Print Fluent Newsletter - MESco
Print Fluent Newsletter - MESco
Print Fluent Newsletter - MESco
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>Fluent</strong> News<br />
APPLIED COMPUTATIONAL FLUID DYNAMICS VOL XV ISSUE 1 • SPRING 2006<br />
ACADEMIC<br />
A Dry Passage<br />
to the Afterlife<br />
AUTOMOTIVE<br />
Citroën CS in<br />
a Crosswind<br />
Flexible<br />
Fliers<br />
ENVIRONMENTAL<br />
Activated Sludge<br />
Basins Get on Track<br />
FOOD<br />
Looking Inside<br />
Dough Mixers<br />
PROCESS INDUSTRIES<br />
SUPPLEMENT INSIDE!
EDITOR’S NOTE<br />
THE TERM “MULTIPHYSICS” has been a part of<br />
simulation engineers’ vocabulary for some time. It<br />
describes any situation where two or more physical<br />
phenomena are coupled together. Often, the term is<br />
used to describe simulations that involve both fluid<br />
and structural mechanics, such as fluid-structure interaction<br />
(FSI). During the past few years, we have run<br />
stories on FSI in <strong>Fluent</strong> News and in the current issue,<br />
several stories are featured that illustrate different<br />
approaches to this difficult engineering problem. The<br />
articles cover applications ranging from aerospace<br />
(p. 5) to healthcare (p. 10). Most describe tightly<br />
coupled interactions between CFD and structural<br />
solvers and one illustrates the use of MpCCI (Meshbased<br />
parallel Code Coupling Interface, from<br />
Fraunhofer SCAI) to manage the coupled calculation<br />
(p. 11). Not all such applications are tightly coupled<br />
however, and the Support Corner (p. 34) describes a<br />
tool, based on user-defined functions in FLUENT, that<br />
can be used to transfer data between fluid and structural<br />
solvers for loosely coupled interactions.<br />
In addition to the FSI stories, several of the articles in<br />
the Process Industries supplement could be given the<br />
multiphysics label as well. Multiphase flows and reacting<br />
flows are examples where strong coupling exists<br />
between either separate fluid phases (p. S13 - S14) or<br />
chemical species through temperature-dependent<br />
reactions (p. S8 - S14). Traditional processes such as<br />
mixing are covered (p. S3 - S6) as are novel processes<br />
that involve the flow through microchannels (p. S7) or<br />
the extrusion of foam products (p. S16).<br />
As is customary with <strong>Fluent</strong> News, the current slate of<br />
articles represents engineering efforts from many corners<br />
of the world. Air conditioning units being manufactured<br />
in Malaysia (p. 13) and aircraft fuselage design<br />
being performed in China (p. 19) are two examples.<br />
There are several articles from the US and Europe,<br />
covering topics such as automotive components (p. 16<br />
- 17), a unique dough conditioner (p. 22), a car in a<br />
crosswind (p. 14), and a two-stroke engine (p. 18). A<br />
few new applications of CFD are also presented. The<br />
flow inside an infant incubator is optimized (p. 24), and<br />
a ventilation system is installed in an Egyptian tomb to<br />
help preserve the ancient wall paintings (p. 28).<br />
Students shine once again with their unbound energy<br />
for innovation. This time, their efforts are used to develop<br />
a human-powered submarine (p. 29).<br />
Last November, a 64-bit version of FLUENT running on<br />
a Windows Cluster was showcased at Supercomputing<br />
2005. The performance of this exciting new capability<br />
is summarized (p. 30). In other product news, the<br />
upcoming releases of GAMBIT 2.3 and TGrid 4.0 are<br />
reviewed (p. 32). These new products and computing<br />
opportunities will make our work more manageable<br />
in the months and years to come. Please continue<br />
to keep us informed of your own efforts to push the<br />
limits of CFD and let us know what CFD has been able<br />
to do for you. <br />
LIZ MARSHALL<br />
fluentnews@fluent.com<br />
<strong>Fluent</strong> News is published by<br />
10 Cavendish Court • Lebanon, NH 03766 USA<br />
603 643 2600 • www.fluent.com<br />
© 2006 <strong>Fluent</strong> Inc. All rights reserved.<br />
ON THE COVER:<br />
Contours of presssure on the wing of the<br />
Aermacchi M346 Advanced Trainer<br />
Courtesy of Politecnico di Milano and Aermacchi SpA<br />
ON THE SUPPLEMENT COVER:<br />
Contours of velocity magnitude on a plane between<br />
two impellers in a stirred tank and vortex structures<br />
near the upper impeller, colored by velocity magnitude<br />
Courtesy of Prague Institute of Chemical Technology<br />
Editor: Liz Marshall<br />
Assistant Editor: Susan Wheeler<br />
Contributing Editors: Erik Ferguson and Keith Hanna<br />
Design: Lufkin Graphic Designs<br />
FLUENT, FiDAP, GAMBIT, POLYFLOW, G/Turbo, MixSim,<br />
FlowLab, Icepak, Airpak, and FloWizard are trademarks of<br />
<strong>Fluent</strong> Inc. Icepak and Airpak are joint developments of<br />
<strong>Fluent</strong> Inc. and ICEM-CFD Engineering. All other products<br />
or name brands are trademarks of their respective holders.
