| Modern modeling and simulation e˙orts are becoming larger and more complex with corresponding evolutionary advances in computation hardware. High-fidelity physics data representation and communication are critical issues to any multiphysics simulation, no matter the scale. Functional expansions have been previously shown to have characteristics desirable for these situations.
In this dissertation, the understanding of functional expansion tallies was advanced through characterization in the Serpent reactor physics Monte Carlo code. New algorithms were developed that significantly improved the computational eÿciency of functional expansion-based data representation methodologies. Next, a figure-of-merit was developed for functional expansion tallies, which was then used to demonstrate their advantage in overall computational time.
Functional expansion tools were then developed as a module in MOOSE, a finite element analysis framework. Developments included classes to generate, store, and reconstruct variable fields. These functional expansion tools were then adapted in MOOSE to couple with Serpent’s multiphysics interface. All these developments were accompanied by testing to demonstrate both the viability and use of these tools.
This work is unique in the sense that it: 1) characterized—and improved upon—the computational eÿciency of functional expansion algorithms, 2) developed a generalized functional expansion-based coupling framework in MOOSE, 3) coupled MOOSE and Serpent using functional expansions, and 4) performed all these functions in a fully-3D fully-multivariate context.
Key Words: functional expansions (FEs), functional expansion tallies (FETs), algorithm optimization, statistical convergence, multiphysics coupling, MOOSE, Serpent |