SCCS Colloquium - Jul 31, 2019
|Date:||July 31, 2019|
|Time:||15:00 - 16:00|
Ji-Ho Yang: Interface Jacobian Substructuring Algorithm for Multi-Component Dynamical Systems
This is a Master's thesis submission talk. Ji-Ho is advised by Tobias Neckel and BMW Research, New Technologies, Innovations.
Modern engineering products are becoming ever-increasingly more complex, involving many subcomponents from different subdomains of engineering. Performing numerical simulations for designing such products are not only computationally expensive, but also challenging due to their multi-physics nature. Although classical domain decomposition methods serve as highly robust and efficient means for design component integration, they are only able to handle Dirichlet-Dirichlet or Neumann-Neumann interface problems. In this thesis, we present an Interface Jacobian-based Substructuring method for general combination of interface constraints, including Dirichlet-Neumann coupling, and apply the method to linear dynamical systems. The core of this method is in generalization of formulation for solving interface problems by introducing both Dirichlet and Neumann constraints via interface Jacobian matrix. The algorithm is formulated and implemented on GPU in order to achieve real-time numerical simulation. Finally, the method is applied to an existing engineering product, and its computational performance is studied.
Keywords: Design Component Integration, Domain Decomposition, Model Order Reduction, GPGPU
Vivian Haller: Evaluation of dimension-wise Error Estimates using the Spatially Adaptive Combination Technique
This is a Bachelor's thesis submission talk. Vivian is advised by Michael Obersteiner.
In a recent paper by M. Obersteiner and H.-J. Bungartz, a novel approach for numerical quadrature has been developed, based on a spatially adaptive variant of the sparse grid combination technique. In this thesis, the resulting procedure, henceforth referred to as the split-extend scheme, has been subject to a slight modification regarding one of its fundamental operations, essentially reducing its consumption of function evaluations per application. This modified procedure, restricted to piecewise linear basis functions, has been put to the test with regard to a selection of widely used functions. In several cases, the corresponding results have either shown a noticeably increased performance or proved to be at least on par with the basic scheme, which indicates that the method itself might be worthy of further improvement.
Keywords: Quadrature, Spatial Adaptivity, Sparse Grid Combination Technique