We are approaching yet another new release of Simcenter STAR-CCM+, this time it is version 2506. And as usual I will give some extra attention to the news that relates to multiphase simulations. This time around, the focus has been on automation, streamlining, and some GPU compatibility.
Volume fraction Sub-stepping made default (VOF/MMP)
I have been writing many times about the Implicit multi-stepping that was implemented a few years ago for Volume of fluid simulation and for the Mixture multiphase model. Where we allow for more inner iterations in relation to the volume fraction solver, to achieve a larger global time step (alternatively a higher resolution). Now the options of single step, Implicit multi-step and Explicit multi-steps are gone from the terminology. Instead, you now specify the number of sub-steps directly. Here the single-step option is a special case where the number of sub-steps is specified to one. All VOF-simulations now have the dynamic sub-step specification when running a fixed time-step. The sub-stepping is now also compatible with 2nd order time.
Multiphase Porous Media: Phasic Porous Media with EMP
Applications such as fuel cells require the accurate modelling of liquid phases in porous materials, for this reason in Simcenter STAR-CCM+ 2506, compatibility between Phasic Porous Media (the most comprehensive porous media model) and EMP (the most comprehensive multiphase model) has been added. The transport of liquid phases in a porous material is very different to traditional multiphase simulations where dispersed liquid droplets are carried by the continuous gas. In porous material, the transport of the liquid phase is determined primarily by interaction with the solid in the porous structure through capillary effects (interaction with the gaseous phases is secondary). For this reason, Eulerian Multiphase (EMP) is better suited to this application as it has separate transport equations for each phase allowing a greater degree of independence than is possible with EMP. Prior to this release, multiphase simulations containing Phasic Porous Media was only compatible with MMP. This implementation for EMP now allow for more correct modelling where phase specific resistances are important.
With this change, a model for Porous capillary pressure is also introduced. This model allows for accurate transport of liquid phases in porous materials. In that type of material, the transport of the dispersed phase is not governed by the interaction, the drag, with the continuous gas phase, but rather by the interaction with the solid porous structure, the capillary effects. Typically, small droplets sit in pores in the structure where they are held in place by the surface tension, the droplets may coalesce, and the wetting phase may advance due to capillary effects. In version 2506 an inbuilt Porous Capillary Pressure model is added to cover applications from fuel cells to oil transport in rocks. Available for both porous media and porous region. The models available are Leverett, Kumbur, Brooks-Coey and a user-defined option.
Further, it is now also possible to specify a Relative Permeability Model for Porous resistance. Permeability is a measure of the ease with which a fluid can pass through a porous media. It removes the need to specify loss data using curve fitting in order to determine the required resistance coefficients. The permeability is a physical property of the material. The example below shows a reservoir dam wall holding back water. The dam wall is porous and permeable, and the relative permeability of the wetting phase is shown in the contour inside the dam.
Overset mass conservation with mass tracking for EMP
This new option for the overset in conjunction with EMP allows for better conservation in EMP simulation with overset mesh regions. Overset mesh is not inherently mass conservative, and this can lead to large errors in domains without any throughflow. This model enforces mass conservation by adding source terms to correct the mass balance across the entire computational domain. It is not yet compatible with interphase mass transfer models like boiling, Fluid film or the AMUSIS population balance.
Improvement to the Lagrangian solver
It is now possible to speed up steady state Lagrangian simulations due to implementations of advanced Lagrangian steady solver options for scalability. What this in reality entails are two new introductions, the first being enablement for Sub-step Load balancing. This allows for particle rebalancing at different Lagrangian sub-step time-levels. If the solver detects an imbalance, repartitioning of the domain will happen. The other new option is Parallel migration Frequency. It enables setting the sub-step time level of parallel parcel migration events. When specifying N>1 integer for this option, parcels entering halo cells will suspend particle tracking at those cells until the next parallel migration event takes place, after which parcel tracking is resumed in a new partition. This is especially effective when the two-way coupling model is used.
Another new feature for the Lagrangian solver is support for stages in combustion simulations. Users can now stage track files, turbulent dispersion, and all secondary droplet breakup models as well as injector parameters.
Another improvement for the Lagrangian solver is the introduction of the isothermal model for Lagrangian phases. Users can leverage this feature when particle temperature variations are insignificant, yet energy interactions with other phases need to be accounted for. Unlike the Temperature option, the Isothermal model does not update particle temperature during the solution process, resulting in a computationally lighter approach.
GPU-native Meshfree DEM
As a first step in transitioning all LMP and DEM capabilities to GPU, it is now possible to run meshfree DEM on a single GPU. With this new feature many applications that require Meshfree DEM workflow can be run on GPGPU hardware faster without compromising accuracy of the results. For instance, when simulating the mixing of 10 million solid spherical particles with a 10 mm diameter in a rotating drum using an NVIDIA A100 GPU card, the process is 1.8 times faster compared to running the simulation on a 64-core computer cluster, as demonstrated in a column plot.
There are some limitations yet, all boundaries must be of wall-type, particles need to be injected a short time after simulation start and the injectors are not ported to GPU. The GPU-native DEM is compatible with rigid body motions and DFBI. See the documentation for exactly which models have been ported to GPU. In the example below we see an example of 1.5 million particles in a rotating drum. The simulation was made on a laptop and 6 seconds of physical time was simulated in under 5 hours.
I hope you are as excited about testing these new features as we are at Volupe. As usual, if you have questions, reach out to support@volupe.com.
Author
Robin Victor
support@volupe.com