Annual Report


Previous Contents

4.2   Numerical Methods and Software Engineering

4.2.1   Finite Element Method in Unstructured Meshes (G. Chevalier, F. Ducros, B. Marquez)

The work on that field has followed three axis.

  • first a work on the Domain Decomposition and preconditioning, with firstly the implementation of the ILDU method in an official version of AETHER, and secondly other improvements such as introduction of explicit Runge-Kutta scheme, variables interpolation, LES modeling. All these developments have been directly delivered to Dassault Aviation in an official version of AETHER;

  • second a work concerning the modeling relative to the LES of complex unsteady flows. This work has involved the development and validation of two formulations of wall functions, validated on standard periodized channel flow. CERFACS has also collaborated with Dassault Aviation through the LESFOIL EC program, sharing expertise and data on the subject ([Ducros, 2000]);

  • following the work on the preconditioning, a study has been conducted on the influence of the node numbering on the robustness and efficiency of the preconditioning. This was done on different types of preconditioning.

4.2.2   Meshing techniques (J. Bohbot, J.-Ch. Jouhaud, M. Montagnac, D. Darracq)

Grid generation is a crucial problem for the computation of complex aircraft configurations using a body fitted structured code. Furthermore, due to the data management of structured grids, the local refinements around the CAD (boundary layers, stagnation lines, wakes) will be propagated through the whole grid even in zones where gradients are expected to be low. This can lead to very large grids, especially for complex geometries. Two approaches permit to reduce this drawback: i) conservative patched grid, ii) automatic mesh refinement.
CERFACS is developing these techniques in order to help Airbus France in reducing the turnaround time and in enhancing the flow accuracy on a given grid size. These two meshing techniques have been implemented and validated in NSMB.

Conservative Patched Grid (CPG) (M. Montagnac)

The purpose is to develop an efficient way to simplify the grid generation and to deal with complex configurations using moderately sized grids. The approach has been chosen to use NSMB on meshes having no coincident interfaces. For this kind of meshes, blocks must have common interfaces but do not need the same location of grid nodes. This approach avoids mesh propagation from a block to another block. The flexibility of this kind of mesh allows mesh refinement and makes it easier to cluster grid points.
A 3-D conservative patched grid algorithm using Jameson centered scheme has been implemented in NSMB. The Multigrid acceleration convergence technique has been extended when using patched grids strategy. The parallel programming has been implemented to maintain the CPU performance of the NSMB code. The efficiency of the method has been demonstrated on complex geometry. The wake vortices behind a civil aircraft (DLR-F11) have been calculated on ten processors using a grid containing 3 millions nodes with non-coincident interfaces (Fig.4.3).
These CPG techniques were also developped in elsA as a product of the NSMB/elsA tightening. Full totally non-coicident patched grid techniques for 3D turbulent flows were implemented in the elsA code. Special care was also taken to ensure the use of these techniques when computing on parallel computers. An example is shown in the Fig.4.4 for a 2D turbulent application : the RAE 2822 test case.

Adaptive Mesh Refinement (AMR) (J.-Ch. Jouhaud, M. Montagnac)

Another approach is to enrich the mesh with a hierarchical grid structure, that is to say using an Adaptive Mesh Refinement Strategy (AMR). It consists in splitting a cluster of grid cells in a characteristics regions (shocks, boundary layers, wake vortices, ...) and to automatically join these cells in new blocks. The activity on AMR has increased over the period with a new algorithm to improve the method. In particular, a parallel implementation has been realized. In the next phase, extension to viscous flows is planned.
The efficiency of the method has been demonstrated on a complex geometry. The wake vortices behind a civil aircraft (F11) have been calculated on ten processors using a grid containing 1.5 millions nodes with two levels of refinement.
The most recent investigations were realized within the elsA tool. Two major objectives were reached. Firstly, the extension of the AMR method to turbulent transonic flows. Secondly, the addtion of local time increment prolongations, as in standard multigrid strategies, in order to increase the convergence rate. The RAE airfoil and the AS28 wing test cases have shown the efficiency of this new AMR extension (Fig. 4.5).



Figure 4.3: Flow simulation behind a civil aircraft during the landing configuration. Application of the AMR technique.





Figure 4.4: elsA/Patched grid technique applied to the RAE2822 test case





Figure 4.5: elsA/AMR technique applied to the AS28G airfoil


4.2.3   Spalart-Allmaras (H. Pascal-Jenny)

In the framework of 'elsA-NSMB' convergence activities, the update of the one equation turbulence model functionnalities for a production-level use has been addressed during year 2001. This topic follows the work previously realized for the algebraic turbulence model of Baldwin-Lomax. The LU-SSOR implicit algorithm extension to the well-known Spalart-Allmaras (SA) turbulence model has been transfered from the NSMB code to the elsA software. Specific treatments (unknown's positivity monitoring, parallelism, ...) have also been implemented leading to a robust and efficient solver for the computation of turbulent flows over full aircraft configurations. The extension of the multigrid algorithm to the SA model is in progress and will be finalized during the beginning of year 2002. These activities are part of a contract with Airbus France with the goal of providing the same level of efficiency with both softwares in order to switch from one to the other in a transparent way inside the design process.