CONTENTS<br />
16<br />
20 13<br />
17<br />
8<br />
29 S3<br />
5<br />
S5 10<br />
31<br />
FEATURES<br />
APPLICATIONS<br />
5<br />
FLUID-STRUCTURE<br />
INTERACTION<br />
Flexible Fliers in the<br />
Transonic Regime<br />
13 HVAC<br />
8<br />
Unconventional Sail Design<br />
10<br />
Artificial Heart Valve Takes<br />
Shape<br />
16<br />
11 FSI Controls Flow Rate<br />
Fan Research Makes<br />
Cool Air Conditioners<br />
14 AUTOMOTIVE<br />
Citroën C5 in a<br />
Crosswind<br />
17<br />
Torque Converters<br />
Get In Gear<br />
Emissions Control<br />
Through Carbon<br />
Canisters<br />
18<br />
POWER TOOLS<br />
Scavenging in a<br />
Stratified, Charged<br />
Two-Stroke Engine<br />
19 AEROSPACE<br />
Nacelle Impact on<br />
Aircraft Wing &<br />
Fuselage<br />
20<br />
Compressors Benefit<br />
from the NASA<br />
Rotor 37<br />
21 FOOD<br />
Image-based Meshing:<br />
Easy as Pie<br />
22<br />
Looking Inside Dough<br />
Mixers<br />
24 HEALTHCARE<br />
CFD Assists Neonatal<br />
Intensive Care<br />
26 ENVIRONMENTAL<br />
Activated Sludge<br />
Basins Get on Track<br />
<strong>Fluent</strong> News · Spring 2006 3
CONTENTS<br />
DEPARTMENTS<br />
28<br />
ACADEMIC NEWS<br />
A Dry Passage to the Afterlife<br />
29<br />
Student Submariners Peddle<br />
Their Way to Victory<br />
14<br />
11<br />
29<br />
Convective Motions Inside a<br />
Gearbox<br />
30<br />
PRODUCT NEWS<br />
<strong>Fluent</strong> & Microsoft Team to<br />
Deliver 64-bit FLUENT on<br />
Windows Clusters<br />
31<br />
Quick Turnaround with<br />
Rapid Flow Modeling<br />
32<br />
Impressive Preprocessing<br />
33 PARTNERSHIPS<br />
Discrete Element Modeling<br />
of Particles for FLUENT<br />
34<br />
SUPPORT CORNER<br />
Mapping Thermal Data<br />
from FLUENT to Structural<br />
Codes Quickly<br />
23<br />
S7 32<br />
36<br />
AROUND FLUENT<br />
<strong>Fluent</strong> Opens Larger Office<br />
in Ann Arbor, MI<br />
17<br />
27<br />
36<br />
Upcoming User Group<br />
Meetings<br />
PROCESS INDUSTRY<br />
SUPPLEMENT<br />
S6<br />
s2 OVERVIEW<br />
Transporting CFD to<br />
Process Engineers<br />
s12 FURNACES<br />
Ultra-low NOx<br />
Burners Get Cracking<br />
S11<br />
S15<br />
s3 MIXING<br />
Tracing Homogenization<br />
s5<br />
Finding the Optimum<br />
Blend Time Calculation<br />
s7 MICROREACTORS<br />
Liquid Mixing<br />
in Microreactors<br />
s8<br />
THERMAL RUNAWAY<br />
Preventing Runaway<br />
Reaction Accidents<br />
s10 EMISSIONS<br />
Scrubbers for<br />
Flue Gas Cleanup<br />
s13 MULTIPHASE<br />
Controlling Droplet<br />
Size Distribution in<br />
Emulsions<br />
s14 Understanding<br />
Fluid-Bed Coating<br />
s15 PUMPS<br />
Pumping out New<br />
Designs More Quickly<br />
s16 EXTRUSION<br />
Polymer Processing<br />
Simulation for Foam<br />
Extrusion<br />
4 <strong>Fluent</strong> News · Spring 2006
FLUID-STRUCTURE INTERACTION<br />
Flexible Fliers<br />
in the Transonic Regime<br />
By L. Cavagna, G. Quaranta, P. Mantegazza, Department of Aerospace Engineering, Politecnico di Milano, Italy<br />
D. Marchetti, M. Martegani, M346 Program, Aermacchi SpA, Varese, Italy<br />
Pressure contours on the surface of the<br />
M346 Advanced Trainer<br />
FOR AN AIRCRAFT IN TRANSONIC FLIGHT, shock<br />
waves appear and move through the flow field. As a<br />
consequence of the unsteady flexible motion of the<br />
aircraft skin, shock waves dynamically modify the<br />
pressure distribution, eventually causing a drop in<br />
the flutter velocity, or speed at which the aircraft<br />
vibrations are no longer damped and instead grow<br />
in amplitude. The premature onset of flutter during<br />
transonic conditions, otherwise known as the “transonic<br />
dip” effect, is usually underpredicted by classical<br />
unsteady potential methods [1]. In a research<br />
project at the Politecnico di Milano [2], a computational<br />
aeroelastic (CA) model for deformable aircrafts<br />
in the transonic regime has been developed. Special<br />
care has been taken to create an environment that<br />
integrates all of the relevant physics, keeping in<br />
mind the high number of analyses that typically have<br />
to be run during the development of an aircraft. The<br />
implemented methods have been applied to the<br />
new generation Aermacchi M346 trainer.<br />
The aeroelastic solver core is implemented in FLUENT<br />
through the Scheme programming language.<br />
Access to flow variables and data manipulation is<br />
permitted through an extensive use of user-defined<br />
functions (UDFs) called at different times by the<br />
master Scheme library. The structural model comes<br />
from the commercial finite-element code MSC-<br />
Nastran, which has a long tradition in the aerospace<br />
industry.<br />
The adoption of a partitioned approach for the solution<br />
of fluid-structure interaction (FSI) problems<br />
requires the definition of an interface method to<br />
The Advanced Trainer M346 developed by Aermacchi<br />
<strong>Fluent</strong> News · Spring 2006 5
FLUID-STRUCTURE INTERACTION<br />
The aeroelastic procedure is illustrated for geometry elements (top) and the<br />
resulting mesh (bottom) for a case of amplified motion<br />
The AGARD 445.6 wing showing two modes of<br />
structural deformation<br />
exchange displacements and velocities from the<br />
structural grid to the aerodynamic wet surfaces of<br />
the CFD grid and to transfer back aerodynamic<br />
forces on the structural nodes. The structural and<br />
CFD models are described in a very different and<br />
often not compatible way; this is especially true in<br />
an industrial environment, where the models usually<br />
come from different departments. For the CA<br />
model developed at Politecnico di Milano, an interfacing<br />
procedure based on a “mesh-free” moving<br />
least squares (MLS) method [3] is used. This method<br />
is suitable for the treatment of geometrically complex<br />
configurations and, unlike many other interface<br />
methods, it ensures the conservation of momentum<br />
and energy transferred between the fluid and structure,<br />
a key factor for stability analysis like flutter. In<br />
fact, the comparison of spurious energy created or<br />
destroyed by the interface scheme may alter the<br />
stability boundary of the system.<br />
The primary assessment necessary for the aeroelastic<br />
certification of an aircraft is related to analyses of the<br />
linearized solutions to catch instabilities and specifically,<br />
flutter conditions. In the case of strong nonlinearities<br />
in the flow field, it is necessary to assess the<br />
stability of each movement associated with each<br />
equilibrium point of the aeroelastic system.<br />
Consequently, any flight configuration could potentially<br />
assume a different stability behavior. However, if<br />
there are no abrupt changes in the fluid flow, it is<br />
reasonable to consider the linearization around a specific<br />
flight condition as representative of the behavior<br />
at nearby flight points, which are characterized by<br />
small differences in the mass and stiffness distributions,<br />
and consequently small variations in the aircraft<br />
attitude. In the work presented by Melville [4] on an<br />
application similar to the one considered here, a<br />
stability investigation was conducted following a<br />
“numerical experiment” technique in which the<br />
stability is determined by analyzing the decaying or<br />
diverging behavior of time responses. This method is<br />
extremely time-consuming because a large number<br />
of analyses are required to bracket the flutter conditions.<br />
An approach based on the adoption of a CFD<br />
solver for generating reduced order models (ROMs) of<br />
transonic aerodynamics is more affordable and<br />
industrially oriented. As in classical flutter analyses,<br />
