COSMOS FloWorks (Fundamentals)

2006
COSMOSFloWorks 2006 Fundamentals i

Solving Engineering Tasks
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-1
Simulating Engineering Tasks with COSMOSFloWorks . . . . . . . . . . . . . . . . . . . . . . 1-4
Solving Engineering Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-6
Frequent Errors and Improper Actions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-11
Advanced Knowledge
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-1
Mesh Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-1
Type of Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-1
Mesh Constructing Stages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3
Basic Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4
Control Planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5
Resolving Small Features by Using the Control Planes . . . . . . . . . . . . . . . . . . . . 2-5
Contracting the Basic Mesh. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-7
Resolving Small Solid Features. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-9
Curvature Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-10
Tolerance Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-12
Contents
ii COSMOSFloWorks 2006 Fundamentals
Narrow Channel Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-12
Local Mesh Settings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-15
Recommendations for Creating the Computational Mesh. . . . . . . . . . . . . . . . . 2-16
Mesh-associated Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-16
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-16
Visualizing the Basic Mesh Before Constructing the Initial Mesh . . . . . . . . . . 2-17
Widened Capabilities of Loading the Results . . . . . . . . . . . . . . . . . . . . . . . . . . 2-17
Viewing the Initial Computational Mesh Saved in the .cpt Files . . . . . . . . . . . 2-18
Viewing the Computational Mesh Cells with the Mesh Option . . . . . . . . . . . . 2-18
Visualizing the Real Computational Geometry . . . . . . . . . . . . . . . . . . . . . . . . . 2-19
Switching-off the Calculation Results Interpolation and Extrapolation . . . . . . 2-21
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-22
Calculation Control Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-23
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-23
Finishing the Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-23
Refining the Computational Mesh During Calculation . . . . . . . . . . . . . . . . . . . 2-25
Flow Freezing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-27
What is Flow Freezing? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-27
How It Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-28
Flow Freezing in a Permanent Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-28
Flow Freezing in a Periodic Mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-30
Meshing Additional Insight
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-1
Initial Mesh Generation Stages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2
Refinements at Interface Between Substances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-7
Local Mesh Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3-11
Irregular Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-12
Postamble . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-13
Glossary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-14
Validation Examples
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-1
Flow through a Cone Valve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-3
Laminar Flows Between Two Parallel Plates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-7
Laminar and Turbulent Flows in Pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-15
COSMOSFloWorks 2006 Fundamentals iii
Flows Over Smooth and Rough Flat Plates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-21
Flow in a 90-degree Bend Square Duct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-25
Flows in 2D Channels with Bilateral and Unilateral Sudden Expansions . . . . . . . . 4-29
Flow over a Circular Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-33
Supersonic Flow in a 2D Convergent-Divergent Channel . . . . . . . . . . . . . . . . . . . . 4-37
Supersonic Flow over a Segmental Conic Body . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-41
Flow over a Heated Plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-47
Convection and Radiation in an Annular Tube . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-51
Heat Transfer from a Pin-fin Heat Sink . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-55
Unsteady Heat Conduction in a Solid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-59
Tube with Hot Laminar Flow and Outer Heat Transfer . . . . . . . . . . . . . . . . . . . . . . 4-63
Flow over a Heated Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-67
Natural Convection in a Square Cavity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-71
Particles Trajectories in Uniform Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-75
Porous Screen in a Non-uniform Stream . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-79
Lid-driven Flows in Triangular and Trapezoidal Cavities . . . . . . . . . . . . . . . . . . . . 4-85
Flow in a Cylindrical Vessel with a Rotating Cover . . . . . . . . . . . . . . . . . . . . . . . . 4-91
Flow in an Impeller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-95
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-101
Technical Reference
Physical Capabilities of COSMOSFloWorks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-1
Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-2
The Navier-Stokes Equations for Laminar and Turbulent Fluid Flows . . . . . . . . 5-2
Laminar/turbulent Boundary Layer Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-5
Constitutive Laws and Thermophysical Properties . . . . . . . . . . . . . . . . . . . . . . . 5-5
Compressible Liquids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-7
Non-Newtonian Liquids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-7
Equilibrium volume condensation of water from steam. . . . . . . . . . . . . . . . . . . . 5-9
Buoyancy Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-9
Conjugate Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-9
Radiation Heat Transfer Between Solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-10
Rotating Reference Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-10
Mass Transfer in Fluid Mixtures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-11
iv COSMOSFloWorks 2006 Fundamentals
Flows in Porous Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5-11
Two-phase (fluid + particles) Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-12
Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-14
Internal Flow Boundary Conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-14
External Flow Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-15
Wall Boundary Conditions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-15
Internal Fluid Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-16
Numerical Solution Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-16
Computational Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-16
Spatial Approximations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-18
Spatial Approximations at the Solid/fluid Interface . . . . . . . . . . . . . . . . . . . . . . 5-18
Temporal Approximations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-19
Form of the Numerical Algorithm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-20
Methods to Resolve Linear Algebraic Systems . . . . . . . . . . . . . . . . . . . . . . . . . 5-21
Iterative Methods for Nonsymmetrical Problems . . . . . . . . . . . . . . . . . . . . . . . . 5-21
Iterative Methods for Symmetric Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-21
Multigrid Method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-21
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-22
COSMOSFloWorks 2006 Fundamentals 1-1
1
Solving Engineering Tasks
Introduction
Engineering problems are problems connected with designing various objects or systems.
There are three general approaches to solving engineering problems:
an experimental approach: a hardware rig or prototype, i.e., the full-scale object
and/or its model, is manufactured and the experiments needed for designing the
object are conducted with this hardware;
a computational approach: the computations needed for designing the object are
performed and their results are directly used for designing the object, without
conducting any experiments;
a computational-experimental approach combines computations and
experiments (with the manufactured full-scale object and/or its model) needed for
designing the object, their sequence and contents depending on the solved problem,
e.g. iterative procedures may be run.
Each of the first two approaches has advantages and disadvantages.
The purely experimental approach, being properly conducted, does not require additional
validations of the obtained results, but it is very expensive, even if it is realized on the
objects models, since testing facilities and hardware are required anyway. Moreover, if
the objects models are tested, the obtained results must be scaled to the full-scale object,
so computations may be required for that.
Chapter Introduction
1-2
The purely computational approach, being properly performed, is substantially less
expensive than the experimental one, both in finances and in time, but it requires
assurance in the obtained computational results adequacy. Naturally, such assurance must
be based on numerous verifications and validations of the used computational codes, both
from mathematical and physical viewpoints, i.e., both on the mathematical accuracy of the
obtained results (the results adequacy to the used mathematical model) and on the
adequacy of the used mathematical model to the governing physical processes, that is
validated by comparing the computations with the available experimental data.
The third approach, if it reasonably combines experiments and computations, joins the
advantages of both of the first two above-mentioned approaches and avoids their
disadvantages. Complex engineering problems are solved mainly in this way. A
computational code validated on available experimental data allows of quickly selecting
the objects optimal design and/or its optimal operating mode. Then necessary
experiments are conducted to verify the selection.
When selecting from the world market a computational code that is most suitable for
solving your problems, it is necessary to take into account the following suggestions. Any
computational code is based, firstly, on a mathematical model of the governing physical
processes, expressed, as a rule, in the form of a set of differential and/or integral equations
derived from physical laws, and include, if necessary, semi-empirical and empirical
constants and/or relationships. Secondly, a method of solving these equations is required.
Since, as a rule, the mathematical models equations cannot be solved analytically, they
are solved in a discrete form on a computational mesh, so the mathematical problems
solution is obtained with a certain degree of accuracy. Naturally, the mathematical
problems solution accuracy depends both on the method of discretising the differential
and/or integral equations and on the method of solving the obtained discrete equations.
Once these methods have been selected, the mathematical problems solution accuracy
depends on resolving with the computational mesh the solutions regions of non-linear
behavior, and, to provide good accuracy, the mesh must be rather fine in these regions.
Moreover, a usual way of estimating the mathematical problems solution accuracy
consists of obtaining the problems solutions on several meshes, both coarser and finer.
So, if beginning from some mesh from this set, the difference in the interesting physical
parameters between the solutions obtained on the finer and coarser meshes becomes
negligible from the engineering problems viewpoint, i.e., the solution flattens, then the
mathematical problems solution accuracy required for solving this engineering problem
is considered to be attained, since the so-called solution mesh convergence is attained.
Naturally, the mathematical problems solution can differ from the experimental values
(i.e., from the physical problems solution, if it is known), and this difference depends,
firstly, from the mathematical models adequacy to the simulated physical processes, and,
secondly, on the errors with which these experimental values have been measured and
which, as a rule, are known and decrease when increasing the number of tests in which
they are measured.
Correspondingly, the computational codes put on the market differ from one another in
not only their cost, but also in accuracy of mathematically simulating the physical
problems, as well as in the procedure of specifying the initial data and the users time
needed for this specification, in the procedure of solving a problem and the computer
memory and CPU time needed for obtaining a solution of the required accuracy, and last
in the procedures of visualizing and processing the obtained results and the user time
needed for that.
Naturally, a highly accurate solution requires a fine computational mesh, and
consequently rather substantial computer memory and CPU time, as well as, in some
cases, increased user time and efforts for specifying the initial data for the calculation. As
a result, if the time needed to solve an engineering problem with a computational code
exceeds some threshold time, then either the engineering problem becomes non-actual,
e.g. because your competitors have out-distanced you by this time, or alternative
approaches, which may not be so accurate, but are certainly faster, are used instead in
order to solve this problem for the given time.
Before getting acquainted with the recommended procedure of obtaining a reliable and
rather accurate solution of an engineering problem with COSMOSFloWorks, it is
expedient to consider COSMOSFloWorks features governing the below-described
strategy of solving engineering problems with COSMOSFloWorks.
Since COSMOSFloWorks is based on solving time-dependent Navier-Stokes equations,
steady-state problems are solved through a steady-state approach. To more quickly obtain
the steady-state solution, a method of local (over the computational domain) time steps is
employed. A multigrid method is used for accelerating the solution convergence and
suppressing parasitic oscillations. The computational domain is designed as a
parallelepiped enveloping the model with planes orthogonal to the SolidWorks models
Cartesian Global coordinate system. The computational mesh is designed by dividing the
computational domain into cubic cells whose sides are orthogonal to the Global
coordinate systems axes. (The cells lying outside the regions filled with a fluid and
outside solids with heat conduction inside do not participate in the subsequent
calculations). Procedures of refining (splitting) the computational mesh are used for better
resolving the models features, such as high-curvature surfaces in contact with a fluid, thin
walls surrounded by a fluid, narrow flow passages (gaps), and the specified insulators
boundaries. During the subsequent calculations during the solving of the problem the
computational mesh can be additionally refined (if that is specified) for better resolving
the high-gradient flow and solid regions revealed in these calculations (Solution-Adaptive
Meshing).
Chapter Simulating Engineering Tasks with COSMOSFloWorks
1-4
Since steady-state problems are solved in COSMOSFloWorks through the steady-state
approach, it is necessary to properly select the moment of finishing the calculation: if the
calculation is finished too early, then the obtained solution can depend on the specified
initial conditions, i.e., the steady state solution has not been attained yet, therefore the
problem has not been solved; whereas if the calculation is finished too late, then some
time has been wasted uselessly. To optimize the moment of finishing the calculation, as
well as to more accurately determine the physical parameter you are interested in and
which oscillates in iterations (e.g. a force acting on a model surface, or a models
hydraulic resistance), you can specify the physical parameters you are interested in as the
calculations goals.
How to correctly and adequately (from the physical viewpoint) simulate an engineering
problem with SolidWorks+COSMOSFloWorks, i.e. to state the corresponding model
problem, as well as how to properly and reliably solve this model problem with
COSMOSFloWorks, is described in the Simulating Engineering Tasks with
COSMOSFloWorks and Solving Engineering Tasks chapters.
1 Simulating Engineering Tasks with COSMOSFloWorks
It is expedient to remember that a fast but inaccurate beginning will be paid by your
efforts and time wasted uselessly not only in specifying the initial data, but, even more
substantially, in the subsequent calculations, until it finally becomes evident. Therefore,
we strongly recommend that you carefully read this section.
1.1 Selecting the Tasks Geometrical and Physical Features
Before you begin to create a SolidWorks model and a COSMOSFloWorks project, it is
necessary to select the engineering problems geometrical and physical features that most
substantially influence this problems solution. First of all, which is necessary and
important for estimating the possibility of solving this problem with COSMOSFloWorks.
For example,
if the problem contains movable parts, then it is necessary to estimate the
importance of taking their motions into account when solving the problem, and, if
these motions are important, then to estimate a possibility of solving this problem
with a quasi-stationary approach, since model parts motions during a calculation
are not handled in COSMOSFloWorks (however, you can specify a translational
and/or rotational motion of the specific wall),
if the problem includes several fluids, or fluid and solid, then it is necessary to
estimate the importance of chemical reactions between them for the problems
solution, and, if the reactions are important, i.e., the reactions rates are rather high
and the reacting fluids are intensely mixed with one another under the problems
conditions, then to estimate a possibility of introducing the reaction products as an
additional fluid when solving this problem, since chemical reactions are not
considered in COSMOSFloWorks,
if the problem includes fluids of different types (for example, a gas and a liquid),
and there is an interface between them or these fluids are mixing, then it is
necessary to estimate the importance of taking it into account, since
COSMOSFloWorks can not consider an interface (free surface in the case of our
example) between and mixing of fluids of different types.
We can present other examples of an clear impossibility of solving some engineering
problems with COSMOSFloWorks, as well as of simplifying the engineering problems for
solving them with COSMOSFloWorks, but it is impossible to envision and describe all the
possible situations in the present document, so you have to make decisions on each case
for yourself.
1.2 Creating the Model and the COSMOSFloWorks Project
If the SolidWorks model has already been created when designing the object, i.e. it is fully
adequate to the object, then, to solve the engineering problem with COSMOSFloWorks, it
may be required:
to simplify the model by removing those parts which do not influence the problems
solution, but consume computer resources, i.e. memory and CPU time (e.g., instead
of specifying a corrugated model surface, resulting in an extremely large number of
mesh cells being required to resolve it with the computational mesh, it is expedient
to specify this model surface as smooth with equivalent wall roughness; or, e.g., if a
model has thin walls, both surfaces of which are in contact with the fluid, then, if
these walls thickness is not substantial for the problems solution, it is expedient to
increase this thickness as much as possible, which will decrease the number of mesh
cells needed to resolve these walls),
to add auxiliary parts to the model, e.g. inlet and outlet tubes needed for solving the
engineering problem with COSMOSFloWorks.
Both these actions, being executed properly, can be very pivotal in obtaining a reliable and
accurate solution. Naturally, adding the auxiliary parts to a model will inevitably cause an
increase of the computational meshs cells and, consequently, the required computer
memory and CPU time, therefore these parts dimensions must be adequate to the stated
problem.
If a model has not been created yet, it is expedient to take all the above-mentioned factors
into account when creating it.
If all effects of these actions are not fully clear, it may be expedient to vary the models
parts and/or their dimensions in a series of calculations in order to determine their effects
on the obtained solution.
Chapter Solving Engineering Tasks
1-6
Then, in accordance with the engineering problems physical features revealed and
adapted to COSMOSFloWorks capabilities, the COSMOSFloWorks projects basic part is
specified, i.e., the problems type (internal or external), the fluids and solids, the physical
features taken into account (e.g. compressible gas or incompressible liquid,
time-dependent analysis, or steady-state analysis by default, gravitational effects, etc.), the
calculation domain boundaries, the boundary and initial conditions, and, if necessary, the
fluid subdomains, the rotating regions, the volume and/or surface heat sources and/or the
fans and/or other conditions and features.
The specified boundary conditions, as well as the heat sources and the fans, and other
conditions and features must correspond to the problems physical statement and must not
conflict with one another.
Eventually, you specify the physical parameters you are interested in as the
COSMOSFloWorks projects goals. They can be local or integral, defined within the whole
computational domain or on some model surfaces, or in some volumes (the local parameters
are determined over a region in the form of their minimum, or maximum, or average values).
This will allow you to substantially increase reliability and accuracy of these physical
parameters determination, since their behavior during the calculation is saved and can be
analyzed after finishing the calculation, whereas all other parameters are stored for a
subsequent analysis at the last iteration and, additionally, at the specified iterations, whose
number is restricted by the computer memory.
2 Solving Engineering Tasks
As soon as the COSMOSFloWorks projects basic part, which will most probably not be
changed in the subsequent calculations, has been specified, the next step consists in
selecting the strategy of solving the engineering problem with COSMOSFloWorks, i.e.,
obtaining the problems reliable and accurate solution.
2.1 Strategy of Solving the Engineering Tasks
As it has been mentioned in Introduction, from performing a set of calculations on a set of
coarser and finer computational meshes we can estimate the mathematical problems
solution accuracy: as soon as the calculation on a finer mesh does not yield a noticeably
different (from the engineering problems viewpoint) solution, i.e. the solution flattens
with respect to the mesh cells number, we can conclude that the mathematical problems
solution has achieved mesh convergence, i.e., the required mathematical solution accuracy
is attained. Naturally, the solution-vs.-mesh threshold change, such that the smaller
solution-vs.-mesh changes are considered as negligible, must be determined. Since this
thresholds determination without its connection with a physical parameter is impossible,
it is natural to connect it with the physical parameters in which you are interested in the
engineering problem being solved; more exactly with these physical parameters
admissible determination errors. Moreover, since steady-state problems are solved with
COSMOSFloWorks through the steady-state approach, monitoring a behavior of these
physical parameters (specified as the calculations goals) during the calculation (i.e., in
iterations) will allow you to, firstly, if these parameters oscillate during the iterations, to
determine these parameters values and determination errors more accurately by averaging
them over a number of iterations and determining their deviation from this average value,
and, secondly, intervene in the process of making a decision on finishing the calculation,
i.e., manually finish the calculation if you see that either the calculation is unacceptable
for you by some reasons, or, vice versa, the solution has already converged, i.e., obtained,
and therefore it is not necessity to continue the calculation.
Therefore, the strategy of solving an engineering problem with COSMOSFloWorks
consists, first of all, in performing several calculations on the same basic project (i.e., with
the same model, the same computational domain, the same boundary and initial
conditions) with varying computational meshes. Since the computational mesh is
automatically designed in COSMOSFloWorks, it is varied by varying the project
parameters governing its design (the initial computational mesh, on which the calculation
starts, and its refinement during the calculation): Result Resolution Level, Minimum Gap
Size, Minimum Wall Thickness.
An additional item of this strategy of solving an engineering problem with
COSMOSFloWorks consists in varying the models auxiliary elements added to the model
as needed for solving the problem with COSMOSFloWorks (e.g. inlet and outlet tubes for
internal problems with inlet and outlet openings), whose dimensions are questionable
from their necessity and sufficiency viewpoint. The engineering problems (the projects)
physical parameters whose values are not known exactly and which, in your opinion, can
influence the problems solution must be varied also. When performing these calculations,
there is no necessity to investigate the solution-vs.-mesh convergence again, since it has
been done before, so it is enough only to perform these calculations with the projects
mesh settings recognized in the solution-vs.-mesh convergence investigation that provided
the solutions satisfactory accuracy. Also this applies to parametric engineering
calculations with changing the models parts and/or flow parameters, but you must not
forget the potential necessity for checking the solution-vs.-mesh convergence, which must
be checked in doubtful cases.
In spite of the apparent simplicity of the proposed strategy, this strategys full realization
is usually hampered by substantial difficulties consisting, first of all, in a substantial rise
of the required computer memory and CPU time when substantially increasing the
computational mesh cells number. Since both the computer memory and the time for
which the engineering problem must be solved are usually restricted, the concrete
realization of this strategy eventually governs the problems solution accuracy, whether it
will be satisfactory or not. Perhaps, a further simplification of the model and/or reducing
the computational domain will be required.
Some concretizations of this strategy are presented in the next sections of this document.
2.2 Settings for Resolving the Models Geometrical Features and for
1-8
Obtaining the Required Solution Accuracy
The computational mesh variation described in Section 2.1 is the key item of the proposed
strategy of solving engineering problems with COSMOSFloWorks.
The result resolution level specified in the Wizard governs the basic meshs cell number,
the criteria for refining (splitting) the basic mesh for resolving the models geometry, i.e.,
creating the initial mesh, as well as the criteria for refining (splitting) the initial mesh
during solving the problem at result resolution levels of 68. The Wizards Result
Resolution is virtually split into Level of initial mesh on the Automatic Settings tab of the
Initial Mesh dialog box, governing the initial mesh only, and Results resolution level in
the Calculation Control Options dialog box, governing finishing the calculation and
refining computational mesh during calculation. As for the Geometry Resolution options,
influencing the initial mesh also, they can be changed on the Automatic Settings tab of the
the Initial Mesh dialog box, and/or their effects can be corrected on the other tabs of the
Initial Mesh dialog box or in the Local Initial Mesh dialog box.
Before creating the initial mesh, COSMOSFloWorks automatically determines the
minimum gap size and the minimum wall thickness for the walls whose both sides are in
contact with a fluid. That is required for resolving the models geometrical features with
COSMOSFloWorks computational mesh. So, when creating the initial mesh, it is taken
into account that the number of the meshs cells along the normal to the models surface
must not be less than a certain number if the distance along this normal from this surface
to the opposite wall is not less than the minimum gap size. Depending on the meshs cell
arrangement, the models flow passages not resolved with the computational mesh either
are automatically replaced with a wall, or increased up to the meshs cell size. If both sides
of a model wall are in contact with a fluid, then, if this walls thickness is smaller than the
projects minimum gap size and the projects minimum wall thickness, this wall can be not
resolved with the computational mesh, so it will have holes in the calculation. In the
automatic mode these mesh parameters are determined from dimensions of the surfaces on
which boundary conditions have been specified, e.g. the models inlet and outlet openings
in an internal analysis, as well as those surfaces and volumes on or in which heat sources,
volume conditions, surface and/or volume goals and some of the other conditions and
features have been specified. Before the calculation, you can see the minimum gap size
and the minimum wall thickness determined in such a manner. If these values cannot
provide an adequate resolution of the models geometry, you can specify them manually.
At that, it is necessary to take into account that the number of the computational meshs
cells generated to resolve the models geometrical features depends on the specified result
resolution level.
Evidently, when creating a COSMOSFloWorks project it is necessary to make sure that
both the minimum gap size and the minimum wall thickness are adequate to the models
geometry. However, if the models geometry is intricate (e.g. non circular flow passages,
sharp edges protruding into the stream, etc.), it can be difficult to determine these
parameters unambiguously. In this case it is expedient to perform several calculations by
varying these parameters within a reasonable range in order to reveal their influence on
the problems solution. In accordance with the strategy of solving engineering problems
these calculations must be performed at different result resolution levels.
The initial mesh created at result resolution levels of 35 is not changed during the
solving of a problem, i.e. is not adapted to the solution being obtained. Result resolution
levels of 57 yield the same initial mesh, but at result resolution levels of 6 and 7 the
mesh is refined during the calculations in the regions of increased physical parameters
gradients. At level 8, a finer initial mesh is generated and refinements during calculation
occur.
It makes sense to perform calculations at the result resolution level of 3 if both the models
geometry and the flow field are smooth. For more complex problems we recommend first
of all to perform the calculation at the result resolution level of 4 or, better, 5 (naturally,
with specifying the required minimum gap size and minimum wall thickness). After that,
if the calculation at the result resolution level of 5 has been performed, we recommend, in
order to ascertain the mesh convergence, to perform the calculation at the result resolution
level of 7 and, if the computer resources allow you to do this, at the result resolution level
of 8.
2.3 Monitoring the Calculation
Monitoring the calculation, i.e., at least, monitoring behavior of the physical parameters
specified by you as the projects goals (you can inspect also physical parameters fields at
the specified sections) is useful for the following reasons:
you can intervene in the process of making the decision on finishing the calculation,
i.e., manually finish the calculation before this is done automatically, if you see that
either the calculation is unacceptable for you for some reasons, e.g.
COSMOSFloWorks has generated warnings from which it follows that continuing
the calculation is senseless, or, vice versa, when solving a steady-state problem (that
concerns some time-dependent problems also), the solution has already converged,
i.e., is obtained, and there is no reason to continue the calculation;
if a steady-state problem is solved, and the physical parameters specified by you as
the projects goals oscillate during the iterations, then inspecting these parameters
behavior during the calculation will allow you to determine their values and
determination errors more accurately by averaging their values over the iterations
and determining their deviations from these average values;
if the physical parameters you are interested in do not substantially change during
the calculation, you can see their intermediate (preliminary) values, which will be
refined in the subsequent iterations;
1-10
if you solve a time-dependent problem, you can immediately see the calculation
results before finishing the calculation.
The first above-mentioned reason is especially useful since it allows you to substantially
reduce the CPU time in some cases. For example, if you do not specify the high Mach
number gas flow in the projects settings, whereas the flow occurs supersonic, or if
COSMOSFloWorks warns about a vortex at the models outlet, that substantially reduces
the calculations accuracy, making it is necessary to lengthen the models outlet tube. If
you solve a steady-state problem at the result resolution level of 7 or 8 and see that the
computational meshs refinements performed during the calculation do not increase the
mesh cells number and, therefore, do not noticeably refine the problems solution (the
projects goals values), you can finish the calculation if at least 12 travels of iterations
have been performed.
2.4 Viewing and Analyzing the Obtained Solution
When viewing and analyzing the obtained solution after finishing the calculation, it is
recommended first of all, in accordance with the above-mentioned suggestions, to plot the
evolutions of the projects goals from the calculation, if you did not monitor them during
the calculation. If a steady-state problem is solved, and you have specified the physical
parameter you are interested in as the projects goal, then, if this parameter has oscillated
during the calculation, you can more accurately determine its value by averaging it over
the last iterations interval in which its steady-state oscillation is seen. At that, this goals
variance, i.e., its deviation from its average value, characterizing this goals determination
error in the obtained solution is determined also.
It is also useful to check to see that there are no vortices at the models outlet, as well as to
see the flow pattern in the model and, if heat transfer in solids has been calculated, the
temperature distribution over the models solids. Naturally, first of all it is expedient to see
the obtained field of the physical parameter you are interesting in, not only in the region
you are interesting in, but in a wider region also, in order to check this field for appearance
of physically incorrect results.
It is also expedient to see the obtained fields of the physical parameters connected with the
physical parameter you are interested in. For example, if you are interested in the total
pressure loss, it is expedient to see the velocity field, whereas if you are interested in the
solids temperature, it is expedient to see the fluid-to-solid heat flux field.
2.5 Estimating the Obtained Solutions Reliability and Adequacy
In accordance with the general approach to estimating a reliability and an accuracy of an
engineering problems solution obtained with a computational code, this estimation
consists of the following two estimations: an estimation of the accuracy of solving the
mathematical problem corresponding to the physical processs mathematical model, and
an estimation of the accuracy of simulating the physical process with this mathematical
model.
The mathematical problems solution accuracy is determined with mathematical methods,
independently of the mathematical models adequacy to the physical process under
consideration. In our case, this accuracys estimation is based on analyzing the mesh
convergence of the problems solutions obtained on different computational meshes (See
Section 2.2). Then, since steady-state problems are solved with COSMOSFloWorks
through a steady-state approach by employing local time steps, it is useful to additionally
verify the obtained solutions accuracy by solving the similar time-dependent problem not
employing local time steps.
As soon as the mathematical problems solution of a satisfactory accuracy has been
obtained, the next step consists of estimating the accuracy of simulating the physical
process under consideration with the mathematical model employed in the computational
code. To do this, the obtained solution is compared with the available experimental data
(taking into account their errors consisting of the measurement errors and the experiment
errors connected with parasitic influences). Naturally, since experimental data are usually
restricted, for this validation it is desirable to select experimental data which are closest to
the engineering problem being solved. To validate the computational code on the available
experimental data, you have to solve the corresponding test problem in addition to the
engineering problem being solved (preferably before solving the engineering problem,
following the above-mentioned strategy), but this so substantially increases the reliability
of estimating the engineering problems obtained solution that the required additional time
and efforts will fully pay for themselves later on, in particular when solving similar
engineering problems.
If after solving the test problem you see that accuracy of its solution obtained with
COSMOSFloWorks is not satisfactory from your viewpoint, check to see that you have
properly specified the COSMOSFloWorks project, that all the engineering problems
substantial features have been taken into account, and, finally, that COSMOSFloWorks
restrictions do not impede solving this engineering problem.
3 Frequent Errors and Improper Actions
Let us consider errors and improper actions frequently done when solving engineering
problems with COSMOSFloWorks.
When Specifying Initial Data:
not taking into account physical features which are important for the engineering
problem under consideration: e.g. high Mach number gas flow (should be taken
into account if M>3 for steady-state or M>1 for transient tasks or supersonic flow
occurs in about a half of the computational domain or greater), gravitational
effects (must be taken into account if either the fluid velocity is small, the fluid
density is temperature-dependent, and a heat source is considered, or several
fluids having substantially different densities are considered in a gravitational
field), necessity of the time-dependent analysis (e.g. at the moderate Reynolds
numbers, at which unsteady vortices are generated);
Chapter Frequent Errors and Improper Actions
1-12
incorrectly specifying symmetry planes as the computational domains
boundaries (e.g. at the moderate Reynolds numbers, at which unsteady vortices
are generated, so they can intersect the proposed symmetry planes);
if symmetry planes have been specified and you click Reset at the Size tab of the
Computational Domain dialog box, please do not forget to replace Symmetry
by Default at the Boundary Condition tab;
if you have specified symmetry planes and intend to specify mass or volume flow
rates at a models inlet or outlet openings, please do not forget to specify only
their parts falling into the computational domain instead of the total flow rates at
these openings;
if you specify integral boundary or volume conditions (heat transfer rates, heat
generation rates, etc.), please remember that their values specified in the
COSMOSFloWorks dialog boxes correspond to the area or volume's part falling
into the computational domain;
if you specify a flow swirl on a models inlet or outlet openings (in the Fans or
Boundary Conditions dialog boxes), please do not forget to properly specify
their swirl axes and the proper coordinate system for that in the Definition tab;
if you specify a Unidirectional or Orthotropic porous medium, please do not
forget to specify their directions;
please do not forget that the specified boundary conditions must not conflict with
one another: e.g. if you deal with gas flows and the models inlet flow is
subsonic, whereas the flow inside the model becomes supersonic, it is incorrect
to specify flow velocity or volume flow rate as a boundary condition at the
models inlet, since they are fully governed by the models flow passage
geometry and the fluids specific heat ratio;
if you solve a time-dependent problem, and this problem has cyclic-in-time
boundary conditions, and therefore has a steady-state cyclic-in-time solution,
which to obtain it is necessary to calculate the flow in the cycle several times,
specifying the solution obtained from the previous calculation as the initial
condition for the next calculation, then there is no necessity to specify the
boundary conditions for several cycles. Instead it is more convenient to specify
them for a cycle and perform a series of calculations, running each calculation
from the zero moment. To do this it is necessary to select Take previous results
in the Run dialog box for each next calculation and specify the cycle ends time
moment in the Start from the time moment box;
when specifying Surface Goals, Volume Goals, Equation Goals, it is better to
give them names unambiguously identifying these goals without the necessity of
finding them in the SolidWorks graphic area;
if you want to monitor the intermediate calculation results at certain sections of
the model during the calculation, it is better beforehand, i.e. before running the
calculation, to determine these sections positions in the Global coordinate
system, since during the calculation it is more difficult and time-consuming to do
so;
When Monitoring a Calculation:
when monitoring intermediate calculation results during a calculation, please do
not forget the solved problems spatiality (3-D nature, if the problem is not 2-D
plane): to see a full results pattern it is expedient to see the results in 2..3
intersecting planes;
When Viewing the Obtained Solution after Finishing a Calculation:
please take into account that all settings made in the View Settings dialog box
concern all Cut Plots, 3D Plots, Surface Plots, Flow Trajectories, Isosurfaces,
which are active in the active SolidWorks graphic area, therefore:
your attempts to open the Flow Trajectories dialog box will be futile if a
parameter defined on wall surfaces only has been selected at the Contours tab
of the View Settings dialog box and the Use from contours option has been
selected at the Flow Trajectories tab of this dialog box;
to simultaneously view different physical parameters in sections and/or
surfaces, and/or along trajectories (by coloring them), it is necessary to split
the SolidWorks graphics area into 2 or 4 graphical areas and view different
physical parameters in different graphical areas through their individual Cut
Plots, 3D Plots, Surface Plots, Flow Trajectories, Isosurfaces defined in
these areas;
if you intend to see integral physical parameters (e.g. area, mass or volume flow
rates, heat generation rates, forces, etc.) with the Surface Parameters dialog
box, please remember that
their shown values are determined over the surfaces parts falling into the
computational domain;
their determination errors include errors of representing these surfaces in
SolidWorks and COSMOSFloWorks, the latter depends on the computational
mesh;
if you want to see a computational mesh in Cut Plots and/or Surface Plots,
please do not forget to select Display mesh in Tools, Options, Third Party
Options, otherwise the Mesh box in the Cut Plots and Surface Plots dialog
boxes will be absent, so you can not view a computational mesh.
Chapter Frequent Errors and Improper Actions
1-14
2
Advanced Knowledge
Introduction
The present document supplies you with our experience of employing the advanced
COSMOSFloWorks capabilities, divided into the following topics:
Manual adjustment of the initial computational mesh settings
Mesh-associated tools (viewing the mesh before and after calculation and advanced
post-processing tools)
Calculation control options (refining the computational mesh during calculation,
conditions of finishing the calculation)
Flow freezing
1 Mesh Introduction
This chapter provides the fundamentals of the COSMOSFloWorks computational mesh,
describes the mesh generation procedure, and explains the use of parameters governing
both automatically and manually controlled meshes.
First, let us introduce a set of definitions.
1.1 Type of Cells
COSMOSFloWorks uses only rectangular cells of the following four types:
Fluid cells cells are entirely in the fluid.
Solid cells cells are entirely in the solid.
Partial cells cells are partly in solid and partly in fluid. For partial cells the
following information is known: coordinates of intersections of cells edges with
the solid body, solid face area within a cell, and normal to the solid face.
Irregular cells partial cells with undefined normal to the solid face.
Chapter Mesh Introduction
2-2
As an illustration lets look at the original model (Fig.1.1) and the generated
computational mesh.
When analyzing the mesh you should pay special attention to the
irregular cells. These cells may appear after meshing of thin walls,
narrow channels or sharp corners. If irregular cells appear it is
recommended that you examine the location of the cells (refer to the
Users Guides Mesh Visualization chapter) to make a decision
whether further resolving of these cells is needed, or if leaving them
unresolved is sufficient. See also Irregular Cells on page 12.
Fig.1.1 The original SolidWorks model.
Fig.1.2 The mesh was expressly made worse to show the different cell types.
Zero level cell (basic cell)
Solid cell
Partial cell
First level cell
Fluid cell
Irregular cell
1.2 Mesh Constructing Stages
Refining (splitting) a cell is split into eight cells by three orthogonal faces passing
through the center of the cell which divide the cells edges in half. The non-split cells are
called basic cells or zero level cells and they compose the basic mesh. Cells derived from
the initial splitting of the basic cells are called first level cells; the next splitting produces
second level cells, and so on. The maximum level of splitting is seven. A seventh level cell
is in 8
7
times smaller (in volume) than the basic cell.
Regarding refining and merging the following rule is applied: Two neighboring cells can
only be cells whose level is the same or differs by one, e.g. a fifth level cell can neighbor
only with fourth, fifth, or sixth level cells.
The mesh is constructed in the following stages:
Constructing the basic mesh, taking into account the Control Planes and the relative
mesh step.
Resolving interface between substances: refining the basic mesh at a solid/fluid and
solid/solid boundaries to resolve the relatively small solid features and solid/solid
interface; tolerance and curvature refinement refining the mesh at a solid/fluid,
solid/porous and a fluid/porous boundary to resolve the interface curvature (e.g.
small-radius circle surfaces, etc.).
Narrow channels refinement - refining the mesh in narrow channels.
Irregular cells refinement refining the irregular cells until the derived daughter cells
become resolved or their level of splitting exceeds the maximum value.
Mesh conservation a set of controlling procedures, e.g. checking the difference in
area of cell facets common for the adjacent cells of different levels.
After passing each stage the number of cells is increased.
In COSMOSFloWorks you can control the following parameters and options that govern
the computational mesh:
1 Nx number of basic mesh cells (zero level cells) along the X axis of the Global
Coordinate System. 1 Nx 1000
2 Ny number of basic mesh cells (zero level cells) along the Y axis of the Global
Coordinate System. 1 Ny 1000.
3 Nz number of basic mesh cells (zero level cells) along the Z axis of the Global
Coordinate System. 1 Nz 1000.
During the solution-adaptive meshing cells can be refined and
merged. See Refining the Computational Mesh During
Calculation on page 25.
If you switch on or off heat conduction in solids, or add/move
insulators then you should rebuild the mesh.
2-4
4 Control planes that govern stretching or contracting the basic mesh by means of the
planes location and the relative mesh step specified for each control plane. Any
analysis has at the minimum 6 control planes which are computational domain
boundaries.
5 Small solid features refinement level. Lets denote it as L
b
. 0 L
b
7.
6 Curvature refinement level. Lets denote it as L
cur
. 0 L
cur
7.
7 Curvature refinement criterion. Lets denote it as C
cur
. 0 C
cur
.
8 Tolerance refinement level. Lets denote it as L
tol
. 0 L
tol
7.
9 Tolerance refinement criterion. Lets denote it as C
tol
. 0 C
tol
.
10 Narrow channels refinement: Characteristic number of cells across a narrow channel,
Narrow channels refinement level, The minimum and maximum height of narrow
channels to be refined.
These options will be described in detail later in this chapter.
1.3 Basic Mesh
The basic mesh is a mesh of zero level cells. If you want to make a 2D calculation you
need to specify one cell along the eliminated direction. This is done automatically when
you specify the 2D plane flow option in the Computational Domain dialog box. By default
COSMOSFloWorks constructs the cells in as nearly a cubic manner as possible.
The number of basic mesh cells could be one or two less than the
user-defined number (Nx, Ny, Nz). There is no limitation on a cells
oblongness, or aspect ratio, but you should carefully check the results
in a particular case having quite oblong, or stretched, cells.
Fig.1.3 Basic mesh examples.
a) 10x12x1 b) 40x36x1
1.4 Control Planes
The Control Planes option is another powerful tool aimed to create an optimal
computational mesh, and the user should certainly become acquainted with this tool if he
interested in optimal meshes resulting in higher accuracy, saving time and computer
resources. Control planes allow you to resolve small features, contracting the basic mesh
locally to resolve a particular region by the denser mesh and stretching the basic mesh to
avoid excessively dense meshes.
1.5 Resolving Small Features by Using the Control Planes
If the level of splitting is not high enough small features may be not resolved properly. In
this case, there are two methods to improve the mesh:
increase the level of splitting. However, this may result in needlessly increasing the
number of cells in other regions, creating a non-optimal mesh, or
set a control plane crossing the relevant small feature. This will allow you to resolve
the feature better without creating an excessively dense mesh elsewhere. It is
especially convenient in cases of thin walls oriented along the Global Coordinate
Systems axes.
Fig.1.4 The original model.
2-6
Fig.1.5 Curvature refinement level is 0; Small solid features refinement level is 0; Total number of
cells 960; No control planes.
Fig.1.6 Curvature refinement level is 3; Small solid features refinement level is 3; Total number
of cells 3865; No control planes.
1.6 Contracting the Basic Mesh
Using control planes you can contract the basic mesh in the regions of interest. To do this,
you need to set control planes surrounding the region and then decrease the control planes
relative mesh step parameters. The Relative Mesh Step defines the distance between the
basic mesh planes with respect to the other control planes mesh steps. For example,
assume there are two control planes CP1 and CP2 and their relative mesh steps are S1=2
and S2=1 respectively. Then the basic mesh cells adjacent to the CP1 will be
approximately two times longer than the basic mesh cells adjacent to the CP2.
It is recommended that you place a control plane slightly submerged
into the solid, and avoid placing it coincident with the solid surface.
cells 1143; 9 control planes set across the each fin.
2-8
Use of control planes is especially recommended for external analyses, where the
computational domain may be substantially larger than the model.
In the Fig.1.10 two custom control planes are set through the center of the body with the
relative mesh step of 1 whereas the default boundary control planes has the relative mesh
step set to 5.
Fig.1.9 Default control planes. Fig.1.10 Two custom control planes.
Fig.1.8 Specifying custom control planes.
1.7 Resolving Small Solid Features
The procedure of resolving small solid features refines only the cells where the solid/fluid
(solid/solid, solid/porous as well as fluid/porous) interface position is not uniquely
determined. There are two different types of such cells: irregular cells and cells where
substance features (small solid feature at the solid/fluid interface or thin layers of different
solids, etc.) are not sufficiently resolved by the mesh.
Such unresolved cells are always split unless they are resolved or their level of splitting
exceeds the maximum value. Here, the maximum level of splitting for irregular cells is
defined as maximum from the Small solid features refinement level, Curvature
refinement level, Tolerance refinement level, Narrow channels refinement level. The
maximum level of splitting for regular cells with unresolved feature is maximum from the
Small solid features refinement level and Narrow channels refinement level.

