Parallel object-oriented algorithms for simulation of multiphysics : application to thermal systems

The present and the future expectation in parallel computing pose a new generational change in simulation and computing. Modern High Performance Computing (HPC) facilities have high computational power in terms of operations per second -today peta-FLOPS (10e15 FLOPS) and growing toward the exascale...

Descripción completa

Detalles Bibliográficos
Autor: López Mas, Joan
Tipo de recurso: tesis doctoral
Fecha de publicación:2016
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/96253
Acceso en línea:https://hdl.handle.net/2117/96253
https://dx.doi.org/10.5821/dissertation-2117-96253
Access Level:acceso abierto
Palabra clave:Àrees temàtiques de la UPC::Informàtica
Descripción
Sumario:The present and the future expectation in parallel computing pose a new generational change in simulation and computing. Modern High Performance Computing (HPC) facilities have high computational power in terms of operations per second -today peta-FLOPS (10e15 FLOPS) and growing toward the exascale (10e18 FLOPS) which is expected in few years-. This opens the way for using simulation tools in a wide range of new engineering and scientific applications. For example, CFD&HT codes will be effectively used in the design phase of industrial devices, obtaining valuable information with reasonable time expenses. However, the use of the emerging computer architectures is subjected to enhancements and innovation in software design patterns. So far, powerful codes for individually studying heat and mass transfer phenomena at multiple levels of modeling are available. However, there is no way to combine them for resolving complex coupled problems. In the current context, this PhD thesis presents the development of parallel methodologies, and its implementation as an object-oriented software platform, for the simulation of multiphysics systems. By means of this new software platform, called NEST, the distinct codes can now be integrated into single simulation tools for specific applications of social and industrial interest. This is done in an intuitive and simple way so that the researchers do not have to bother either on the coexistence of several codes at the same time neither on how they interact to each other. The coupling of the involved components is controlled from a low level code layer, which is transparent to the users. This contributes with appealing benefits on software projects management first and on the flexibility and features of the simulations, later. In sum, the presented approaches pose a new paradigm in the production of physics simulation programs. Although the thesis pursues general purpose applications, special emphasis is placed on the simulation of thermal systems, in particular on buildings energy assessment and on hermetic reciprocating compressors.