CORDIS Archive
This page has been archived. It will no longer be updated.
Parallel Solution for Solving 3D Flows
Aerodynamic analysis and the simulation of complex 3D viscous flows around aircraft components has been made more cost-effective with the use of parallel NSMB
In 1992 the Navier-Stokes Multi-Block (NSMB) code was designed for the efficient solution of turbulent Navier-Stokes problems. This code can be used to solve air flows of various speeds over complete and component parts of aircraft. However, the extremely high computing cost of simulating such complex flow situations prevents the code from being used as a design tool on a routine basis. The PARALLEL-AERO consortium, comprising two aerospace companies, Aérospatiale and Saab, and three research institutions, EPFL, KTH and CERFACS, have focused on migrating the NSMB code to parallel platforms to improve efficiency and cost-effectiveness. Test cases have demonstrated both the high scalability of parallel NSMB and its capability to solve problems too complex for conventional computer systems.
Solving the turbulent Navier-Stokes problems is critical to the competitiveness of the aerospace industry, as it allows faster aircraft design and the calculation of aerodynamic forces and air flows over aeroplanes. NSMB is therefore already playing a major role in the research and development activities of the European aerospace industry. Its parallelisation is likely to further strengthen the industry, by allowing the solution of more complex problems, and achieving performances equal to those of its US competitors.
NSMB was originally designed with its future parallelisation in mind. As such its structure is well suited to dividing large problems into smaller blocks which can be solved concurrently. Parallelisation has been acheived by grid partitioning, that is, 
dividing large meshes into submeshes using a special tool MB-Split.
Three test cases of different complexities have been chosen to demonstrate the performance of parallel NSMB. The first, the simulation of subsonic air flow over the so-called A-airfoil, typically takes a whole weekend to solve on a single workstation. Using the parallel version of NSMB, it can be solved much faster depending on the number of processors. When eight processors are used, performance surpasses that of the sequential NSMB run on a Cray YMP (1 processor). The second benchmark test refers to the 3D transonic flow around the so-called F5 wing inside a wind tunnel, which is a widely used validation test case. The results showed that parallel NSMB on 16 processors (i.e. Meiko nodes) solves this problem in the same time as the sequential NSMB on 1 processor of a Cray YMP. The most demanding test case refers to the transonic flow around the complete AS28G aircraft model (wind, body, pylon and nacelle), a problem traditional vector machines were unable to solve because of its complexity. Using parallel NSBM, the problem can be resolved in around 7hours using 32 processors of a thin-node IBM SP2. This compares to over 40 hours required by the sequential code on the Cray YMP.
tel +33-61-19-30-34 -- fax +33-61-19-30-00
e-mail poinot@cerfacs.fr
Research Area High Performance Computing and Networking
Related Results EUROPORT 1-HPCN3S; EUROPORT 1-Polyflow; EUROPORT 1-Samcef; EUROPORT 2-Paramation; EUROPORT 2-Maxhom; EUROPORT 2-Pepse; EUROPORT 2-Pulsar
Keywords aircraft design; computational fluid dynamics, parallel codes for; parallelisation;
| Project Participants |
|---|
| Aerospatiale FR |
| CERFACS FR |
| CIRA IT |
| EPFL CH |
| INPT ENSEEIHT FR |
| KTH SE |
| Saab SE |
| Universität Stuttgart DE |
home page | Results Zone | IT Solutions | Application Areas | Research Areas | High Performance Network Computing The URL for this page is /esprit/src/results/res_area/hpcn/hpcn11.htm
This page was last updated on 22 November 1996, and is maintained by esprit@dg3.cec.be