Skip to main content

High performance simulation of internal combustion engines

Searching for OpenAIRE data...

Exploitable results

Simulations that are performed on massively parallel computers result in huge amounts of data that usually cannot be handled by workstations. The COVISE environment provides the possibility to do the first visualization steps directly on the parallel computer and thus reduce the amount of data that has to be transferred to the workstation for the final visualization significantly. The integration of the visualization with the numerical simulation makes it easy to view the results of intermediate results of the ongoing simulation. The increasing need for computing power to solve increasingly complex problems (eg combustion engines or crash simulation) makes massively parallel computers very attractive. The main market will be the users of extreme computing power like car manufacturers and the aerospace industry. The main emphasis in the HPS-ICE project is on the overall speedup of the development cycle of a combustion engine. To achieve this, the computational fluid dynamics (CFD) simulation and parts of the visualization are done on a MPP system (a fast internal network massively parallel computer), focusing on the optimal usage of the parallelization. The main technical achievement is the integration of a distributed visualization environment with the simulation code on a massively parallel computer system. Numerical simulations of internal combustion engines are one of the most complex computations executed today. The benefit of using MPP systems for simulation and visualization can be in one of two ways: the users solve problems that are much bigger and more detailed than those that can be solved on standard supercomputers; the user solves the same problem in a shorter time. Both ways offer the user a competitive advantage in that he either has a better solution in the same amount of time, or the same solution in a shorter time.
STAR-HPC/PROICE is a suite of engineering software tools for rapid analysis of flow, mixing and combustion in reciprocating engines, using computational fluid dynamics (CFD) methodology and contemporary message-passing parallel computers. The software will be of interest to automobile and engine manufacturing companies and their suppliers, as well as research institutes and consulting companies involved in reciprocating engine analysis and design. The novel features of the software are the ability to quickly set up, perform and display CFD engine simulations using parallel computers and automated meshing techniques. The main benefit from using the software is the ability to perform engine simulations in a much shorter period and with less specialized skill requirements than hitherto, increasing the utility of the tool for engine design There are two important technical aspects of the STAR-HPC/PROICE combination which give it the improved performance. One is the ability to rapidly generate the body-fitted moving meshes required for engine simulations. This is done in PROICE by use of pre-constructed templates, which can be fitted to computer aided design (CAD) descriptions of the actual engine geometry and already contain the necessary controls for mesh movement. The second aspect is the ability of STAR high performance computing (HPC) to perform moving-mesh engine simulations on parallel computers. Here the main novelty is dynamic load balancing, in which the topology changes which the moving mesh undergoes are catered for by periodic redistribution over the computer processors to maintain uniform load and hence near-optimal performance. Neither of these aspects is available in competing software.