Committee Chair

Karman, Steve L., Jr.

Committee Member

Newman, James C. III; Jones, William T.; Tanis, Craig R.

Department

Dept. of Computational Engineering

College

College of Engineering and Computer Science

Publisher

University of Tennessee at Chattanooga

Place of Publication

Chattanooga (Tenn.)

Abstract

The CFD (Computational Fluid Dynamics) community has long used the Overset Grid method to enable dynamic simulations with bodies in relative motion. In Overset simulations, information is transferred between overlapping grids via interpolation. Domain Assembly is the process that governs the location of intergrid boundaries and how the solution is interpolated across grids at those boundaries. Performing Domain Assembly in a distributed environment is computationally expensive and inherently poorly load balanced due to the solver partitioning. Dynamic load balancing is therefore required to alleviate the imbalance and make very large Overset problems feasible. In this work, a radically different parallel domain assembly method is introduced. The new method takes a fundamentally different approach to load balancing, concurrency, and communication patterns. A detailed discussion is provided that describes the method's implementation in YOGA (Yoga is an Overset Grid Assembler). Finally, several case studies are analyzed and preliminary performance and scaling results are provided.

Acknowledgments

I would like to acknowledge my colleagues at NASA Langley for providing me a supportive and growth oriented environment. I would like to thank my committee for seeing my project through to the end, despite being spread across three states. I am particularly grateful for Dr. James Newman III for all of his hard work to make sure that so many students could graduate at the same time.

Degree

Ph. D.; A dissertation submitted to the faculty of the University of Tennessee at Chattanooga in partial fulfillment of the requirements of the degree of Doctor of Philosophy.

Date

8-2016

Subject

Numerical grid generation (Numerical analysis)

Document Type

Doctoral dissertations

Extent

xiii, 110 leaves

Language

English

Rights

Under copyright.

License

http://creativecommons.org/licenses/by/3.0/

Share

COinS