Dissertation: Implementation techniques for the lattice Boltzmann method

26.08.2010

The PhD thesis of Keijo Mattila (University of Jyväskylä) representing the field of information technology deals with research into computational fluid dynamics simulations based on the lattice Boltzmann method. The Louhi and Murska clusters, as well as other computing resources of CSC – IT Center for Science Ltd. have been used in this research.

Mattila’s thesis work is a subfield of natural sciences, aiming to explain real fluid flows by simulating flow phenomena with a computer. Mattila’s research results have been utilized when simulation software for computational fluid dynamics (CFD) research was developed in the collaboration project by the University of Jyväskylä and Åbo Akademi University. This software is vying for the title of the world’s fastest software of its kind. This software was used, for example, in the decision-making for the acquisition of the most powerful supercomputer in Finland to date: the run time on CSC’s Louhi supercomputer was one of the evaluation criteria of computing power. The simulation software has also been used in industrial applications. At the Department of Physics of the University of Jyväskylä, the software has been used to simulate liquid flow within paper, cardboard, and other complex porous materials. This type of research has practical applications in, for example, paper manufacturing technologies.

This branch of research is known as high-performance computation (HPC). With every decade computers’ computing power capabilities have grown immensely. For many jobs in our daily routines the theoretical computing power maxima are already well in excess of needs. In spite of this, new and ever more powerful computers are still being developed at a hectic pace. Hence, there is good reason to ask why this development work is being continued. There are two main reasons: firstly, the idea is to make the day to day tasks performed on the computer more user-friendly and more secure. More powerful computers are required by the new operating systems and the ever-more comprehensive data security software. Secondly, there are special fields where the computing power levels available currently or within the near future are not enough. In addition to the entertainment industry and industrial product development, this group of fields includes academic research. Therefore, increasing computing powers and providing applications that better utilize the computing power are continuously in demand.

Simulations in scientific research are always based on “numerical methods”. In his doctoral thesis Mattila focuses on the lattice Boltzmann method. The computing method determines the main algorithm solutions used in programming. One of the research topics in HPC research is how to solve the remaining technical details relating to programming. These individual solutions can have an important effect on the speed of the simulation program, depending on the operating computing system.

Relating to the software engineering solutions, the thesis introduces a novel, more efficient algorithm for the lattice Boltzmann method. The work also contains an improved data structure utilizing cache memory better than previous solutions. In addition to HPC, the thesis addresses other matters relating to method development. For example, a new type of boundary condition for the lattice Boltzmann method is introduced; the meaning of this boundary condition is to improve simulation of fluids in complex structures.
- One possible application area might be for carrying drugs through deformed blood vessels towards a cancerous tumor, Mattila speculates.

Additional information:

Keijo Mattila, keijo.mattila(at)jyu.fi

CSC

CSC - IT Center for Science Ltd. is a non-profit limited company administered by the Ministry of Education, Science and Culture. Having core competences in modeling, computing and information services, CSC provides versatile IT services, support and resources for academia, research institutes, and companies. CSC provides Finland's widest selection of scientific software and databases and Finland's most powerful supercomputers that researchers can use via the Funet network.