the result from this type of approach is a linearized<br />
model in either the frequency or time domain that<br />
can be used as an efficient tool for stability boundary<br />
assessment and dynamic loading response analysis.<br />
In order to avoid any possible source of error, a<br />
backup procedure should also be available to run<br />
the coupled nonlinear analysis. This alternative<br />
procedure can be used to verify and validate key<br />
instability points obtained by using the simplified<br />
linearized approach. A direct time integration can<br />
be easily implemented for this purpose using a<br />
loosely coupled, partitioned algorithm.<br />
6 <strong>Fluent</strong> News · Spring 2006
FLUID-STRUCTURE INTERACTION<br />
To correctly represent the structural deformation of<br />
the aircraft, the CFD computational grid must be<br />
modified at each time step in order to be compatible<br />
with the new aircraft shape. Grid deformation<br />
has a significant impact on the time required by<br />
CFD for aeroelastic simulations. For the sake of overall<br />
computational efficiency, nonlinear models for<br />
grid deformation have been avoided. Instead, an<br />
elastic analogy is used to represent the grid as a linear<br />
elastic continuum with a local Young’s modulus<br />
proportional to the minimal dimension of each element.<br />
The structural analogy provides ample freedom<br />
for choosing the material constitutive properties<br />
that will rule the grid deformation. Furthermore,<br />
by dividing the computational domain into different<br />
fluid zones, the deformation process can be customized<br />
by choosing different structural properties<br />
for the adjacent wall zones. The result is a flexible<br />
and extremely robust grid deformation algorithm.<br />
The structural model is represented as a set of<br />
modal shapes. To accelerate the solution time, the<br />
linearity of the problem is exploited, allowing the<br />
CFD grid deformation to be expressed as a superposition<br />
of grid deformations computed for each<br />
modal shape. All of the deformed grids associated<br />
with each modal shape are computed only once<br />
and stored in a database. Special care must be<br />
taken in the treatment of the governing surface<br />
movements since surface rotation can modify the<br />
CFD domain topology by creating cuts and new<br />
wall surfaces at the boundaries. As a result, the correct<br />
treatment of these conditions ideally requires<br />
the creation of a new mesh during the transient<br />
simulation, significantly increasing the computational<br />
burden. To efficiently overcome this hurdle, a<br />
solution based on a non-conformal mesh topology<br />
is used. This allows distinct cell domains to be<br />
defined that are associated with the movable surfaces<br />
that may be deformed independently from<br />
the rest of the grid. The implemented solver<br />
has been applied to the AGARD 445.6 wing, a<br />
classic aeroelastic benchmark case with published<br />
data [5]. The results are in good agreement with<br />
measurements in the transonic regime and correctly<br />
predict the transonic dip. <br />
References:<br />
1 Isogai, K.: On the Transonic-dip Mechanism of Flutter<br />
of a Sweptback Wing. AIAA Journal, 17:735-795, 1979.<br />
2 Cavagna, L., Quaranta, G., Ghiringhelli, G.L. and<br />
Mantegazza, P.: Efficient Application of CFD Aeroelastic<br />
Methods Using Commercial Software. In International<br />
Forum on Aeroelasticity and Structural Dynamics<br />
IFASD-2005, Munich, Germany, June 28 - July 1 2005.<br />
3 Quaranta, G., Masarati, P. and Mantegazza, P.: A<br />
Conservative Mesh-free Approach for Fluid-structure<br />
Interface Problems. Papadrakakis, M., Oñate, E. and<br />
Schrefler, B., Editors, in International Conference on<br />
Computational Methods for Coupled Problems in Science<br />
and Engineering, CIMNE, Santorini, Greece, 2005.<br />
4 Melville, R.: Nonlinear Mechanisms of Aeroelastic<br />
Instability for the F-16. AIAA Paper 2002-0871, January<br />
2002.<br />
5 Yates, E.C.: AGARD Standard Aeroelastic Configurations<br />
for Dynamic Response. I wing 445.6. R 765, AGARD,<br />
1985.<br />
V F<br />
Modal amplitude<br />
Modal amplitude<br />
0.6<br />
0.5<br />
0.4<br />
0.3<br />
Experiment<br />
FLUENT<br />
0.2<br />
0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2<br />
Mach number<br />
The transonic dip, or decrease in flutter velocity index<br />
in the transonic regime, is nicely captured by the<br />
computational aeroelastic model<br />
.002<br />
.001<br />
.000<br />
-.001<br />
Mode 1<br />
Mode 2<br />
-.002<br />
0 8 16 24 32 40 48<br />
Structural dimensionless time<br />
.15<br />
.10<br />
.05<br />
.0<br />
-.05<br />
-.10<br />
Mode 1<br />
-.15<br />
Mode 2<br />
-.20<br />
0 4 8<br />
12 16<br />
Structural dimensionless time<br />
Modal amplitude histories for the AGARD 445.6<br />
wing for the case with Mach number 0.678 and<br />
flutter velocity index 0.34 (top) and 0.5 (bottom)<br />
Different fluid domains and a non-conformal mesh are used to control the motion of the various surfaces<br />
<strong>Fluent</strong> News · Spring 2006 7
FLUID-STRUCTURE INTERACTION<br />
Unconventional<br />
Sail Design<br />
By Paolo Conti, Marco Argento, Matteo Scarponi, Dipartimento Ingegneria Industriale, Università degli Studi di Perugia, Italy<br />
Tornado class catamaran<br />
Photo Courtesy of HoldenSailing<br />
IN THE WORLD OF DINGHY AND YACHT<br />
racing, the sail design process is largely<br />
based on the experienced eye of the sailmaker;<br />
with few exceptions, the refinement<br />
of a given sail shape is a time-consuming<br />
process and requires a great amount of fullscale<br />
tests at sea. One of the most relevant<br />
issues in sail design is the correlation of a<br />
nominal sail shape with its wind-loaded<br />
shape and its actual performance. The aim<br />
of the present work is to provide a partial<br />
response to such a problem through the<br />
development of a systematic procedure for<br />
analysis and comparison of different sail<br />
designs. In particular, the goals are to<br />
estimate how given design factors affect<br />
the sail’s performance and to provide an<br />
analytical model of the phenomenon. The<br />
method used is based on coupled structural<br />
and CFD simulations and the Design of<br />
Experiments (DOE) technique.<br />
According to Richter et al. [1], computational<br />
fluid dynamics “has demonstrated<br />
the ability to predict sail and appendage<br />
forces under upwind conditions.” In their<br />
work, the authors integrated FLUENT and<br />
a proprietary software for structural analysis<br />
to estimate the flying shape of sails.<br />
Thanks to CFD, an adequate degree of<br />
accuracy can also be obtained when predicting<br />
the behavior of a yacht sailing<br />
downwind [2].<br />
Regression techniques can be regarded as<br />
well-established tools in the naval architecture<br />
domain [3]. Furthermore, DOE has<br />
proved to be a valuable tool in sail design<br />
[4]. However, the potential of DOE techniques<br />
has not been fully explored. DOE<br />
offers three advantages [5]. First, a number<br />
of parameters (both discrete and continuous)<br />
and responses can be involved.<br />
Second, regression models can be provided<br />
with reasonably low computational effort.<br />
Third, possible design trends can be identified<br />
by means of appropriate representation<br />
of responses, or response surfaces, over a<br />
given design space.<br />
The selected case study is the design<br />
of asymmetrical spinnakers used by the<br />
Tornado, the Olympic class catamaran. A<br />
set of parameters representing a full 3D<br />
spinnaker shape has been identified and<br />
their domains evaluated, in order to perform<br />
systematic perturbations on a given<br />