If you want to resolve irregular cells it is recommended that you
increase the Small solid features refinement level since increasing
the other two levels may increase the number of cells where it is not
necessary.
cells 211, No irregular cells.
2-10
1.8 Curvature Refinement
The curvature refinement level is the maximum level to which the cells will be split during
the curvature refinement procedure refining of the mesh at a solid/fluid boundary and a
fluid/porous boundary to resolve the solid/fluid, fluid/porous interface curvature in
accordance with the specified curvature criterion (C
cur
).
The curvature refinement procedure has the following stages:
1 Each solid surface is triangulated: COSMOSFloWorks gets triangles that make up the
SolidWorks surfaces.
2 For each partial cell, COSMOSFloWorks finds triangles crossing the cell.
3 A local (for each cell) surface curvature is defined as the maximum difference in
normals to the triangles that compose the surface enclosed by the cell.
4 If the difference exceeds the specified C
cur
, and the curvature refinement level is not
reached then the cell is split.
The performance settings do not govern the triangulation
performance.
cells 568, No irregular cells.
The curvature refinement is a powerful tool, whose competent use allows you to obtain
proper and optimal computational mesh. Look at the following examples as illustration for
the curvature refinement.
Fig.1.13 Curvature refinement level is 0;
Total number of cells 64.
Curvature criterion 0.317;
Curvature criterion 0.1;
2-12
1.9 Tolerance Refinement
Tolerance refinement allows you to control how well (with what tolerance) the mesh
polygons resolves the real interface. The tolerance refinement may affect the same cells
those affected by the small solid features refinement and the curvature refinement. It more
effectively resolves the interface's curvature then the small solid feature refinement, and,
in contrast to the curvature refinement, discerns small and large features of equal
curvature, thus avoiding refinements in regions of less importance (see images below).
Any surface is approximated by set of polygons whose vertices are surface's intersection
points with the cells' edges. This approach accurately represents flat faces though
curvature surfaces are approximated with some precision (e.g. as a circle can be
approximated by a polygon). The tolerance refinement criterion controls this precision. A
cell will be split if the distance (h, see below) between the outermost interface's point
within the cell and the polygon approximating this interface is larger than the specified
criterion value.
1.10 Narrow Channel Refinement
The narrow channel refinement is applied to each flow passage within the computational
domain (or a region, in cases where local mesh settings are specified) unless you specify
for COSMOSFloWorks to ignore passages of a specified height. The Narrow Channels
term is conventional and used for the definition of the models flow passages in the
normal-to-solid/fluid-interface direction.
The basic concept of narrow channel refinement is to resolve the models narrow channels
with a sufficient number of cells to produce a higher level of accuracy in the solution. It is
especially important to have narrow channels resolved in analyses of low Reynolds
numbers or analyses with long channels, i.e. in such analyses where the boundary layer
thickness becomes comparable to the size of the partial cells where the layer is developed.
The following narrow channel settings are available in COSMOSFloWorks:
Narrow channels refinement level the maximum level of refining cells in narrow
channels with respect to the basic mesh cell.
Characteristic number of cell across a narrow channel the number of cells
(including partial cells) that COSMOSFloWorks will try to set across the models
flow passages in the normal-to-solid/fluid-interface direction. If possible, the
number of cells across narrow channels will be equal to the specified characteristic
number, otherwise it will be close to the characteristic number. The Characteristic
number of cells across a narrow channel (lets denote it as Nc) and the Narrow
channels refinement level (lets denote it as L) both influence the mesh in narrow
channels in the following manner: the basic mesh in narrow channels will be split to
have Nc number per channel, if the resulting cells satisfy the specified L. In other
words, whatever the specified Nc, a narrow channels cells cannot be smaller in 8
L
(2
L
in each direction of the Global Coordinate System) times that of the basic mesh
cell. This is necessary to avoid undesirable mesh splitting in very fine channels that
may cause the number of cells to increase to an unreasonable value.
The minimum height of narrow channels, The maximum height of narrow
channels the minimum and maximum height bounds outside of which a flow
passage will not be considered as a narrow channel and thus will not be refined by
the narrow channel resolution procedure.
For example, if you specify the minimum and maximum height of narrow channels, the
cells will be split only in those fluid regions where the distance between the flow
passages opposite walls in the normal-to-wall direction lies between the specified
minimum and maximum heights.
The narrow channel refinement operates as follows: the normal to the solid surface known
for each partial cell is extended up to the next solid surface, which will be considered to be
the flow passages opposite wall. If the number of cells per this normal-to-wall direction is
less than the specified Nc, the cells will be split to satisfy the narrow channel settings
described above.
Although it depends on a particular task as to which settings produce an optimal mesh,
here are some general recommendations for narrow channel settings:
1 Set the number of cells across narrow channel to a minimum of 5.
2 Use the minimum and maximum heights of narrow channels to localize only interested
regions.
3 If possible avoid setting a high narrow channels refinement level, since it may cause a
significant increase in the number of cells where it is not necessary.
Fig.1.17 Curvature refinement level is 3; Small solid features refinement level is 3; Narrow channel
refinement is disabled; Total number of cells 6476.
2-14
refinement is on: 5 cells across narrow channels, Narrow channels refinement level is 2; Total number of
cells 8457.
refinement is on: 5 cells across narrow channels, Narrow channels refinement level is 7; Total number of
cells 33293.
1.11 Local Mesh Settings
The local mesh settings option is one more tool to help create an optimal mesh. Use of
local mesh settings is especially beneficial if you are interested in resolving a particular
region within a complex model.
The local mesh settings can be applied to a component, face, edge or vertex. You can
apply local mesh settings to fluid regions and solid bodies. To apply the local mesh
settings to a fluid region you need to specify this region as a solid part or subassembly and
then disable this component in the Component Control dialog box. The local mesh settings
are applied to the cells intersected with the selected component, face, edge, or a cell
enclosing the selected vertex. However, cells adjacent to the local regions cell may be
also affected due to the refinement rule described in the Mesh Constructing Stages
chapter.
Fig.1.20 The local mesh settings used: The rhombic channel is refined into 4
th
level cells, and two narrow
channels are refined to have 10 cells across each channel.
Chapter Mesh-associated Tools
2-16
1.12 Recommendations for Creating the Computational Mesh
1 At the beginning create the mesh using the default (automatic) mesh settings. Start
with the Level of initial mesh of 3. On this stage it is important to recognize the
appropriate values of the minimum gap size and minimum wall thickness, which will
provide you with the suitable mesh. The default values of the minimum gap size and
minimum wall thickness are calculated using information about the overall model
dimensions, the Computational Domain size, and area of surfaces where conditions
(boundary conditions, sources, etc.) and goals are specified.
2 Closely analyze the obtained automatic mesh total numbers of cells, irregular cells,
resolving of the interested regions and narrow channels. If the automatic mesh does not
satisfy you and changing of the minimum gap size and minimum wall thickness
values do not give the desired effect you can proceed with the custom mesh.
3 Start to create your custom mesh with the disabled narrow channel refinement, while
the Small solid features refinement level and the Curvature refinement level are
both set to 0. This will produce only zero level cells (basic mesh only). Use control
planes to optimize the basic mesh.
4 Next, adjust the basic mesh by step-by-step increasing of the Small solid features
refinement level and the Curvature refinement level. Then, enable the narrow
channels refinement.
5 Finally, try to use the local mesh settings.
2 Mesh-associated Tools
2.1 Introduction
Since the mesh settings tool is an indirect way of constructing the computational mesh, to
better visualize the resulting mesh various post-processing tools are offered by
COSMOSFloWorks. In particular, these tools allow to visualize the mesh in detail before
the calculation, saving substantial CPU and user time.
The computational mesh constructed by COSMOSFloWorks or other CFD codes cannot
resolve the model geometry at the mesh cell level exactly. A discrepancy can lead to
prediction errors. To facilitate an analysis of these errors and/or to avoid their appearance,
COSMOSFloWorks offers various options for visualizing the real computational
geometry corresponding to the computational mesh used in the analysis.
Since the numerical solution is obtained inevitably in the discrete form, i.e., in the
computational mesh cells centers, it is interpolated and extrapolated by the post-processor
to present the results in a smooth form, which is typically more convenient to the user. As
a result, some prediction errors can stem from these interpolations and extrapolations. To
facilitate an analysis of such errors and/or to avoid their appearance, COSMOSFloWorks
offers various options for visualizing the physical parameters values calculated at the
computational mesh cells centers. So, when presenting results by coloring the area with a
palette, the results are considered constant within the mesh cells.
2.2 Visualizing the Basic Mesh Before Constructing the Initial Mesh
Using this option the user can inspect the Basic mesh and its Control planes corresponding
to the mesh settings, which can be made manually or retained by default. The plot appears
as soon as these settings have been made or changed, so you immediately see the resulting
mesh. (See Help or Users Guide defining the Basic mesh and its Control planes).
To enable this option, select the Show basic mesh option in the FloWorks, Project menu,
or in the Initial Mesh dialog box. The option is accessible both before and after the
calculation.
This option is useful for shifting the Control planes to desired positions to assure that
certain features of the model geometry are captured by the computational mesh.
2.3 Widened Capabilities of Loading the Results
COSMOSFloWorks allows for viewing not only the calculation results and the current
computational mesh on which they have been obtained, but also the initial (i.e., on which
the calculation begins) computational mesh separately. The latter can be viewed either
before or after the calculation, allowing the user to compare the initial and current (i.e.,
refined during the calculation) computational meshes.
To view the various meshes, the user must open the corresponding file with the Load
results dialog box. The calculation results, including the current computational mesh, are
saved in the .fld files, whereas the initial computational mesh is saved separately in the
.cpt files. Both of the files are saved in the project folder, whose numerical name is formed
by COSMOSFloWorks and must not be changed. The .cpt files and the final (i.e., with the
solution obtained at the last iteration) .fld files have the project folders name, whereas the
solutions obtained during the calculation at the previous iterations (corresponding to
certain physical time moments, if the problem is time-dependent) are saved in the .fld files
with names r_<iteration number>, e.g. the projects initial data are saved in the
r_000000.fld file.
Do not try to load the calculation results obtained in another project
with a different geometry; the effect is unpredictable.
Fig.2.1 The Basic mesh (left) and the Initial mesh (right).
2-18
2.4 Viewing the Initial Computational Mesh Saved in the .cpt Files
To optimize the process of solving an engineering problem and to save time, sometimes it
is useful to view the initial computational mesh before performing the calculation,
particularly to be sure that the models features are well resolved by this mesh. To view the
initial computational mesh after loading the .cpt file, COSMOSFloWorks offers you Cut
Plots, Surface Plots, and the Mesh option (see below), which are also used for viewing
the calculation results.
2.5 Viewing the Computational Mesh Cells with the Mesh Option
To view the computational mesh cells lying fully in the fluid (named fluid cells), lying
fully in the solid (named solid cells), and the partial cells partly lying in both the fluid and
the solid, COSMOSFloWorks offers you the Mesh option.
Different colors can be used to better differentiate between the computational mesh cells
of each of the above-mentioned types. To see the cells in a certain parallelepiped region,
the user must specify the region boundaries coordinates in the SolidWorks models
Global Coordinate System.
Using the Mesh option, you can also save the information on the mesh cells, including the
problems solution obtained in their centers, in ASCII or Excel files.
Visualization of a large amount of computational mesh cells (e.g. all
fluid cells in the whole computational domain) may be impractical,
since it could require substantial time and memory, and ultimately
you might not be able to see all the visualized cells because a large
part of them may be screened from view by a small part of them.
Fig.2.2 The Load Results dialog box.
2.6 Visualizing the Real Computational Geometry
Since the SolidWorks model geometry, especially its high-curvature parts, cannot be
resolved exactly at the mesh cell level by the rectangular (parallelepiped) computational
mesh, the real computational geometry corresponding to the computational mesh used in
the analysis can be viewed after the calculation to avoid or estimate the prediction errors
stemming from this discrepancy. If no solution-adaptive meshing occurs during the
calculation, the real computational geometry can be viewed just after the mesh generation.
This capability is realized by clearing the Use CAD geometry option in Cut Plots, 3D
Plots, Surface Plots, Flow Trajectories, Point Parameters and XY Plots, being especially
clear when colored Contours are used for visualizing a physical parameter values (see
Fig.2.3).
This capability is especially useful for revealing the model surface regions resolved
inadequately by the computational mesh. Let us consider Fig.2.4 with the temperature Cut
Plots as an example. The white detail is a heat insulator, with the orange-colored heat
source being mounted to its top surface, so the heat from the heat source cannot penetrate
through the heat insulator to the metallic case in contact with the heat insulator from
below. Nevertheless, in Fig.2.4 (left) we see the gray region (i.e., with the undetermined
temperature) and the supposedly heated region under the heat source. Switching to the
computational geometry, shown in Fig.2.4 (right), reveals that these effects are caused by
the computational mesh not resolving the metallic case well, so these regions have been
attributed to the heat insulator.
Fig.2.3 Surface Plots on a SolidWorks model inner surface (left) and on its computational
realization (right).
Fig.2.4 Cut plots using a model geometry (left) and its computational realization (right).
2-20
On the other hand, this option can substantially save both time and memory when creating
Surface Plots for SolidWorks models containing rippled or ribbed surfaces whose ripples
have not been resolved by the computational mesh and are not substantial from the
problem solution viewpoint, since the simplified solid/fluid boundary of the
computational domain is colored instead of coloring the SolidWorks models faces.
If the computational mesh has resolved the SolidWorks model well,
so the obtained computational results are adequate, then enable the
Use CAD geometry option before performing the final Cut Plots and
Surface Plots to obtain smooth pictures which are more convenient
for analysis.
When creating a Surface Plot under the switched-off Use CAD
geometry option, the computational mesh partial cells solid/fluid
interfaces are colored only. As a result, when a Surface Plot is
created in the Use all faces mode, all the partial cells solid/fluid
interfaces are colored. However, when a Surface Plot is created on a
selected surface, the solid/fluid interfaces are colored only in the
partial cells intersected by the SolidWorks models surface
approximated by triangles inside SolidWorks, which can be different
from the SolidWorks models surface approximation, so some partial
cells can exist that are not intersected by the surfaces triangle
approximation, and therefore their solid/fluid interfaces would not be
colored (see gray strips in Fig.2.5). Naturally, this circumstance
concerns the picture only and does not affect the calculations results.
Fig.2.5
2.7 Switching-off the Calculation Results Interpolation and Extrapolation
Since the numerical solution is obtained inevitably in the discrete form, i.e., in the
computational mesh cells centers in COSMOSFloWorks, it is interpolated and
extrapolated by the post-processor to present the results in a smooth form, which is
typically more convenient to the user. As a result, prediction errors can stem from and/or
be hidden by these interpolations and extrapolations that smooth the calculation results. To
facilitate revealing, analyzing, and eliminating such errors, COSMOSFloWorks offers an
option of visualizing the physical parameter values calculated at the computational mesh
cells, i.e., at their centers, when presenting calculation results in Cut Plots and Surface
Plots (isolines and isosurfaces can not be created without interpolating and extrapolating
the calculation results), so when coloring a surface with a palette, the results are
considered constant within the mesh cells (see Fig.2.6).
Since the mesh cells centers used in coloring the surface can lie at
different distances from the surface, this can introduce an additional
variegation into the picture, if the coloring parameters value
noticeably depends on this distance (see Fig.2.6).
Fig.2.6 The fluid velocity Surface Plots created with (left) and without (right) interpolating and
extrapolating the calculation results.
2-22
2.8 Conclusion
The presented mesh-associated tools of COSMOSFloWorks are additional tools for
obtaining reliable and accurate results with this code. These tools are summarized in the
table:
Option
Application
Reason
Basic
mesh
Initial
mesh
After the
calculation
Visualizing the
Basic mesh
+ + +
Inspecting the Basic mesh and
setting its Control planes
Widened
capabilities of
loading the results
+ +
To view the Initial mesh and
the calculation results
Viewing the Initial
mesh
+ +
Analyzing the Initial mesh
Viewing mesh
cells of different
type
+ +
Viewing mesh cells and
saving their physical
parameters
Visualizing the
computational
geometry
+ +
Analyzing inadequate results
and quickly post-processing
the results of intricate models
Switching-off the
results
interpolation
+
Analyzing inadequate results
3 Calculation Control Options
3.1 Introduction
The Calculation Control Options dialog box has been introduced into COSMOSFloWorks
to easily and effectively control:
finishing the calculation,
saving the results during the calculation,
refining the computational mesh during the calculation,
freezing the flow calculation,
adjusting the time step for a time-dependent analysis (PE only),
controlling the number of rays traced from the surface if radiating heat transfer is
enabled (PE only).
This dialog box is accessible both before the calculation and during the calculation (in
which case this boxs settings are applied starting from the next iteration).
The main information on employing the options of Finishing the calculation and
Refining the computational mesh during calculation is presented in this document.
3.2 Finishing the Calculation
COSMOSFloWorks solves the time-dependent set of equations for all problems, including
steady-state cases. For such cases it is necessary to recognize the moment when a
steady-state solution is attained and therefore stop the calculation. To that end a set of
independent finishing conditions is offered by COSMOSFloWorks, allowing the user to
select the most appropriate conditions and criteria for the calculation. The following
finishing conditions are offered by COSMOSFloWorks:
the maximum number of refinements;
the maximum number of iterations;
the maximum physical time (for time-dependent problems only);
the maximum CPU time;
the maximum number of travels;
the Goals convergence.
In COSMOSFloWorks you can select the finishing conditions that are most appropriate
from your viewpoint for solving the problem under consideration, and specify their
values. For the latter two conditions (i.e., for the maximum number of travels and the
Travel is the number of iterations required for the propagation of a
disturbance over the computational domain. The current number of
iterations in a travel is presented in the Info box of the Calculation
monitor.
Chapter Calculation Control Options
2-24
Goals convergence settings) COSMOSFloWorks presents their default values (details are
described below). You can also select the general finishing condition in the Finish
Conditions value cell: either to finish the calculation when all the selected finishing
conditions are satisfied or to finish it when one of them is satisfied.
In any case, information on the finishing conditions due to which the calculation has
finished is shown in the Monitors Log box.
The Goals convergence finishing condition is complex since it consists of satisfying all
the specified Goals criteria. A specified Goal criterion is a specified dispersion, which is
the criterion difference between the Goals maximum and minimum values, and a
specified analysis interval calculated from the last iteration, over which this difference
(i.e., the dispersion) is determined. The analysis interval is the same for all the specified
Goals and is actuated after an automatically specified initial calculation period (in travels),
and, if the option to refine the computational mesh during calculation is enabled, after an
automatically or manually specified relaxation period is realized (in travels or iterations)
from the last mesh refinement. As soon as the Goal dispersion obtained in the calculation
becomes lower than the specified dispersion, the Goal is considered achieved, i.e., having
converged. As soon as all the Goals included in the Goals convergence finishing condition
(by selecting them in the On/Off column) have converged, this condition is considered
satisfied. The Goals not included into the Goals convergence finishing condition are used
for information only, i.e., with no influence on finishing the calculation.
Let us consider the COSMOSFloWorks default values for the maximum number of travels
and the Goals convergence settings in detail. These default (or recommended by
COSMOSFloWorks) values depend on the Result resolution level, specified either in the
Wizard or in the Calculation Control Options dialog boxs Reset dialog box. For higher
Result resolution levels the finishing conditions are tighter.
The default maximum number of travels depends on
the type of the specified Goal (i.e., dynamic or diffusive, see below);
the specified Result resolution level;
the problem's type (i.e., incompressible liquid or compressible gas, low or high
Mach number gas flow, time-dependent or steady-state).
Dynamic goals: Static Pressure, Dynamic Pressure, Total Pressure,
Mass Flow Rate, Forces, Volume Flow Rate, and Velocity.
Diffusive goals: Temperature, Density, Mass in Volume, Heat flux,
Heat transfer rate, Concentrations, Mass Flow Rate of species, and
Volume Flow Rate of species.
The default Goals convergence settings are the default analysis interval, which is shown
in the Finish tab of the Calculation Control Options dialog box, and the default Goals
criterion dispersion values, which are not shown in the Calculation Control Options dialog
box, but, instead, are shown in the Monitors Goal Table or Goal Plot table (in the Criteria
column), since they depend on the Goal physical parameters values calculated in the
computational domain, and therefore are not known before the calculation and, moreover,
can change during it. In contrast, the Goals criterion dispersion values specified manually
do not change during the calculation.
As for the automatically specified initial calculation period (measured in travels), it
depends on the problems type, the Goals type, and the specified Result resolution level.
3.3 Refining the Computational Mesh During Calculation
Refining the computational mesh during calculation consists of splitting or merging the
computational mesh cells in high-gradient solution regions. This option has the following
governing parameters:
refinement level,
splitting/merging criteria (named, accordingly, refinement/unrefinement criteria),
allowance to refine cells in fluid and/or solid regions,
approximate maximum number of cells,
strategy of refinements during the calculation.
The first four parameters are described in COSMOSFloWorks Help and Users Guide.
Here, let us consider the Refinement Strategy in detail. The following three strategies are
available:
Periodic refinement;
Tabular Refinement;
Manual Only refinement.
the manually specified analysis interval for the Goals
convergence finishing criteria must be substantially longer than
the typical flow field oscillation period (if it occurs);
the Goals determined on solid/fluid interfaces or model
openings, as well as the Post-processors Surface Parameters,
yield the most accurate and correct numerical information on
flow or solid parameters, especially integral ones;
Global Goals yield the most reliable information on flow or solid
parameters, although they may be too general;
the CPU time depends slightly on the number of the specified
Goals, but, in some cases, substantially on the presence of a
Surface Goal;
Surface Goals are exactly equal to Surface Parameters.
Chapter Calculation Control Options
2-26
In the first two strategies the refinement moment is known beforehand. The solution
gradients are analyzed over iterations belonging to the Relaxation interval, which is
calculated backward from the refinement moment and specified by default or manually.
As a result, only steady-state gradients are taken into account. On the other hand, this
analysis must not occur within the same relaxation interval defined from the calculation
beginning, in order to avoid taking the initial highly unsteady period into account.
Therefore, at least a period equal to two relaxation intervals is recommended before the
first refinement. If the first assigned refinement is performed within this period, the
gradients are analyzed over a shorter period, which can be equal to the current iteration in
the limit. If you initiate a refinement manually within this period, the gradients are
analyzed in the current iteration only. Naturally, such short periods for analyzing the
gradients are not good for solving the problem using the minimum CPU time and memory.
The figure below illustrates these concepts. Here, the letter r denotes the Relaxation
interval. This figure involves both Periodic and Tabular refinements. Case 1 is the normal,
recommended approach. In Case 2 the refinement is too close to the calculation
beginning, and as a result the gradients are analyzed over the shorter interval. (which
could even be as short as simply the current iteration in the limiting case). Case 3 is a
particular case in which a refinement is initiated manually before a previously assigned
refinement. As a result, the manual refinement is well determined, since the gradients have
been analyzed for it over almost the entire relaxation interval. However, when the
previously assigned refinement is performed then the gradients are analyzed over the now
substantially shorter interval, and therefore its action can be incorrect. Case 3
demonstrates the possible error of performing manual and previously assigned
refinements one after another.
4 Flow Freezing
4.1 What is Flow Freezing?
Sometimes it is necessary to solve a problem dealing with different processes developing
at substantially different rates. If the rates difference is substantial (10 or more times)
then the CPU time required to solve the problem is governed by the slowest process. To
reduce the CPU time, a reasonable approach is to stop the calculation of the process that
has fully developed and does not change further and use its results to continue the
calculation of the slower processes. Such an approach is called freezing.
In the case of problems solved with COSMOSFloWorks the processes of convective mass,
momentum, and energy transport are the fastest processes to develop or converge, whereas
the processes of mass, momentum, and energy transfer by diffusion are the slowest ones.
Accordingly, COSMOSFloWorks offers the Flow Freezing option, allowing for the
freezing, or fixing, of the pressure and velocity field while continuing the temperature and
species calculations. This option is especially useful in solving steady-state problems
Fig.3.1 Refinement strategy.
Collecting of the statistics is prohibited
Statistics are collected
Refinement
Case 1
Auto ref.
r r
r
r
0
0
Ref. point 1
Ref. point 2
Case 2 Case 3
Ref. point 1
r r
Manual ref.
Chapter Flow Freezing
2-28
involving diffusion processes that are important from the users viewpoint, e.g. species or
heat propagation in dead zones. Time-dependent analyses with nearly steady-state velocity
fields and diffusion processes developing with time are also examples of this class of
problems. As a result, the CPU time for solving such problems can be substantially
reduced by applying the Flow Freezing option.
4.2 How It Works
To access the Flow Freezing option, open the Calculation Control Options dialog box,
then the Advanced tab. This option has three modes: Disabled (by default), Periodic, and
Permanent.
Flow Freezing in a Permanent Mode
As an example of applying the Flow Freezing option, let us consider a plane flow (2D)
problem of heating the vortex core in a vessel (Fig.4.1).
At the beginning the entire fluid region is filled with a cold (T=300 K) liquid. A hot
(T=400K) liquid enters the vessel through the lower channel (the upper channel is the
exit). As a result, a vortex having a cold core is generated in the vessel. The vortex core
temperature is changed mainly due to heat diffusion. To measure it, a small body, disabled
in the Component Control dialog box, so it is treated by COSMOSFloWorks as a fluid
region, is placed at the vortex center. Its minimum temperature (i.e., the minimum fluid
temperature in this region) is the calculations Volume Goal.
First of all, let us consider Flow Freezing operating in the Permanent mode. The
Permanent modes only parameter is the starting moment of enabling the Flow Freezing
option. Until this moment the calculation runs in the usual manner. After this moment the
fluid velocity field becomes frozen, i.e., it is no longer calculated, but is taken from the
last iteration performed just before the Flow Freezing Start moment. For the remainder of
the run only the equations terms concerning heat conduction and diffusion are calculated.
As a result, the CPU time required per iteration is reduced.
Fig.4.1 Heating the vortex core in a vessel.
The starting moment of the Flow Freezing option should not be set too early such that the
flow field has not had a chance to fully develop. As a rule, a minimum 0.25 initial period
travels is required to satisfy this condition. In most problems the 0.5 travel initial period is
sufficient, but there are problems requiring a longer initial period.
When first solving the problem under consideration we set the maximum number of
travels to 10. The calculation performed without applying the Flow Freezing option then
required about 10 travels (the CPU time of 13 min. 20 s on a 500 MHz PIII computer) to
reach the project Goals convergence (the steady-state minimum fluid temperature in the
vortex core). However, the steady-state fluid velocity field was reached in about 0.5
travels, i.e., substantially earlier. So, by applying the Flow Freezing option in the
Permanent mode (just after 0.5 travels) the same calculation only required a CPU time of 7
min. 25 s on the same computer to reach the project Goals convergence. Both the Goal
convergence histories are plotted in Fig.4.2.
The Flow Freezing Start moment, as well as other parameters of the
Calculation Control Options dialog box can be changed during a
calculation.
As soon as the Flow Freezing option is invoked, only the slowest
processes are calculated. As a result, the convergence and finishing
criteria can become non-optimum. Therefore, to avoid obtaining
incorrect results when enabling the Flow Freezing option, it is
recommended to increase the maximum number of travels specified
at the Finish tab of the Calculation Control Options dialog box by
1.55 times compared to the number that was set automatically or
required for the calculation performed without the Flow Freezing
option.
Fig.4.2 The Goals (which is the minimum fluid temperature in the vortex core) convergence history
with and without applying the Flow Freezing option.
2-30
If it is necessary to perform several calculations with the same fluid velocity field, but
different temperatures and/or species concentrations, it is expedient to first calculate this
fluid velocity field without applying the Flow Freezing option. Then, clone the
COSMOSFloWorks project into several projects (including copying the calculations
results), make the required changes to these projects, and perform the remaining
calculations for these projects using the calculated results as initial conditions and
applying the Flow Freezing option in the Permanent mode with a zero Start period.
Flow Freezing in a Periodic Mode
In some problems the flow field depends on temperature (or species concentrations), so
both the velocity and the temperature (concentrations) simultaneously change throughout
the calculation. Nevertheless, since they change in a different manner, i.e., the velocity
field changes faster than the temperature (concentrations) field, therefore approaching its
steady state solution earlier, the Flow Freezing option can be used in a Periodic mode to
reduce the CPU time required for solving such problems. The Periodic mode of the Flow
Freezing option consists of calculating the velocity field not in each of the iterations (time
steps), but in each n-th one only, where n is changed linearly from 2 at the Periodic Flow
Freezing Start moment (prior to this moment the Flow Freezing option does not act, i.e. n
= 1) to a specified maximum N (N - 1 is equal to the Maximum freezing period value
specified in the Flow Freezing tab), whereas the temperatures and concentrations are
calculated in each iteration. Once N is reached the velocity field is calculated in each N-th
iteration only for the remainder of the run. Examples include channel flows with
specified mass flow rates and pressures, so the fluid density and, therefore, velocity
depend on the fluid temperature, or flows involving free convection, where due to the
buoyancy the hot fluid rises, so the velocity field depends on the fluid temperature.
As an example, let us consider a 3D external problem of an air jet outflow from a body
face into still air (see Fig.4.3, in which the jet outflow face is marked by a red line). Here,
the wire frame is the computational domain. The other body seen in this figure is
introduced and disabled in the Component Control dialog box (so it is a fluid region) in
order to see the air temperature averaged over its face (the projects Goal), depending on
the air temperature specified at the jet outflow face.
If you forget to use the calculated results as initial conditions, then
the saved fluid velocity field will be lost in the cloned project, so the
project must be created again. To use the calculated results as initial
conditions for the current project, select the Transferred type of
Parameter definition for the initial conditions in the General Settings
dialog box.
This problem is solved in several stages. At the first stage, the calculation is performed for
the cold (T = 300 K, which is equal to the environment temperature) air jet. This
calculation requires a CPU time of 12 min. 14 s. Then we clone the project including
copying the results. Next, we set the outlet air temperature to T = 400 K, specify the
Periodic mode of the Flow Freezing option by its Start moment of 0.25 travels (in order
for the heat to have time to propagate along the jet to the measuring face) and retain the
default (N 1) = 10 (see above), since it yields adequate results in solving various
problems. Then perform the calculation on the same computational mesh with the Run
boxs option Take the previous results. This calculation takes 8 min. 29 s of CPU time.
The similar calculation without applying the Flow Freezing option takes 12 min. 19 s of
CPU time.
Fig.4.3 Air jet outflow from a body face into a still air.
2-32
3
Meshing Additional Insight
Introduction
COSMOSFloWorks considers the real model designed in SolidWorks and generates a
rectangular mesh automatically distinguishing the fluid and solid domains. Depending on
the characteristics of the task at hand, the corresponding computational domain is
generated in the form of rectangular prism enclosing the model. In the mesh generation
process, the computational domain is divided into uniform rectangular cells, which form a
so-called basic mesh. Then, using information about the model geometry,
COSMOSFloWorks further constructs the mesh by means of various refinements, which
split the basic mesh cells into smaller rectangular cells, thus better representing the model.
The mesh from which the calculation starts, the so-called initial mesh, is fully defined by
the generated basic mesh and the refinements settings.
Each refinement has its criterion and level. The refinement criterion denotes which cells
have to be split, and the refinement level denotes the smallest size to which the cells can
be split. Regardless of the refinement considered, the smallest cell size is always defined
with respect to the basic mesh cell size so the constructed basic mesh is of great
importance for the resulting computational mesh.
The main types of refinements are:
Small Solid Features Refinement
Curvature Refinement
Tolerance Refinement
Narrow Channel Refinement
Square Difference Refinement
In addition, the following two types of refinements can be invoked locally:
Cell Type Refinement
Chapter Initial Mesh Generation Stages
3-2
Solid Boundary Refinement
During the calculation, the initial mesh can be further refined by the
Solution-Adaptive Refinement.
Though it depends on a refinement which criterion or level is available for user control,
we will consider all of them (except for the Solution-Adaptive Refinement) to give you a
comprehensive understanding of how COSMOSFloWorks meshing works.
In the chapter below the most important conclusions are marked with the blue italic font.
For abbreviation list refer to the Glossary paragraph.
1 Initial Mesh Generation Stages
1.1 Basic Mesh Generation and Resolving the Interface
1.1.1) Create a basic mesh cell whose size is defined by the computational domain size,
specified Control Planes and the number of the basic mesh cells. [Nx, Ny, Nz, Control
Planes - parameters which act on the stage are shown in square brackets]
1.1.2) Analyze triangulation for each basic mesh cell at interface between substances
(such as solid/fluid, solid/porous, solid/solid and porous/fluid interfaces), finding the
maximum angle between normals to the triangles, which compose the solid surface within
the cell.
1.1.3) Depending on the maximum angle found, the decision is made whether to split the
cell or not in accordance with the specified Small solid features refinement level (SSFRL),
Narrow channel refinement level (NCRL), Curvature refinement level (CRL) and
Curvature criterion (CRC), Tolerance refinement level (TRL) and Tolerance Refinement
Criterion (TRC) (see the Refinements at Interface Between Substances paragraph).
[SSFRL, NCRL, CRL and CRC]
1.1.4) If a basic mesh cell is split, the resulting daughter cells are analyzed as described in
1.1.2 and 1.1.3 items and, if necessary, split further. The cell splitting will proceed until
the interface resolution satisfies the specified SSFR criterion, CRC and TRC, or the
corresponding level of splitting reaches its specified value.
1.1.5) The operations 1.1.2 to 1.1.4 are applied for the next basic mesh cell and so on,
taking into account the following Cell Mating rule: two neighboring cells (cells with
common side) can be only cells whose level is the same or differs by one. The rule is
priority and necessary for simplifying numerical algorithm in solver.
If a cell belongs to a local initial mesh area the corresponding local
refinement levels will be applied (see the Local Mesh Settings
paragraph).
The specified levels of splitting denote the maximum admissible
splitting, i.e. they show to which level a basic mesh cell can be split
if it is required for resolving the solid/fluid interface within the cell.
The mesh at this stage is called the primary mesh. The primary mesh implies the complete
basic mesh with the resolution of the solid/fluid (as well as solid/solid, solid/porous, etc.)
interface by the small solid features refinements and the curvature refinement also taking
into account the local mesh settings.
1.2 Narrow Channel Refinement
After the primary mesh has been created, the narrow channel refinement is actuated. The
Narrow Channels term is conventional and used for the definition of the models flow
passages in the normal-to-solid/fluid-interface direction.
Regardless of the real solid curvature, the mesh approximation is that the solid boundary is
always represented by a set of flat elements, whose nodes are the points where the model
intersects with the cell edges. Thus, whatever the model geometry, within a partial cell
there is always a flat element and the normal to this element denotes the
normal-to-solid/fluid-interface direction for this partial cell. The only exception is an
irregular cell where it is impossible to construct the flat element unambiguously. See the
Irregular Cells paragraph for details.
The narrow channel refinement operates as follows:
1.2.1) For each partial cell COSMOSFloWorks calculates the local narrow channel
width as the distance between the partial cell and the first next partial cell lying in the
normal-to-the-solid/fluid-interface direction (normal to the cells flat element).
The fourth-level red cells, appearing
after resolving the cog, cause the
neighboring cells to be split up to
third level (yellow cells) that, in
turn, actuates the subsequent
refinement producing second level
cells (green cells) and first level
cells (blue cells). The white zero
level cell (basic mesh cell) remains
unsplit since it borders with first
level cells only, thus satisfying the
rule.
The rule is strict and has higher priority than the other cell
operations. The Cell Mating rule is also enforced for the cells that are
entirely in a solid.
If a normal-to-solid/fluid-interface direction crosses a local initial
mesh area the corresponding local narrow channel refinement
settings will be applied to the cells in this direction.
Fig.1.1
Fluid cell refinement due to the Cell Mating rule.
3-4
1.2.2) If the distance value is within the range defined by the Minimum height of narrow
channel (NCHmin) and Maximum height of narrow channel (NCHmax) options, the
number of cells per this distance is calculated including both partial cells while taking into
account which part of the partial cell is in fluid. [NCHmin, NCHmax]
1.2.3) In fact, the number of cells per the direction is calculated as N = N
f
+ n
p1
+ n
p2
,
where N
f
number of fluid cells in the direction, n
p1
and n
p2
are relative fluid parts of the
partial cells in the direction. In the Fig. 1.2 N
f
= 2; n
p1
= n
p2
= 2/5, and N = 2+2/5+2/5 =
2.8. This value is compared with the specified Characteristic number of cells across a
narrow channel (CNC). If N is less than the specified CNC then the cells in this direction
will be split. In the Fig.1.3 the partial cells become smaller so the relative fluid parts of the
partial cells are increased (n
p1
= n
p2
= 9/10) and the criterion is satisfied (N > CNC).
The narrow channel refinement is symmetrical with respect to the mid point of the
distance and proceeds from the partial cells toward the mid point. [CNC, NCRL].
In the Fig. 1.4 the specified Characteristic number of cells across a channel is 5 but only
two cells were generated since the maximum refinement level of one allows only basic
mesh cells and first-level cells to be generated.

