With
the advent of robust, accurate flow solvers and automatic grid
generators, the
task of defining quickly a flow domain and the required boundary
conditions has
become a key bottleneck for numerical simulations. For so-called body
fitted
grids, the surface definition must be water-tight, and any kind of
geometrical
singularity, as well as small angles, should be avoided in order to
generate a
mesh of high quality. This typically presents no problems if the
definition of
the structure emanates from a CAD package. However, in many cases first
the CSD
models (i.e. abstractions of the real geometry) are built, and the
`wetted
surface' is exported, serving, in part, as the definition of the flow
domain.
This data may not be water-tight, may have small strips and geometrical
singularities. Therefore, even with
sophisticated software toolkits, manual cleanup in most cases takes
days for
complex geometries (e.g. a complete building).
An alternative is to use grids that are not body-conforming, and
simply
`embed' the triagulations of the wetted surfaces of the structures in
them.
Techniques of this kind are also known as immersed, embedded,
fictitious domain
or Cartesian methods. The treatment of points in the vicinity of the
embedded
CSD triagulations or CSD bodies is modified in such a way that the
required
kinetic or kinematic boundary conditions are properly imposed. The talk will review the implementation of
embedded mesh and immersed body techniques in the context of adaptive
unstructured grids. Comparison of
results for a wide variety of flows using the same solvers with the
body-fitted, embedded and immersed options will be given.
Dr.
Lohner earned an MSc in
Mechanical Engineering from the Technische Universitat Braunschweig (