SC²S Colloquium - December 2, 2015
|Date:||December 2, 2015|
|Time:||3:00 pm, s.t.|
Paul Cristian Sarbu: Fluid Simulation with Moving Grids for Medical Training
Many fluid dynamics applications, such as those in medical training simulators, require high resolution close to the visual center of the domain view, with the rest of the coarser simulation in the background. The aim of this thesis is to modify an existing grid-based Eulerian solver in a minimal way such that it allows the coupling between a coarse background grid and a finer grid on top, maintaining consistency across the visual domain. A simple coupling technique is proposed for the advection step of the simulation, while for the pressure solver step two coupling approaches are presented, one velocity based and the other one pressure based. We bring theoretical arguments based on operator splitting techniques for the validity of the methods and we test the implementations against the original solver. Simulations show that the velocity based approach has an inherent instability that can only be reduced through an unrealistically strong coupling. The pressure based approach is far more stable and accurate. Tests in 2D and 3D settings show also a possible attainable speedup compared to the original solution, due to parallelism in the coupling algorithm. However, the maximum performance of the method is dependent on the extent of the modifications brought to the code of the original solver.
Tamas Borbath: Enhancing Software Tools in Magnetic Resonance Spectroscopy
Magnetic Resonance Spectroscopy(MRS) is an in vivo health care diagnostic tool used to diagnose several diseases originating in the chemical composition of the tissues. The recent developments of the technology have brought it to a wider use both in clinics and research. To improve the technology, there is a need to create tools which allow a faster improvement of the reconstruction algorithms by scientists. This thesis aims at developing a Graphical User Interface for an existing Software Development Kit which provides several algorithmic steps used for the standard reconstructions of the data. It presents the software engineering design using patterns of software engineering to address the complexity and multi-dimensionality of the MRS reconstruction and visualization. Software enhancements to create a Matlab interface and custom recorded functions (i.e. Macros), as well as code optimizations were also a focus of the work.