Fig.1.2
NCRL = 2; CNC = 3;
N = 2.8 < CNC
Fig.1.3
NCRL = 3; CNC = 3;
N = 3.8 > CNC
Like the other refinements, the Narrow channel refinement level
(NCRL) denotes the maximum level to which the cells can be split to
satisfy the CNC criterion. The NCRL has more priority than the
CNC, so the refinement will proceed until the CNC criterion is
satisfied or all the cells reach the Narrow channel resolution level.

Fig.1.4
CNC = 5; NCRL = 1
Fig.1.5
CNC = 5; NCRL = 3
In the Fig. 1.5 the specified Narrow channel refinement level is great enough to satisfy the
five cells per the channel requirement.
Note that the partial cells in the beginning and end of the channel were not split like the
other partial cells along the channel. This happened because the right angle is
approximated by the flat element sloping to both the channels sides. Therefore, the
normal-to-solid\fluid direction taken from these corner cells is not orthogonal to the
channel surface as for the other partial cells, so the number of cells per this direction
satisfies the criterion preventing the cells from splitting (see Fig. 1.6).
1.2.4) Refinement on Openings This refinement is intended to force the splitting of the
partial cells which were not refined due to the chamfer approximation of the right angles
(Fig. 1.6), if these cells are cells at the boundary condition surface.
In Fig. 1.7 the boundary condition, specified at the channel end-wall, actuates the
Refinement on Opening procedure that splits the red partial cells to the level of adjacent
partial cells.
1.2.5) Next, for all the fluid cells within the entire computational domain the following
Fluid Cell Leveling procedure is applied: if a fluid cell is between two cells of lower level,
the fluid cell will be split to equalize the level of neighboring cells.
1.3 Square Difference Refinement
The Square Difference Refinement checks the neighboring partial cells of different levels
for the difference between the cells fluid passage areas. If the difference between the
greater-level cells fluid passage area and the total fluid passage areas of the adjacent
lower-lever cells exceeds the Square Difference Refinement Criterion (SDRC) than the
greater-level cell is split to the level of adjacent cells in order to equalize the fluid passage
areas (see Fig.1.8). The Square Difference Refinement is always enabled and cannot be
disabled since it is a strict solver requirement. As with the Cell Mating rule, this is another
condition imposed by the solver to provide stability for the convergence processes.
Though you cannot turn off the Square Difference Refinement, you can control its
criterion, which is directly proportional to the Curvature refinement criterion. [CRC].

Fig.1.6
Normal-to-solid\fluid-interface direction in the
corner cells.
Fig.1.7
CNC = 5; NCRL = 3; Inlet velocity at the
channel end-wall.
3-6
Fig. 1.8 shows two neighboring partial cells of different levels at the cylinder's solid/fluid
interface. The first-level cell's fluid passage area is the ABDE polygon. The total
second-level cell's fluid passage area is the ABCDE polygon. The difference between the
fluid passages is the yellow BCD polygon. In this example we increased the curvature
refinement criterion to , thereby increasing the Square Difference Refinement Criterion
so the fluid passage difference (BCD) is smaller than the criterion, and thus, there is no
need to split the first-level cell.
Note that the Square Difference Refinement may cause a domino-like refinement when
one splitting produces cells which become lower-level cells for the neighboring cell
causing this cell to be split and so on resulting in an increased number of cells.

Fig.1.8
Two adjacent partial cells of
different levels at the cylinder
surface.
Fig.1.9
Cut plot of the cylinder. The concerning cells are blue.
SSFRL = 2; CRL = 0; CRC = 3.14; NCRL = 1.
Fig.1.10
SSFRL = 3, CRL = 0; CRC = 0.45
Total cells = 49391.
Fig.1.11
SSFRL = 3, CRL = 2; CRC = 0.50;
Total cells = 41376.
In the Fig. 1.10 the total number of cells is almost 20% more than in the Fig. 1.11 in spite
of the fact that the Curvature refinement is disabled (CRL = 0) in the first case. Here, the
model geometry is such that before the Square Difference Refinement in both cases the
mesh is practically the same and mostly denoted by the Small Solid Features Refinement
when the SSFRL exceeds the CRL, i.e. changing the CRL from 0 to 3 would not
dramatically change the number of cells generated. However, in the first case the
curvature criterion is less, resulting in more a stringent criterion of the Square Difference
Refinement. So the smaller Square Difference Refinement criterion, leads to a greater
number of cells being subjected to the Square Difference Refinement. In the Fig.1.10 you
can see a stripe of the third level cells along the cylinder that is the result of the Square
Difference Refinement and the domino effect when a cell on the cylinder edge involves
the neighboring cell in the refinement procedure and so forth along the cylinder.
If in the first case we specify the same CRC as in the second case (0.5054 rad), the total
number of cells decreases to 40963.
1.4 Irregular Cells Refinement
If irregular cells appear, they are split to the maximum level among all the refinement
levels specified for the region of irregular cells or until the cells become regular. See
Irregular Cells.
1.5 Mesh Diagnostic
The mesh diagnostic is intended to inspect the resulting initial mesh and do not change the
total number of cells.
2 Refinements at Interface Between Substances
Different interface types (solid/fluid, solid1/solid2, solid/porous or porous/fluid) are
checked on different refinement criteria: solid/fluid and solid/porous interfaces - small
solid features criterion, curvature refinement criterion, narrow channel refinement
criterion and irregular cells refinement; solid1/solid2 - small solid features criterion;
porous/fluid - small solid features criterion and curvature refinement criterion. Whereas
the specified refinement levels are equally applied to any interface type.
Increasing the curvature criterion will increase the Square Difference
Refinement Criterion, and, in turn, decrease the number of cells in
both the cases.
Chapter Refinements at Interface Between Substances
3-8
2.1 Small Solid Features Refinement
The small solid features refinement acts on the cells where the maximum angle between
normals to the surface-forming triangles is strictly greater than 120. To make this
120-degree criterion easier to understand, let us consider simple small solid features
formed of planar faces only. The planar face-forming triangles have the same normal that
is normal to the planar face. Therefore, instead of considering the normals to the triangles
we can consider normals to faces, or better the angle between faces, proceeding on the
assumption that the faces normals form the maximum angle over all normals within the
cell.
In the Fig. 2.1 the cells with the cogs of 150 and 60 degrees were not split by the small
solid features refinement because the maximum angle between the faces (i.e. between
normals to the triangles enclosed by the cell) is 30 and 120 correspondingly. If the angle
between the normals becomes greater than 120 (121 for the 59-cog) the cell is split.
The cell with the square spike was split because the spikes lateral faces have their
normals at the angle of 180, thus satisfying the 120-degree criterion.
Note that 90-degree corners (the rightmost cell) do not satisfy the criterion and therefore
will not be resolved by the small solid features refinement.
Remember that if the Narrow channel refinement is enabled, the
maximum level to which the small solid features refinement can split
the cells is set as the maximum level from the specified SSFRL and
Narrow channel refinement level (NCRL). In other words, if the
Narrow channel refinement is enabled, the SSFRL is idle if it is
smaller than the NCRL.
Fig.2.1
SSFRL = 1, CRL = 0, NCRL = 0
Fig.2.2 SSFRL = 0, CRL = 0, NCRL = 1
From the Fig. 2.2 it seen that the cells are split by the 120-degree criterion but to the level
one as defined by the narrow channel refinement level.
For the information about how the NCRL influences the narrow channel refinement see
the Narrow Channel Refinement paragraph.
2.2 Curvature Refinement
The curvature refinement works the same as the small solid features refinement with the
difference being that the critical angle between the normals can be specified by the user as
curvature refinement criterion (CRC) in radians. Here, the smaller the criterion, the better
resolution of the solid curvature. To give a more exact and descriptive explanation, the
following table is given considering both the angle between normals and the angle
between planar faces, proceeding on the assumption that the faces normals form the
maximum angle over all normals within the cell.
The table states that if the CRC is equal to 0.4510268 rad, then all the cells where the
angle between normals to the surface-forming triangles is more than 25 degrees will be
split.
You can see that the curvature criterion set to 0.4510268 rad splits the cells with the
150-degrees cog.
Table 2.1: Influence of the curvature criterion on the solid curvature resolution.
Curvature
criterion, rad
0.31
76
0.45
10
0.55
48
0.64
35
1.04
72
1.57
08
2.09
44
3.14
16
' between
normals,
[degrees]
>19 >25 >31 >36 >60 >90 >120 180
between faces,
[degrees]
<161 <154 <148 <143 <120 <90 <60 0
Fig.2.3 CRL = 1, CRC = 0.5548,
SSFRL = 0, NCRL = 0
Fig.2.4 CRL = 1, CRC = 0.451,
SSFRL = 0, NCRL = 0
Note that the curvature refinement works as the small solid features
refinement when the curvature criterion is equal to 2.0944 rad (2/3).
Chapter Refinements at Interface Between Substances
3-10
However, the default curvature criterion values are small enough to resolve obtuse angles
and curvature well. Decreasing the curvature criterion is reasonable if you want to avoid
superfluous refinement but it is recommended that you try different criteria to find the
most appropriate one.
The curvature criterion also denotes the criterion of the Square Difference Refinement.
The square difference refinement criterion is directly proportional to the CRC, so the
smaller CRC, the smaller square difference refinement criterion, resulting in a greater
number of cells appearing after the Square Difference Refinement.
2.3 SSFRL or CRL
Why is it necessary to have two criteria? As you can see the curvature refinement has
more priority than the small solid features refinement if the curvature criterion is smaller
than 2/3 . Note that COSMOSFloWorks specified curvature criterion values are always
smaller than 2/3 .
Nevertheless, the advantage of the small solid features refinement is that being sensitive to
relatively small geometry features it does not notice the large-scale curvatures, thus
avoiding refinements in the entire computational domain but resolving only the areas of
small features. At the same time, the curvature refinement can be used to resolve the
large-scale curvatures. So both the refinements have their own coverage providing a
flexible tool for creating an optimal mesh.
In other words, if you did not set the CRC greater than 2/3 p, the
small solid feature refinement would be idle if the SSFRL and NCRL
are smaller than the CRL.
2.4 Tolerance Refinement
Any surface is approximated by set of polygons whose vertices are surface's intersection
points with the cells' edges. This approach accurately represents flat faces though
curvature surfaces are approximated with some precision (e.g. as a circle can be
approximated by a polygon). The tolerance refinement criterion controls this precision. A
cell will be split if the distance between the outermost interface's point within the cell and
the polygon approximating this interface is larger than the specified criterion value.
3 Local Mesh Settings
The local mesh settings influence only the initial mesh and do not affect the basic mesh in
the local area, but are basic mesh sensitive in that all refinement levels are set with respect
to the basic mesh cell.
The local mesh settings are applied to the cells intersected with the local mesh region
which can be represented by a component, face, edge or vertex.
In case a cell intersects with different local mesh setting regions, the refinement settings in
this cell will be used to achieve the maximum refinement.
Cell Type Refinement. The level of refining cells of a specific type (all cells, fluid and
partial cells, solid and partial cells, or only partial cells) denotes the minimum level to
which the corresponding cells must be split if it doesnt contradict the Cell Mating rule.
The minimum level means the lower bound to which it is obligatory
to split cells, though the cells can be split further if it is required to
satisfy the other criteria such as Small solid features refinement,
Curvature refinement, Narrow channels refinement or Solid
Boundary Refinement.
Tolerance Refinement Small Solid Feature Refinement
Tolerance criterion = 0.1
Tolerance
criterion = 0.08

