Version 3.0, released in March 2013, is a "long-term support" version, and the 3.0 branch has undergone a full validation process. We recommend users who prefer to use the most current version to switch from version 2.3 to version 3.0. The documentation has been updated.
The current bug-fix release is 3.0.3. It was released December 11, 2013.
As with any version, in case you detect bugs, we appreciate your feedback on the forum and bug-tracker, which will help us to provide you with patch releases for this stable series, as well as improvements for future developement versions.
Reminder: current Code_Saturne developpment cycles are based on a release approximately every 6 months, including a fully validated, "long-term support" version every 2 years.
Version 3.1 was first released in May 2013, and 3.2 December 2013, shifting the release cycle by a few months to synchronize with the SALOME platform, so 4.0 should be released early 2015.
The main improvements provided by Code_Saturne 3.0 as compared to the previous version (namely 2.0) are the following:
Lagrangian particles tracking
New deposition model
New unified model for coal and fuel combustion for the gas phase
New dilatable model for fire modelling (idilat = 4)
New Low-Mach algorithm (idilat = 3)
New humid atmosphere model
New soil model
New solar radiation model
New v2f (bl-v2k) low Reynolds turbulence model (Billard, Uribe, & Laurence) (iturb = 51)
New Rij-EBRSM second order low Reynolds turbulence model (Manceau & Hanjalic, 2002) (iturb = 32)
New Spalart-Allmaras high Reynolds turbulence model (iturb = 70)
Correction models for rotation and curvature for eddy viscosity turbulence models
New turbulent thermal flux models: generalized gradient (GGDH, iturt = 10), algebraic model (AFM, iturt = 20) and transport model (DFM, ITURT = 30)
Synthetic Eddy Method (SEM) inlet boundary condition for LES
New free surface boundary condition
New porosity model for incompressible flows (iporos)
New dilatable algorithm (idilat = 2)
New algorithm for stratified flows (iphydr = 2)
Mesh modification tools
Unwarp smoother (non coplanar faces)
Adaptation of the algebraic multigrid algorithm for a vectorial Poisson equation
New linear solver GMRES (iresol = 3)
Coupling of velocity components (momentum equations) and coupling of the mesh velocity components in ALE (ivelco = 1)
New option (iswdyn(ipr) = 1, 2) for an automatic management of sweeps and relaxation coefficient on pressure equation.
A new residue is printed in the listing file. It’s now a residue on the complete resolution of the considered equation more representative of the resolution precision (it is no longer the final residual given by the linear solver - be careful: new residue cannot be compared to the previous one -).
Rewriting of the temperature equation: Cp is now out of the divergence term (to be consistent with the enthalpy equation). This modification has an influence only when Cp is variable. Be careful: now the diffusion term must not be divided by Cp.
New formulation for boundary conditions (split between the gradient and for the diffusive flux boundary conditions)
Wall post-treatment simplified to get Nusselt, strain coefficient, etc.
Improvement of the robustness of the eddy viscosity turbulence models
New numerical methods to solve more rigorously anisotropic diffusion equations (GGDH, head losses, ...).
SYRTHES 4.0(parallel version) with volumetric coupling
Code_Aster 11.0 via SALOME 6.6 (stabilisation of SALOME module FSI_COUPLING)
CFD_STUDY SALOME module
Unification of packages: bft, fvm, mei, ecs, and ncs are now regrouped in a single directory. The code installation and packaging is thus simplified.
The Fortran 95 standard (instead of fortran 90/77 format)
Memory dynamic allocation in Fortran files
The runcase (calculation launch script) has been simplified, advanced options for launch and pre-processing can be controlled with python scripts and C user files.
Factorisation and cleanings in Navier-Stoke solver subroutines
Migration to C language
Gradients computation is now in C
Pre-processing and post-processing user functions
New field structure to store solution field has been added. To help the addition of new variables and the migration to C (Ex.: rtp and rtpa arrays).
New post-processing structure (“writers” and “mesh”) to easily configure sub-mesh and advanced post-processing
Wrapping Fortran <-> C with the iso-c-binding features. To simplify the interaction between C and Fortran code and facilitate the migration to C.
Renaming and merging of the most common user subroutines.
Parallelisation of mesh joining algorithm (integrated to the kernel).
The mesh partitioning is done in parallel (integrated to the kernel).
New HPC performance management tool (performance tuning)
Automatic choice of the best options (case and machine dependent) to obtain the smaller computational cost (benchmark mode)
Simplified interface to select and test HPC options.
Parallelisation of Lagrangian particles tracking module
Hybrid parallelism MPI/OpenMP (work in progress)
Mesh renumbering tools to optimize cache and memory access
Graphical User Interface (GUI):
The following functionalities have been added or improved in the GUI:
Improvement in other advanced modelling management
Improvement of pre-processing management
Batch system calculation management
Rewriting of the theory guide
A new documentation of source code and user subroutines/functions with Doxygen is now available, including user examples and key words description.
Creation of new tutorials and improvements on the previous ones.
New default options:
Velocity components coupling is activated by default (ivelco = 1)
Improvement of symmetric boundary condition for Rij is activated by default (iclsyr = 1)
The previous validation campaign shown that this option is highly recommended to obtain good result with Rij turbulence models in axi-symmetric configurations. Moreover the iclsyr = 1 option is, theoretically, a more exact formulation than the previous one.
Relaxation on k-ω has been deactivated (relaxv(iomg) = 1)
The rewriting of eddy viscosity turbulence models previously mentioned to increase the robustness led to remove this relaxation.
Relaxation in steady algorithm is set to 0.7 (0.9 in 2.0)
This increases the robustness of the algorithm (but it could be more costly on a range of cases). Besides, 0.7 is the standard value in CFD codes.
k-ε with linear production turbulence model is suggested by default (in the GUI and in the cs_user_parameters.f90 file)
This model was already partially validated on the previous V&V stage and has been fully validated on this one. k-ε with linear production gives equivalent results than k-ε on most of test cases and gives significant better results on a part them (ex. jet impingement). We suggest using this model instead of standard k-ε.
For more details, the reader may refer either to the release notes of the version or to the svn ChangeLog and/or to the NEWS file.