sail design and set up a DOE-driven analysis.<br />
The surface mesh on the sail and wind<br />
tunnel boundaries<br />
A mesh sizing function whose source is<br />
the sail<br />
8 <strong>Fluent</strong> News · Spring 2006
FLUID-STRUCTURE INTERACTION<br />
Streamlines coloured by velocity magnitude; the<br />
spinnaker is viewed from the leeward side<br />
Comparison between a design shape (left) (input to the iterative process) and the corresponding flying shape<br />
(right) after convergence is reached following four fluid-structure iterations<br />
Although such parameters refer to a sail section at a<br />
given mast height, their perturbations have been<br />
propagated to the rest of the sail in order to obtain<br />
results consistent with common sailmaking practices.<br />
The selected process factors are the sail section<br />
camber, its position over the chord, and the<br />
section twist angle with respect to the sail foot<br />
chord (or bottom edge of the sail).<br />
Because of the need to sample a significant region<br />
of the design space, the various designs evaluated<br />
are considerably different from each other; notable<br />
differences have been observed in terms of the sail<br />
angle of attack at given reference heights. The<br />
structural behavior of the design shapes can’t be<br />
represented as that of rigid bodies. In fact, the<br />
overall sail shape undergoes large displacements<br />
and a fluid-structure interaction approach<br />
becomes necessary.<br />
A “numerical wind tunnel” consisting of a coupled,<br />
iterative procedure involving structural and fluiddynamic<br />
calculations has been developed to estimate<br />
the wind-loaded shapes of given sails and their<br />
performance in terms of driving force, side force,<br />
and heeling moment. At each iteration, FLUENT<br />
evaluates the pressure field over a given sail surface<br />
and provides it to a structural module that yields a<br />
new loaded shape. For the purposes of the structural<br />
analysis, the sail is modeled as an infinitely flexible<br />
web with fixed elongation [6]. Each structural<br />
step consists of a constrained optimization routine<br />
whose objective function is the work of aerodynamic<br />
forces acting on the sail.<br />
Since a wind tunnel facility suitable for the testing of<br />
model sails is available at the University of Perugia,<br />
the computational domain for the CFD analysis has<br />
been scaled accordingly, for validation purposes.<br />
The numerical wind tunnel underwent preliminary<br />
tests showing good agreement with actual flow field<br />
data. In order to be consistent with the actual<br />
process of sail trim, the ideal flow incidence with<br />
respect to a given reference section of the sail<br />
should be preserved until the iterative method converges.<br />
As an example, critical conditions may occur<br />
at low angles of attack, where the leading edge of<br />
the sail collapses easily.<br />
A proper module was therefore implemented<br />
through GAMBIT scripting language, which allowed<br />
the sail to be correctly positioned with respect to<br />
the flow field at the beginning of each FLUENT run,<br />
simulating actions taken by the helmsman. The<br />
problem just described implies that a unique meshing<br />
strategy must be suitable for several sail shapes<br />
(those generated at each step of the iterative<br />
process). The use of sizing functions allowed these<br />
difficulties to be overcome and, at the same time,<br />
provided a satisfactory mesh quality independent of<br />
the sail shape considered.<br />
Results of all simulations were collected and<br />
processed through an analysis of variance tools, in<br />
order to select the geometric parameters affecting<br />
sail performance. Regression models and response<br />
surfaces were then evaluated, in order to provide<br />
guidelines for future improvements of a given<br />
design. A validation of the present approach has<br />
been carried out by means of wind tunnel tests and<br />
consists of two stages. First, the performance in<br />
terms of driving force coefficient was evaluated.<br />
Second, the predicted and actual flying shapes of a<br />
given sail were compared. An optical measurement<br />
system [7] allowed the wind loaded shapes of<br />
model sails during wind tunnel runs to be digitized.<br />
In both cases, reasonable agreement with numerical<br />
results was found, which suggests the possibility<br />
of using the methodology described above as a<br />
valuable tool for sail designers. <br />
Response surface showing the influence of camber<br />
and position, while twist is fixed to its mid-range<br />
value, obtained from the final regression model<br />
References<br />
1 Richter H.J., Horrigan K.C. and Braun J.C.:<br />
Computational Fluid Dynamics for Downwind Sails. In<br />
Proc. of The 16th Chesapeake Sailing Yacht<br />
Symposium, 2003.<br />
2 Jones, P. and Korpus, R.: International America’s Cup<br />
Class Yacht Design Using Viscous Flow CFD. In Proc. of<br />
The 15th Chesapeake Sailing Yacht Symposium, pp.<br />
27-34, 2001.<br />
3 Battistin, D., Peri, D., and Campana, E.F.: Geometry<br />
and Resistance of the IACC Systematic Series Il Moro di<br />
Venezia. In Proc. of The 17th Chesapeake Sailing Yacht<br />
Symposium, pp. 33-51, 2005.<br />
4 Lasher, W.C., Sonnenmeier, J.R., Forsman, D.R., Zhang,<br />
C. and White, K.: Experimental Force Coefficients for a<br />
Parametric Series of Spinnakers. In Proc. of The 16th<br />
Chesapeake Sailing Yacht Symposium, 2003.<br />
5 Myers, R.H. and Montgomery, D.C.: Response Surface<br />
Methodology. John Wiley & Sons, New York, 2002.<br />
6 Le Maitre, O., Huberson, S.G. and Souza de Cursi, J.E.:<br />
Application of a Non-convex Model of Fabric<br />
Deformations to Sail Cut Analysis. J. Wind Engineering<br />
and Industrial Aerodynamics, Vol. 63, pp. 77-93, 1997.<br />
7 Barone, S., Bianconi, F., Conti, P., Razionale, A. and<br />
Scarponi, M.: Acquisition and Modelling of 3D Sail<br />
Shapes. CAD 2004, International CAD Conference and<br />
Exhibition, Pattaya Beach, Thailand, 2004.<br />
more.info@<br />
matteo.scarponi@unipg.it<br />
<strong>Fluent</strong> News · Spring 2006 9
FLUID-STRUCTURE INTERACTION<br />
t = 1.00 sec<br />
Artificial<br />
Heart Valve<br />
Takes Shape<br />
t = 1.04 sec<br />
t = 1.14 sec<br />
By Jan Vierendeels, Joris Degroote, Kris Dumont, Lieve Lanoye, Pascal Verdonck,<br />
Ghent University, Belgium<br />
AT GHENT UNIVERSITY, BELGIUM, research on fluid-structure interaction<br />
(FSI) modeling has been ongoing since 1996. At the beginning, in-house codes<br />
were developed to compute the interaction between the motion of the cardiac<br />
wall and the blood flow in the heart [1]. During the past ten years, CFD and CSD<br />
(computational structural dynamics) packages have made significant progress<br />
and nowadays, the best approach for coupling fluid motion and structural displacement<br />
is to use existing software packages as partitioned solvers. When using<br />
a partitioned approach, however, it can be difficult to obtain convergence, especially<br />
when the interaction between the different motions is strong. This is the<br />
case for blood flow in the heart and for blood flow through a heart valve with<br />
flexible leaflets. Both of these types of blood flow have been simulated using the<br />
FSI approach developed at Ghent University. The blood flow in the moving<br />
geometries is computed with FLUENT 6.2, and the structural problem is solved<br />
using in-house software. The coupled calculation is robust and does not cause<br />
undue convergence difficulty, even for strongly interacting system responses.<br />
During the calculation, an implicit coupling method is used. For each new time,<br />