Tolerance Refinement Curvature Refinement
Refines cells taking into account the
curvature only.
Tolerance criterion = 0.1
Refines cells only if the solid part cut by
the polygon is large enough (h > 0.1)
Tolerance
criterion = 0.03

Chapter Irregular Cells
3-12
In case different cell types are to be refined, the level of refining partial cells is set as the
maximum level among all selected levels.
The local mesh settings have higher priority than the initial mesh settings. Therefore, the
local mesh cells will be split to the specified local refinement levels regardless of the
general SSFRL, CRL and NCRL (specified in the Initial Mesh dialog box). This, however,
may produce refining of cells that are out of the local region due to imposing the Cell
Mating rule.
4 Irregular Cells
An irregular cell is a computational mesh cell lying at the solid/fluid interface (or
solid/solid interface in cases where two or more different solids are within the cell), partly
in one substance and partly in another substance, and characterized by the impossibility of
defining the solid/fluid interface position within the cell, given the cells nodes positions
relative to solid region and the intersections of the solid/fluid interface with the cell.
The example of an irregular cell at the solid/fluid interface is given below.
The model's rectangular cut is represented with the computational mesh as shown in the
Fig. 4.2. The central cell is irregular since the cell position is such that the solid/fluid
interface within the cell cannot be uniquely constructed. In the Figs. 4.3, 4.4 two possible
variants of resolving the cell are shown. In both cases the information about the body
position is the same because it is fully dependent on the cell's nodes and intersection
points. However, the intersection points can be joined together differently, resulting in
completely different geometry as seen from Figs 4.3, 4.4. Because of the lack of
information about the body position, the cell is considered as irregular.
Another example of irregular cell is shown in the Figs. 4.5 4.8.
Fig.4.1 The
geometry of the
model (white region is
fluid, yellow is solid).
Fig.4.2 The real
computational
geometry. Irregular
cell is red.
Fig.4.3 First way of
possible irregular cell
resolution.
Fig.4.4 Second
way of possible
irregular cell
resolution.
Nevertheless irregular cells are used in the calculation, irregular cells at the solid/fluid
interface are treated as fluid cells. That is why if an irregular cell exists you should analyze
its possible influence on the calculation taking into account that it is a fluid cell. Thus, in
the first example (Figs 4.1-4.4) the irregular cell doesnt dramatically worsen the results,
only compromising somethat the resolution of the cut. On the contrary, in the second
example (Figs 4.5-4.8) the irregular cell may cause inadequate results as the fluid can flow
through the solid neck.
The automatic resolution of irregular cells would require a time and memory consuming
meshing algorithm. In most cases a simple refining of irregular cells eliminates its
irregularity.
Use the Results Summary to find out whether irregular cells are present and use the
Mesh Visualization tool to detect where they are located.
5 Postamble
The task of how a model should be resolved with the computational mesh is always model
specific. In general, a denser mesh will provide better accuracy but you should avoid
redundant refining and tend to create an optimal mesh.
Fig.4.5
The geometry of the
model (white region is
fluid, yellow is solid).
Fig.4.6
The real
computational
geometry. Irregular
cell is red.
Fig.4.7
First way of possible
irregular cell
resolution.
Fig.4.8
Second way of
possible irregular cell
resolution.
All irregular cells are always split to the maximum level among all
the refinement levels specified for the region of irregular cells or
until the cells become regular. Thus, if you want to get rid of
irregular cells, you should increase the refinement levels, starting
with increasing of the Small solid features refinement level, because
it will change the existing mesh in other regions to a lesser degree
than the other refinement levels.
Chapter Glossary
3-14
When making a calculation, try different mesh settings and carefully analyze the obtained
results to understand where it is necessary to refine the mesh and where a rough resolution
or even irregular cells are acceptable for the desired accuracy.
6 Glossary
Nx, Ny, Nz Number of basic mesh cells per X, Y and Z direction.
SSFRL Small solid feature refinement level.
CRL Curvature refinement level.
CRC Curvature refinement criterion.
TRL Tolerance refinement level.
TRC Tolerance refinement criterion.
NCRL Narrow channel refinement level.
CNC Characteristic number of cells across a narrow channel.
NCHmin The minimum height of narrow channels.
NCHmax The maximum height of narrow channels.
SDRC Square difference refinement criterion.
4
Validation Examples
Introduction
A series of calculation examples presented below validate the ability of
COSMOSFloWorks to predict the essential features of various flows, as well as to solve
conjugate heat transfer problems (i.e. flow problems with heat transfer in solids). In order
to perform the validation accurately and to present clear results, which the user can check
independently, simple examples have been selected. For each of the following examples,
exact analytical or well-documented experimental results exist. Each of the examples
focus on one or two particular physical phenomena such as: laminar flows with and
without heat transfer, turbulent flows including vortex development, boundary-layer
separation and heat transfer, and compressible gas flows with shock and expansion waves.
Therefore, these examples validate the ability of COSMOSFloWorks to predict
fundamental flow features accurately. The accuracy of predictions can be extrapolated to
typical industrial examples (encountered every day by design engineers using
COSMOSFloWorks), which may include a combination of the above-mentioned physical
phenomena and geometries of arbitrary complexity.
Chapter Introduction
4-2
1 Flow through a Cone Valve
Let us see how COSMOSFloWorks predicts incompressible turbulent 3D flows in a 3D
cone valve taken from Ref.14 (the same in Ref.2) and having a complex flow passage
geometry combining 3D sudden contractions and 3D sudden expansions at different
turning angles (Fig.1.1). Following the Refs.2 and 14 recommendations on determining
a valves hydraulic resistance correctly, i.e. to avoid any valve-generated flow
disturbances at the places of measuring the flow total pressures upstream and downstream
of the valve, the inlet and outlet straight pipes of the same diameter D and of enough
length (we take 7D and 17D) are connected to the valve, so constituting the experimental
rig model (see Fig.1.2). As in Ref.14, a water flows through this model. Its temperature of
293.2 K and fully developed turbulent inlet profile (see Ref.1) with mass-average velocity
U 0.5 m/s (to yield the turbulent flows Reynolds number based on the pipe diameter
Re
D
= 10
5
) are specified at the model inlet, and static pressure of 1 atm is specified at the
model outlet.
The corresponding model used for these predictions is shown in Fig.1.2. The valves
turning angle, , is varied in the range of 055
o
(the valve becomes closed at = 8230).
The flow predictions performed with COSMOSFloWorks are validated by comparing the
valves hydraulic resistance,
v
, and the dimensionless coefficient of torque M (see
Fig.1.1) acting on the valve, m, to the experimental data of Ref.14 (Ref.2).
Fig.1.1 The cone valve under consideration: D = 0.206 m, D
ax
= 1.515D, = 1340.
Fig.1.2 The model for calculating the 3D flow in the cone valve.
Outlet static pressure
P = 101325 Pa
Inlet velocity
profile
Chapter Flow through a Cone Valve
4-4
Since Ref.14 presents the valves hydraulic resistance (i.e. due to the flow obstacle, which
is the valve),
v
, whereas the flow calculations in the model (as well as the experiments on
the rig) yield the total hydraulic resistance including both
v
and the tubes hydraulic
resistance due to friction,
f
, i.e. =
v
+
f
, then, to obtain
v
from the flow predictions
(as well as from the experiments),
f
is calculated (measured in the experiments)
separately, at the fully open valve ( = 0); then
v
= -
f
.
In accordance with Ref.14, both and
f
are defined as (P
o inlet
- P
o outlet
)/(U
2
/2), where
P
o inlet
and P
o outlet
are the flow total pressures at the models inlet and outlet, accordingly,
is the fluid density. The torque coefficient is defined as m = M/[D
3
(U
2
/2)(1+
v
)],
where M is the torque trying to slew the valve about its axis (vertical in the left picture in
Fig.1.1) due to a non-uniform pressure distribution over the valves inner passage
(naturally, the valves outer surface pressure cannot contribute to this torque). M is
measured directly in the experiments and is integrated by COSMOSFloWorks over the
valves inner passage.
The COSMOSFloWorks predictions have been performed at result resolution level of 5
with manually setting the minimum gap size to the valves minimum passage in the Y = 0
plane and the minimum wall thickness to 3 mm (to resolve the valves sharp edges).
COSMOSFloWorks has predicted
f
= 0.455,
v
shown in Fig.1.3, and m shown in Fig.1.4.
It is seen that the COSMOSFloWorks predictions well agree with the experimental data.
This cone valve's 3D vortex flow pattern at = 45
o
is shown in Fig.1.5 by flow
trajectories colored by total pressure. The corresponding velocity contours and vectors at
the Y = 0 plane are shown in Fig.1.6.
The predictions have been performed on a computational mesh consisting of about
150000 cells and have required about 270 MB memory and about 7 hours to run on a
1GHz PIII platform for each specified .

0.1
1
10
100
15 20 25 30 35 40 45 50 55
() () () ()

v
Experimental data
Calculation
Fig.1.3 Comparison of the COSMOSFloWorks predictions with the Ref.14 experimental data on the
cone valves hydraulic resistance versus the cone valve turning angle.
Fig.1.4 Comparison of the COSMOSFloWorks predictions with the Ref.14 experimental data on the
cone valves torque coefficient versus the cone valve turning angle.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
15 20 25 30 35 40 45 50 55
()
m
Experimental data
Calculation
Chapter Flow through a Cone Valve
4-6
Fig.1.5 Flow trajectories colored by total pressure at = 45
o
.
Fig.1.6 The cone valves velocity contours and vectors at = 45
o
.
2 Laminar Flows Between Two Parallel Plates
Let us consider two-dimensional (plane) steady-state laminar flows of Newtonian,
non-Newtonian, and compressible liquids between two parallel stationary plates spaced at
a distance of 2h (see Fig.2.1).
In the case of Newtonian and non-Newtonian liquids the channel has a 2h = 0.01 m height
and a 0.2 m length, the inlet for these liquids have standard ambient temperature (293.2 K)
and a uniform inlet velocity profile of u
average
= 0.01 m/s (entrance disturbances are
neglected). The inlet pressure is not known beforehand, since it will be obtained from the
calculations in accordance with the specified channel exit pressure of 1 atm. (The fluids
pass through the channel due to a pressure gradient.)
Since the Reynolds number based on the channel height is equal to about Re
2h
=100, the
flow is laminar.
As for the liquids, let us consider water as a Newtonian liquid and four non-Newtonian
liquids, having the same density of 1000 kg/m
3
, the same specific heat of 4200 J/(kgK)
and the same thermal conductivity of 10 W/(mK), but obeying the different
non-Newtonian liquid models available in COSMOSFloWorks.
The considered non-Newtonian liquids' models and their governing characteristics are
presented in Table 4.1. These models are featured by the function connecting the flow
shear stress () with the flow shear rate ( ), i.e. , or, following Newtonian
liquids, the liquid dynamic viscosity () with the flow shear rate ( ), i.e. :
1 the Herschel-Bulkley model: , where K is the consistency
coefficient, n is the power-law index, and is the yield stress (this model's n = 1 case
is the Bingham model);
2 the power-law model: , i.e., , is the Herschel-Bulkley
model's = 0 case;
3 the Carreau model: , , where
is the liquid dynamic viscosity at infinite shear rate, i.e. the minimum dynamic
viscosity, is the liquid dynamic viscosity at zero shear rate, i.e. the maximum
Fig.2.1 Flow between two parallel plates.
u
average
or m
( ) f =
( ) =
( )
o
n
K + =
o
( )
n
K = ( )
1
=
n
K
o
=
( ) ( ) [ ]
( ) 2 / 1
2
1
1

+ + =
n
o
K

Chapter Laminar Flows Between Two Parallel Plates

4-8
dynamic viscosity, K
1
is the time constant, n is the power-law index (this model is a
smooth version of the power-law model).
In accordance with the well-known theory presented in Ref.1, after some entrance length,
the flow profile u(y) becomes fully developed and invariable. It can be determined from
the
Navier-Stokes x-momentum equation corresponding to this
case in the
coordinate system shown in Fig.2.1 (y = 0 at the channel's center plane, is the
longitudinal
pressure gradient along the channel, in the flow under consideration).
As a result, the fully developed u(y) profile for a Newtonian fluid has the following form:
u(y) = - ,
where - fluid dynamic viscosity, - half height of the channel,
,
where u
average
is the flow's mass-average velocity defined as the flow's volume flow rate
divided by the flow passage cross-sectional area.
Non-Newtonian liquid No. 1 2 3 4
Non-Newtonian liquid model Herschel-Bulkley Binghem Power-law Carreau
Consistency coefficient, K (Pas
n
) 0.001 0.001 0.001 -
Power-law index, n 1.5 1 0.6 0.4
Yield stress, o
(Pa) 0.001 0.001 - -
Minimum dynamic viscosity,

(Pas)
- - - 10
-4
Maximum dynamic viscosity, o
(Pas)
- - - 10
-3
Time constant, K
1
(s) - - - 1
h
const
dy
d
dx
dP
w

= = =
dx
dP
dy
du
=
) (
2
1
2 2
y h
dx
dP
2
3
h
u
dx
dP
average
=

For a non-Newtonian liquid described by the power-law model the fully developed u(y)
profile and the corresponding pressure gradient can be determined from the following
formulae:
, .
For a non-Newtonian liquid described by the Herschel-Bulkley model the fully developed
u(y) profile can be determined from the following formulae:
at ,
at ,
where the unknown wall shear stress is determined numerically by solving the
nonlinear equation
,
e.g. with the Newton method, as described in this validation. The corresponding pressure
gradient is determined from .
For a non-Newtonian liquid described by the Carreau model the fully developed u(y)
profile can not be determined analytically in an explicit form, so in this validation example
it is obtained by solving the following parametric equation:
,
( )
|
|
|
.
|
\
|
|
.
|
\
|
+
+
=
+
n
n
average
h
y
n
n
u y u
1
1
1
1 2
n
average
n
n
h
u
h
K
dx
dP
|
|
.
|
\
| +
=
1 2
( ) ( )
n
n
o w
w
n
n
n
K
h
u y u
1
/ 1
max
1
+
+
= =
h y
w
o
<
( )
|
|
|
|
|
.
|
\
|
|
|
|
|
.
|
\
|
=
+
n
n
o w
o w
h
y
u y u
1
max
1

h y
w
o
>
w
( )
|
|
.
|
\
|
+

+
=
+
w
o w
n
n
o w
w
n
average
n
n
n
n
K
h
u

1 2
1
1
1
/ 1
h dx
dP
w
=
( ) ( )
2 / ) 1 (
2 2
0
1 ) (

+ + =
n
w
p p
h
y
( )
( 1) / 2
2 2 2 2 0 0
max
2 2
( ) ( ) 1
1
2 ( 1) ( 1)
n
w w w
h h h
u u p p np
n n

| |
= +
|
+ +
\ .
4-10
where p is a free parameter, varied within the p
max
range,
,
.
The p value is varied to satisfy .
The corresponding pressure gradient is equal to .
The SolidWorks model for the 2D calculation is shown in Fig.2.2. The boundary
conditions are specified as mentioned above and the initial conditions coincide with the
inlet boundary conditions. The results of the calculations performed with
COSMOSFloWorks at result resolution level 5 are presented in Figs.2.3 - 2.8. The channel
exit u(y) profile and the channel P(x) profile were obtained along the sketches shown by
green lines in Fig.2.2.
( ) |
.
|
\
|
+ + =

2 / ) 1 (
2
max
2
0 max
1 ) (
n
w
p p
( )
( 1) / 2
2 2 2 2 0 0
max max max max
2 2
( ) ( ) 1
1
2 ( 1) ( 1)
n
w w w
h h h
u p p np
n n

| |
= + + +
|
+ +
\ .
max
0 0
h p
average
dy
hu udy u dp
dp
= =

h dx
dP
w
=
Fig.2.2 The model for calculating 2D flow between two parallel plates with COSMOSFloWorks.
Fig.2.3 The water and liquid #3 velocity profiles u(y) at the channel exit.
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
-0.005 -0.003 -0.001 0.001 0.003 0.005
Y, m
U, m/s
Water, theory
Water, calculation
Liquid #3, theory
Liquid #3, calculation
From Figs.2.4, 2.6, and 2.8 you can see that for all the liquids under consideration, after
some entrance length of about 0.03 m, the pressure gradient governing the channel
pressure loss becomes constant, nearly the same as predicted theoretically. From Figs.2.3,
2.5, and 2.7 you can see that the fluid velocity profiles at the channel exit obtained from
the calculations are close to the theoretical profiles.
Fig.2.4 The water and liquid #3 longitudinal pressure change along the channel, P(x).
101325
101325.05
101325.1
101325.15
101325.2
101325.25
101325.3
0 0.05 0.1 0.15 0.2
X, m
P, Pa
Water, theory
Water, calculation
Liquid #3, theory
Liquid #3,
calculation
Fig.2.5 The liquids #1 and #2 velocity profiles u(y) at the channel exit.
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
-0.005 -0.003 -0.001 0.001 0.003 0.005
Y, m
U, m/s
Liquid #1, theory
Liquid #1,
calculation
Liquid #2, theory
Liquid #2,
calculation
4-12
In the case of compressible liquids the channel has the 2h = 0.001 m height and the 0.5 m
length, at its inlet these liquids have standard ambient temperature (293.2 K) and a
uniform inlet velocity profile corresponding to the specified mass flow rate of
= 0.01 kg/s.
The inlet pressure is not known beforehand, since it will be obtained from the calculations
as providing the specified mass flow rate under the specified channel exit pressure of 1
atm. (The fluids pass through the channel due to the pressure gradient).
Fig.2.6 The liquids #1 and #2 longitudinal pressure change along the channel, P(x).
101325
101325.1
101325.2
101325.3
101325.4
101325.5
101325.6
0 0.05 0.1 0.15 0.2
X, m
P, Pa
Liquid #1, theory
Liquid #1,
calculation
Liquid #2, theory
Liquid #2,
calculation
Fig.2.7 The liquid #4 velocity profile u(y) at the channel exit.
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
-0.005 -0.003 -0.001 0.001 0.003 0.005
Y, m
U, m/s
Liquid #4, theory
Liquid #4,
calculation
m
Let us consider two compressible liquids whose density obeys the following laws:
the power law:
, where
0
, P
0
, B and n are specified:
0
is the liquid's density
under the reference pressure P
0
, B and n are constants,
the logarithmic law:
, where
0
, P
0
, B and C are specified:
0
is the liquid's
density under the reference pressure P
0
, B and C are
constants.
In this validation example these law's parameters values have been specified as
0
=10
3

kg/m
3
, P
0
= 1 atm, B = 10
7
Pa, n = 1.4, C = 1, and these liquids have the 1Pas dynamic
viscosity.
Since this channel is rather long, the pressure gradient along it can be determined as
, where is the liquids' dynamic viscosity, is the liquid mass
flow rate, S is the channel's width, is the liquid density.
Therefore, substituting the compressible liquids' (P) functions, we obtain the following
equations for determining P(x) (along the channel):
for the power-law liquid:
its solution is
Fig.2.8 The liquid #4 longitudinal pressure change along the channel, P(x).
101325
101325.02
101325.04
101325.06
101325.08
101325.1
101325.12
101325.14
0 0.05 0.1 0.15 0.2
X, m
P, Pa
Liquid #4, theory
Liquid #4,
calculation
0 0
n
P B
P B
| |
+
=
|
+
\ .
0
1 *ln
B P
C
B P
=
+
+
2
3 P m
x h S
m
1/
0
2
0
3
n
P B P m
x h S P B
+ | |
=
|
+
\ .

4-14
, where C
1
is a constant
determined from the boundary
conditions;
for the logarithmic-law liquid: , this equation
is solved numerically.
Both the theoretical P(x) distributions and the corresponding distributions computed
within COSMOSFloWorks on a 5*500 computational mesh are presented in Figs.2.10 and
2.9. It is seen that the COSMOSFloWorks calculations agree with the theoretical
distributions.
( )
1 1
1
0 1 2
0
3
( )
1
n n
n m
P B P B x C
n h S
+
+ = + +
+
2
0 0
3
1 ln
P m P B
C
x h S P B
| |
+
=
+
\ .
Fig.2.9 The logarithmic-law compressible liquid's longitudinal pressure change along the channel,
P(x).
0
2000000
4000000
6000000
8000000
10000000
12000000
14000000
16000000
18000000
0 0.1 0.2 0.3 0.4 0.5
X, m
P, Pa
LN, theory
LN, calculation
Fig.2.10 The power-law compressible liquid's longitudinal pressure change along the channel, P(x).
0
5000000
10000000
15000000
20000000
25000000
30000000
35000000
0 0.1 0.2 0.3 0.4 0.5
X, m
P, Pa
Power, theory
Power, calculation
3 Laminar and Turbulent Flows in Pipes
Having been encouraged by the 2D results presented in the previous example, let us now
see how the 3D flow through a straight pipe is predicted. Let us consider water (at
standard 293.2 K temperature) flowing through a long straight pipe of circular cross
section of d = 0.1 m (see Fig.3.1). At the pipe inlet the velocity is uniform and equal to
u
inlet
. At the pipe outlet the static pressure is equal to 1 atm.
The SolidWorks model used for all the 3D pipe flow calculations is shown in Fig.3.2. The
initial conditions have been specified to coincide with the inlet boundary conditions. The
computational domain is reduced to domain (Z 0, Y 0) with specifying the flow
symmetry planes at Z = 0 and Y = 0.
According to theory (Ref.1), the pipe flow velocity profile changes along the pipe until it
becomes a constant, fully developed profile at a distance of L
inlet
from the pipe inlet.
According to Ref.1, L
inlet
is estimated as:
Fig.3.1 Flow in a pipe.
u
inlet
Fig.3.2 The SolidWorks model for calculating 3D flow in a pipe with COSMOSFloWorks.
Chapter Laminar and Turbulent Flows in Pipes
4-16
where Re
d
= Ud/ is the Reynolds number based on the pipe diameter d, U is the
mass-average flow velocity, is the fluid density, and is the fluid dynamic viscosity.
Therefore, to provide a fully developed flow in the pipe at Re
d
under consideration, we
will study the cases listed in Table.3.1. Here, L
pipe
is the overall pipe length. All the
COSMOSFloWorks predictions concerning the fully developed pipe flow characteristics
are referred to the pipe section downstream of the inlet section.
*) the lengths in brackets are for the rough pipes.
The flow regime in a pipe can be laminar, turbulent, or transitional, depending on Re
d
.
According to Ref.1, Re
d
= 4000 is approximately the boundary between laminar pipe flow
and turbulent one (here, the transitional region is not considered).
Theory (Refs. 1 and 4) states that for laminar fully developed pipe flows
(Hagen-Poiseuille flow) the velocity profile u(y) is invariable and given by:
where R is the pipe radius, and dP/dx is the longitudinal pressure gradient along the pipe,
which is also invariable and equal to:
The COSMOSFloWorks predictions of dP/dx and u(y) of the laminar fully developed pipe
flow at Re
d
= 100 performed at result resolution level 6 are presented in Fig.3.3 and
Fig.3.4. The presented predictions relate to the smooth pipe, and the same ones not
presented here have been obtained for the case of the rough tube with relative sand
roughness of k/d=0.20.4 %, that agrees with theory (Ref.1).

Table.3.1 Pipe inlet velocities and lengths.
Re
d
u
inlet
, m/s L
inlet
, m L
pipe
, m
0.1 10
-6
0.3 0.45
100 0.001 0.3 0.45
1000 0.01 3 4.5
10
4
0.1 4 (5)* 6 (10)*
10
5
1 4 (5)* 6 (10)*
10
6
10 4 (5)*6 6 (10)*
=
=
=
6
10 ... 6000 Re , 40
6000 ... 2500 Re , 100
2500 ... 1 . 0 Re , 3 Re 03 . 0
d
d
d d
d
d
d d
L
inlet
=
) (
4
1
) (
2 2
y R
dx
dP
y u =
,
2
8
R
u
dx
dP
inlet
=

.

From Fig.3.3 one can see that after an entrance length of about 0.15 m the pressure
gradient predicted by COSMOSFloWorks coincides with the one predicted by theory.
Therefore, the prediction of pipe pressure loss is excellent. As for local flow features, from
Fig.3.4 one can see that the fluid velocity profiles predicted at the pipe exit are rather close
to the theoretical profile.
101325.0000
101325.0002
101325.0004
101325.0006
101325.0008
101325.0010
101325.0012
101325.0014
101325.0016
101325.0018
101325.0020
0 0.1 0.2 0.3 0.4 0.5
X (m)
Pressure (Pa)
Theory
Calculation
Fig.3.3 The longitudinal pressure change (pressure gradient) along the pipe at Re
d
100.
Fig.3.4 The fluid velocity profile at the pipe exit for Re
d
100.
0
0.0005
0.001
0.0015
0.002
0.0025
0 0.01 0.02 0.03 0.04 0.05
Y (m)
Velocity (m/s)
Theory
Calculation
4-18
The velocity profile and longitudinal pressure distribution in a smooth pipe at Re
d
= 10
5
,
i.e., in a turbulent pipe flow regime, predicted by COSMOSFloWorks at result resolution
level 6 are presented in Figs.3.5 and 3.6 and compared to theory (Ref.1, the Blasius law of
pressure loss, the 1/7-power velocity profile).
Then, to stand closer to engineering practice, let us consider the COSMOSFloWorks
predictions of the pipe friction factor used commonly and defined as:
where L is length of the pipe section with the fully developed flow, along which pressure
loss P is measured.
101200
101400
101600
101800
102000
102200
102400
0 2 4 6 8 10
X (m)
Pressure (Pa)
Theory
Calculation
Fig.3.5 The longitudinal pressure change (pressure gradient) along the pipe at Re
d
= 10
5
.
L
d
u
P
f
inlet
=
2
2
,
In Figs. 3.7 and 3.8 (scaled up) you can see the COSMOSFloWorks predictions performed
at result resolution level 5 for the smooth pipes in the entire, both laminar and turbulent,
Re
d
range and compared with the theoretical and empirical values determined from the
following formulae which are valid for fully-developed flows in smooth pipes (Refs.1, 2,
and 4):
It can be seen that the friction factor values predicted for smooth pipes, especially in the
laminar region, are fairly close to the theoretical and empirical curve.
As for the friction factor in rough pipes, the COSMOSFloWorks predictions for the pipes
having relative wall roughness of k/d=0.4% (k is the sand roughness) are presented and
compared with the empirical curve for such pipes (Refs.1, 2, and 4) in Fig.3.8. The
underprediction error does not exceed 13%.
Additionally, in the full accordance with theory and experimental data the
COSMOSFloWorks predictions show that the wall roughness does not affect the friction
factor in laminar pipe flows.
Fig.3.6 The fluid velocity profile at the pipe exit at Re
d
= 10
5
.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.01 0.02 0.03 0.04 0.05
Y (m)
Velocity (m/s)
Theory
Calculation
1 4 5
2
5
64
, Re 2300
Re
0.316 Re , 4000 Re 10
, Re 10
d
d
d d
d
d
laminar flows,
f turbulent flows,
Re
1.8 log turbulent flows
6.9
= < <
| |
\ .

4-20

Fig.3.7 The friction factor predicted by COSMOSFloWorks for smooth pipes in comparison with the
theoretical and empirical data (Refs.1, 2, and 4).
1.E-03
1.E-02
1.E-01
1.E+00
1.E+01
1.E+02
1.E+03
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Re
d
Friction factor
Smooth pipes,
theoretical and
empirical data
Calculation
0.010
0.100
1.E+03 1.E+04 1.E+05 1.E+06
Re
d
Friction Factor
Smooth wall,
theoretical and
empirical data
Smooth wall,
calculation
Rough wall,
k/d=0.4%, empirical
data
Rough wall,
k/d=0.4%,
calculation
Fig.3.8 The friction factor predicted by COSMOSFloWorks for smooth and rough pipes in
comparison with the theoretical and empirical data (Refs.1, 2, and 4).
4 Flows Over Smooth and Rough Flat Plates
In the previous example we have presented a validation for laminar and turbulent flows in
smooth and rough pipes for a wide range of Reynolds number. We will now consider
uniform flows over smooth and rough flat plates with laminar and turbulent boundary
layers, so COSMOSFloWorks predictions of a flat plate drag coefficient are validated.
We consider the boundary layer development of incompressible uniform 2D water flow
over a flat plate of length L (see Fig.4.1). The boundary layer develops from the plates
leading edge lying at the upstream computational domain boundary. The boundary layer at
the leading edge is considered laminar. Then, at some distance from the plate leading edge
the boundary layer automatically becomes turbulent (if this distance does not exceed L).
The SolidWorks model is shown in Fig.4.2. The problem is solved as internal in order to
avoid a conflicting situation in the corner mesh cell where the external flow boundary and
the model wall intersect. In the internal flow problem statement, to avoid any influence of
the upper model boundary or wall on the flow near the flat plate, the ideal wall boundary
condition has been specified on the upper wall. The plates length is equal to 10 m, the
channels height is equal to 2 m, the walls thickness is equal to 0.5 m.

