Extended Abstract (964K)J19.5
Integrating a complete Lagrangian dispersion model into an urban GPU-based simulation system
Balwinder Singh, University of Utah, Salt Lake City, UT; and E. R. Pardyjak, A. Norgren, and P. Willemsen
Our team has utilized the highly parallel and inexpensive data paths provided by computer graphics video hardware to enhance the performance of an urban Lagrangian dispersion model. In previously reported validation studies, our team has shown that the GPU (Graphics Processing Unit) based Lagrangian dispersion model reduces simulation time by two orders of magnitude compared to the CPU version of the code. GPUs are a highly programmable component of the computer graphics hardware that can be used for solving engineering problems and rapidly visualizing results. One immediate benefit of using the GPU is the real time rendering of data (present on the graphical server) onto the screen. Real time data rendering is particularly important for applications involving real time visualization such as those typically associated with virtual environments. The Lagrangian dispersion model that we used in previously reported studies was based on a simplified version of the Langevin equations with idealized boundary layer assumptions. Recently, we have integrated the generalized form of the Langevin equations (without the boundary layer assumption) in the Lagrangian dispersion model following the work of Yee and Wilson (2006). In the present paper, the challenges associated with the integration of the generalized form of the Langevin equations on the GPU architecture will be discussed. Validation simulations for the simple test cases of uniform flow and boundary layer flow along with a validation study for a more realistic urban geometry will be presented.
Joint Session 19, Urban Transport and Dispersion Modeling—Part II (Joint with the Meteorological Aspects of Air Pollution Committee)
Wednesday, 14 January 2009, 4:00 PM-5:30 PM, Room 124A
Previous paper Next paper
Recorded presentation