readme

采用有限容积法、非正交结构化网格（H型、O型及C型）求解REYNOLDS时均方程的程序

CODES FOR RANSE
===============

This directory contains two versions of the CAFFA-code which
include turbulence models for high-Reynolds-number flows. 
One version includes the standard k-epsilon model with wall
functions; the corresponding files are in the sub-directory
"keps". The other version includes the k-omega model of Wilcox;
the files are in the sub-directory "komega". The codes are
otherwise very similar to the laminar version, which is now 
in the directory "2dgl"; both models were implemented by
Martin Schmid, PhD student at the Institute of Shipbuilding
in Hamburg. See the readme-file in the directory "2dgl" for
more information on code structure and how to run it.

The extensions are described through comments in the code;
otherwise see Chap. 9 for a description of the k-epsilon
model (or the original references cited therein) and the
literature listed in the "readme"-file in the sub-directory
"komega" for the k-omega model.

In the sub-directory "examples" there are input files for
some of the test cases already used for laminar flow
computations. I have tested the "channel", "damdif", and
"bfstep", but I can not guarantee that the codes are bug-free,
since it is not easy to test codes with turbulence models.
In the case of laminar flows, there is a number of test cases
with very accurate solutions, so we know exactly what we must
get. With turbulence models, we may get similar results to
someone else's in one or two test cases and still have a bug
in the code... We shall keep testing these codes and I'll
appreciate any reports on problems experienced by other users.

The parameters used in the input files for the test cases listed
above are those I suggest for general applications. They have not 
been optimized in any sense and are rather conservative; the user
may achieve a substantial improvement of convergence rates by 
playing with the under-relaxation factors, but note that these
are problem-dependent. In general, if the grid is orthogonal,
one can use higher values of under-relaxation parameters for 
U, V, K, etc. (but use always 0.1 to 0.2 for pressure). Also,
central differencing will not work in general for k, eps
equations due to strong peaks; here one would need a local
blending with UDS to avoid oscillations. This is a well-known
technique which I have been too lazy to implement (only the
global blending with a fixed blending factor is implemented).
See papers from aero-experts using TVD, MUSCLE, ENO and other
bounded schemes; also a number of Leonard's papers on this topic
(e.g. in Advances in Numerical Heat Transfer, pp. 1-57,
W.J. Minkowycz and E.M. Sparrow, Eds., Taylor and Francis,
New York, 1997).

Questions specific to the implementation of the models may
be directed either to me or to Martin Schmid; Email
addresses are listed at the beginning of the code.