Water
L
Fig.4.1 Flow over a flat plate.
Fig.4.2 The model for calculating the flow over the flat plate with COSMOSFloWorks.
Outlet
Rough or smooth wall
Ideal wall Inlet
Chapter Flows Over Smooth and Rough Flat Plates
4-22
To solve the problem, an incoming uniform water flow of a certain velocity (see below),
temperature of 293.2 K, turbulence intensity of 1%, and turbulence length of 0.01 m is
specified at the channel inlet, whereas the water static pressure of 1 atm is specified at the
channel outlet.
The flow computation is aimed at predicting the flat plate drag coefficient, defined as (see
Refs. 1 and 4):
where F is the plate drag force, A is the plate surface area, is fluid density, and V is the
fluid velocity.
According to Refs.1 and 4, the plate drag coefficient value is governed by the Reynolds
number, based on the distance L from the plates leading edge (Re
L
= VL/, where is
the fluid density, V is the incoming uniform flow velocity, and is fluid dynamic
viscosity), as well as by the relative wall roughness L/k, where k is the sand roughness. As
a result, Refs.1 and 4 give us the semi-empirical flat plate C
D
(Re
L
) curves obtained for
different L/k from the generalized tubular friction factor curves and presented in Fig.4.3
(here, k). If the boundary layer is laminar at the plates leading edge, then the wall
roughness does not affect C
D
until the transition from the laminar boundary layer to the
turbulent one, i.e., the C
D
(Re
L
) curve is the same as for a hydraulically smooth flat plate.
The transition regions boundaries depend on various factors, the wall roughness among
them, but here the theoretical transition region is shown for a hydraulically smooth flat
plate. The transition region's boundary corresponding to fully turbulent flows (i.e., at the
higher Re
L
) is marked in Fig.4.3 by a dashed line. At the higher Re
L
, the semi-empirical
theoretical curves have flat parts along which Re
L
does not affect C
D
at a fixed wall
roughness. These flat parts of the semi-empirical theoretical curves have been obtained by
a theoretical scaling of the generalized tubular friction factor curves to the flat plate
conditions under the assumption of a turbulent boundary layer beginning from the flat
plate leading edge.
To validate the COSMOSFloWorks flat plate C
D
predictions within a wide Re
L
range, we
have varied the incoming uniform flow velocity at the model inlet to obtain the Re
L
values
of 10
5
, 310
5
, 10
6
, 310
6
, 10
7
, 310
7
, 10
8
, 310
8
, 10
9
.To validate the wall roughness
influence on C
D
, the wall roughness k values of 0, 50, 200, 10
3
, 510
3
, 10
4
m have been
considered. The COSMOSFloWorks calculation results obtained at result resolution level
5 and compared with the semi-empirical curves are presented in Fig.4.3.
As you can see from Fig.4.3, C
D
(Re
L
) of rough plates is somewhat underpredicted by
COSMOSFloWorks in the turbulent region, at L/k 1000 the C
D
(Re
L
) prediction error
does not exceed about 12%.
A
V
F
C
d
2
2
=
,
Fig.4.3 The flat plate drag coefficient predicted with COSMOSFloWorks for rough and hydraulically
smooth flat plates in comparison with the semi-empirical curves (Refs.1 and 4).
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
1.00E+05 1.00E+06 1.00E+07 1.00E+08 1.00E+09
Re
C
D
L/k=1e3
L/k=2e3
L/k=1e4
L/k=5e4
L/k=2e5
Smooth
Chapter Flows Over Smooth and Rough Flat Plates
4-24
5 Flow in a 90-degree Bend Square Duct
In the previous examples we considered laminar and turbulent flows over flat plates and in
straight pipes. Let us now see how COSMOSFloWorks predicts 3-dimensional
incompressible flow in a 90
o
-bend square duct.
Following Ref.8, we will consider a steady-state flow of water (at 293.2 K inlet
temperature and U
inlet
= 0.0198 m/s inlet uniform velocity) in a 4040 mm square
cross-sectional duct having a 90-angle bend with r
i
= 72 mm inner radius (r
o
= 112 mm
outer diameter accordingly) and attached straight sections of 1.8 m upstream and 1.2 m
downstream (see Fig.5.1). Since the flow's Reynolds number, based on the duct's
hydraulic diameter (D=40 mm), is equal to Re
D
= 790, the flow is laminar.
The COSMOSFloWorks prediction was performed at result resolution level 7.
The predicted dimensionless (divided by U
inlet
) velocity profiles are compared in Figs.5.2,
5.3 with the ones measured with a laser-Doppler anemometry at the following duct cross
sections: X
H
= -5D, -2.5D, 0 (or =0
o
) and at the =30
o
, 60
o
, 90
o
bend sections. The z
and r directions
are represented by coordinates and , where z
1/2
= 20 mm. The dimensionless
velocity isolines (with the 0.1 step) at the duct's = 60
o
and 90
o
sections, both measured in
Ref.8 and predicted with COSMOSFloWorks, are shown in Figs.5.4 and 5.5.
It is seen that the COSMOSFloWorks predictions are close to the Ref.8 experimental data.
Fig.5.1 The 90
o
-bend square duct's configuration indicating the velocity measuring
stations and the dimensionless coordinates used for presenting the velocity profiles.
o
i o
r r
r r
1/ 2
z
z
Chapter Flow in a 90-degree Bend Square Duct
4-26
Fig.5.2 The duct's velocity profiles predicted by COSMOSFloWorks (red lines) in comparison with
the Ref.8 experimental data (circles).

Fig.5.3 The duct's velocity profiles predicted by COSMOSFloWorks (blue lines) in comparison
with the Ref.8 experimental data (circles).
z/z1/2=0.5 z/z1/2=0
z/z
1/2
=0.5 z/z
1/2
=0
Chapter Flow in a 90-degree Bend Square Duct
4-28
Fig.5.4 The duct's velocity isolines at the = 60
o
section predicted by COSMOSFloWorks (left)
in comparison with the Ref.8 experimental data (right).
Fig.5.5 The duct's velocity isolines at the =90
o
section predicted by COSMOSFloWorks
(left) in comparison with the Ref.8 experimental data (right).
6 Flows in 2D Channels with Bilateral and Unilateral Sudden Expansions
In this example we will consider both turbulent and laminar incompressible steady-state
flows through 2D (plane) channels with bilateral and unilateral sudden expansions and
parallel walls, as shown in Figs.6.1 and 6.2. At the 10 cm inlet height of the
bilateral-sudden-expansion channels a uniform water stream at 293.2 K and 1 m/s is
specified. The Reynolds number is based on the inlet height and is equal to Re = 10
5
,
therefore (since Re > 10
4
) the flow is turbulent. At the 30 mm height inlet of the
unilateral-sudden-expansion channel an experimentally measured water stream at 293.2 K
and 8.25 mm/s mean velocity is specified, so the Reynolds number based on the inlet
height is equal to Re = 250, therefore the flow is laminar. In both channels, the sudden
expansion generates a vortex, which is considered in this validation from the viewpoint of
hydraulic loss in the bilateral-expansion channel (compared to Ref.2) and from the
viewpoint of the flow velocity field in the unilateral-expansion channel (compared to
Ref.13).
In accordance with Ref.2, the local hydraulic loss coefficient of a bilateral sudden
expansion (the so-called total pressure loss due to flow) for a turbulent (Re > 10
4
) flow
with a uniform inlet velocity profile depends only on the expansion area ratio and is
determined from the following formula:
where A
0
and A
1
are the inlet and outlet cross sectional areas respectively, P
0
and P
1
are
the inlet and outlet total pressures, and u
0
2
/2 is the inlet dynamic head.
Fig.6.1 Flow in a 2D (plane) channel with a bilateral sudden expansion.
outlet
X
Y
inlet
Water
1 m/s
Fig.6.2 Flow in a 2D (plane) channel with a unilateral sudden expansion.
400 mm 20 mm
3
0

m
m
h

=

1
5

Y
X
0
Inlet experimental velocity profile
recirculation
Lr
2
1
0
2
0
1 0
1
2
|
|
.
|
\
|
=
=
A
A
u
P P
s
,
Chapter Flows in 2D Channels with Bilateral and Unilateral Sudden Expansions
4-30
In a real sudden expansion the flow hydraulic loss coefficient is equal to =
f
+
s
, where
f
is the friction loss coefficient. In order to exclude
f
from our comparative analysis, we
have imposed the ideal wall boundary condition on all of the channel walls.
In this validation example the channel expansion area ratios under consideration are: 1.5,
2.0, 3.0, and 6.0. To avoid disturbances at the outlet due to the sudden expansion, the
channel length is 10 times longer than its height. The 1 atm static pressure is specified at
the channel outlet.
The
s
values predicted by COSMOSFloWorks at result resolution level 8 for different
channel expansion area ratios A
0
/A
1
are compared to theory in Fig.6.3.
From Fig.6.3, one can see that COSMOSFloWorks overpredicts
s
by about 4.5...7.9 %.
The model used for the unilateral-sudden-expansion channel's flow calculation is shown
in Fig.6.4. The channel's inlet section has a 30 mm height and a 20 mm length. The
channel's expanded section (downstream of the 15 mm height back step) has a 45 mm
height and a 400 mm length (to avoid disturbances of the velocity field compared to the
experimental data from the channel's outlet boundary condition). The velocity profile
measured in the Ref.13 at the corresponding Re
h
= 125 (the Reynolds number based on
the step height) is specified as a boundary condition at the channel inlet. The 10
5
Pa static
pressure is specified at the channel outlet.
Fig.6.3 Comparison of COSMOSFloWorks calculations to the theoretical values (Ref.2) for the
sudden expansion hydraulic loss coefficient versus the channel expansion area ratio.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.2 0.4 0.6 0.8 1
A
0
/A
1

s
Theory
Calculation

The flow velocity field predicted by COSMOSFloWorks at result resolution level 8 is
compared in Figs.6.5, 6.6, and 6.7 to the values measured in Ref.13 with a laser
anemometer. The flow X-velocity (u/U, where U = 8.25 mm/s) profiles at several X =
const (-20 mm, 0, 12 mm, 150 mm) cross sections are shown in Fig.6.5. It is seen that
the predicted flow velocity profiles are very close to the experimental values both in the
main stream and in the recirculation zone. The recirculation zone's characteristics, i.e. its
length L
R
along the channel's wall, (plotted versus the Reynolds number Re
h
based on the
channel's step height h, where Re
h
=125 for the case under consideration), the separation
streamline, and the vortex center are shown in Figs.6.6 and 6.7. It is seen that they are very
close to the experimental data.
Fig.6.4 The SolidWorks model for calculating the 2D flow in the
unilateral-sudden-expansion channel with COSMOSFloWorks.
Inlet velocity profile
Chapter Flows in 2D Channels with Bilateral and Unilateral Sudden Expansions
4-32
As one can see, both the integral characteristics (hydraulic loss coefficient) and local
values (velocity profiles and recirculation zone geometry) of the 2D sudden expansion
channels' turbulent and laminar flows under consideration are adequately predicted by
COSMOSFloWorks.
Fig.6.5 The unilateral-sudden-expansion channel's velocity profiles predicted by COSMOSFloWorks
(red lines) in comparison with the Ref.13 experimental data (black lines with dark circles).
0
2
4
6
8
10
12
0 50 100 150 200 250
Reh
L
/
h

-

r
e
c
i
r
c
u
l
a
t
i
o
n

z
o
n
e

l
e
n
g
t
h
Fig.6.6 The unilateral-sudden-expansion channel's recirculation zone length predicted by
COSMOSFloWorks (red square) in comparison with the Ref.13 experimental data (black signs).
0
5
10
15
0 0.2 0.4 0.6 0.8 1
separation streamlines,
calculation
vortex center, calculation
Fig.6.7 The unilateral-sudden-expansion channel recirculation zone's separation streamlines and
vortex center, both predicted by COSMOSFloWorks (red lines and square) in comparison with the
Ref.13 experimental data (black signs).
7 Flow over a Circular Cylinder
Until now we have only considered internal flows. Let us now consider an external
incompressible flow example. In this example, water at a temperature of 293.2 K and a
pressure of 1 atm flows over a cylinder of 0.01 m or 1 m diameter. The flow pattern of this
example substantially depends on the Reynolds number which is based on the cylinder
diameter. At low Reynolds numbers (4 < Re < 60) two steady vortices are formed on the
rear side of the cylinder and remain attached to the cylinder, as it is shown schematically
in Fig.7.1 (see Refs.3).
At higher Reynolds numbers the flow becomes unstable and a von Karman vortex street
appears in the wake past the cylinder. Moreover, at Re > 60100 the eddies attached to
the cylinder begin to oscillate and shed from the cylinder (Ref.3). The flow pattern is
shown schematically in Fig.7.2.
To calculate the 2D flow (in the X-Y plane) with COSMOSFloWorks, the model shown in
Fig.7.3 has been created. The cylinder diameter is equal to 0.01 m at Re 10
4
and 1 m at
Re>10
4
. The incoming stream turbulence intensity has been specified as 1%. To take the
flows physical instability into account, the flow has been calculated by
COSMOSFloWorks using the time-dependent option. All the calculations have been
performed at result resolution level 7.
Fig.7.1 Flow past a cylinder at low Reynolds numbers (4 < Re < 60).
Fig.7.2 Flow past a cylinder at Reynolds numbers Re > 60100.
Chapter Flow over a Circular Cylinder
4-34
In accordance with the theory, steady flow patterns have been obtained in these
calculations in the low Re region. An example of such calculation at Re=41 is shown in
Fig.7.4 as flow trajectories over and past the cylinder in comparison with a photo of such
flow from Ref.9. It is seen that the steady vortex past the cylinder is predicted correctly.
Fig.7.3 The SolidWorks model used to calculate 2D flow over a cylinder.
Fig.7.4 Flow trajectories over and past a cylinder at Re=41 predicted with COSMOSFloWorks
(above) in comparison with a photo of such flow from Ref.9 (below).
The unsteady vortex shedding from a cylinder at Re > 60..100, yields oscillations of both
drag and lateral forces acting on the cylinder and a von Karman vortex street is formed
past the cylinder. An X-velocity field over and past the cylinder is shown in Fig.7.5. The
COSMOSFloWorks prediction of the cylinder drag and lateral force oscillations'
frequency in the Strouhal number form (Sh = D/(tU), where, D is the cylinder diameter, t
is the period of oscillations, U is the incoming stream velocity) in comparison with
experimental data for Re10
3
is shown in Fig.7.6.
Fig.7.5 Velocity contours of flow over and past the cylinder at Re=140.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 Re
Sh
Fig.7.6 The cylinder flow's Strouhal number predicted with COSMOSFloWorks (red triangles) in
comparison with the experimental data (blue line with dashes, Ref.4).
Chapter Flow over a Circular Cylinder
4-36
The time-averaged cylinder drag coefficient is defined as
where F
D
is the drag force acting on the cylinder, U
2
/2 is the incoming stream dynamic
head, D is the cylinder diameter, and L is the cylinder length. The cylinder drag
coefficient, predicted by COSMOSFloWorks is compared to the well-known C
D
(Re)
experimental data in Fig.7.7.
DL U
F
C
2
D
D
2
1
=
0.1
1
10
100
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07
Re
C
D
Fig.7.7 The cylinder drag coefficient predicted by COSMOSFloWorks (red rhombs) in comparison
with the experimental data (black signs, Ref.3)
8 Supersonic Flow in a 2D Convergent-Divergent Channel
Until now we have only considered incompressible flows, so now we will study a
compressible, supersonic flow.
The first example is a supersonic flow of air in a 2D (plane) convergent-divergent channel
whose schematic is shown in Fig.8.1.
A uniform supersonic stream of air, having a velocity of Mach number M=3, static
temperature of 293.2 K, and static pressure of 1 atm, is specified at the channel inlet
between two parallel walls. In the next convergent section (see Fig.8.2) the stream
decelerates through two oblique shocks shown schematically in Fig.8.1 as lines separating
regions 1, 2, and 3. Since the convergent section has a special shape adjusted to the inlet
Mach number, so the shock reflected from the upper plane wall and separating regions 2
and 3 comes to the section 3 lower wall edge, a uniform supersonic flow occurs in the next
section 3 between two parallel walls. In the following divergent section the supersonic
flow accelerates thus forming an expansion waves fan 4. Finally, the stream decelerates in
the exit channel section between two parallel walls when passing through another oblique
shock.
The SolidWorks model of this 2D channel is shown in Fig.8.3.
Fig.8.1 Supersonic flow in a 2D convergent-divergent channel.
Fig.8.2 Dimensions (in m) of the 2D convergent-divergent channel including a reference line for
comparing the Mach number.
Chapter Supersonic Flow in a 2D Convergent-Divergent Channel
4-38
Since the channel was designed for the inviscid flow of an ideal gas, the ideal wall
boundary condition has been specified and the laminar only flow has been considered
instead of turbulent. The computed Mach number along the reference line and at the
reference points (1-5) are compared with the theoretical values in Fig.8.4.
To obtain the most accurate results possible with COSMOSFloWorks, the calculations
have been performed at result resolution level 8. The predicted Mach number at the
selected channel points (1-5) and along the reference line (see Fig.8.2), are presented in
Table.8.1 and Fig.8.4 respectively.
Table.8.1 Mach number values predicted with COSMOSFloWorks with comparison
to the theoretical values at the reference points.
From Table.8.1 and Fig.8.4 it can be seen that the COSMOSFloWorks predictions are very
close to the theoretical values. In Fig.8.4 one can see that COSMOSFloWorks properly
predicts the abrupt parameter changes when the stream passes through the shock and a fast
parameter change in the expansion fan.
Fig.8.3 The model for calculating the 2D supersonic flow in the 2D convergent-divergent
channel with COSMOSFloWorks.
Point 1 2 3 4 5
X coordinate of point, m 0.0042 0.047 0.1094 0.155 0.1648
Y coordinate of point, m 0.0175 0.0157 0.026 0.026 0.0157
Theoretical M 3.000 2.427 1.957 2.089 2.365
COSMOSFloWorks
prediction of M 3.000 2.429 1.965 2.106 2.380
Prediction error,% 0.0 0.1 0.4 0.8 0.6
To show the full flow pattern, the predicted Mach number contours of the channel flow are
shown in Fig.8.5.
This example illustrates that COSMOSFloWorks is capable of capturing shock waves
with a high degree of accuracy. This high accuracy is possible due to the
COSMOSFloWorks solution adaptive meshing capability. Solution adaptive meshing
automatically refines the mesh in regions with high flow gradients such as shocks and
expansion fans.
Fig.8.4 Mach number values predicted with COSMOSFloWorks along the reference line (the
reference points on it are marked by square boxes with numbers) in comparison with the theoretical
1.9
2
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
3
3.1
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18x, m
M
Theory
Calculation
1
2
3
4
5
Fig.8.5 Mach number contours predicted by COSMOSFloWorks.
Chapter Supersonic Flow in a 2D Convergent-Divergent Channel
4-40
9 Supersonic Flow over a Segmental Conic Body
We will now consider an external supersonic flow of air over a segmental conic body
shown in Fig.9.1. The general case is that the body is tilted at an angle of with respect to
the incoming flow direction. The dimensions of the body whose longitudinal (in direction
t, see Fig.9.1) and lateral (in direction n) aerodynamic drag coefficients, as well as
longitudinal (with respect to Z axis) torque coefficient, were investigated in Ref.5 are
presented in Fig.9.2. They were determined from the dimensionless body sizes and the
Reynolds number stated in Ref.5.
The model of this body is shown in Fig.9.3.
Fig.9.1 Supersonic flow over a segmental conic body.
Center of
mass
n
y
x
a
t
External air flow
M
= 1.7
Fig.9.2 Model sketch dimensioned in centimeters.
Chapter Supersonic Flow over a Segmental Conic Body
4-42
To compare the COSMOSFloWorks predictions with the experimental data of Ref.5, the
calculations have been performed for the case of incoming flow velocity of Mach number
1.7. The undisturbed turbulent incoming flow has a static pressure of 1 atm, static
temperature of 660.2 K, and turbulence intensity of 1%. The flow Reynolds number of
1.710
6
(defined with respect to the body frontal diameter) corresponds to these
conditions, satisfying the Ref.5 experimental conditions.
To compare the flow prediction with the experimental data of Ref.5, the calculations have
been performed for the body tilted at = 0, 30, 60, 90, 120, 150 and 180 angles. To
reduce the computational resources, the Z = 0 flow symmetry plane has been specified in
all of the calculations. Additionally, the Y = 0 flow symmetry plane has been specified at
= 0 and 180.
The calculations have been performed at result resolution level 6.
The comparison is performed
on the longitudinal aerodynamic drag coefficient,
where F
t
is the aerodynamic drag force acting on the body in the t direction (see Fig.9.1),
U
2
/2 is the incoming stream dynamic head, S is the body frontal cross section (being
perpendicular to the body axis) area;
Fig.9.3 The SolidWorks model for calculating the 3D flow over the 3D segmental conic
body with COSMOSFloWorks.
S U
F
C
2
t
t
2
1
=
,
on the lateral aerodynamic drag coefficient,
where F
n
is the aerodynamic drag force acting on the body in the n direction (see Fig.9.1),
U
2
/2 is the incoming stream dynamic head, S is the body frontal cross section (being
perpendicular to the body axis) area;
on the longitudinal (with respect to Z axis) aerodynamic torque coefficient,
where M
z
is the aerodynamic torque acting on the body with respect to the Z axis (see
Fig.9.1), U
2
/2 is the incoming stream dynamic head, S is the body frontal cross section
(being perpendicular to the body axis) area, L is the reference length.
The calculation results are presented in Figs.9.4 and 9.5.
From Fig.9.4, it is seen that the COSMOSFloWorks predictions of both C
n
and C
t
are
excellent.
S U
F
C
2
n
n
2
1
=
,
SL U
M
m
2
z
z
2
1
=
,
Fig.9.4 The longitudinal and lateral aerodynamic drag coefficients predicted with
COSMOSFloWorks and measured in the experiments of Ref.5 versus the body tilting angle.
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0 30 60 90 120 150 180
Attack angle
(degree)
C
t , C
n
Ct experiment
Ct calculation
Cn experiment
Cn calculation
4-44
As for the longitudinal aerodynamic torque coefficient (m
z
) prediction, it is also close to
the experimental data of Ref.5, especially if we take into account the measurements error.
To illustrate the quantitative predictions with the corresponding flow patterns, the Mach
number contours are presented in Figs. 9.6, 9.7, and 9.8. All of the flow patterns presented
in the figures include both supersonic and subsonic flow regions. The bow shock consists
of normal and oblique shock parts with the subsonic region downstream of the normal
shock. In the head subsonic region the flow gradually accelerates up to a supersonic velocity
and then further accelerates in the expansion fan of rarefaction waves. The subsonic wake
region past the body can also be seen.
Fig.9.5 The longitudinal aerodynamic torque coefficient predicted with COSMOSFloWorks and
measured in the experiments (Ref.5) versus the body tilting angle.
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0 30 60 90 120 150 180
Attack angle
(degree)
m
z
Experiment
Calculation
Fig.9.6 Mach number contours at = 0
o
.
As the forward part becomes sharper, the normal part of the bow shock and the
corresponding subsonic region downstream of it become smaller. In the presented
pictures,

the nose shock, correspondingly its subsonic region, is smallest at a = 60
o
.
o
.
o
.
4-46
10 Flow over a Heated Plate
Until now we have only considered flow in or around bodies with adiabatic walls. We will
now consider flows with other thermal boundary conditions.
The first example, is a uniform 2D flows with a laminar boundary layer on a heated flat
plate, see Fig.10.1. The incoming uniform air stream has a velocity of 1.5 m/s, a
temperature of 293.2 K, and a static pressure of 1 atm. Thus, the flow Reynolds number
defined on the incoming flow characteristics and on the plate length of 0.31 m is equal to
3.110
4
, therefore the boundary layer beginning from the plates leading edge is laminar
(see Ref.6).
Then, let us consider the following three cases:
Case #1: the plate over its whole length (within the computational domain) is 10
o
C
warmer than the incoming air (303.2 K), both the hydrodynamic and the thermal boundary
layer begin at the plate's leading edge coinciding with the computational domain
boundary;
Case #2: the upstream half of the plate (i.e. at x 0.15 m) has a fluid temperature of 293.2
K, and the downstream half of the plate is 10
o
C warmer than the incoming air (303.2 K),
the hydrodynamic boundary layer begins at the plate's leading edge coinciding with the
computational boundary;
Case #3: the same plate temperature, as in case #1, the thermal boundary layer begins at
the inlet computational domain boundary, whereas the hydrodynamic boundary layer at
the inlet computational domain boundary has a non-zero thickness which is equal to the
one in case #2 at the thermal boundary layer starting.
The calculation goal is to predict the local coefficient of heat transfer from the wall to the
fluid, as well as the local skin-friction coefficient.
The SolidWorks model used for calculating the 2D flow over the heated flat plate with
COSMOSFloWorks is shown in Fig.10.2. The problem is solved as internal in order to
avoid the conflicting situation when the external flow boundary with ambient temperature
conditions intersects the wall with a thermal boundary layer.
Fig.10.1 Laminar flow over a heated flat plate.
Heated plate
T = 303.2 K
Air flow
V = 1.5 m/s
T = 293.2 K
Computational domain
Chapter Flow over a Heated Plate
4-48
To avoid any influence of the upper wall on the flow near the heated lower wall, the ideal
wall boundary condition has been specified on the upper wall. To solve the internal
problem, the incoming fluid velocity is specified at the channel inlet, whereas the fluid
static pressure is specified at the channel exit. To specify the external flow features, the
incoming stream's turbulent intensity is set to 1%, the turbulent length is set to 0.01 m, i.e.,
these turbulent values are the same default values as in external flow problems.
The heat transfer coefficient, h, and the skin-friction coefficient, C
f
, are
COSMOSFloWorks output flow parameters. The theoretical values for laminar flow
boundary layer over a flat plate, in accordance with Ref.6 can be determined from the
following equations:
where
k - fluid thermal conductivity,
x - distance along the wall from the start of the hydrodynamic boundary layer,
Nu
x
- Nusselt number, defined on a heated wall and determined from the following
formulae:
for a laminar boundary layer, if its starting point coincides with the thermal
boundary layer starting point, and
Fig.10.2 The SolidWorks model used for calculating the 2D flow over heated flat plate with
COSMOSFloWorks.
Inlet velocity U
x
Ideal wall
Heated wall
Static pressure opening
x
k
h
x
Nu
=
,
2 / 1
x
3 / 1
x
Re Pr 332 . 0 Nu =
for a laminar boundary layer, if the thermal boundary layer begins at point x
0
lying
downstream of the hydrodynamic boundary layer starting point, in this case Nu
x
is
defined at x>x
0
only;
where is the Prandtl number, is the fluid dynamic viscosity, C
p
is the
fluid specific heat at constant pressure, is the Reynolds number
defined on x, is the fluid density, and V is the fluid velocity;
at , i.e., with a laminar boundary layer.
As for the hydrodynamic boundary layer thickness, , needed for specification at the
computational domain boundary in case #3, in accordance with Ref.6, has been
determined from the following equation: , so = 0.00575 m in this
case. For these calculations all fluid parameters are determined at the outer boundary of
the boundary layer.
The COSMOSFloWorks predictions of h and C
f
performed at result resolution level 7, and
the theoretical curves calculated with the formulae presented above are shown in Figs.10.3
and 10.4. It is seen that the COSMOSFloWorks predictions of the heat transfer coefficient
and the skin-friction coefficient are in excellent agreement with the theoretical curves.
3 4 / 3
0
2 / 1
x
3 / 1
) x / x ( 1
Re Pr 332 . 0
=
x
Nu
Pr
C
p
k
---------- =
Re
x
Vx
---------- - =
0, 664
Re
fx
C =
5
Re 5 10
x

0.5
4.64 / Re
x
x =
Chapter Flow over a Heated Plate
4-50

Fig.10.3 Heat transfer coefficient change along a heated plate in a laminar boundary layer:
COSMOSFloWorks predictions compared to theory.
0
5
10
15
20
25
0 0.05 0.1 0.15 0.2 0.25 0.3
X (m)
h (W/m^2/K)
Case #1, theory
Case #1, calculation
Case #2, theory
Case #3, theory
Fig.10.4 Skin-friction coefficient change along a heated plate in a laminar boundary layer:
COSMOSFloWorks predictions compared to theory.
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0.016
0.018
0.02
0 0.05 0.1 0.15 0.2 0.25 0.3
X (m)
C
f
Cases #1 and #2,
theory
Case #3, theory
Case#3, calculation
11 Convection and Radiation in an Annular Tube
We will now consider incompressible laminar flow in a portion of an annular tube, whose
outer shell is a heat source of a specified positive heat generation rate Q
1
with a
heat-insulated outer surface, and whose central body fully absorbs the heat generated by
the tubes outer shell (i.e. the negative heat generation rate Q
2
is specified in the central
body); see Fig.11.1. (The tube model is shown in Fig.11.2). We will assume that this tube
is rather long, so the tube's L=1 m portion under consideration has fully developed fluid
velocity and temperature profiles at the inlet, and, since the fluid properties are not
temperature-dependent, the velocity profile will also not be temperature-dependent.
To validate the COSMOSFloWorks capability for solving conjugate heat transfer
problems both with and without radiation, let us solve the following three problems: 1) a
conjugate heat transfer problem with convection only, 2) radiation heat transfer only
problem, and 3) a conjugate heat transfer problem with both convection and radiation.
Fig.11.1 Laminar flow in a heated annular tube.
X
Y
1.2m
0.4m
1.4 m
X
Y
1m
Q1 or T1
Q2 or T2
P = 1 atm
U, T
Fig.11.2 A model created for calculating 3D flow within a heated annular tube using
COSMOSFloWorks.
Chapter Convection and Radiation in an Annular Tube
4-52
In the first problem we specify Q
2
= -Q
1
, so the convective heat fluxes at the tube inner
and outer walls are constant along the tube. The corresponding laminar annular pipe flow's
fully developed velocity and temperature profiles are derived from the theory presented in
Ref.6 in the following form:
u(r) = ,
T(r)= ,
where = ,
u velocity, T fluid temperature,
r radial coordinate (r = r
2
is the tubes central body radius, r = r
1
is the tube outer shells
inner radius),
the mass-average velocity, defined as the mass flow rate divided by the tubes
cross-sectional area,
q
2
the heat flux from the fluid to the tubes central body,
k fluid thermal conductivity,
T
2
the tubes central body surface temperature.
The heat flux from the fluid to the tube's central body (is negative, since the heat comes
from the fluid to the solid) is equal to
Let Q
1
= - Q
2
= 107.235 W and 58 m/s ( = -10 m/s), the fluid has the following
properties: k = 0.5 W/(mK), C
p
= 500 J/(kgK), = 0.002 Pas, = 0.1 kg/m
3
. Since the
corresponding tube (defined on the equivalent tube diameter) Reynolds number of Re
d

