Geometry Query Optimizations in CAD-based Tessellations for Monte Carlo Radiation Transport

Download or read book Geometry Query Optimizations in CAD-based Tessellations for Monte Carlo Radiation Transport written by Patrick C. Shriwise and published by . This book was released on 2018 with total page 0 pages. Available in PDF, EPUB and Kindle. Book excerpt: The performance of direct CAD-based Monte Carlo Radiation Transport (MCRT) relies heavily on its ability to return geometric queries robustly via ray tracing methods. Current applications of ray tracing for MCRT are robust given that certain requirements are met \cite{Smith_2011}, but cause simulations to run much longer than native code geometry representations. This work explores alternate geometry query methods aimed at reducing the complexity of these operations as well as algorithmic optimization by adapting recent developments in CPU ray tracing for use in engineering analysis. A preconditioning scheme is presented aimed at avoiding unnecessary ray queries for volumes with high collision densities. A model is also developed to inform the application of the preconditioning data structure based on a \textit{post facto} analysis. Next, a specialized ray tracing kernel for MCRT is presented. As new ray tracing kernels are developed for real-time, photo-realistic rendering, algorithmic approaches have appeared which are demonstrated to be advantageous when applied in radiation transport. In particular, the application of data parallelism in ray tracing for Monte Carlo is demonstrated - resulting in significant performance improvements. Finally, model features resulting in systematic performance degradation commonly found in CAD models for MCRT are studied. Methods are proposed and demonstrated to improve performance of ray tracing kernels in models with these features. The combination of this work is shown to provide improvement factors ranging from 1.1 to 9.54 in simulation run time without loss of robustness for several production analysis models. The final impact of this work is the alleviation of concern for additional computational time in using CAD geometries for MCRT while maintaining the benefit of reduced human time and effort in model preparation and design.