RANS model in passive mode within the LES zone, meaning the model is solved on the unsteady velocity field, but the resulting eddy-viscosity is overwritten by the LES model eddy-viscosity for the momentum equations. The RANS model is then reactivated at the LES-RANS interface. It turns out that this approach fails for standard RANS models, as they severely over-predict the turbulent kinetic energy in the LES zone. The reason lies in the high strain rates computed from the unsteady velocity field. These strain rates enter into the production term Pk and generate very high turbulence levels in the LES zone. However, this option is sensible in combination with the SST-SAS model. This model recognizes the resolved scale and adjusts turbulence levels accordingly. When reactivated at the downstream interface, the model gradually converts back to RANS as the grid coarsens, or stays in LES-mode if time step and grid resolution permits.

4.17.3.3. Internal Interface Without LES Zone

There is a final option, where only a single hybrid model is run in the entire domain, that is, there is no change in the model between the RANS and LES zone. The choice is between the (rke- or SST-) DDES, the (SST-) SAS model, the (SST-) SDES model, and the (SST-) SBES model, all of which can be run in RANS and LES mode. At the interface, again, the Vortex Method or the Spectral Synthesizer can be selected to generate resolved turbulence. When using the (rke- or SST-) DDES model, one has to consider that the model will not be in RANS mode for free shear flows (flows away from walls where the DDES shielding function is not active) if the grid is sufficiently fine to activate the DES limiter in the RANS zone. The combination of the (I)DDES model with the interface option is therefore mostly limited to wall-bounded internal flows. For more information about DDES, see Detached Eddy Simulation (DES), and for more information about IDDES, see Improved Delayed Detached Eddy Simulation (IDDES).

The current option of using an internal interface works more naturally in combination with the SST-SAS model, where little attention is required to carry the model in RANS mode to the interface. The current interface treatment can therefore be used in situations, where the internal flow instability is not strong enough to convert the SAS model into scale-resolving (‘LES’) mode at a desired location. Downstream of the interface, the SAS model will then remain in LES mode if the grid and time step size permit, or will gradually turn back to RANS otherwise.

If a DDES, SAS, SDES, or SBES model is run in the entire domain, the interface treatment ensures that the turbulent kinetic energy that is converted from RANS to LES is subtracted from the RANS model. This avoids a double-accounting of the turbulence.

4.17.3.4. Grid Generation Guidelines

The grids used in the RANS and LES zones have to be conforming to the resolution requirements of the underlying turbulence model. Figure 4.14: Typical Grid for ELES for Backward Facing Step shows an example of an ELES grid. Frequently, the LES zone will be connected to the RANS zones by Non-Conformal Interfaces (NCI) to allow a more refined grid in the LES zone.

Figure 4.14: Typical Grid for ELES for Backward Facing Step

Typical Grid for ELES for Backward Facing Step