2248 is rather low, the flow has to be laminar. We specify the corresponding velocity and
temperature profiles as boundary conditions at the model inlet and as initial conditions,
and P
out
= 1 atm as the tube outlet boundary condition.
To reduce the computational domain, let us set Y=0 and X=0 flow symmetry planes
(correspondingly, the specified Q
1
and Q
2
values are referred to the tube section's quarter
lying in the computational domain). The calculation have been performed at result
resolution level 7.
(
(
|
|
.
|
\
|
|
|
.
|
\
|
|
|
.
|
\
|
1
) / ln(
) / ln(
1
2 1
2
2
2
1
2
2
r r
r r
r
r
r
r
|
|
.
|
\
|
2
2
2
2
ln
r
r
r
k
q
T
(
(
|
|
.
|
\
|
|
|
.
|
\
|
|
|
.
|
\
|
1 ) / ln( / 1 / 2
2
2
1
2 1
2
2
1
r
r
r r
r
r
u
u
2
r r
r
T
=
|
.
|
\
|
L r
Q

2
2
2
q
2
= k =
u
The fluid temperature profile predicted at 0.75 m from the tube model inlet is shown in
Fig.11.3 together with the theoretical curve.
It is seen that this prediction practically coincides with the theoretical curve.
Before solving the third problem coupling convection and radiation, let us determine the
radiation heat fluxes between the tube's outer and inner walls under the previous problem's
wall temperatures. In addition to holding the outer shell's temperature at 450 K and the
central body's temperature at 300 K as the volume sources, let us specify the outer shell's
emissivity of
1
= 0.95 and the central body's emissivity of
2
= 0.25.
Let J
2
denote the radiation rate leaving the central body, and G
2
denotes the radiation rate
coming to the central body, therefore Q
2r
= J
2
- G
2
(the net radiation rate from the central
body). In the same manner, let J
1
denote the radiation rate leaving the outer shell's inner
surface, and G
1
denote the radiation rate coming to the outer shell's inner surface,
therefore Q
1r
= J
1
- G
1
(the net radiation rate from the outer shell's inner surface). These
radiation rates can be determined by solving the following equations:
J
2
= A
2
2
T
2
4
+ G
2
(1-
2
),
G
2
= J
1
F
1-2
,
J
1
= A
1
1
T
1
4
+ G
1
(1-
1
),
G
1
= J
2
F
2-1
+ J
1
F
1-1
,
where =5.66910
-8
W/m
2
K
4
is the Stefan-Boltzmann constant, F
1-2
, F
2-1
, F
1-1
are these
surfaces' radiation shape factors, under the assumption that the leaving and incident
radiation fluxes are uniform over these surfaces, Ref.6 gives the following formulas:
F
1-2
=(1/X) - (1//X){arccos(B/A) - (1/2/Y)[(A
2
+ 4A - 4X
2
+ 4)
1/2
arccos(B/X/A) +
+ Barcsin(1/X)-A/2]},
Fig.11.3 Fluid temperature profiles across the tube in the case of convection only, predicted with
COSMOSFloWorks and compared to the theoretical curve.
250
275
300
325
350
375
400
425
450
475
500
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Y, m
T
Theory
Calculation
Chapter Convection and Radiation in an Annular Tube
4-54
F
1-1
=1-(1/X)+(2//X)arctan[2(X
2
-1)
1/2
/Y]-(Y/2//X){[(4X
2
+Y
2
)
1/2
/Y]arcsin{[4(X
2
-1)+
+ (Y/X)
2
(X
2
-2)]/[Y
2
+4(X
2
-1)]}-arcsin[(X
2
-2)/X
2
]+(/2)[(4X
2
+Y
2
)
1/2
/Y-1]}
F
2-1
= F
1-2
A
1
/A
2
, where X=r
1
/r
2
, Y=L/r
2
, A=X
2
+Y
2
-1, B=Y
2
-X
2
+1.
These net and leaving radiation rates (over the full tube section surface), both calculated
by solving the equations analytically and predicted by COSMOSFloWorks at result
resolution level 7, are presented in Table.11.1.
Table.11.1 Radiation rates predicted with COSMOSFloWorks with comparison to
the theoretical values.
It is seen that the prediction errors are rather small.To validate the COSMOSFloWorks
capabilities on the third problem, which couples convection and radiation, let us add the
theoretical net radiation rates, Q
1 r
and Q
2 r
scaled to the reduced computational domain,
i.e., divided by 4, to the Q
1
and Q
2
values specified in the first problem. Let us specify Q
1

= 1108.15 W and Q
2
= -203.18 W, so theoretically we must obtain the same fluid
temperature profile, as in the first considered problem.
The fluid temperature profile predicted at 0.75 m from the tube model inlet at the result
resolution level 7 is shown in Fig.11.4 together with the theoretical curve. It is seen, that
this prediction practically coincides with the theoretical curve.
Value, W Prediction error,%
Q2 r -383.77 -388.30 1.2%
J2 r 1728.35 1744.47 0.9%
Q1 r 4003.68 3931.87 -1.8%
J1 r 8552.98 8596.04 0.5%
Parameter Theory (Ref.6), W
COSMOSFloWorks predictions
275
300
325
350
375
400
425
450
475
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Y, m
T, K
Theory
Calculation
Fig.11.4 Fluid temperature profiles across the tube in the case coupling convection and radiation,
predicted with COSMOSFloWorks and compared to the theoretical curve.
12 Heat Transfer from a Pin-fin Heat Sink
Heat sinks play an important role in electronics cooling. Following the experimental work
presented in Ref.19, let us consider heat transfer from an electrically heated thermofoil
which is mounted flush on a plexiglass substrate, coated by an aluminum pin-fin heat sink
with a 99 pin fin array, and placed in a closed plexiglass box. In order to create more
uniform ambient conditions for this box, it is placed into another, bigger, plexiglass box
and attached to the heat-insulated thick wall, see Figs. 12.1, 12.2. Following Ref.19, let us
consider the vertical position of these boxes, as it is shown in Fig.12.1 (c) (here, the
gravity acts along the Y axis).
Fig.12.1 The pin-fin heat sink nestled within two plexiglass boxes: l
cp
=L
s
=25.4 mm, h
cp
=0.861 mm, H
p
=5.5
mm, H
b
=1.75 mm, S
p
=1.5 mm, S
ps
= Ls/8, L=127 mm, H=41.3 mm, Hw=6.35 mm (from Ref.19).
Chapter Heat Transfer from a Pin-fin Heat Sink
4-56
The corresponding model used in the calculations is shown in Fig.12.2. In this model's
coordinate system the gravitational acceleration vector is directed along the X axis. The
computational domain envelopes the external box's outer surface, and the Z=0 symmetry
plane is used to reduce the required computer resources.
According to Ref.19, both the heat sink and the substrate are coated with a special black
paint to provide a surface emissivity of 0.95 (the other plexiglass surfaces are also opaque,
diffuse and gray, but have an emissivity of 0.83).
The maximum steady-state temperature T
max
of the thermofoil releasing the heat of
known power Q was measured. The constant ambient temperature T
a
was measured at the
upper corner of the external box. As a result, the value of
R
ja
= (T
max
- T
a
)/Q (12.1)
was determined at various Q (in the 0.1...1W range).
The ambient temperature is not presented in Ref.19, so, proceeding from the suggestion
that the external box in the experiment was placed in a room, we have varied the ambient
temperature in the relevant range of 15...22C. Since R
ja
is governed by the temperature
difference T
max
- T
a
, (i.e. presents the two boxes thermal resistance), the ambient
temperature range only effects the resistance calculations by 0.6C/W at Q = 1W, (i.e. by
1.4% of the experimentally determined R
ja
value (43C/W)). As for the boundary
conditions on the external boxs outer surface, we have specified a heat transfer
coefficient of 5.6 W/m
2
K estimated from Ref.20 for the relevant wind-free conditions and
an ambient temperature lying in the 15...22C range (additional calculations have shown
varying the constant ambient temperature on this boundary, yield practically the same
results). As a result, at Q = 1W (the results obtained at the other Q values are shown in
Ref.21) and T
a
=20C we have obtained R
ja
= 41C/W, i.e. only 5% lower than the
experimental value.
Fig.12.2 A model created for calculating the heat transfer from the pin-fin heat sink through the two
nested boxes into the environment: (a) the internal (smaller) box with the heat sink; (b) the whole model.
a b
As for the flow streamlines, visualized in Ref.19 using smoke and obtained in the
calculations, are shown in Fig.12.3.
Fig.12.3 Flow streamlines visualized by a smoke in the Ref.19 experiments (left) and
obtained in the calculations (their colors show the flow velocity values) (right).
Chapter Heat Transfer from a Pin-fin Heat Sink
4-58
13 Unsteady Heat Conduction in a Solid
Until now, we have studied various flow problems including those with heat flow from the
model walls to the fluid, but we have not considered heat conduction in solids (i.e., a
conjugate heat transfer). To validate this capability, let us consider unsteady heat
conduction in a solid. To compare the COSMOSFloWorks predictions with the analytical
solution (Ref.6), we will solve a one-dimensional problem.
A warm solid rod having the specified initial temperature and the heat-insulated side
surface suddenly becomes and stays cold (at a constant temperature of T=300 K) at both
ends (see Fig.13.1). The rod inner temperature evolution is studied. The constant initial
temperature distribution along the rod is considered: T
initial
(x)=350K.
The problem is described by the following differential equation:
where , C, and k are the solid material density, specific heat, and thermal conductivity,
respectively, and is the time, with the following boundary condition: T=T
0
at x = 0 and
at x = L.
In the general case, i.e., at an arbitrary initial condition, the problem has the following
solution:
where coefficients C
n
are determined from the initial conditions (see Ref.6).
With the uniform initial temperature profile, according to the initial and boundary
conditions, the problem has the following solution:
Fig.13.1 A warm solid rod cooling down from an initial temperature to the temperature at the
ends of the rod.
T
initial
= 350 K
T = 300 K
T = 300 K
X
L
T
k
C
x
T
2
2
,
+ =
1
) /( ) / (
0
sin
2
n
C k L n
n
L
x n
e C T T

,
[ ]
+ =
1
) /( /
) sin(
1 4
50 300
2
n
C k L n
L
x n
e
n
T

(K).
Chapter Unsteady Heat Conduction in a Solid
4-60
To perform the time-dependent analysis with COSMOSFloWorks, a SolidWorks model
representing a solid parallelepiped with dimensions 10.20.1 m has been created (see
Fig.13.2).
The evolution of maximum rod temperature, predicted with COSMOSFloWorks and
compared with theory, is presented in Fig.13.3. The COSMOSFloWorks prediction has
been performed at result resolution level 5. One can see that it is close to the theoretical
curve.
The temperature profiles along the rod at different time moments, predicted by
COSMOSFloWorks, are compared to theory and presented in Fig.13.4. One can see that
the COSMOSFloWorks predictions are very close to the theoretical profiles. The
maximum prediction error occurs at the rod center and does not exceed 1K.
Fig.13.2 The SolidWorks model used for calculating heat conduction in a solid rod with
COSMOSFloWorks (the computational domain envelopes the rod).
Fig.13.3 Evolution of the maximum rod temperature, predicted with COSMOSFloWorks and
compared to theory.
305
315
325
335
345
355
0 2000 4000 6000 8000 10000
Physical time (s)
T
e
m
p
e
r
a
t
u
r
e

(
K
)
Theory
Calculation
Fig.13.4 Evolution of the temperature distribution along the rod, predicted with COSMOSFloWorks
and compared to theory.
300
305
310
315
320
325
0 0.2 0.4 0.6 0.8 1
X (m)
T
e
m
p
e
r
a
t
u
r
e

(
K
)
t=5000s/theory
t=5000s/calculation
t=10000s/theory
t=10000s/calculation
Chapter Unsteady Heat Conduction in a Solid
4-62
14 Tube with Hot Laminar Flow and Outer Heat Transfer
Let us now consider an incompressible laminar flow of a hot fluid through a circular tube
cooled externally (Fig.14.1). The fluid flow has fully developed velocity and temperature
profiles at the tube inlet, whereas the heat transfer conditions specified at the tube outer
surface surrounded by a cooling medium sustain the self-consistent fluid temperature
profile throughout the tube.
In accordance with Ref.6, a laminar tube flow with a fully developed velocity profile has a
self-consistent fully developed temperature profile if the following two conditions are
satisfied: the fluid's properties are temperature-independent and the heat flux from the
tube inner surface to the fluid (or vise versa) is constant along the tube. These conditions
provide the following fully developed tube flow temperature profile:
T(r, z) = T(r=0, z=z
inlet
) - ,
where
T - fluid temperature,
r - radial coordinate (r = 0 corresponds to the tube axis, r = R
i
corresponds to the tube
inner surface, i.e., R
i
is the tube inner radius),
z - axial coordinate (z = z
inlet
corresponds to the tube inlet),
q
w
- a constant heat flux from the fluid to the tube inner surface,
k - fluid thermal conductivity,
- fluid density,
C
p
- fluid specific heat under constant pressure,
Fig.14.1 Laminar flow in a tube cooled externally.
Liquid
Laminar flow
Polystyrene
Te(z)
e
= const
r
( )
2 4
4
1
4
w inlet
w i
i i p max i
q z z
q R r r
k R R C u R
(
| | | |
( +
| |
(
\ . \ .

Chapter Tube with Hot Laminar Flow and Outer Heat Transfer
4-64
u
max
- the maximum fluid velocity of the fully developed fluid velocity profile
= u
max
.
Since the tube under consideration has no heat sinks and is cooled by a fluid medium
surrounding it, lets assume that the fluid medium surrounding the tube has a specified
temperature, T
e
(distributed over the tube's outer surface), and the heat transfer between
this medium and the tube outer surface is featured by a specified constant heat transfer
coefficient
e
.
By assuming a constant thermal conductivity of the tube material, k
s
, specifying an
arbitrary
e
, and omitting intermediate expressions, we can obtain T
e
from the following
formula:
,
where R
o
is the tube outer radius.
In the validation example under consideration (Fig.14.2) the following tube and fluid
characteristics have been specified: R
i
= 0.05 m, R
o
= 0.07 m, z - z
inlet
= 0.1 m, the tube
material is polystyrene with thermal conductivity k
s
= 0.082 W/(mK), u
max
= 0.002 m/s,
T(r=0, z=z
i
) = 363 K, q
w
= 147.56 W/m
2
, k = 0.3 W/(mK), C
p
= 1000 J/(kgK), fluid
dynamic viscosity = 0.001 Pas, = 1000 kg/m
3
(these fluid properties provide a
laminar flow condition since the tube flow Reynolds number defined on the tube diameter
is equal to Re
d
= 100). The T(r,z
inlet
) and u(r) profiles at the tube inlet, the T
e
(z)
distribution along the tube,
e
= 5 W/(m
2
K), and tube outlet static pressure P
out
= 1 atm
have been specified as the boundary conditions.
The inlet flow velocity and temperature profiles have been specified as the initial
conditions along the tube.
To reduce the computational domain, the calculations have been performed with the Y=0
and X=0 flow symmetry planes. The calculations have been performed at result resolution
level 7.
The fluid and solid temperature profiles predicted at z = 0 are shown in Fig.14.3 together
with the theoretical curve. It is seen that the prediction practically coincides with the
theoretical curve (the prediction error does not exceed 0.4%).
( )
u r
1
2
r
R
i
| |
| |
|
|
|
|
|
\ .
\ .
( ) + + = = =
|
|
.
|
\
|
o
R
i
R
s
k
o
R
e
k
i
R
w
q
inlet
z z r T z
e
T ln
1 1
4
3
) , 0 (
i
R
max
u
p
C
inlet
z z
w
q
|
.
|
\
|

+
4

Outlet static
pressure opening
Computational
domain
Inlet velocity
opening
Sketch line for
temperature profile
determination
Fig.14.2 The model used for calculating the 3D flow and the conjugate heat transfer in the tube
with COSMOSFloWorks.
Fig.14.3 Fluid and solid temperature profiles across the tube, predicted with COSMOSFloWorks
and compared with the theoretical curve.
290
300
310
320
330
340
350
360
370
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
R, m
T, K
Theory liquid
Theory solid
Calculation
Chapter Tube with Hot Laminar Flow and Outer Heat Transfer
4-66
15 Flow over a Heated Cylinder
Let us now return to the earlier validation example of incompressible flow over a cylinder,
but now we will also specify a heat generation source inside the cylinder (see Fig.15.1).
The cylinder is placed in an incoming air stream and will acquire a certain temperature
depending on the heat source power and the air stream velocity and temperature.
Based on experimental data for the average coefficient of heat transfer from a heated
circular cylinder to air flowing over it (see Ref.6), the corresponding Nusselt number can
be determined from the following formula:
where constants C and n are taken from the following table:
Here, the Nusselt number, Nu
D
= (hD)/k (where h is the heat transfer coefficient
averaged over the cylinder, and k is fluid thermal conductivity), the Reynolds number,
Re
D
= (UD)/ (where U is the incoming stream velocity, and is fluid dynamic
viscosity), and the Prandtl number, Pr=C
p
/k (where is fluid dynamic viscosity, C
p
is
fluid specific heat at constant pressure, and k is fluid thermal conductivity) are based on
the cylinder diameter D and on the fluid properties taken at the near-wall flow layer.
According to Ref.6, Pr = 0.72 for the entire range of Re
D
.
Re
D
C n
0.4 - 4 0.989 0.330
4 - 40 0.911 0.385
40 - 4000 0.683 0.466
4000 - 40000 0.193 0.618
40000 - 400000 0.0266 0.805
Fig.15.1 2D flow over a heated cylinder.
External air flow
Heat source q
Y
X
( )
3 1
Pr Re =
n
D D
C Nu
,
Chapter Flow over a Heated Cylinder
4-68
To validate the COSMOSFloWorks predictions, the air properties have been specified to
provide Pr = 0.72: k = 0.0251375 W/(mK), = 1.810
-5
Pas, specific heat at constant
pressure C
p
= 1005.5 J/(kgK). Then, the incoming stream velocity, U, has been specified
to obtain Re
D
= 1, 10, 100, 10
3
, 10
4
, 510
4
, 10
5
, 210
5
, and 310
5
for a cylinder
diameter of D = 0.1 m. (see Table.15.1)
This validation approach consists of specifying the heat generation source inside the
cylinder with a power determined from the desired steady-state cylinder temperature and
the average heat transfer coefficient, h = (Nu
D
k)/D. Nu
D
is determined from the
empirical formula presented above versus the specified Re
D
. Also required for specifying
the heat source power Q (see Table.15.1) is the final cylinder surface temperature, which
is assumed to be 10
o
C higher than the incoming air temperature. The initial cylinder
temperature and the incoming air temperature are equal to 293.15 K. The cylinder material
is aluminum. Here, the heat conduction in the solid is calculated simultaneously with the
flow calculation, i.e., the conjugate heat transfer problem is solved.
As a result of the calculation, the cylinder surface has acquired a steady-state temperature
differing from the theoretical one corresponding to the heat generation source specified
inside the cylinder. Multiplying the theoretical value of the Nusselt number by the ratio of
the obtained temperature difference (between the incoming air temperature and the
cylinder surface temperature) to the specified temperature difference, we have determined
the predicted Nusselt number versus the specified Reynolds number. The values obtained
by solving the steady-state and time dependent problems at result resolution level 5 are
presented in Fig.15.2 together with the experimental data taken from Ref.6.
Table.15.1 The COSMOSFloWorks specifications of U and Q for the problem under
consideration.
From Fig.15.2, it is seen that the predictions made with COSMOSFloWorks, both in the
time-dependent approach and in the steady-state one, are excellent within the whole Re
D

range under consideration.
1 1. 510
-4
0. 007
10 1. 510
-3
0. 016
10
2
0. 015 0. 041
10
3
0. 15 0. 121
10
4
1. 5 0. 405
10
5
15 1. 994
Re D U , m/ s Q ,W
Fig.15.2 Nusselt number for air flow over a heated cylinder: COSMOSFloWorks predictions and the
experimental data taken from Ref.6.
0.1
1
10
100
1000
1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Re D
NuD
Calculation,
steady-state
Calculation,
time-dependent
Chapter Flow over a Heated Cylinder
4-70
16 Natural Convection in a Square Cavity
In the previous validation examples we have considered a forced convection, i.e., heat
transfer between a wall and a fluid flow specified directly through the flow parameters at a
model's openings (internal problems) or at far-field boundaries (external problems). Let us
now consider a heat transfer due to a heat-induced natural convection within a closed
cavity.
Here we will consider a 2D square cavity with a steady-state natural convection, for which
a highly-accurate numerical solution has been proposed in Ref.10 and used as a
benchmark for about 40 computer codes in Ref.11, besides it well agrees with the
semi-empirical formula proposed in Ref.12 for rectangular cavities. This cavity's
configuration and imposed boundary conditions, as well as the used coordinate system,
are presented in Fig.16.1. Here, the left and right vertical walls are held at the constant
temperatures of T
1
= 305 K and T
2
= 295 K, accordingly, whereas the upper and bottom
walls are adiabatic. The cavity is filled with air.
The square cavity's side dimension, L, is varied within the range of 0.0111...0.111 m in
order to vary the cavity's Rayleigh number, governing the characteristics of the natural
convection inside the cavity,
, within the range of 10
3
10
6
,
where - the air's volume expansion coefficient,
g - gravitational acceleration,
C
p
- the air's specific heat at constant pressure,
T=T
1
- T
2
= 10 K - the temperature difference between the walls,
k - the air's thermal conductivity,
- the air's dynamic viscosity.
Fig.16.1 An enclosed 2D square cavity with natural convection.
Adiabatic walls
L
305 K
295 K
Y
g
X

k
T L C g
Ra
p

=
3 2
1
T
=
Chapter Natural Convection in a Square Cavity
4-72
The cavity's model is shown in Fig.16.2.
Due to gravity and the differential temperatures of the cavity's vertical walls, a
steady-state natural convection flow (vortex) with a vertical temperature stratification
forms inside the cavity. The Ra = 10
5
flow's prediction performed with
COSMOSFloWorks is shown in Fig.16.3.
Fig.16.2 The model created for calculating the 2D natural convection
flow in the 2D square cavity using COSMOSFloWorks.
Fig.16.3 The temperature, X-velocity, Y-velocity, the velocity vectors, and the streamlines,
predicted by COSMOSFloWorks in the square cavity at Ra = 10
5
.
A quantitative comparison of the COSMOSFloWorks predictions performed at result
resolution level 8 with References.10, 11 (computational benchmark) and 12
(semi-empirical formula) for different Ra values is presented in Figs.16.4 - 16.6. The
Nusselt number averaged over the cavity's hot vertical wall (evidently, the same value
must be obtained over the cavity's cold vertical wall) , where
q
w av
is the heat flux from the wall to the fluid, averaged over the wall, is considered in
Fig.16.4.
Here, the dash line presents the Ref.12 semi-empirical formula
,
where D is the distance between the vertical walls, L is the cavity's height (D=L in the
case under consideration). One can see that the COSMOSFloWorks predictions practically
coincide with the benchmark at Ra 10
5
and are close to the semi-empirical data.
The natural convection flow's dimensionless velocities in the X and Y directions,
and (which are maximum along the cavity's
mid-planes,
i.e., along the vertical mid-plane and along the horizontal mid-plane) are
considered
in Fig.16.5. The dimensionless coordinates, and , of these maximums'
locations
/( )
av wav
Nu q L T k =
1/ 4 1/ 4
0.28 ( / )
av
Nu Ra L D

=
Fig.16.4 The average sidewall Nusselt number vs. the Rayleigh number.
0
1
2
3
4
5
6
7
8
9
10
1.E+03 1.E+04 1.E+05 1.E+06
Ra
Nu
av
Refs.10, 11
Ref.12
Calculation
p
U L C
U
k

=
p
V L C
V
k

=
max
U
max
V
L
x
x =
L
y
y =
Chapter Natural Convection in a Square Cavity
4-74
(i.e., for and for ) are presented in Fig.16.6. One can see that the
COSMOSFloWorks predictions of the natural convection flow's local parameters are
fairly close to the benchmark data at Ra 10
5
.

y
max
U x
max
V
Fig.16.5 Dimensionless maximum velocities vs. the Rayleigh number.
1
10
100
1000
1.E+03 1.E+04 1.E+05 1.E+06
Ra
D
i
m
e
n
s
i
o
n
l
e
s
s

U

m
a
x
,

V
m
a
x
Vmax, Refs.10, 11
Vmax, calculation
Umax, Refs.10, 11
Umax, calculation
Fig.16.6 Dimensionless coordinates of the maximum velocities' locations.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.E+03 1.E+04 1.E+05 1.E+06
Ra
D
i
m
e
n
s
i
o
n
l
e
s
s

X

V
m
a
x
,

Y
U
m
a
x
Y Umax, Refs.10, 11
Y Umax, calculation
X Vmax, Refs.10, 11
X V max,
17 Particles Trajectories in Uniform Flows
Let us now consider the COSMOSFloWorks capability to predict particles trajectories in a
gas flow (i.e. two-phase flow of fluid + liquid droplets or solid particles).
In accordance with the particles motion model adopted in COSMOSFloWorks, particle
trajectories are calculated after completing a fluid flow calculation (it can be steady or
time-dependent). That is, the particles mass and volume flow rates are assumed
substantially lower than those of the fluid stream, so particle motions and temperatures
have a negligible influence on the fluid flow parameters, and a particle motion obeys the
following equation:
where m is the particle mass, t is time, V
p
and V
f
are the particle and fluid velocities
(vectors), accordingly,
f
is the fluid density, C
d
is the particle drag coefficient, A is the
particle frontal surface area, and F
g
is the gravitational force.
Particles are treated as non-rotating spheres of constant mass and specified (solid or
liquid) material, whose drag coefficient is determined from Hendersons semi-empirical
formula (Ref.7). At a very low velocity of particles with respect to fluid-carrier (i.e., at the
relative velocitys Mach number M 0) this formula becomes
where Reynolds number
d is the particle diameter, is the fluid dynamic viscosity.
To validate COSMOSFloWorks, let us consider three cases of injecting a particle
perpendicularly into an incoming uniform flow, Fig.17.1. Since both the fluid flow and the
particle motion in these cases are 2D (plane), we will solve a 2D (i.e. in the XY-plane)
flow problem.
g d
p f p f f
p
F A C
V V V V
dt
dV
m +

=
2
) (
,
38 0.
Re 48 0 Re 03 0 1
12 4.
Re
24
C
d
+
+ +
+ =
. .

d V V
p f f

= Re
,

Uniform
fluid flow

Particle injection

Fig.17.1 Injection of a particle into a uniform fluid flow.
Chapter Particles Trajectories in Uniform Flows
4-76
Due to the same reason, as in the previous validation examples with flow over flat plates,
we will solve this validation as an internal problem. The corresponding SolidWorks model
is shown in Fig.17.2. Both of the walls are ideal, the channel has length of 0.233 m and
height of 0.12 m, all the walls have thickness of 0.01 m. We specify the uniform fluid
velocity V
inlet
, the fluid temperature of 293.2 K, and the default values of turbulent flow
parameters with the laminar boundary layer at the channel inlet, and the static pressure of
1 atm at the channel outlet. All the fluid flow calculations are performed at a result
resolution level of 5.
To validate calculations of particles trajectories by comparing them with available
analytical solutions of the particle motion equation, we consider the following three cases:
a) the low maximum Reynolds number of Re
max
= 0.1 (air flow with V
inlet
=
0.002 m/s, gold particles of d = 0.5 mm, injected at the velocity of 0.002
m/s perpendicularly to the wall),
b) the high maximum Reynolds number of Re
max
= 10
5
(water flow with
V
inlet
= 10 m/s, iron particles of d = 1 cm, injected at the velocities of 1, 2,
3 m/s perpendicularly to the wall),
c) a particle trajectory in the Y-directed gravitational field (gravitational
acceleration g
y
= -9.8 m/s
2
, air flow with V
inlet
= 0.6 m/s, an iron particle
of d = 1 cm, injected at the 1.34 m/s velocity at the angle of 63.44
o
with the
wall).
In the first case, due to small Re values, the particle drag coefficient is close to C
d
=24/Re
(i.e., Stokes law). Then, neglecting gravity, we obtain the following analytical solution
for the particle trajectory:
Fig.17.2 The model.
Inlet
Outlet
Ideal Wall
Origin and particle
injection point
)
18
exp( ) (
18
) (
2
0
2
0
t
d
V V
d
t V X t X
p
fx
t
px
p
fx
t
+ + =
=
=
,
)
18
exp( ) (
18
) (
2
0
2
0
t
d
V V
d
t V Y t Y
p
fy
t
py
p
fy
t
+ + =
=
=
,
where V
fx
, V
px
, V
fy
, V
py
are the X- and Y-components of the fluid and particle velocities,
accordingly,
p
is the particle material density. The COSMOSFloWorks calculation and
the analytical solution are shown in Fig.17.3. It is seen that they are very close to one
another. Special calculations have shown that the difference is due to the C
D
assumptions
only.
In the second case, due to high Re values, the particle drag coefficient is close to C
d
=0.38.
Then, neglecting the gravity, we obtain the following analytical solution for the particle
trajectory:
The COSMOSFloWorks calculations and the analytical solutions for three particle
injection velocities, V
py
(t=0) = 1, 2, 3 m/s, are shown in Fig.17.4. It is seen that the
COSMOSFloWorks calculations coincide with the analytical solutions. Special
calculations have shown that the difference is due to the C
D
assumptions only.
In the third case, the particle trajectory is governed by the action of the gravitational force
only, the particle drag coefficient is very close to zero, so the analytical solution is:
Fig.17.3 Particle trajectories in a uniform fluid flow at Re
max
= 0.1, predicted by COSMOSFloWorks
and obtained from the analytical solution.
0.000
0.005
0.010
0.015
0.020
0.025
0.030
0.035
0.00 0.05 0.10 0.15 0.20 0.25 X (m)
Y (m)
Analytical
solution
Calculation
)
285 . 0
1 ln( ) (
285 . 0
) (
0
0
t
d
V V
d
t V Y t Y
p
fy
t
py
p
fy
t
+ + + =
=
=
,
)
.
ln( ) (
.
) ( t
d
285 0
1 V V
285 0
d
t V X t X
p
fx
0 t
px
p
fx
0 t
+ + + =
=
=
.
2
0
0
0
2
1
|
|
.
|
\
|

+ + =
=
=
=
px
t
t
py
t
V
X X
y
g t V Y Y
.
Chapter Particles Trajectories in Uniform Flows
4-78
The COSMOSFloWorks calculation and the analytical solution for this case are presented
in Fig.17.5. It is seen that the COSMOSFloWorks calculation coincides with the analytical
solution.
I
Fig.17.4 Particle trajectories in a uniform fluid flow at Re
max
= 10
5
, predicted by COSMOSFloWorks
and obtained from the analytical solution.
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.00 0.05 0.10 0.15 0.20
X (m)
Y (m)
Vp = 1 m/s,
analytical solution
Vp = 1 m/s,
Calculation
Vp = 2 m/s,
analytical solution
Vp = 2 m/s,
Calculation
Vp = 3 m/s,
analytical solution
Vp = 3 m/s,
Calculation
Fig.17.5 Particle trajectories in the Y-directed gravity, predicted by COSMOSFloWorks and obtained
from the analytical solution.
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.00 0.03 0.06 0.09 0.12 0.15
X (m)
Y (m)
Calculation
Theory
18 Porous Screen in a Non-uniform Stream
Let us now validate the COSMOSFloWorks capability to calculate fluid flows through
porous media.
Here, following Ref.2, we consider a plane cold air flow between two parallel plates,
through a porous screen installed between them, see Fig.18.1. At the channel inlet the air
stream velocity profile is step-shaped (specified). The porous screen (gauze) levels this
profile to a less-non-uniform profile. This effect depends on the screen drag, see Ref.2.
The SolidWorks model used for calculating the 2D (in XY-plane) flow is shown in
Fig.18.2. The channel has height of 0.15 m, the inlet (upstream of the porous screen) part
of the 0.3 m length, the porous screen of the 0.01 m thickness, and the outlet (downstream
of the porous screen) part of the 0.35 m length. All the walls have thickness of 0.01 m.

