Generation of Computational Mesh for Haemodynamics Analysis - Computational Hemodynamics: Theory, Modelling and Applications - page 165

Some drawbacks of unstructured mesh include the treatment of the nodal points
which requires additional computational memory as they can't be treated simply by
a double of indices ( i , j ) in two dimensions or a triple of indices ( i , j , k ) in three di-
mensions; additional computations are needed to connect up arbitrary neighbouring
nodes. This typically results in increased computational times to obtain a solution
and erode the gains in computational efficiency compared with a structured mesh.
An unstructured mesh is ineffective in resolving wall boundary layers since tri-
angles and tetrahedrals do not deform (stretch or bend) during local refinement
to make the cells very small. Tetrahedral cells are also prone to high aspect ratios
which affect the skewness of the cell. Additionally these cells are difficult to align
with the flow direction. These two problems can impede solution convergence and
lead to artificial errors in the solution known as numerical diffusion . It is a common
source of error and is also called false diffusion because it is a numerical error and
does not represent a physical diffusion process. A few techniques may be applied to
minimise the likelihood of false diffusion such as choosing higher order discretisa-
tion schemes (see Chap. 5) and increasing the resolution of the mesh.

6.2.9 Mesh Terminology

When developing a mesh, there are some terminologies that a new user will encoun-
ter. In this section we describe some of the terminology associated with meshing,
and with examples shown in Fig. 6.14 .

Structured Mesh
• any mesh containing an ordered set of locally orthogonal lines
• quadrilateral cells are used

Fig. 6.14 Examples of mesh configurations using both structured, unstructured, and hybrid
meshes