a new position of the boundary and the corresponding load distribution on that<br />
boundary is sought. This information is the result of simultaneous solutions of the<br />
CFD and CSD solvers, and could be obtained by alternating the solver calls. For<br />
strongly coupled cases however, such an approach can only be stabilized by<br />
reducing the underrelaxation factors or by using an Aitken-like technique, both<br />
of which tend to slow the rate of convergence.<br />
t = 1.36 sec<br />
t = 2.00 sec<br />
The figures show a flexible leaflet being opened by<br />
an accelerating flow; the stiffness of the leaflet is very<br />
small, so that it undergoes a huge displacement from<br />
the closed to the fully open position and back again<br />
To speed the convergence, a reduced order model is used [2]. If the Jacobian of<br />
the fluid solver were known when solving the structural problem, one would<br />
know how the load would change when a certain boundary displacement is<br />
applied, so the load change would not need to be calculated with the fluid solver.<br />
The boundary condition for the structural solver (the load) would not need to be<br />
assumed constant during this call, but could be written as a function of the<br />
unknown boundary position itself, resulting in a much better prediction for the<br />
boundary position. Since the Jacobian of the fluid solver is not available, a<br />
reduced order model of the solver can be built for which a Jacobian can be<br />
constructed. This approximation of the Jacobian of the fluid solver is then used<br />
for the coupled calculation. During each time step, whenever FLUENT is called<br />
a boundary displacement mode is applied and the corresponding load change<br />
at the boundary is retrieved through user defined functions (UDFs). This<br />
information is used to build up the reduced order model for the fluid problem.<br />
The structural solver is then called with the variable load boundary condition.<br />
FLUENT is called again and the reduced order model is updated. The loop is<br />
executed until a residual drop of four to five orders of magnitude is obtained.<br />
This new technique for FSI problems has proved to be robust and powerful, and<br />
is ideally suited for FSI problems with strong coupling. <br />
References:<br />
1 Vierendeels, J.A., Riemslagh, K., Dick, E. and Verdonck, P.R.: Computer Simulation of<br />
Intraventricular Flow and Pressure Gradients During Diastole. Journal of Biomechanical<br />
Engineering – Transactions of the ASME, 122(6):667-674, 2000.<br />
2 Vierendeels, J.A.: Implicit Coupling of Partitioned Fluid-structure Interaction Solvers using<br />
a Reduced Order Model. In Proc. of the 35th AIAA Fluid Dynamics Conference and<br />
Exhibit, June 6-9 2005, Toronto, Ontario, Canada. AIAA-2005-5135. AIAA Meeting<br />
papers on disc, vol. 10, nr 11-12. ISBN 1-56347-763-7.<br />
10 <strong>Fluent</strong> News · Spring 2006
FLUID-STRUCTURE INTERACTION<br />
FSI Controls<br />
Flow Rate<br />
By Subham Sett, ABAQUS, Inc., Providence, Rhode Island, USA<br />
FLOW CONTROL DEVICES are commonly employed in<br />
the automotive, biomedical, and consumer appliances<br />
industries to maintain a constant bulk flow rate for varying<br />
inlet pressures. Fluid pressure variations may result<br />
from pipe friction loss, downstream restrictions, distance<br />
from the water tower, or elevation of the water tap, for<br />
example. Minimizing the impact of inlet pressure variation<br />
on the flow is essential for the proper performance of<br />
fluid systems.<br />
A schematic cross<br />
section of a typical<br />
VernaFlo device<br />
before (top) and<br />
after (bottom) the<br />
rubber component<br />
is deformed<br />
Vernay VernaFlo ® flow controls are custom-designed flow<br />
management devices used in a wide range of applications<br />
where consistent, reliable fluid flow is essential. This study<br />
evaluates the performance of a custom VernaFlo device,<br />
using a fully coupled fluid-structure interaction (FSI)<br />
analysis with ABAQUS and FLUENT. The computational<br />
results compare favorably with available experimental<br />
data.<br />
In a typical VernaFlo device, an elastomeric rubber component<br />
is housed inside the flow path. This rubber insert<br />
rests on a rigid seat and deforms under the influence of<br />
the incoming flow. At low operating pressures the rubber<br />
component undergoes very little deformation and allows<br />
flow<br />
The CFD sub-domain where the fluid flows (red) and the ABAQUS sub-domain<br />
which, for the VernaFlo valve, is a rubber insert that deforms for different flow rates<br />
(white)<br />
Static flow pressure (top) and Mises stress (bottom) at 20, 40, 60, and 90 psi; at high pressures, the rubber insert moves closer to the axis,<br />
stabilizing the flow rate<br />
<strong>Fluent</strong> News · Spring 2006 11
FLUID-STRUCTURE INTERACTION<br />
2.5<br />
Flow rate (L/min)<br />
2.0<br />
1.5<br />
1.0<br />
0.5 ABAQUS-FLUENT<br />
Experiment<br />
0.0<br />
0 20 40 60 80 100<br />
Inlet pressure (psi)<br />
120<br />
Experimental validation of the computation results<br />
the flow to develop. With increasing upstream pressure, the deformation increases, restricting<br />
the orifice diameter and thus limiting the fluid flow. Capturing the interaction between<br />
the fluid flow and the structural deformation is critical to accurately predicting the shape of<br />
the rubber component and the subsequent flow behavior.<br />
The FSI analysis of the valve with ABAQUS and FLUENT makes use of a co-simulation technique<br />
managed by MpCCI (Mesh-based parallel Code Coupling Interface). The fluid and<br />
structural domains are modeled and solved separately, and solution information is<br />
exchanged at the fluid-structure interface. The ABAQUS sub-domain includes the rubber<br />
component, which is modeled using reduced order hybrid axisymmetric elements. The<br />
rubber component presses against a rigid seat. Penalty contact with a friction coefficient of<br />
0.5 is defined between the rubber and the rigid seat. The simulation includes the effect<br />
of geometrical and material nonlinearities. The CFD sub-domain models the flow path,<br />
including a short upstream section, the section around the rubber component, and a long<br />
downstream section. The upstream variation in pressure is accounted for using a pressureinlet<br />
boundary condition, and a zero gauge pressure is used at the outlet. To enable local<br />
remeshing, the flow path is modeled using triangular elements. The water is modeled as a<br />
turbulent, incompressible fluid, and the cavitation model is enabled.<br />
Velocity magnitude contours on a 2D model of the<br />
cross flow fan; the flow is from bottom to top<br />
The fluid-structure interface is defined using MpCCI. During the simulation, the pressures<br />
acting on the rubber component in the fluid sub-domain are mapped and transferred to the<br />
structural sub-domain in ABAQUS via MpCCI. ABAQUS computes the deformations and the<br />
resulting stress state in the structure. The interface deformation quantities are then mapped<br />
and transferred from the structural sub-domain to the fluid sub-domain in FLUENT via MpCCI.<br />
This process of exchanging solution quantities continues incrementally until the analysis is<br />
complete.<br />
The analysis results identify how variations in inlet pressure affect the bulk fluid flow rate<br />
through the device, and how the fluid flow affects the deformation of the rubber component.<br />
Contour plots of the fluid pressures at inlet pressure levels of 20, 40, 60 and 90 psi and<br />
the corresponding deformed shapes show that with an increase in the inlet pressure, the rubber<br />
component deforms and undergoes increased contact with the rigid seat. Partial contact<br />