Air

L
Porous screen
X
Y
Fig.18.1 Leveling effect of a porous screen (gauze) on a non-uniform stream.
Fig.18.2 The SolidWorks model used for calculating the 2D flow between two parallel plates and
through the porous screen with COSMOSFloWorks.
Chapter Porous Screen in a Non-uniform Stream
4-80
Following Ref.2, we consider porous screens (gauzes) of different drag, :
= 0.95, 1.2, 2.8, and 4.1, defined as:
where P is the pressure difference between the screen sides, V
2
/2 is the incoming
streams dynamic pressure (head).
Since in COSMOSFloWorks a porous mediums resistance to flow is characterized by
parameter k = - gradP/V, then for the porous screens k = V /(2L), where V is the fluid
velocity, L is the porous screen thickness. In COSMOSFloWorks, this form of a porous
mediums resistance to flow is specified as k = (AV+B)/, so A = /(2L), B = 0 for the
porous screens under consideration. Therefore, taking L = 0.01 m and = 1.2 kg/m
3
into
account, we specify A = 57, 72, 168, and 246 kg/m
-4
for the porous screens under
consideration. In accordance with the screens nature, their permeability is specified as
isotropic.
According to the experiments presented in Ref.2, the step-shaped velocity profiles V(Y)
presented in Fig.18.3 have been specified at the model inlet. The static pressure of 1 atm
has been specified at the model outlet.
The air flow dynamic pressure profiles at the 0.3 m distance downstream from the porous
screens, both predicted by COSMOSFloWorks at result resolution level 5 and measured in
the Ref.2 experiments, are presented in Fig.18.4 for the = 0 case (i.e., without screen)
and Figs.18.5-18.8 for the porous screens of different .
2
2
V
P

= ,
Fig.18.3 Inlet velocity profiles.
0
5
10
15
20
25
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Y (m)
V (m/s)
=0, 0.95 =0, 0.95 =0, 0.95 =0, 0.95
=1.2 =1.2 =1.2 =1.2
=2.8 =2.8 =2.8 =2.8
=4.1 =4.1 =4.1 =4.1
It is seen that the COSMOSFloWorks predictions agree well, both qualitatively and
quantitatively, with the experimental data both in absence of a screen and for all the
porous screens (gauzes) under consideration, demonstrating the leveling effect of the
gauze screens on the step-shaped incoming streams. The prediction error in the dynamic
pressure maximum does not exceed 30%.
Fig.18.4 The dynamic pressure profiles at = 0, predicted by COSMOSFloWorks and compared to
the Ref.2 experiments.
0
50
100
150
200
250
300
350
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Y (m)
D
y
n
a
m
i
c

P
r
e
s
s
u
r
e

(
P
a
)
Calculation
Experiment
Fig.18.5 The dynamic pressure profiles at = 0.95, predicted by COSMOSFloWorks and compared
to the Ref.2 experiments.
0
50
100
150
200
250
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Y (m)
D
y
n
a
m
i
c

P
r
e
s
s
u
r
e

(
P
a
)
Calculation
Experiment
4-82

0
50
100
150
200
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Y (m)
D
y
n
a
m
i
c

P
r
e
s
s
u
r
e

(
P
a
)
Calculation
Experiment
0
20
40
60
80
100
120
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Y (m)
D
y
n
a
m
i
c

P
r
e
s
s
u
r
e

(
P
a
)
Calculation
Experiment
0
10
20
30
40
50
60
70
80
90
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Y (m)
D
y
n
a
m
i
c

P
r
e
s
s
u
r
e

(
P
a
)
Calculation
Experiment
4-84
19 Lid-driven Flows in Triangular and Trapezoidal Cavities
In the previous validation examples we have considered COSMOSFloWorks predictions
of pressure-driven incompressible flows in various channels with stationary (motionless)
walls. Since COSMOSFloWorks allows the motion of walls, let us now see how
COSMOSFloWorks predicts lid-driven (i.e., shear-driven) 2D recirculating flows in
closed 2D triangular and trapezoidal cavities with one or two moving walls (lids) in
comparison with the calculations performed in Refs.15 and 16.
These two cavities are shown in Fig.19.1. The triangular cavity has a moving top wall, the
trapezoidal cavity has a moving top wall also, whereas its bottom wall is considered in two
versions: as motionless and as moving at the top wall velocity. The no-slip conditions are
specified on all the walls.
Shown in Refs. 15 and 16, the shear-driven recirculating flows in these cavities are fully
governed by their Reynolds numbers Re = U
wall
h/, where is the fluid density, is
the fluid dynamic viscosity, U
wall
is the moving wall velocity, h is the cavity height. So,
we can specify the triangular cavitys height h = 4 m, the trapezoidal cavitys height h = 1
m, U
wall
= 1 m/s for all cases under consideration, the fluid density = 1 kg/m
3
, the fluid
dynamic viscosity =0.005 Pas in the triangular cavity produces a Re = 800, and = 0.01,
0.0025, 0.001 Pas in the trapezoidal cavity produces a Re = 100, 400, 1000, respectively.
Fig.19.1 The 2D triangular (left) and trapezoidal (right) cavities with the moving walls (the
motionless walls are shown with dashes).
2
h
U
1
1
2
U
U
U
wa
Chapter Lid-driven Flows in Triangular and Trapezoidal Cavities
4-86
The cavities models are shown in Fig.19.2. The COSMOSFloWorks calculation of flow
in the triangular cavity has been performed on the 4896 computational mesh. The results
in comparison with the Ref.15 calculation results are presented in Fig.19.3 (streamlines)
and in Fig.19.4 (the fluid velocity X-component along the central vertical bisector shown
by a green line in Fig.19.2). A good agreement of these calculations is seen.

Fig.19.2 The models for calculating the lid-driven 2D flows in the triangular (left) and trapezoidal
(right) cavities with COSMOSFloWorks.
Fig.19.3 The flow streamlines in the triangular cavity, calculated by COSMOSFloWorks (right) and
compared to the Ref.15 calculation (left).
The COSMOSFloWorks calculations of flows in the trapezoidal cavity with one and two
moving walls at different Re values have been performed with the 10050 computational
mesh. Their results in comparison with the Ref.16 calculation results are presented in
Fig.19.5-19.10 (streamlines) and in Fig.19.11 (the fluid velocity X-component along the
central vertical bisector shown by a green line in Fig.19.2). A good agreement of these
calculations is seen.
Fig.19.5 The flow streamlines in the trapezoidal cavity with a top only moving wall at Re = 100,
calculated by COSMOSFloWorks (right) and compared to the Ref.16 calculation (left).
Fig.19.4 The triangular cavitys flow velocity X-component along the central vertical bisector,
calculated by COSMOSFloWorks (red line) and compared to the Ref.15 calculation (black line with
circlets).
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
0 0.5 1 1.5 2 2.5 3 3.5 4
Y, m
Vx/Uwall
Calculation
4-88
Fig.19.8 The flow streamlines in the trapezoidal cavity with two moving walls at Re = 100,
Fig.19.9 The flow streamlines in the trapezoidal cavity with two moving walls at Re = 400,
Fig.19.10The flow streamlines in the trapezoidal cavity with two moving walls at Re = 1000, calculated
by COSMOSFloWorks (right) and compared to the Ref.16 calculation (left).
Fig.19.11The flow velocity X-component along the central vertical bisector in the trapezoidal cavity
with two moving walls at Re = 400, calculated by COSMOSFloWorks (red line) and compared to the
Ref.16 calculation (black line).
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Y, m
Vx/Uwall
Calculation
4-90
20 Flow in a Cylindrical Vessel with a Rotating Cover
In the previous COSMOSFloWorks validation example we considered a lid-driven 2D
recirculating flow in a 2D cavity. Since COSMOSFloWorks allows for wall rotation, let
us now see how COSMOSFloWorks predicts a 3D recirculating flow in a cylindrical
vessel closed by a rotating cover (see Fig.20.1) in comparison with the experimental data
presented in Ref.17 (also in Ref.18). This vessel of R = h = 0.144 m dimensions is filled
with a glycerol/water mixture. The upper cover rotates at the angular velocity of . The
other walls of this cavity are motionless. The default no-slip boundary condition is
specified for all walls.
Due to the cover rotation, a shear-driven recirculating flow forms in this vessel. Such
flows are governed by the Reynolds number Re = R
2
/, where is the fluid density,
is the fluid dynamic viscosity, is the rotating covers angular velocity, R is the rotating
covers radius. In the case under consideration the 70/30% glycerol/water mixture has =
1180 kg/m
3
, = 0.02208 Pas, the cover rotates at = 15.51 rpm, so Re = 1800.
The COSMOSFloWorks calculation has been performed on the 824182 computational
mesh. The formed flow pattern (toroidal vortex) obtained in this calculation is shown in
Fig.20.2 using the flows velocity vectors projected onto the XY-plane. The calculated
flow velocitys tangential and radial components along four vertical lines arranged in the
XY-plane at different distances from the vessel axis in comparison with the Ref.17
experimental data are presented in Figs.20.3-7 in the dimensionless form (the
Y-coordinate is divided by R, the velocitys components are divided by R). There is
good agreement with the calculation results and the experimental data shown.
Fig.20.1 The cylindrical vessel with the a rotating cover.
rotating cover
R
h
Chapter Flow in a Cylindrical Vessel with a Rotating Cover
4-92
Fig.20.2 The vessel's flow velocity vectors projected on the XY-plane.
Fig.20.3 The vessel's flow tangential and radial velocity components along the X = 0.6 vertical,
calculated by COSMOSFloWorks (red) and compared to the Ref.17 experimental data.
Chapter Flow in a Cylindrical Vessel with a Rotating Cover
4-94
21 Flow in an Impeller
In the previous validation example we had considered COSMOSFloWorks predictions of
a recirculating flow in a cylindrical vessel with a rotating cover. The calculation had been
performed in the coordinate system related to the stationary object (vessel), and a part
(cover) rotated in this coordinate system. Let us now validate the COSMOSFloWorks
ability to perform calculations in a rotating coordinate system related to a rotating solid.
Following Ref.22, we will consider the flow of water in a 9-bladed centrifugal impeller
having blades tilted at a constant 60 angle with respect to the intersecting radii and
extending out from the 320 mm inner diameter to the 800 mm outer diameter (see Fig.
21.1). The water in this impeller flows from its center to its periphery. To compare the
calculation with the experimental data presented in Ref.22, the impeller's angular velocity
of 32 rpm and volume flow rate of 0.00926 m
3
/s are specified.
Since the impeller's inlet geometry and disk extension serving as the impeller's vaneless
diffuser have no exact descriptions in Ref.22, to perform the validating calculation we
arbitrarily specified the annular inlet as 80 mm in diameter with an uniform inlet velocity
profile perpendicular to the surface in the stationary coordinate system.The impeller's
disks external end was specified as 1.2 m diameter, as shown in Fig. 21.2.
The above-mentioned volume flow rate at the annular inlet and the potential pressure of 1
atm at the annular outlet are specified as the problem's flow boundary conditions.
Fig. 21.1The impeller's blades geometry.
Fig. 21.2The model used for calculating the 3D flow in the impeller.
Chapter Flow in an Impeller
4-96
The COSMOSFloWorks 3D flow calculation is performed on the computational mesh
using a specified result resolution level of 5 and a minimum wall thickness of 2 mm (since
the blades have a constant thickness). To further capture the curvature of the blades a local
initial mesh was also used in the region from the annular inlet to the blades' periphery. As
a result, the computational mesh has a total number of about 1,000,000 cells.
Following Ref.22, let us compare the passage-wise flow velocities (w
s
, see their definition
in Fig. 21.3, = 60) along several radial lines passing through the channels between the
blades (lines g, j, m, p in Fig. 21.4) at the mid-height between the impeller's disks.
The passage-wise flow velocities divided by r
2
, where is the impeller's angular
velocity and r
2
= 400 mm is the impeller's outer radius, which were measured in Ref.22
and obtained in the performed COSMOSFloWorks calculations, are shown in Fig. 21.5, 6,
7, and 8. In these figures, the distance along the radial lines is divided by the line's length.
The COSMOSFloWorks results are presented in each of these figures by the curve
obtained by averaging the corresponding nine curves in all the nine flow passages between
the impeller blades. The calculated passage-wise flow velocity's cut plot covering the
whole computational domain at the mid-height between the impeller's disks is shown in
Fig. 21.9. Here, the g, j, m, p radial lines in each of the impeller's flow passages are shown.
A good agreement of these calculation results with the experimental data is seen.
Fig. 21.3Definition of the passage-wise flow velocity.
Fig. 21.4Definition of the reference radial lines along which the passage-wise flow
velocity was measured in Ref.22 (from a to s in the alphabetical order).
Fig. 21.5The impeller's passage-wise flow velocity along the g (see Fig. 21.4) radial line, calculated
by COSMOSFloWorks and compared to the experimental data.
0
0.25
0.5
0.75
0 0.2 0.4 0.6 0.8 1
relative distance along the radial line
r
e
l
a
t
i
v
e

p
a
s
s
a
g
e
w
i
s
e

v
e
l
o
c
i
t
y

(
a
v
e
r
a
g
e
d
)
calculation
experiment
Fig. 21.6The impeller's passage-wise flow velocity along the j (see Fig. 21.4) radial line, calculated
0
0.25
0.5
0.75
0 0.2 0.4 0.6 0.8 1
r
e
l
a
t
i
v
e

p
a
s
s
a
g
e
w
i
s
e

v
e
l
o
c
i
t
y

(
a
v
e
r
a
g
e
d
)
calculation
experiment
4-98
Fig. 21.7The impeller's passage-wise flow velocity along the m (see Fig. 21.4) radial line, calculated
0
0 .2 5
0 .5
0 .7 5
0 0 .5 1
re lative dis tanc e alo ng the rad ial line
r
e
l
a
t
i
v
e

p
a
s
s
a
g
e
w
i
s
e

v
e
l
o
c
i
t
y

(
a
v
e
r
a
g
e
d
)
calculation
experiment
Fig. 21.8The impeller's passage-wise flow velocity along the p (see Fig. 21.4) radial line, calculated
0
0.25
0.5
0.75
0 0.2 0.4 0.6 0.8 1
r
e
l
a
t
i
v
e

p
a
s
s
a
g
e
w
i
s
e

v
e
l
o
c
i
t
y

(
a
v
e
r
a
g
e
d
)
calculation
experiment
Fig. 21.9A cut plot of the impeller's passage-wise flow velocity calculated by COSMOSFloWorks.
4-100
References
1 Schlichting, H., Boundary Layer Theory. 7th ed., McGraw Hill, New York, 1979.
2 Idelchik, I.E., Handbook of Hydraulic Resistance. 2nd ed., Hemisphere, New York,
1986.
3 Panton, R.L., Incompressible Flow. 2nd ed., John Wiley & Sons, Inc., 1996.
4 White, F.M., Fluid Mechanics. 3rd ed., McGraw-Hill, New York, 1994.
5 Artonkin, V.G., Petrov, K.P., Investigations of aerodynamic characteristics of
segmental conic bodies. TsAGI Proceedings, No. 1361, Moscow, 1971 (in Russian).
6 Holman, J.P., Heat Transfer. 8th ed., McGraw-Hill, New York, 1997.
7 Henderson, C.B. Drag Coefficients of Spheres in Continuum and Rarefied Flows.
AIAA Journal, v.14, No.6, 1976.
8 Humphrey, J.A.C., Taylor, A.M.K., and Whitelaw, J.H., Laminar Flow in a Square
Duct of Strong Curvature. J. Fluid Mech., v.83, part 3, pp.509-527, 1977.
9 Van Dyke, Milton, An Album of Fluid Motion. The Parabolic press, Stanford,
California, 1982.
10 Davis, G. De Vahl: Natural Convection of Air in a Square Cavity: a Bench Mark
Numerical Solution. Int. J. for Num. Meth. in Fluids, v.3, p.p. 249-264 (1983).
11 Davis, G. De Vahl, and Jones, I.P.: Natural Convection in a Square Cavity: a
Comparison Exercise. Int. J. for Num. Meth. in Fluids, v.3, p.p. 227-248 (1983).
12 Emery, A., and Chu, T.Y.: Heat Transfer across Vertical Layers. J. Heat Transfer, v.
87, p. 110 (1965).
13 Denham, M.K., and Patrick, M.A.: Laminar Flow over a Downstream-Facing Step in a
Two-Dimensional Flow Channel. Trans. Instn. Chem. Engrs., v.52, p.p. 361-367
(1974).
14 Yanshin, B.I.: Hydrodynamic Characteristics of Pipeline Valves and Elements.
Convergent Sections, Divergent Sections, and Valves. Mashinostroenie, Moscow,
1965.
15 Jyotsna, R., and Vanka, S.P.: Multigrid Calculation of Steady, Viscous Flow in a
Triangular Cavity. J. Comput. Phys., v.122, No.1, p.p. 107-117 (1995).
16 Darr, J.H., and Vanka, S.P.: Separated Flow in a Driven Trapezoidal Cavity. J. Phys.
Fluids A, v.3, No.3, p.p. 385-392 (1991).
17 Michelsen, J. A., Modeling of Laminar Incompressible Rotating Fluid Flow, AFM
86-05, Ph.D. Dissertation, Department of Fluid Mechanics, Technical University of
Denmark, 1986.
18 Sorensen, J.N., and Ta Phuoc Loc: Higher-Order Axisymmetric Navier-Stokes Code:
Description and Evaluation of Boundary Conditions. Int. J. Numerical Methods in
Fluids, v.9, p.p. 1517-1537 (1989).
Chapter References
4-102
19 Enchao Yu, Yogendra Joshi: Heat Transfer Enhancement from Enclosed Discrete
Components Using Pin-Fin Heat Sinks. Int. J. of Heat and Mass Transfer, v.45, p.p.
4957-4966 (2002).
20 Kuchling, H., Physik, VEB FachbuchVerlag, Leipzig, 1980.
21 Balakin, V., Churbanov, A., Gavriliouk, V., Makarov, M., and Pavlov, A.: Verification
and Validation of EFD.Lab Code for Predicting Heat and Fluid Flow, In: CD-ROM
Proc. Int. Symp. on Advances in Computational Heat Transfer CHT-04, April 19-24,
2004, Norway, 21 p.
22 Visser, F.C., Brouwers, J.J.H., Jonker, J.B.: Fluid flow in a rotating low-specific-speed
centrifugal impeller passage. J. Fluid Dynamics Research, 24, pp. 275-292 (1999).
5
Technical Reference
1 Physical Capabilities of COSMOSFloWorks
With COSMOSFloWorks it is possible to study a wide range of fluid flow and heat
transfer phenomena that include the following:
External and internal fluid flows
Steady-state and transient fluid flows
Compressible gas and incompressible fluid flows (either in different projects or
simultaneously in different regions not in contact with each other)
Free, forced, and mixed convection
Fluid flows with boundary layers, including wall roughness effects
Laminar and turbulent fluid flows
Multi-species fluids and multi-component solids
Heat transfer in fluid and solid media with/without conjugate heat transfer and/or
contact heat resistance between solids and/or radiation heat transfer between
opaque solids (some solids can be considered as transparent for radiation)
Fluid flows in porous media
Flows of Non-Newtonian liquids
Flows of compressible liquids
Two-phase (fluid + particles) flows
Equilibrium volume condensation of water from steam and its influence on fluid
flow and heat transfer
Fluid flows in models with moving/rotating surfaces and/or parts
Chapter Governing Equations
5-2
2 Governing Equations
2.1 The Navier-Stokes Equations for Laminar and Turbulent Fluid Flows
COSMOSFloWorks solves the Navier-Stokes equations, which are formulations of mass,
momentum and energy conservation laws for fluid flows. The equations are supplemented
by fluid state equations defining the nature of the fluid, and by empirical dependencies of
fluid viscosity and/or thermal conductivity on temperature. Inelastic non-Newtonian fluids
are considered by introducing a dependency of their dynamic viscosity on flow shear rate
and temperature, compressible liquids are considered by introducing a dependency of their
density on pressure and temperature. A particular problem is finally specified by the
definition of its geometry, boundary and initial conditions.
COSMOSFloWorks is capable of predicting both laminar and turbulent flows. Laminar
flows occur at low values of the Reynolds number, which is defined as the product of
representative scales of velocity and length divided by the kinematic viscosity. When the
Reynolds number exceeds a certain critical value, the flow becomes turbulent, i.e. flow
parameters become randomly fluctuating.
Most of the fluid flows we encounter in engineering practice are turbulent, so
COSMOSFloWorks was mainly developed to simulate and study turbulent flows. To
predict turbulent flows, we use the Reynolds averaged Navier-Stokes equations, where
time-averaged effects of the flow turbulence on the flow parameters are considered,
whereas the other, i.e. large-scale, time-dependent phenomena are taken into account
directly. Through this procedure, extra terms known as the Reynolds stresses appear in the
equations for which additional information must be provided. To close this system of
equations, COSMOSFloWorks employs transport equations for the turbulent kinetic
energy and its dissipation rate, the so-called k- model.
COSMOSFloWorks employs one system of equations to describe both laminar and
turbulent flows. Moreover, transition from a laminar to turbulent state and/or vice versa is
possible.
Flows in models with moving walls (without changing the model geometry) are computed
by specifying the corresponding boundary conditions. Flows in models with rotating parts
are computed in coordinate systems attached to the models rotating parts, i.e. rotating with
them, so the models' stationary parts must be axisymmetric with respect to the rotation
axis.
The mass, momentum and energy conservation laws in a Cartesian coordinate system
rotating at the angular velocity about an axis passing through the coordinate system's
origin can be written in the conservation form as follows:
where u is the fluid velocity, is the fluid density, is a mass-distributed external force
per unit mass due to a porous media resistance (Si
porous
), a buoyancy (Si
gravity
), and the
coordinate systems rotation (Si
rotation
), i.e., Si = Si
porous
+ Si
gravity
+ Si
rotation
, E is the
total energy per unit mass, is a heat source or sink per unit volume, is the
viscous shear stress tensor, is the diffusive heat flux, and the subscripts are an
expression to denote summation over the three coordinate directions.
Following Boussinesq's assumption, the Reynolds stresses are defined in
this model as
For Newtonian fluids the viscous shear stress tensor is defined as
where
Here is the Kronecker delta function (it is equal to unity when i = j, and zero
otherwise), is the dynamic viscosity coefficient, is the turbulent eddy viscosity
coefficient and k is the turbulent kinetic energy. Note that and k are zero for laminar
( )
i
i
ik k i
k
i
S
x
P
u u
x t
u
=
( )
( ) ( )
H k k i ik k k
k
Q u S u q u P E
x t
E
+ = + +
( ) 0 =
k
k
u
x t

(5.1)
(5.2)
(5.3)
S
i
Q
H

i k
q
i
j i
R
ij
u u
ij ij
l
l
i
j
j
i
t
R
ij
k
x
u
x
u
x
u

3
2
3
2
|
|
.
|
\
|
=
(5.4)
(5.5)
ij ij
l
l
i
j
j
i
ij
k
x
u
x
u
x
u

3
2
3
2
|
|
.
|
\
|
=

l

t
+ =
(5.6)
i j
l

t
t
5-4
flows. In the frame of the k- turbulence model, is defined using two basic turbulence
properties, namely, the turbulent kinetic energy k and the turbulent dissipation ,
Here is a turbulent viscosity factor. It is defined by the expression
and y is the distance from the wall. This function allows us to take into account
laminar-turbulent transition.
Two additional transport equations are used to describe the turbulent kinetic energy and
dissipation,
where the source terms and are defined as
Here represents the turbulent generation due to buoyancy forces and can be written as
2
k C
f
t
=
(5.7)
f
( ) [ ]
|
|
.
|
\
|
+ =
T
y
R
R f
5 , 20
1 025 . 0 exp 1
2
(5.8) ,
where

l
T
k
R
2
=
,
l
y
y k
R
=
( )
k
k k
t
l
k
k
k
S
x
k
x
k u
x t
k
+
|
|
.
|
\
|
|
|
.
|
\
|
+
( )

