Book Description

Adaptive mesh refinement is a powerful tool for obtaining the highest solution accuracy for a given computational effort. Over the past decade, many adaptive techniques have been developed and applied to a variety of fluid flow problems. Results obtained for compressible flows, and to an extent, 2D incompressible flows have been impressive, however, similar progress has not been noted for 3D incompressible flows, particularly in complicated geometries. The objective of this thesis was to develop and test an adaptive solution methodology for 3D incompressible flow simulations in domains of arbitrary complexity. To characterize the finite element solution error, the Zienkiewicz-Zhu patch recovery error estimator (LPR) was adopted. An enhanced version of the LPR error estimator was formulated and implemented using 10-noded tetrahedral elements. The enhanced estimator (LPRC) resulted in significantly improved gradient recovery (and consequently, improved error estimates) at virtually no additional computational cost. For mesh refinement, an elemental subdivision procedure was implemented. To enable refinement in complex geometries, a procedure for preserving the boundary integrity of a refined mesh was developed. This methodology can be used for geometric data from any solid modeling (CAD) system provided the data can be exported in the IGES format. A benchmark study of the AMR procedure, in which steady flow over a three-dimensional backward-facing step was simulated, showed that the cumulative computational effort required in the adaptive analysis was lower than that required in a non-adaptive analysis of the same problem. In the second phase of this work, the AMR procedure was applied to modeling flow through two arterial geometries. Specifically, flows in an idealized end-to-side anastomosis and in a human right coronary artery were examined. Both studies assessed whether an AMR analysis could achieve more accurate solutions than conventional analyses that utilize high-resolution meshes whose gradation is based on 'a priori' knowledge of the flow field. It was noted that mesh-independent velocity fields were not very difficult to obtain even in the absence of an adaptive methodology. However, wall shear stress fields were much more difficult to absolutely resolve non-adaptively. Given that shear stresses occurring on arterial walls are widely believed to be a key factor governing the development of arterial disease, it is very important to accurately resolve wall shear stress fields if confidence can be placed in the results of numerical simulations of arterial flow phenomena. These results indicate that wall shear stress is an extremely sensitive measure of spatial resolution, and that the systematic solution-adaptive methodology developed in this thesis is very effective in producing accurately resolved wall shear stress fields.