is observed at 20 psi, and full contact is established at 60 psi. The constriction path narrows<br />
during the inlet pressure rise, resulting in an increased resistance to the fluid flow and a<br />
region of very high stresses. At a given deformation state, the increased material stiffness<br />
helps maintain the bulk flow rate at a fairly constant level. As the flow quickens through the<br />
narrow constriction region, the fluid pressure drops significantly, resulting in a dramatic drop<br />
in the absolute pressure of the liquid. Cavitation occurs at higher upstream pressures.<br />
Contours of stream function for the cross flow fan<br />
The computational flow rate results were found to be in good agreement with experimental<br />
data. The flow rate increases from 0 to 2.1 liters/min during the initial pressure ramp-up from<br />
0 to 20 psi and has nearly constant flow in the operating pressure range of 20 to 120 psi. <br />
Acknowledgments<br />
ABAQUS, Inc. would like to thank Jim Bailey at Vernay Labs for providing the model and experimental<br />
data for the Vernay VernaFlo device, and David Schowalter at <strong>Fluent</strong> Inc. for assisting with the cavitation<br />
modeling.<br />
Velocity vector detail shows recirculation regions in<br />
one area of the cross flow fan<br />
12 <strong>Fluent</strong> News · Spring 2006
HVAC<br />
Fan Research Makes<br />
Cool Air Conditioners<br />
By Teng-Kiat Lim Ph.D. and Chee-Onn Chan, Department of Research and Application, O.Y.L. R&D Center, Selangor, Malaysia<br />
SIGNIFICANT TECHNICAL CHALLENGES face<br />
companies in the HVAC industry today. In an<br />
increasingly competitive global market, HVAC systems<br />
need to make use of advanced technologies<br />
and improved component performance. At the<br />
O.Y.L. R&D Center, CFD is used for the analysis and<br />
optimization of fans, compressors, heat exchangers,<br />
and ducts, and it plays an important role in HVAC<br />
product lifecycle management (PLM) as well. CFD<br />
helps reveal flow structures and comprehensive flow<br />
field information where experimental techniques<br />
cannot provide adequate resolution.<br />
Consider, for example, a split-type air conditioner<br />
that consists of one outdoor condensing unit (housing<br />
an air-cooled condenser, propeller fan, and<br />
refrigerant compressor) and one or more indoor<br />
units (housing an evaporating coil and cross flow<br />
fan). Split-type air conditioners are popular in residential<br />
buildings because of their simplicity and<br />
flexibility. If the air flow in the outer unit is not<br />
properly controlled, however, poor recirculation can<br />
cause the coil temperature to rise and the system<br />
performance to suffer. A series of simulations has<br />
been performed to study the various components<br />
of this type of system, and based on the results,<br />
modifications have been made.<br />
A 3D model of a 3-bladed fan in the outdoor condensing<br />
unit was created and used to visualize the<br />
air flow near the housing and in the discharge. The<br />
steady multiple reference frames (MRF) model in<br />
FLUENT was used along with the standard k-ε<br />
turbulence model and standard wall functions. A<br />
tetrahedral mesh of about 1 million cells was generated<br />
using GAMBIT and TGrid. The triangular surface<br />
mesh was refined in the regions around the<br />
hub, tip, and blade edges. A uniform air velocity<br />
was imposed at the condenser inlet boundary, and<br />
a constant pressure was applied to the outlet. The<br />
results illustrated that when the steady incoming air<br />
flow enters the rotating zone of the fan, a strong<br />
radial component develops, driving the flow<br />
towards the side walls. This behavior is enhanced by<br />
the existence of an outlet front panel that partially<br />
blocks the flow. The CFD predictions were found to<br />
be in good agreement with air flow test measurements.<br />
Studies on new front panel designs are<br />
currently being conducted so that better air flow<br />
performance can be achieved.<br />
A two-bladed fan has also been studied. For this<br />
case, an unstructured mesh of about 1.2 million elements<br />
was needed to give a mesh-independent<br />
result. The 3D, double precision simulation also<br />
made use of the MRF and standard k-ε models. The<br />
results indicated that even though the front panel<br />
impacts the fan performance, a more uniform air<br />
flow distribution can be achieved, along with a<br />
lower pressure drop.<br />
The cross flow fan (CFF) operates in a fundamentally<br />
different way than axial or centrifugal fans. Flow<br />
enters across the full width of the rotating impeller<br />
and exits on the opposite side. The flow structure<br />
inside the impeller can be divided into two regions:<br />
an eccentric vortex or recirculation region having<br />
closed streamlines, and a through-flow region connecting<br />
the inflow and outflow sectors. The blades<br />
typically have a uniform thickness and a circular arc<br />
profile. The performance and stability of a CFF are<br />
governed by the geometrical parameters of the<br />
impeller and casing. Using the sliding mesh model,<br />
a transient 2D analysis was performed to study the<br />
effect of parameters such as blade angle, radius<br />
ratio, and number of blades. Maximum energy<br />
transfer through the impeller takes place in the<br />
region where the flow follows the blade curvature.<br />
Radial velocity is not uniform through the blade<br />
channels. More flow leakage is observed through<br />
the tongue clearance portion at low flow coefficients<br />
and static pressure is always negative in and<br />
around the impeller region <br />
Absolute pressure on the 3-bladed fan (top) and<br />
2-bladed fan (bottom)<br />
<strong>Fluent</strong> News · Spring 2006 13
AUTOMOTIVE<br />
Citroën<br />
Vortex structures colored by streamwise velocity show<br />
the instantaneous and highly three-dimensional flow;<br />
note the interaction between the streamwise vortices<br />
formed at separation on the front fender and the<br />
A-pillar on the leeward side<br />
Postprocessed by Ensight from CEI and Distene<br />
AS PASSENGER VEHICLES BECOME LARGER, their stability<br />
and safety characteristics become more sensitive to aerodynamic<br />
forces. Such effects are particularly important for vehicles exposed<br />
to a crosswind [1]. For instance, all drivers are familiar with the<br />
effect of cross-wind gusts, like those that may occur when a vehicle<br />
emerges from a tunnel or during overtaking. CFD can provide<br />
detailed information for understanding the changes in car behavior,<br />
handling, and performance under these conditions.<br />
Simulations are routinely performed at PSA using engineering<br />
approaches that are based on the solution of the Reynolds-averaged<br />
Navier-Stokes (RANS) equations. While this approach is often<br />
adequate for steady attached flows with no recirculation regions,<br />
numerous studies have shown the limitations of the RANS<br />
approach to accurately predict flows with massive separations and<br />
that are fundamentally unsteady.<br />
Large eddy simulation (LES) can have greater success than RANS<br />
methods in predicting separation and large scale unsteady flows,<br />
particularly in the wake and on the leeward side of vehicles under<br />
crosswind conditions. In recent years, this type of unsteady<br />
calculation has been increasingly used in industry and FLUENT has<br />
made a special effort to develop industrial-strength validation<br />
examples for LES modeling techniques [2,3].<br />
PSA Peugeot Citroën has developed a strong expertise in CFD for<br />
external aerodynamics, cabin modeling, underhood thermal management,<br />
and more. In-house methodologies have been developed<br />
