X hits on this document

20 views

0 shares

0 downloads

0 comments

2 / 8

TIC2003-07970

fluid dynamics and heat transfer phenomena taking advantage of the new computing resources in parallel computation using loosely coupled parallel computers: Beowulf clusters.

Even though the project has a fundamental character, and the main achievements will be in the academic area (publications, doctorate thesis...), it will be essential for the applied studies of the Group in the development of high performance software for the simulation of specific thermal system and equipment of industrial and social interest.

1.2-Strategy for parallel code development According to the most appropriate techniques for their parallelization, the codes/software to be developed within this project for the treatment of the different phenomena of interest have been grouped into three main areas:

1)

Detailed

resolution

of

turbulent

incompressible

flows

based

on

DNS

and

LES

methodologies

2)

Resolution of flows with (approximately) parabolic spatial structure (cases of combustion, free surfaces and flow between plates will be of special interest)

3)

Resolution of flows with elliptic structure (turbulent flow in general with RANS models and moving surfaces).

The possibilities of using the most appropriate algorithms in every area without restrictions will compensate the large scope of the project in respect to the implementation, verification, validation and maintenance of the codes. The design of the algorithms and the structure of the three mentioned codes will assure their portability and efficient use in other architectural types, such as share memory and NUMA (Non-Uniform Memory Access) systems. The recent installation of the supercomputer Mare Nostrum, a distributed memory tightly coupled computer has confirmed that our initial strategy of total portability was appropriated.

1.3-Methodology The general methodology of the work for the development of these algorithms and codes can be divided into the following levels (which present high inter-relationships):

  • 1)

    Mathematical formulation of the physical phenomena.

  • 2)

    Algorithms and techniques for the numerical resolution.

  • 3)

    Development and verification of parallel software.

  • 4)

    Resolution of different phenomena/configurations and analysis of the quality of the

5)

numerical solutions (discretization errors and convergence errors). Validation of the mathematical formulations developed (modelization errors).

Although the project has essentially a basic/fundamental nature, the potential applications are addressed to the simulation of thermal systems and equipment of industrial and social interest in which the milestones achieved in this basic research line are applicable. Thus, the experience acquired will allow a technology transfer to the companies with which the Group collaborates in the framework of research contracts and/or national and European research projects. This will allow not only to maintain and improve these collaborations, but also to increment the possibilities of application to other companies and industrial and social sectors.

Document info
Document views20
Page views20
Page last viewedMon Dec 05 18:56:01 UTC 2016
Pages8
Paragraphs253
Words2996

Comments