TransAT© is a multiscale, multi-physics, finite-volume code solving the Navier-Stokes fluid-flow equations. Compressibility is pressure based (Projection Type) and is used for low Mach-number flows. The code relies on a flexible, multi-block meshing approach used in connection with MPI parallel protocol for HPC infini-band systems. Grid generation can be achieved using the Immersed Surfaces method described next, for which the TransAT Suite has a specific grid generator. Expand
TransAT runs implicit, depending on the test problems considered, for both single and multiphase flows, depending on the turbulence models employed. In implicit solutions, the user can resort to steady state or transient time marching schemes. TransAT uses high order schemes for convection and diffusion processes.
TransAT HPC uses a velocity-pressure coupling algorithm based on the SIMPLE approach, with SIMPLEC and SIMPLEST variants used for multiphase flows and compressible flows (low Ma number flows). A wide range of pressure solvers are possible: including SIP, GMRES with pre-conditioners, AMG for all equations using either the PETSc Library or the genuine TransAT AMG solver.
TransAT is parallelized using MPI and domain decomposition on non-shared memory supercomputers. The code can run explicit as well as implicit; both variants have been tested for scaling, for single and multiphase flow. The scaling efficiency studies on 3D turbulent flows conducted on the DOE supercomputer Titan show that the code breaks up the 90-100% scaling on 10,000 processors and an allocation of 14,000 cells per processor.
TransAT products rely on a flexible meshing strategy using multiblock Cartesian and Body-fitted coordinates (BFC), or coupled Immersed Surfaces Technology (IST) with Block Mesh Refinement (BMR). Expand
Immersed Surfaces Technique (IST)
The Immersed Surfaces Technology (IST) has been developed by ASCOMP, although other similar approaches have been developed in parallel. The underpinning idea is inspired from Interface Tracking techniques for two-phase flows (Level Sets), where free surfaces are described by a hyperbolic convection equation advecting the phase colour function. In the IST the solid is described as the second ‘phase’, with its own thermo-mechanical properties. The technique differs substantially from the Immersed Boundaries method of Peskin, in that the jump condition at the solid surface is implicitly accounted for, not via direct momentum forcing using the penalty approach. It has the major advantage to solve conjugate heat transfer problems, in that conduction inside the body is directly linked to external fluid convection. The solid is defined by its external boundaries using the solid level set function. Like in fluid-fluid flows, this function represents a distance to the wall surface; is zero at the surface, negative in the fluid and positive in the solid.
Block-Mesh Refinement (BMR)
The BMR technique was developed in the TransAT code to help better solve the boundary layer zone when use is made of the IST technique discussed above. In BMR, more refined sub-blocks are automatically generated around solid surfaces; with dimensions made dependent on the Reynolds number (based on the boundary layer thickness). An unlimited number of sub-blocks of various refinements can be generated, with connectivity between the blocks matching up to 1-to-8 cells. This method can save up to 75% grid cells in 3D, since it prevents clustering grids where unnecessary. The combined IST/BMR technique has various major advantages over traditional methods; these are enumerated below (including ongoing developments):
- Rapid gridding of complex geometries & set-up of flow boundary conditions
- Suitable for moving bodies, fluid-structure interactions, and conjugate heat transfer
- Retain high-order scheme accuracy (Cartesian grids)
- Fully conservative
- Scalable parallelization; save up to 70% cells in 3D
Sharp Solid Interface Treatment (SSIT) (non available in V5.2 yet)
The IST algorithm described above may have limitations in some case (thin solid plates), in that objects have to be resolved on the grid, using a minimum of 2-3 grid cells inside the solid to fully block the flow. If there are less cells inside the solid, the flow perceives the presence of the solid only partially. To resolve this issue, the code TransAT has been recently upgraded using the so-called sharp solid interface formulation, which modifies the original IST algorithm by decoupling the equations at surfaces based on surface normal’s that are dependent on the solid level set function (gradients). With this, even an infinitely thin solid perceives the full blocking effect on the oncoming flow.
TransAT can be employed using RANS (from EVM to EASM‘s, for all type of wall resolutions) as well as Scale-Resolving methods, including LES, V-LES and DNS, for all class of grids. Expand
Linear Eddy Viscosity Models (EVM)
The Eddy-viscosity models implemented in TransAT are essentially based on the k-? model, coded for steady & unsteady flows, using adaptive wall functions and low-Re variants (3 models). The base RANS model is augmented with important modifications to cope with complex flow scenarios.
- Kato & Launder modification
- Yap correction
TransAT resorts to the dynamic two-layer approach, whereby the core flow is solved using the 2-equation k-? model, whereas the near-wall region is treated using a one-equation model. Two-layer schemes are preferred to low-Re models for practical applications, and encompass the following variants:
- TLK: k-based one equation
- TLV : v’2 based one equation
Two-Layer EVM (TLK and TLV)
TransAT resorts to the dynamic two-layer approach, whereby the core flow is solved using the 2-equation k-? model, whereas the near-wall region is treated using a one-equation model. Two-layer schemes are preferred to low-Re models for practical applications, and encompass two variants: a k-based one and a v’ based one.
Explicit Algebraic Stress Models (EASM)
To alleviate the difficulty of treating complex fluid flows featuring rotation, anisotropy, buoyancy effects, TransAT adopts the Explicit Algebraic Stress Modelling (EASM) concept, with different variants, ranging from the quadratic up to the quartic versions. The models are all coded for steady & unsteady flows, using adaptive wall functions, two-layer, or low-Re variants. EASM are shown to often return similar results to full RSM’s, with lower computational costs and higher stability.
Algebraic Heat Flux Models (AHFM)
In several situations where the heat (or scalar) transfer does not scale with the flow motion, be it for non-unity Prandtl number fluids, for natural convection, etc., the base EVM models are known to return low-quality results. TransAT adopts a series of high-order models for convective heat transfer to alleviate this difficulty
- GGDH (anisotropic used with EVM or EASM)
- WET (anisotropic used with EVM or EASM)
- AFM (anisotropic used with EVM or EASM)
LES (Large Eddy Simulation)
ASCOMP has decided to build an added-value modelling strategy based on LES, in complement of RANS. This observation applies in particular to industries in which reactive and multi-phase flow control the design. Many other industries are in search for similar new developments, too, namely oil & gas, thermal hydraulics of nuclear reactors and chemical and process engineering. The LES strategy built in TransAT is very stable and highly accurate and follows the guidelines of top literature in terms of wall modelling, gridding, unsteady inflow boundary conditions, numerical schemes. Various SGS models can be employed in connection to LES.
- Smagorinsky Model
- Dynamic Model
- WALE Model
V-LES (Very Large Eddy Simulation)
VLES is an excellent compromise between efficiency and precision as to capturing unsteady turbulence, and may thus be used for industrial problems for which LES remains computationally expensive. The method is a sort of blend between U-RANS and LES, in that it resolves very large structures – way larger than the grid size – and models turbulence subscale using a two-equation model. The method is computationally efficient and numerically robust. The computational cost decreases with increasing filter width, which in contrast to LES is not dependent on the grid size, but based on the characteristics length of the flow.