Introduction to Fluent
What CFD packages do
Aim is to solve, numerically, the equations of
motion (continuity and motion) for a fluid for a
given flow geometry:
Plus transport equations for any additional models
such as heat transfer and/or turbulence models
What CFD packages provide
Offer a CAD/GUI based facility for generating meshes
or grids for a flow geometry
Provide a GUI based facility to set up flow problems
using the grid
Can enable established models without hard coding
Provide reasonably robust solvers to get converged
numerical solutions
Provide post processing tools to analyse results
Provide a facility to implement user defined models
The UQ licence is for Fluent
Why are you doing CFD?
If you have an accurate CFD simulation then
– You will gain a detailed understanding of the flow
physics. This should enable you to understand
why particular things happen, which may be
difficult or impossible to verify experimentally
– You can use the simulations to quickly assess
geometric and flow modifications for design or
R&D purposes without resorting to prototyping
– This is CFD from an industrial perspective as it
should be
Why are you doing CFD?
If you don’t have an accurate simulation then
– You are doing CFD research
– You need to think about experimental verification
– You need to get a competent grasp of the physical
assumptions behind the models so that you find
out what is going wrong and propose
modifications
– This is a PhD project
– This is also as it should be
In Practice
Industry may see CFD as a quick fix
Some dope who doesn’t really
understand the physics of flows will
cook up an apparent solution that looks
pretty but results in a total stuff up
when implemented. This is easy to do
The aim of this course
To teach you not to be that dope
(hopefully)
A warning
CFD is a very young branch of engineering
Even though the modeling capabilities of CFD
packages appear to be immense, the modeling still
grossly simplifies the behavior of turbulent flows
Most (All??) turbulence models are empirical and it
may come as a surprise but are based on some very
smelly assumptions (particularly the k-e model!!!!)
Advanced turbulence modeling (RSM or LES) is still
under development
CFD is only now being applied to multiphase flows
Questionable assumptions of the
Standard k-e turbulence model
Makes the Bousinessq approximation
treats the turbulence only as a turbulent kinetic energy
and makes no attempt to model scales of turbulence
this intrinsically assumes that turbulence is iso-tropic
the model is an empirical model tuned for high Re flows
but it wasn’t until the mid 1990’s that computers came
into wide spread use that could even use this model for
engineering flows !!!!
but for a lot of problems it gives accurate results
for others it gives poor results
Example of what to expect
Cyclone separator for Overflow
mineral processing
Feed
concentration of mineral
slurries on density or particle
size
swirl induced due to
tangential feed
%UF
flow reversal
air core forms due to
negative pressure induced by
swirl m
Increasing particle density
Underflow
CFD modeling of cyclone
separators
Swirl and flow reversal results in highly
strained flows
air core makes modeling at least a two
phase flow problem
if you consider the solids it is a three
phase flow problem
Air core for
a DMC with
a poor
mesh
t-grid mesh
Air core
for a DMC
with a
better
mesh
cooper
mesh
but…..
Even with the better mesh the flow out the
underflow is too large
Experimental underflows are ~15% of feed
flow rate. I am getting ~40% reporting to
underflow
This happens with the RSM turbulence model
which is supposed to be the best model for
swirling flows if you survey the literature
so...
I have have to think fairly hard about
the physics over the next few months
and also think about the funding for
project
Basic Steps in CFD are
Mesh Generation - Gambit
Problem set up - partly Gambit, partly
Fluent
Initialisation and solution - Fluent
Validation
Maybe start again
Gambit
This is the meshing tool used by Fluent
Geometric model of the flow domain is
drawn using the CAD functionality (points,
lines, faces, volumes)
Mesh is generated (line, face and volume
meshes)
and mesh is checked
Nature of the flow domain is specified
Gambit
Boundary condition types are specified
(wall, velocity inlet, outlet)
Solver type is specified (Fluent 4, 5, 6)
Mesh is exported to a file *.msh
Mesh and geometry is stored in a
database *.dbs
Fluent
Mesh is read into Fluent
Mesh is checked by Fluent and scaled
Type of fluid is specified in Fluent (eg water air)
If flow is turbulent then specify a turbulence
model
Boundary conditions are chosen
Other models are enabled depending on the
nature of the problem
Fluent
Implement user designed functions
Discretization methods are chosen
Problem is initialized
Solver is started
Once convergence is reached (non-trivial),
the solution is checked for validity
Intermediate cases should be saved along the
way (you’ll learn this the hard way)