4.2.4   MAEVA (H. Pascal-Jenny)

During year 2001, CERFACS has seen a growing involvement in the developpement of the elsA software through the MAEVA project (Modelisations Aerothermiques des Ecoulements en Ventilation Avion) between Airbus France and ONERA. Indeed, CERFACS has been in charge of implementing a new turbulence model dedicated to strongly thermaly coupled flows (impinging jets for instance), namely the (k, e, v'2, f) model of Durbin. This model is described by 3 additionals transport equations (k, e, v'2) and another one for f which is elliptic. An explicit version of the model has been implemented in the elsA software this year and validations are in progress. Extensions are scheduled for the year 2002.

4.2.5   Development and implementation of a wake vortex model in flight simulators (H. Moet, G. Grondin, D. Darracq)

In the framework of the S-Wake program, the CERFACS is involved in the safety aspects of aircraft wake vortices. An engineering wake vortex behaviour model (called VORTEX) developed with the purpose to do probabilistic investigations has been improved by integrating several decay models which govern the decay of circulation due to turbulence, stratification, viscosity and crosswind. The effect of crosswind on wake vortices has been investigated in order to incorporate the effect properly in the VORTEX model. A new decay model accounting for aircraft parameters and atmospheric turbulence characteristics has been developed at CERFACS  [Möet, 2001]. An analytical model has been developed which is integrated in industrial (Airbus Deustschland) and research (TU-Berlin, NLR, TU-Braunschweig) flight simulator environments to perform parametric studies for wake vortex encounters. These parametric studies (including variation of vortex parameters) have the purpose to obtain the safety hazard for an encounter, to obtain handling criteria and to determine the influence of cockpit motion on piloted simulations. Furthermore, a numerical model is developed which uses a more realistic vortex velocity distribution is used for the same purposes as the analytical model and to simulate a more realistic wake vortex encounter and to investigate the effects of 'low-vortex'- designs of future very large transport aircraft.

4.2.6   Software engineering (J. Bohbot, S. Champagneux, J.-Ch. Jouhaud, Ph. Piras)

Dealing with the CFD code NSMB, the CERFACS is more than ever considered as the official provider as much from the Airbus France point of view as the one of EADS Launch Vehicles . Integration, management and validation processes have consequently been improved in order to face this responsibility in a reliable way (control versioning with both CVS (integration) and SCCS (development), testing over a wide range a representative test cases). Transferring, optimizing and benchmarking the CFD code NSMB on various parallel platforms have also been carried out and continued, motivated by both Airbus France as well as EADS Launch Vehicles for the renewal of their own computing facilities. Algorithm performance improvements and various optimizations have been realized leading to an increased efficiency of the code both on vector and RISC architectures. Finally, the harmonization between computational tools coming from CERFACS and Airbus France has been continued with the delivery and the use of new software (JEDM, GAME, LAMA3D) so that developers at CERFACS can work within an industrial environment in order to reduce the cost of the know how transfer when a development is completed.

4.2.7   Object Oriented CFD with elsA (M. Montagnac, G. Chevalier, S. Champagneux, J.-Ch. Jouhaud)

In the year 2000 the collaboration with ONERA has started on the object-oriented code named elsA ensemble logiciel de simulation Aérodynamique. The elsA software comes along with procedures to enhance productivity in a multi-user and multi-platform environment: validation database, unitary test cases, cvs management tools, software quality program. This work takes fully part of the framework wanted by Airbus France for the harmonization between NSMB and elsA.

The CERFACS elsA team was in charge of the conception and the development of the implicit time-integration LU-SSOR method as well as of the Baldwin-Lomax algebraic turbulence model with generalized distances. It has also contributed to write the specification and the user guide manuals. Some validation and unitary tests have been integrated in the corresponding databases too.

All developments have been tested on usual cases as 2D-3D multi-blocks sequential or parallel square nozzle, flat plates, NACA and RAE profiles. These developments were integrated in the official version of elsA.

The code has been installed on FUJITSU VPP700 and VPP5000 vector machines to run the S3CH configuration (see Fig.4.6) for Airbus France.




Figure 4.6: Cp distribution around the S3CH (wing+pylon+nacelle) configuration


4.2.8   CFD Team Support (S. Champagneux, M. Labadens, F. Dabireau, V. Roche)

This year, the CFD Team was involved in improving its communication by redesigning its website [Roche, 2000] (http://www.cerfacs.fr/cfd). The most important work consisted in updating information about the team that now fit its current structure. A wide modernization has been conducted to make it more attractive, simpler to navigate and easier to maintain. A specific effort has been made to rationalize the management of the CFD Team bibliography that now consists of more than 600 references from year 1986 to 2001 from which some are available on-line.

   Return to TOP
webmaster@cerfacs.fr
Last Updated: Apr 11, 2008
Copyright  © 2002-2007 CERFACS   
All rights reserved.