S
x x
u
x t
k
t
l
k
k
k
+
|
|
.
|
\
|
|
|
.
|
\
|
+
,
,
(5.9)
(5.10)
S
k
S
B t
j
i R
ij k
P
x
u
S +
=
k
f C P C
x
u
f
k
C S
B B t
j
i R
ij
2
2
2
1
1

|
|
.
|
\
|
+
=
.
(5.11)
(5.12)
P
B
i B
i
B
x
g
P
=

1
(5.13)
where is the component of gravitational acceleration in direction , the constant
B
= 0.9, and constant is defined as: C
B
= 1when , and C
B
= 0 otherwise;
The constants , , , , are defined empirically. In COSMOSFloWorks
the following typical values are used:
C
= 0.09, C
1
= 1.44, C
2
= 1.92,
= 1.3,
The diffusive heat flux is defined as
Here the constant
c
= 0.9, and Pr is the Prandtl number, C
P
is specific heat at constant
pressure.
These equations describe both laminar and turbulent flows. Moreover, transitions from
one case to another and back are possible. The parameters k and are zero for purely
laminar flows.
Laminar/turbulent Boundary Layer Model
A laminar/turbulent boundary layer model is used to describe flows in near-wall regions.
The model is based on the so-called Modified Wall Functions approach. This model is
employed to characterize laminar and turbulent flows near the walls, and to describe
transitions from laminar to turbulent flow and vice versa. The model provides accurate
velocity and temperature boundary conditions for the above mentioned conservation
equations. Note that the model is not based on the widely used "law of the wall".
Constitutive Laws and Thermophysical Properties
The system of Navier-Stokes equations is supplemented by definitions of thermophysical
properties for the fluids. COSMOSFloWorks provides simulations of gas and liquid flows
with density, viscosity, thermal conductivity, specific heats, and species diffusivities as
functions of pressure, temperature and species concentrations in fluid mixtures, as well as
equilibrium volume condensation of water from steam can be taken into account when
simulating steam flows.
The state equation is:
g
i
x
i
C
B
P
B
0 >
3
1
05 . 0
1
|
|
.
|
\
|
+ =
f
f
,
( )
2
2
exp 1
T
R f =
(5.14)
C
C
1
C
2

k

k
1 =
(5.15)
k
P
c
t l
k
x
T
c q
|
|
.
|
\
|
+ =

Pr
, k = 1, 2, 3.
(5.16)
t
(5.17)
,
( ) y , ,T p =
5-6
For gas flows the equation of state of a perfect gas is employed,
where R is the gas constant and is equal to the universal gas constant R
univ
divided by the
molecular mass of the fluid M. For fluid mixtures R is defined as
where , m=1, 2, ...,M, are concentrations of mixture components, and is the
molecular mass of the m-th mixture component.
For liquid flows the state equation is defined as a dependency of density on temperature
and species concentrations in fluid mixtures,
where y = (y
1
, y
2
, ... y
M
).
As an example of the state equation, written for the particular case of incompressible
liquid flows, we have
For incompressible liquid mixtures:
Another example of the state equation refers to the case when heat transfer is involved in
the problem, and the density of a liquid is temperature-dependent. In this case the isobaric
density variation equation is commonly employed as the state equation,
Here
T
is the thermal volumetric expansion coefficient and
0
is the density at the
reference temperature T
0.
RT
P
=
(5.18)
,
=
m
m
m
univ
M
y
R R
(5.19) ,
y
m
M
m
(5.20)
,
( ) y , T =
(5.21)
const =
.
1
|
|
.
|
\
|
=

m
m
m
y
(5.22) .
( )
0
0
1 T T
T
+
=
(5.23)
For fluid mixtures the isobaric density variation equation can be written as
where
0
,m
is the density at the reference temperature T
0
and
T,m
is the thermal
volumetric expansion coefficient of the m-th component.
If necessary, both gas and liquid flows can be computed simultaneously, but in the
computational domain's regions partitioned by walls, i.e. gases-with-liquid mixtures are
not considered.
Compressible Liquids
Compressible liquids whose density depends on pressure and temperature can be
considered:
obeying the logarithmic law:
where
o
is the liquid's density under the reference pressure Po, C, B are
coefficients, here
o
, C, and B can depend on temperature, P is the calculated
pressure;
obeying the power law:
where, in addition to the above-mentioned designations, n is a power index, which
can depend on temperature.
Non-Newtonian Liquids
COSMOSFloWorks is capable to compute laminar flows of inelastic non-Newtonian
liquids. In this case the viscous shear stress tensor is defined, instead of Eq.(5.5), as
( ) ( )
0 ,
, 0
1
1
T T
y
m T
m
m
m
+
=
(5.24) ,
|
|
.
|
\
|
+
+
=
0
0
ln 1 /
P B
P B
C
,
n
B P
B P
/ 1
0
0
|
|
.
|
\
|
+
+
=
,
, (5.25)
( )
|
|
.
|
\
|
=
i
k
k
i
ik
x
u
x
u

5-8
where shear rate,
and for specifying a viscosity function the following three models of inelastic
non-Newtonian viscous liquids are available in COSMOSFloWorks:
The Herschel-Bulkley model:
where K is the liquid's consistency coefficient, n is the liquid's power-law index, and
is the liquid's yield stress. This model includes the following special cases:
n = 1, = 0 describes Newtonian liquids, in this case K is the liquid's dynamic
viscosity;
n = 1, > 0 describes the Bingham model of non-Newtonian liquids, featured by a
non-zero yield stress ( ), below of which the liquid behaves as a solid, so to
achieve a flow this threshold shear stress must be exceeded. (In
COSMOSFloWorksthis threshold is modeled by automatically equating K, named
plastic viscosity in this case, to a substantially high value at );
0 < n < 1, = 0 describes the power-law model of shear-thinning non-Newtonian
liquids (see also below).
n > 1, = 0 describes the power-law model of shear-thickening non-Newtonian
liquids (see also below).
The power-law model: ,
in contrast to the Herschel-Bulkley model's special case, the values are restricted:
min

max
;
The Carreau model:
where is the liquid's dynamic viscosity at infinite shear rate, i.e., the minimum
dynamic viscosity, is the liquid's dynamic viscosity at zero shear rate, i.e., the
maximum dynamic viscosity, K
1
is the time constant, n is the power-law index. This
model is a smooth version of the power-law model with the restrictions. In these
models, all parameters, with the exception of the dimensionless power-law index, are
temperature-dependent in a general case.
i
j
j
i
ij jj ii ij
x
u
x
u
d d d d
= = ,
2
( )
( ) ( )

o
n
K + =
1
,
o
o
<
o
( ) ( )
1
=
n
K
,
( ) ( ) [ ]
( ) 2 / 1
2
1
1

+ + =
n
o
K
,

Equilibrium volume condensation of water from steam
If the gas whose flow is computed includes steam, COSMOSFloWorks can predict an
equilibrium volume condensation of water from this steam (without any surface
condensation) taking into account the corresponding changes of the steams temperature,
density, enthalpy, specific heat, and sonic velocity. In accordance with the equilibrium
approach, the condensed waters local mass fraction in the local total mass of the steam
and the condensed water is determined from the fluids local temperature, pressure, and, if
a multi-component fluid is considered, the steams local mass fraction. Since this
condensation model is at equilibrium, the condensation has no history, i.e. it is a local fluid
property only.
In addition, it is assumed that
the condensed waters volume is equal to zero, so this prediction is correct only if
the condensed waters relative volume does not exceed 5%,
the steams temperature lie in the range of 283...610 K and the pressure does not
exceed 10 MPa.
Buoyancy Force
The mass-distributed force arising from buoyancy is built into COSMOSFloWorks. The
relevant expression is
where g
i
is the gravitational acceleration component (i.e., along the coordinate system's
i-th axis).
2.2 Conjugate Heat Transfer
COSMOSFloWorks allows predictions of simultaneous heat transfer in solid and fluid
media with energy exchange between them. Heat transfer in fluids is described by the
energy conservation equation (5.3) where the heat flux is defined by (5.16). The heat
conductivity phenomenon in isotropic solid media is described by the following equation:
where e is the specific internal energy, e = cT, c is specific heat, Q
H
is specific heat
release per unit volume, and is the thermal conductivity.
i i
g S =
(5.26) ,
H
i i
Q
x
T
x t
e
+
|
|
.
|
\
|
(5.27) ,
5-10
If a solid consists of several solids attached one to another, then the thermal contact
resistances between them (on their contact surfaces), specified in the Engineering database
in the form of contact conductance (as m
2
K/W), can be taken into account when
calculating the heat conduction in solids. As a result, a solid temperature step appears on
the contact surfaces. In the same manner, i.e. as a thermal contact resistance, a very thin
layer of another material between solids or on a solid in contact with fluid can be taken
into account when calculating the heat conduction in solids, but it is specified by this
layers material (its thermal conductivity taken from the Engineering database) and
thickness.
When predicting conjugate heat transfer, the continuity of energy flux is enforced at the
interface of fluid and solid media.
Radiation Heat Transfer Between Solids
In addition to heat transfer in solids, COSMOSFloWorks is capable to calculate radiation
heat transfer between solids whose surface emissivity is specified. If necessary, a heat
radiation from the computational domain's far-field boundaries or the model's openings to
the model surfaces can be defined and considered either as from solid surfaces, i.e. by
specifying these boundaries' emissivity and temperature, or as a solar radiation having a
specified constant or time-dependent direction and intensity. At that, the fluid is
considered as non-radiating and transparent for heat radiation. If necessary, some of solid
surfaces can be specified as non-radiating, and some of solids can be specified as
transparent for heat radiation. The radiative solid surfaces can be specified through their
emissivity as either blackbody, or whitebody, or ideal gray-body (i.e. having a continuous
emissive power spectrum similar to the blackbody one, so their monochromatic emissivity
is independent of the emission wavelength). In any case, the total radiation integrated over
all wavelengths is considered only, so the emitted heat is equal to Q
R
=
0
T
w
4
, where
is the surface emissivity,
0
is the Stefan-Boltzmann constant, T
w
is the surface
temperature. When distributing the radiation heat over the directions, the heat radiation
from a solid surface is assumed diffuse, i.e. obeying the Lambert law, according to which
the radiation intensity per unit area and per unit solid angle is the same in all directions.
As a result, the net radiation heat exchange between all the model's radiative surfaces is
calculated.
2.3 Rotating Reference Frame
The coordinate system's rotation is taken into account through the following
mass-distributed force:
i k j ijk rotation i
r u e S
2
2 + = ,
where e
ijk
is the Levy-Chivita symbols (function), is the rotation's angular velocity, r is
the radius-vector coming to the point under consideration from the nearest point lying on
the rotation axis. If necessary, several non-intersecting rotating reference frames can be
calculated within one analysis. Such rotating frames must be represented by an
axisymmetric region. A special boundary condition is automatically applied at the
boundaries of such region to allow correct interface between stationary and rotating
frames.
2.4 Mass Transfer in Fluid Mixtures
The mass transfer in fluid mixtures is governed by species conservation equations. The
equations that describe concentrations of mixture components can be written as
Here D
mn
, are the molecular and turbulent matrices of diffusion, S
m
is the rate of
production or consumption of the m-th component.
In case of Fick's diffusion law:
The following obvious algebraic relation between species concentrations takes place:
2.5 Flows in Porous Media
Porous media are treated in COSMOSFloWorks as distributed resistances to fluid flow, so
they can not occupy the whole fluid region. Therefore, they act on the flow through the Si
term in Eq.(5.2):
where k is the porous medium's resistance vector (see below), other designations are
defined in Section 1.
To perform a calculation in COSMOSFloWorks, you have to specify the following porous
medium properties: the effective porosity of the porous medium, defined as the volume
fraction of the interconnected pores with respect to the total medium volume; the porous
mediums permeability type from the following:
isotropic (i.e., the medium permeability is independent of direction),
( ) ( ) M 1,2,..., , = +
|
|
.
|
\
|
m S
x
y
D D
x
y u
x t
y
m
k
n
t
mn mn
k
m k
k
m
(5.28)
D
mn
t

t
mn
t
mn mn mn
D D D = = ,
(5.29)
=
m
m
y 1
.
(5.30)
j ij porous i
u k S =
,
5-12
unidirectional (i.e., the medium is permeable in one direction only),
axisymmetrical (i.e., the medium permeability is fully governed by its axial and
transversal components with respect to a specified direction),
orthotropic (i.e., the general case, when the medium permeability varies with
direction and is fully governed by its three components determined along three
principal directions);
the porous mediums resistance to fluid flow, i.e., vector k defined as k = - grad(P)/(V),
where P, , V are fluid pressure, density, and velocity respectively. In COSMOSFloWorks
you can specify the k vector components using one of the following four formulas:
k = PS/(mL), where P is the pressure difference between the opposite sides of a
sample parallelepiped porous body, m is the mass flow rate through the body, S and
L are, respectively, the body cross-sectional area and length, in the selected
direction. You can specify P as a function of m, whereas S and L are constants.
Instead of mass flow rate you can specify volume flow rate, v. In this case
COSMOSFloWorks calculates m = v. All these values do not specify the porous
body for the calculation, but the porous mediums k only.
k = (AV+B)/, where V is the fluid velocity, A and B are constants, is the fluid
density. Here, only A and B are specified, since V and are calculated.
k= /(D
2
), where and are the fluid dynamic viscosity and density, D is the
reference pore size determined experimentally. Here, only D is specified, since
and are calculated.
k= /(D
2
)f(Re), differing from the previous formula by the f(Re) factor, yielding
a more general formula. Here, additionally to D, f(Re) is specified as a formula.
To define a certain porous body, you specify both the body position in the model and, if
the porous medium has a unidirectional or axisymmetrical permeability, the reference
directions in the porous body.
2.6 Two-phase (fluid + particles) Flows
COSMOSFloWorks calculates two-phase flows as a motion of spherical liquid particles
(droplets) or spherical solid particles in a steady-state flow field. COSMOSFloWorks can
simulate dilute two-phase flows only, where the particles influence on the fluid flow
(including its temperature) is negligible (e.g. flows of gases or liquids contaminated with
particles).This are the cases where the particles mass flow rate fraction is lower than about
30% of the mixture mass flow rate.
The particles of a specified (liquid or solid) material and constant mass are assumed to be
spherical. Their drag coefficient is calculated with Hendersons formula (Ref.1), derived
for continuum and rarefied, subsonic and supersonic, laminar, transient, and turbulent
flows over the particles, and taking into account the temperature difference between the
fluid and the particle. The particle/fluid heat transfer coefficient is calculated with the
formula proposed in Ref.2. If necessary, the gravity is taken into account. Since the
particle mass is assumed constant, the particles cooled or heated by the surrounding fluid
change their size. The particles' interaction with the model surfaces is taken into account
by specifying the particle's either full absorption (that is typical for liquid droplets
impinging on surfaces at low or moderate velocities) or ideal or non-ideal reflection (that
is typical for solid particles). The ideal reflection denotes that in the impinging plane,
defined by the particle velocity vector and the surface normal at the impingement point,
the particle velocity's tangent-to-surface component is conserved, whereas the particle
velocity's normal-to-surface component changes its sign only. A non-ideal reflection is
specified by the two particle velocity restitution (reflection) coefficients, e
n
and e
,
determining values of these particle velocity components after reflection, V
2,n
and V
2,
, as
these values' ratio to the ones before the impingement, V
1,n
and V
1,
:
As a result of particles impingement on a solid surface, the total erosion mass rate,
R
erosion
, and the total accretion mass rate, R
accretion
, are determined from the following
formulae:
,
,
where:
N the number of the particles fractions specified by user as injections in
COSMOSFloWorks (all particles can be divided by user into fractions, i.e. injections, so
that each of these fractions unambiguously presents the particles properties, i.e.
temperature, velocity, diameter, mass flow rate, and material, at their inlet),
i the particle fractions number,
M
p i
the i-th particle fractions mass impinging on the model walls per a time unit,
K
i
the i-th particle fractions impingement erosion coefficient specified by user,
V
p i
the i-th particle fractions impingement velocity,
b the user-specified velocity exponent (b = 2 is recommended),
f
1 i
(
p i
) the user-specified dimensionless function of the particle impingement angle
p i
,
f
2 i
(d
p i
) the user-specified dimensionless function of the particle diameter d
p i
.
n , 1
V
n , 2
V
n
e =
, 1
V
, 2
V
e =
i p i p i 2 i p i 1
b
i p
N
1 i
M
i erosion
m d ) d ( f ) ( f V K R
i p
=

=
=
N
1 i
i p accretion
M R
5-14
2.7 Boundary Conditions
Internal Flow Boundary Conditions
For internal flows, i.e., flows inside models, COSMOSFloWorks offers the following two
options of specifying the flow boundary conditions: manually at the model inlets and
outlets (i.e. model openings), or to specify them by transferring the results obtained in
another COSMOSFloWorks calculation in the same coordinate system (if necessary, the
calculation can be performed with another model, the only requirement is the flow regions
at the boundaries must coincide).
With the first option, all the model openings are classified into "pressure" openings,
"flow" openings, and "fans", depending on the flow boundary conditions which you
intend to specify on them.
A "pressure" openings flow boundary condition, which can be static pressure, or total
pressure, or environment pressure is imposed in the general case when the flow direction
and/or magnitude at the model opening are not known a priori, so they are to be calculated
as part of the solution. Which of these parameters is specified depends on which one of
them is known. In most cases the static pressure is not known, whereas if the opening
connects the computational domain to an external space with known pressure, the total
pressure at the opening is known. The Environment pressure condition is interpreted by
COSMOSFloWorks as a total pressure for incoming flows and as a static pressure for
outgoing flows. If, during calculation, a vortex crosses an opening with the Environment
pressure condition specified at it, this pressure considered as the total pressure at the part
of opening through which the flow enters the model and as the static pressure at the part of
opening through which the flow leaves the model.
Note that when inlet flow occurs at the "pressure" opening, the temperature, fluid mixture
composition and turbulence parameters have to be specified also.
A "flow" openings flow boundary condition is imposed when dynamic flow properties
(i.e., the flow direction and mass/volume flow rate or velocity/ Mach number) are known
at the opening. If the flow enters the model, then the inlet temperature, fluid mixture
composition and turbulence parameters must be specified also. The pressure at the
opening will be determined as part of the solution. For supersonic flows the inlet pressure
must be specified also.
A "fan" condition simulates a fan installed at a model opening. In this case the
dependency of volume flow rate on pressure drop over the fan is prescribed at the opening.
These dependencies are commonly provided in the technical documentation for the fans
being simulated.
With the second option, you specify the boundary conditions by transferring the results
obtained in another COSMOSFloWorks calculation in the same coordinate system. If
necessary, the calculation can be performed with another model, the only requirement is
the flow regions at the boundaries must coincide. At that, you select the created boundary
conditions type: either as for external flows (so-called "ambient" conditions, see the next
Section), or as for "pressure" or "flow" openings, see above. If a conjugate heat transfer
problem is solved, at the boundarys part lying in a solid body, the temperature is
transferred from the other calculation.
Naturally, the flow boundary conditions specified for an internal flow problem with the
first and/or second options must be physically consistent with each other, so it is expedient
to specify at least one "pressure"-type boundary condition and at least one "flow"-type
boundary condition, if "ambient" boundary conditions are not specified only.
External Flow Boundary Conditions
For external problems such as flow over an aircraft or building, the parameters of the
external incoming flow (so-called "ambient" conditions) must be defined. Namely the
velocity, pressure, temperature, fluid mixture composition and turbulence parameters
must be specified. Evidently, during the calculation they can be partly violated at the flow
boundary lying downstream of the model.
Wall Boundary Conditions
In COSMOSFloWorks the default velocity boundary condition at solid walls corresponds
to the well-known no-slip condition. The solid walls are also considered to be
impermeable. In addition, the wall surface's translation and/or rotation (without changing
the model's geometry) can be specified. If a calculation is performed in a rotating
coordinate system, then some of the wall surfaces can be specified as stationary, i.e. a
backward rotation in this coordinate system (without changing the model's geometry).
COSMOSFloWorks also provides the "Ideal Wall" condition that corresponds to the
well-known slip condition. For example, Ideal Walls can be used to model planes of flow
symmetry.
If the flow of non-Newtonian liquids is considered, then the following slip condition at
solid walls can be specified: if the shear stress exceeds the yield stress value
0,slip
, then
a slip velocity v
slip
determined from , where C
1
and C
2
, as well
as
0,slip
, if it is absent in the non-Newtonian liquids model, it then can be specified by
user.
If conjugate heat transfer in fluid and solid media is not considered, one of the following
boundary conditions can be imposed at solid walls: either the wall temperature
2
C
slip 0, 1 slip
) ( C v =
w
T T =
(5.31)
,
Chapter Numerical Solution Technique
5-16
or the heat flux,
being positive for heat flows from fluid to solid, equal to zero for adiabatic
(heat-insulated) walls, and negative for heat flows from solid to fluid.
When considering conjugate heat transfer in fluid and solid media, the heat exchange
between fluid and solid is calculated by COSMOSFloWorks, so heat wall boundary
conditions are not specified at the walls.
Internal Fluid Boundary Conditions
If one or several non-intersecting axisymmetric rotating regions (local rotating reference
frames) are specified, the following internal boundary condition is applied at the
boundaries of the regions: the flow parameter values must be axisymmetric.
3 Numerical Solution Technique
The numerical solution technique employed in COSMOSFloWorks is robust and reliable,
so does not require any user knowledge about the computational mesh and the numerical
methods employed. But sometimes, if the model and/or the problem being solved are too
complicated, so that the COSMOSFloWorks standard numerical solution technique
requires extremely high computer resources (memory and/or CPU time) which are not
available, it is expedient to employ COSMOSFloWorks options which allow the
adjustment of the numerical solution techniques governing parameters, specified
automatically. To employ these options properly and successfully, take into account the
information presented below about COSMOSFloWorks numerical solution technique.
Briefly, COSMOSFloWorks solves the governing equations with the finite volume (FV)
method on a spatially rectangular computational mesh designed in the Cartesian
coordinate system with the planes orthogonal to its axes and refined locally at the
solid/fluid interface and, if necessary, additionally in specified fluid regions, at the
solid/solid surfaces, and in the fluid region during calculation. Values of all the physical
variables are stored at the mesh cell centers. Due to the FV method, the governing
equations are discretized in a conservative form. The spatial derivatives are approximated
with implicit difference operators of second-order accuracy. The time derivatives are
approximated with an implicit first-order Euler scheme. The numerical schemes viscosity
is negligible with respect to the fluid viscosity.
3.1 Computational Mesh
COSMOSFloWorks computational mesh is rectangular everywhere in the computational
domain, so the mesh cells sides are orthogonal to the specified Cartesian coordinate
systems axes and are not fitted to the solid/fluid interface. As a result, the solid/fluid
interface cuts the near-wall mesh cells. Nevertheless, due to special measures, the mass
and heat fluxes are treated properly in these cells named partial.
w
q q =
(5.32)
The rectangular computational domain is automatically constructed (may be changed
manually), so it encloses the solid body and has the boundary planes being orthogonal to
the specified Cartesian coordinate systems axes. Then, the computational mesh is
constructed in the following several stages.
First of all, a basic mesh is constructed. For that, the computational domain is divided into
slices by the basic meshs planes, which are evidently orthogonal to the Cartesian
coordinate systems axes. The user can specify the number and spacing of these planes
along each of the axes. The so-called control planes whose position is specified by User
can be among these planes also. The basic mesh is governed by the computational domain
only and does not depend on the solid/fluid interfaces.
Then, the basic meshs cells cut by the solid/fluid interface are split uniformly into smaller
cells in order to capture the solid/fluid interface with mesh cells of the specified size
(with respect to the basic meshs cells). The following procedure is employed: each of the
basic meshs cells cut by the solid/fluid interface is split uniformly into the 8 daughter
cells; in its turn, each of the daughter cells cut by the interface is split further into its 8
daughter cells, and so on, until the specified cell size is attained.
At the next meshing stage, the mesh obtained at the solid/fluid interface with the previous
procedure is refined in accordance with the solid/fluid interface curvature and the mesh
splitting and merging criteria: if the angle between the normals dropped from the
neighboring daughter cells centers to the solid/fluid interface exceeds the higher
criterions value, each of these daughter cells is split into 8 cells, whereas if this angle is
less than the lower criterions value, all 8 daughter cells are merged into the parental cell.
Finally, the mesh obtained with these procedures is refined in the computational domain to
satisfy the so-called narrow channel criterion: for each cell lying at the solid/fluid
interface the number of the mesh cells (including the partial cells) lying in the fluid region
along the normal dropped to the solid/fluid interface from this cells center must not be
less than the criterions value. If this number is less, then each of the mesh cells in this
direction is split into the 8 daughter cells.
As a result of all these meshing procedures, a locally refined rectangular computational
mesh is obtained and used then for solving the governing equations on it.
Since all the above-mentioned meshing procedures are performed before the calculation,
the obtained mesh is unable to resolve all the solution features well. To overcome this
disadvantage, the computational mesh can be refined further at the specified moments
during the calculation in accordance with the solution spatial gradients (both in fluid and
in solid, see Users Guide for details). As a result, in the low-gradient regions the cells are
merged, whereas in the high-gradient regions the cells are split. The moments of refining
the computational mesh during the calculation are prescribed either automatically or by
the user.
5-18
3.2 Spatial Approximations
The cell-centered finite volume (FV) method is used to obtain conservative
approximations of the governing equations on the locally refined rectangular mesh. The
governing equations are integrated over a control volume which is a grid cell, and then
approximated with the cell-centered values of the basic variables. The integral
conservation laws may be represented in the form of the cell volume and surface integral
equation:
are replaced by the discrete form
The second-order upwind approximations of fluxes F are based on the implicitly treated
modified Leonard's QUICK approximations (Ref. 3) and the Total Variation Diminishing
(TVD) method (Ref. 4).
In COSMOSFloWorks, especially consistent approximations for the convective terms, div
and grad operators are employed in order to derive a discrete problem that maintains the
fundamental properties of the parent differential problem in addition to the usual
properties of mass, momentum and energy conservation.
Spatial Approximations at the Solid/fluid Interface
Considering equation (5.34) for partial mesh cells (i.e., for the mesh cells cut by the
solid/fluid interface), we introduce the additional boundary faces and the corresponding
boundary fluxes taking the boundary conditions and geometry into account (see Fig.3.1),
as well as use a special calculation procedure for them. As a result, the solid/fluid interface
influence on the problem solution both in the fluid and in the solid is calculated very
accurately.

= +
dv ds dv Q F U
t
(5.33)
( ) Qv S F Uv = +
faces cell
t

(5.34)
3.3 Temporal Approximations
Time-implicit approximations of the continuity and convection/diffusion equations (for
momentum, temperature, etc.) are used together with an operator-splitting technique
(Refs. 5, 6, and 7). This technique is used to efficiently resolve the problem of
pressure-velocity decoupling. Following the SIMPLE-like approach (Ref. 8), an elliptic
type discrete pressure equation is derived by algebraic transformations of the originally
derived discrete equations for mass and momentum, and taking into account the boundary
conditions for velocity.
Fig.3.1. Computational mesh cells at the solid/fluid interface.
5-20
3.4 Form of the Numerical Algorithm
Let index 'n' denotes the time-level, and '*' denotes intermediate values of the flow
parameters. The following numerical algorithm is employed to calculate flow parameters
on time-level (n+1) using known values on time-level (n):
* = (p
n
+p,T*,y*),
Here U = (u, T, , , y)
T
is the full set of basic variables excluding pressure p,
u=(u
1
,u
2
,u
3
)
T
is the velocity vector, y = (y
1
, y
2
, ..., y
M
)
T
is the vector of component
concentrations in fluid mixtures, and p = p
n+1
- p
n
is an auxiliary variable that is called
a pressure correction. These parameters are discrete functions stored at cell centers. They
are to be calculated using the discrete equations (5.35)-(5.40) that approximate the
governing differential equations. In equations (5.35)-(5.40) A
h
, div
h
, grad
h
and L
h
=
div
h
grad
h
are discrete operators that approximate the corresponding differential operators
with second order accuracy.
Equation (5.35) corresponds to the first step of the algorithm when fully implicit discrete
convection/diffusion equations are solved to obtain the intermediate values of momentum
and the final values of turbulent parameters, temperature, and species concentrations.
The elliptic type equation (5.36) is used to calculate the pressure correction p. This
equation is defined in such a way that the final momentum field u
n+1
calculated from
(5.35) satisfies the discrete fully implicit continuity equation. Final values of the flow
parameters are defined by equations (5.37-5.40).
( )
n n
h
S p A
t
= +
*
*
U U
U - U
,
n
n
,
( )
,
t t t
p L
n
h
h
* *
1 u div
(5.35)
(5.36)
, p p p
n n
+ =
+1
, , , , T T
1 n n 1 n n * * 1 * * 1
y y = = = =
+ + + +
( ) y
1 1 1 1
, ,
+ + + +
=
n n n n
T p
(5.37)
(5.38)
(5.39)
(5.40)
,
1
p t
h
n
grad =
+ *
u u
.
3.5 Methods to Resolve Linear Algebraic Systems
Iterative Methods for Nonsymmetrical Problems
To solve the asymmetric systems of linear equations that arise from approximations of
momentum, temperature and species equations (5.35), a preconditioned generalized
conjugate gradient method (Ref. 9) is used. Incomplete LU factorization is used for
preconditioning.
Iterative Methods for Symmetric Problems
To solve symmetric algebraic problem for pressure-correction (5.36), an original
double-preconditioned iterative procedure is used. It is based on a specially developed
multigrid method (Ref. 10).
Multigrid Method
The multigrid method is a convenient acceleration technique which can greatly decrease
the solution time. Basic features of the multigrid algorithm are as follows. Based on the
given mesh, a sequence of grids (grid levels) are constructed, with a decreasing number of
nodes. On every such grid, the residual of the associated system of algebraic equations is
restricted onto the coarser grid level, forming the right hand side of the system on that
grid. When the solution on the coarse grid is computed, it is interpolated to the finer grid
and used there as a correction to the result of the previous iteration. After that, several
smoothing iterations are performed. This procedure is applied repeatedly on every grid
level until the corresponding iteration meets the stopping criteria.
The coefficients of the linear algebraic systems associated with the grid are computed
once and stored.
Chapter References
5-22
References
1 Henderson, C.B. Drag Coefficients of Spheres in Continuum and Rarefied Flows.
2 Carlson, D.J. and Hoglund, R.F. Particle Drag and Heat Transfer in Rocket Nozzles.
3 Roache, P.J., (1998) Fundamentals of Computational Fluid Dynamics, Hermosa
Publishers, Albuquerque, New Mexico, USA.
4 Hirsch, C., (1988). Numerical computation of internal and external flows. John Wiley
and Sons, Chichester.
5 Glowinski, R. and P. Le Tallec, (1989). Augmented Lagrangian Methods and
Operator-Splitting Methods in Nonlinear Mechanics. SIAM, Philadelphia.
6 Marchuk, G.I., (1982). Methods of Numerical Mathematics, Springer-Verlag, Berlin.
7 Samarskii, A.A., (1989). Theory of Difference Schemes, Nauka, Moscow, (in Russian).
8 Patankar, S.V., (1980). Numerical Heat Transfer and Fluid Flow, Hemisphere,
Washington, D.C.
9 Saad, Y. (1996). Iterative methods for sparse linear systems, PWS Publishing
Company, Boston.
10 Hackbusch, W., (1985). Multi-grid Methods and Applications, Springer-Verlag, NY.

COSMOS FloWorks (Fundamentals)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

COSMOS FloWorks (Fundamentals)

Uploaded by

Copyright:

Available Formats

2006

COSMOSFloWorks 2006 Fundamentals i

Chapter Laminar Flows Between Two Parallel Plates

Chapter Laminar Flows Between Two Parallel Plates

Chapter Laminar and Turbulent Flows in Pipes

COSMOSFloWorks 2006 Fundamentals 5-9

You might also like