and are regularly improved, in order to fulfill new project<br />
requirements, which are getting more and more demanding, both<br />
in terms of accuracy and turn-over time. Computations are done<br />
in conjunction with wind tunnel experiments, either on full<br />
scale prototypes or, as in the case presented here, on a 1:5 scale<br />
model. Different measurement techniques are used, such as<br />
particle-image velocimitry (PIV), pressure taps, and tomography.<br />
The different methodologies also benefit from a collaboration<br />
between PSA and <strong>Fluent</strong>, and this was the basis of a recent joint<br />
project to address the potential benefit and cost of LES compared<br />
to the current RANS approach for semi-realistic configurations.<br />
Fujitsu/Siemens provided access to a large-scale high performance<br />
computing (HPC) facility for this project.<br />
Model Exp SST k-ω LES WALE LES WALE RSM v 2 -f<br />
(35M cells) (65M cells)<br />
Drag (SCx) 0.70 0.66 0.69 0.68 0.71 0.73<br />
14 <strong>Fluent</strong> News · Spring 2006<br />
Detailed force and torque components<br />
acting on the vehicle; under crosswind<br />
conditions, the side force (SCy) and<br />
yaw moment (SCn) are the most<br />
influential for vehicle stability<br />
Forces<br />
Moments<br />
Side (SCy) 2.22 2.00 2.19 2.18 2.30 2.10<br />
Lift (SCz) 1.40 1.66 1.27 1.30 1.82 1.77<br />
Yawing (SCn) -0.64 -0.60 -0.57 -0.59 -0.47 -0.47<br />
Rolling (SCl) -0.42 -0.36 -0.49 -0.49 -0.46 -0.41<br />
Pitching (SCm) 0.12 0.10 0.21 0.23 0.03 0.07
AUTOMOTIVE<br />
C5 in a Crosswind<br />
By Sylvain Lardeau, PSA Peugeot Citroën Automobile, Vélizy-Villacoublay, FRANCE, and Fabrice Mathey and Nicolas Vallée, <strong>Fluent</strong> France<br />
The RANS and LES simulations were performed on a<br />
1:5 scale simplified model of a Citroën C5. The crosswind<br />
effect was produced with the model placed at a<br />
yaw angle of 20°. The Reynolds number of the flow,<br />
based on the incoming velocity and car height, H,<br />
was 6.19 x 10 5 . The surface mesh (around 730,000<br />
elements) was generated with ANSA and was locally<br />
refined close to the edges and on the leeward faces of<br />
the vehicle. The boundary layers were generated with<br />
the preprocessor TGrid, using 5 layers of prisms. The<br />
remaining part of the domain consisted of tetrahedral<br />
cells and was generated with the preprocessor<br />
GAMBIT. Sizing functions were used to control the<br />
growth rate and the cell size of the mesh in the region<br />
of separation on the leeward side of the car. The<br />
minimum resolution required to resolve the relevant<br />
turbulent length scales was estimated from prior RANS<br />
simulations. Two grids (of 35 million and 65 million<br />
cells) were created to address the sensitivity of the<br />
results to the mesh resolution. The body was placed in<br />
an open channel with a cross section of 14.5H x 7H.<br />
The channel inlet was located 5H from the front face<br />
of the body, and the channel outlet was located 8.5H<br />
from the rear. The boundary layers were modeled with<br />
the Werner-Wengle wall-function approach. The average<br />
y + for the first cell off the wall was less than 50.<br />
The simulation benefited from the recent enhancements<br />
in numerics and sub-grid scale (SGS) modeling<br />
in FLUENT 6.2. The bounded central differencing<br />
scheme was used to prevent unphysical wiggles<br />
numerically introduced by the pure central differencing<br />
scheme. The non-iterative time advancement, or<br />
NITA algorithm significantly increased the speed of<br />
the transient calculation. The wall-adapting local<br />
eddy-viscosity (WALE) SGS model, which returns the<br />
correct wall asymptotic variation of the turbulent<br />
viscosity in a cost-effective manner, was used as<br />
well. The simulations were performed on a 96 CPU<br />
Fujistu cluster. For the 35M case, statistics were<br />
gathered during 16,000 iterations for an elapsed<br />
time of 70 hours. Statistics were gathered during<br />
6,000 iterations for the 65M case, for an elapsed<br />
time of 50 hours.<br />
The full mean aerodynamic forces and torques were<br />
compared with RANS methods and experimental<br />
data. Under crosswind conditions, the side force and<br />
yaw moment are the most influential for vehicle stability.<br />
The LES results showed very good agreement<br />
with the experimental data for these quantities. The<br />
difference between measurements and calculations<br />
was 2% for the side force coefficient (SCy), and was<br />
in the range of 10% for the yawing moment (SCn).<br />
The LES runs also gave consistent results for the<br />
other aerodynamic force components; the drag<br />
force (SCx) was predicted with a 1% error, for<br />
instance. It is noteworthy that even if some RANS<br />
simulations are able to predict side force and yawing<br />
moment with reasonable accuracy, none of the turbulence<br />
models could predict consistent results for<br />
all of the force components and moments. Surface<br />
pressure coefficients also confirm the differences<br />
between the different computational methods used.<br />
On the A-pillar on the leeward side, especially, the<br />
pressure drop is too strong with the RANS model.<br />
The two LES runs gave comparable results, indicating<br />
that reasonable mesh independence was<br />
reached with the 35 M grid, as far as global force<br />
coefficients are concerned.<br />
The visualization of the mean velocity streamlines<br />
indicates the formation of two streamwise vortices on<br />
the leeward side of the vehicle. The first one is formed<br />
on the front fender and the second is formed on the<br />
A-pillar and along the side window. Animation and<br />
instantaneous flow visualization demonstrate the<br />
highly three-dimensional, unsteady nature of these<br />
vortices. The predominant peak frequencies calculated<br />
from a Fourier transform of the pressure signal at<br />
several locations on the leeward side is equal to<br />
110 Hz, a value very close to the experimental peak<br />
frequency of 100 Hz.<br />
These simulations demonstrate how RANS and LES<br />
modeling approaches can complement each other<br />
to study car aerodynamics. Current RANS-based<br />
methodologies are able to correctly predict drag<br />
force components for attached flow and mild separation<br />
on realistic car geometries under uniform wind<br />
conditions. LES is a better tool for studying vehicle<br />
aerodynamics in crosswind conditions, since it provides<br />
a more accurate drag force component and<br />
pressure distribution along the vehicle. In addition,<br />
LES gave insight into the unsteady behavior of the<br />
flow, an important feature for both vehicle stability<br />
studies and aeroacoustic applications. <br />
References:<br />
1 Ryan, A. and Dominy, R.G.: The Aerodynamic Forces<br />
Induced on a Passenger Vehicle in Response to a<br />
Transient Cross-wind Gust at a Relative Incidence of<br />
30 Degrees. SAE Paper 980392, 1998.<br />
2 Kim, S.E.: Large Eddy Simulation using Unstructured<br />
Meshes and Dynamic Subgrid-scale Turbulence Models.<br />
AIAA Paper no. 2004-2548, 2004.<br />
3 Mathey, F. and Cokljat, D.: Zonal Multi-domain<br />
RANS/LES Simulation of Air Flow over the Ahmed Body.<br />
In: Engineering Turbulence Modelling and Experiments,<br />
6, pp. 647-656, Elsevier Ltd., 2005.<br />
0<br />
-0.5<br />
-1.0<br />
Cp<br />
-1.5<br />
-2.0<br />
Experiment<br />
LES<br />
RSM high-Re<br />
v 2 -f<br />
SST k-ω<br />
-2.5<br />
-1.0 -0.5 0.0 0.5 1.0 1.5 2.0<br />
X (m)<br />
Surface mesh of the vehicle generated with ANSA<br />
Surface pressure distribution on the windward side on<br />
the “eye plane”, z=1.250 m<br />
<strong>Fluent</strong> News · Spring 2006 15
AUTOMOTIVE<br />
