Type
Text
Type
Dissertation
Advisor
Deng, Yuefan | Glimm, James | Harrison, Robert | Bluestein, Danny.
Date
2015-12-01
Keywords
Applied mathematics | GPGPU acceleration, multiple time stepping, multiscale modeling, performance, platelet simulation
Department
Department of Applied Mathematics and Statistics.
Language
en_US
Source
This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree.
Identifier
http://hdl.handle.net/11401/76516
Publisher
The Graduate School, Stony Brook University: Stony Brook, NY.
Format
application/pdf
Abstract
This work presents multiscale models and efficient numerical algorithms for analyzing the activation mechanisms of platelets under blood flow conditions at disparate spatiotemporal scales on supercomputers, with applications in initial thrombogenicity study and medical device optimization. Modeling the multiscale structures of platelets and the dynamics of their motion in viscous blood plasma require multiple time stepping (MTS) algorithm to optimally utilize the computing resources. This MTS algorithm improves the computational efficiency while maintaining stability and prescribed precisions. Our study of the dynamic properties of flipping platelets adapts the hybridized dissipative particle dynamics and coarse-grained molecular dynamics methods, which resolve the appropriate spatial scales of the platelet and the blood flow, respectively. In addition to the algorithmic strategies, general-purpose graphics processing units are also introduced to speed up the computationally intensive force field evaluations. Examinations of the implementation of the double-punch speedup strategy, i.e. | algorithmic MTS and hardware acceleration, reveal significant speedups over single time stepping algorithms and CPU-only solutions. Detailed performance analysis on three representative supercomputers affords the possibility of simulating the millisecond-scale hematology at resolutions of nanoscale platelet and mesoscale bio-flow using millions of particles, the state-of-the-art for the field at the present time. | 112 pages
Recommended Citation
Zhang, Na, "Design and Analysis of Parallel Algorithms for Multiscale Modeling of Platelets" (2015). Stony Brook Theses and Dissertations Collection, 2006-2020 (closed to submissions). 2423.
https://commons.library.stonybrook.edu/stony-brook-theses-and-dissertations-collection/2423