Parallel methods for static and dynamic simulation of flexible pavement systems.

摘要

Flexible pavement systems represent a significant portion of the national infrastructure and constitute a major economic investment for federal and local agencies. More accurate modeling and better quantitative understanding of their behavior have the promise to produce more effective pavement designs and better rehabilitation strategies for increasing the life of existing pavements. This document presents work on the development of high-fidelity, parallel numerical simulations of the behavior of flexible pavement systems.; Under static loading conditions, the finite element models are developed that allow us to capture the three dimensional behavior of flexible pavement systems, including complex geometries, cracks in the pavement surface, inter-layer slip under non-uniform truck tire loads. A parallel iterative solution strategy is implemented to allow simulations of large flexible pavement systems on distributed memory computers and achieve solution scalability. These models are verified with classic elasticity problems and used to assess the effect of non-uniform tire loads on surface initiated cracking.; For dynamic loading conditions, finite volume methods are used to model the behavior of layered flexible pavement systems and are shown to be effective in simulating the response to a Falling Weight Deflectometer load. The parallel implementations of the dynamic simulations are explored for uniform and non-uniform grid solutions including the Adaptive Mesh Refinement algorithm using the data parallel programming language ZPL. The results show that the performance of both static and dynamic simulations is scalable.