TORQUE CONVERTERS ARE FLUID-COUPLING DEVICES that<br />
transmit power from vehicle engines to the wheels. They are used<br />
in automobiles with automatic transmissions to smoothly control<br />
the torque supplied to the wheels at all speeds.<br />
An exploded view of the torque converter showing<br />
the pump (top), stator (middle), and turbine (bottom)<br />
Torque<br />
Converters<br />
Get In Gear<br />
By Vamshi Korivi, DaimlerChrysler, Auburn Hills, Michigan, USA<br />
Sandeep Sovani and Pepi Maksimovic, <strong>Fluent</strong> Inc.<br />
Torque converters are essentially chambers filled with transmission<br />
fluid, with the primary active parts being the pump (or impeller),<br />
stator, and turbine. The impeller blades rotate at the engine<br />
speed and impart angular momentum to the fluid as they force it<br />
radially outward. The turbine blades receive the fluid from the<br />
pump and turn it radially inward. The angular momentum thus<br />
absorbed by the turbine is passed to the transmission as torque.<br />
The flow exits from the turbine blade passages with little remaining<br />
angular momentum and encounters the stator blades, which<br />
accelerate the fluid back to the pump rotational speed. This action<br />
creates a torque on the stator, which is arrested by a non-rotating<br />
shaft built into the transmission housing. Torque multiplication<br />
results from the fact that the turbine torque is larger than that of<br />
the pump.<br />
Numerical simulations of torque converters are important because<br />
the complex topology and rotation of these devices make detailed<br />
experimental investigation virtually impossible. Simulations help<br />
identify optimal designs that yield improved performance and fuel<br />
economy. Using the multiple reference frames (MRF) model in<br />
FLUENT, a DaimlerChrysler torque converter has been studied.<br />
The MRF model allows the different parts of the torque converter<br />
(the pump, stator, and turbine) to be simulated using one relative<br />
orientation, but separate rotating frames. The model is well suited<br />
for torque converters, since the number of blades is chosen to<br />
prevent the amplification of harmonics and possible structural<br />
damage to the parts. This means that any single relative position<br />
of the blades is a good representation of the averaged behavior of<br />
the system as a whole. For the torque converter considered here,<br />
the number of blades for the pump, turbine, and the stator is 31,<br />
29, and 20 respectively.<br />
Contours of total pressure on the blade surfaces of<br />
the three elements; large pressure losses can be seen<br />
across the stator blades (in the middle)<br />
K-factor<br />
A hexahedral mesh of about 2.9 million cells was used for the initial<br />
calculation. A pressure gradient-based mesh adaption was<br />
subsequently performed to enhance the solution accuracy. Using<br />
transmission fluid and a pump speed of 2000 rpm, the turbulent<br />
flow was captured using the realizable k-ε turbulence model.<br />
Custom postprocessing tools were created for the automatic<br />
extraction of torque converter quantities of interest, such as the<br />
mass-averaged flow angles at the inlet and outlet of each element,<br />
the efficiency, and the torque imbalance, for example.<br />
The results illustrated several flow features. Total pressure contours<br />
were used to illustrate the pressure loss across the stator blades.<br />
Velocity vectors were used to check for separation regions. The<br />
k-factor, the ratio of pump torque to the square root of pump<br />
speed and an indicator of the torque converter capacity, was<br />
found to be within 2% of measured values at all speed ratios (the<br />
ratio of turbine speed to pump speed) studied.<br />
Efficiency<br />
FLUENT<br />
Experiment<br />
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8<br />
Speed ratio<br />
Predictions of efficiency and k-factor as a function of<br />
speed ratio show good agreement with measured values<br />
0.9<br />
Predictions of efficiency, defined as the product of torque ratio<br />
and speed ratio, were found to be within 1.6% of the data at a<br />
speed ratio of 0.7. The torque ratio, defined as the ratio of turbine<br />
torque to pump torque, was found to be within 5% for speed<br />
ratios above 0.2. The ability to predict efficiency and torque ratios<br />
at high speed ratios is important for improving fuel economy. <br />
16 <strong>Fluent</strong> News · Spring 2006
AUTOMOTIVE<br />
Emissions Control Through<br />
Carbon Canisters<br />
By Ranjit Singh, Expert Corporation, and Pepi Maksimovic, <strong>Fluent</strong> Inc.<br />
CARBON CANISTERS ARE DEVICES commonly<br />
used to control the emissions of volatile hydrocarbons.<br />
Hydrocarbons (HC) are hazardous to human<br />
health and the environment. For automobiles, HC<br />
emissions are produced during the filling of the fuel<br />
tank, and during vehicle operation. When the<br />
engine is off, evaporation from the vehicle fuel<br />
system still occurs, even at ambient temperatures<br />
typical of the diurnal cycle. Allowable HC emission<br />
limits are set by government regulations; for example,<br />
the LEV II (Low Emitting Vehicle-II) standard<br />
allows a certain amount of hydrocarbon emissions<br />
for a specific range of gross vehicle weight.<br />
Carbon canisters are part of the evaporative emission<br />
control system, which includes the fuel tank,<br />
vent and purge valves, and fuel lines. The role of the<br />
carbon canister is to store the fuel vapor generated<br />
in the system instead of having it escape into the<br />
atmosphere. The HCs are then burned off by purging<br />
the canister into the intake manifold when the<br />
engine is running. An optimum design includes a<br />
high working capacity, minimal pressure drop in the<br />
evaporative emissions system, space and size restrictions,<br />
and the ability to meet the mandated emission<br />
standards. Furthermore, the carbon utilization<br />
in the canister should be uniform; when less carbon<br />
material is needed, savings are realized. In today’s<br />
competitive market, carbon canister manufacturers<br />
must be able to predict canister performance without<br />
having to build and test various prototypes.<br />
desorbing), until breakthrough is achieved, i.e.<br />
when a set cumulative amount of butane passes<br />
through the canister outlet. Arrhenius reaction rates<br />
for the adsorption and desorption reactions were<br />
prescribed. The carbon pellets were modeled as<br />
a porous region. A mesh of 184,000 hexahedral<br />
elements was used for the laminar simulation.<br />
The CFD results were used to assess the design.<br />
Using pathlines, a region with little or no flow was<br />
identified in the first chamber. Ideally, the canister<br />
design should provide uniform utilization of the<br />
carbon media in all of the chambers, and this result<br />
suggests that it does not. Mass fraction contours<br />
were used to illustrate the distribution of butane at<br />
several times during the canister loading process.<br />
The growth of butane inside the canister is the<br />
result of the ongoing surface reactions with the carbon<br />
pellets, so this provides another measure of<br />
how well the unit is operating as a whole. A plot of<br />
the butane mass fraction at the outlet as a function<br />
of loading time can also be used to indicate the<br />
time when breakthrough is achieved. The carbon<br />
canister capacity is computed by integrating under<br />
this curve. <br />
The geometry of the carbon canister, showing the<br />
inflow port (blue), 3 chambers packed with carbon<br />
pellets (gray) and vent port (yellow)<br />
Pathlines show a dead zone in the upper left corner<br />
in the first carbon-filled chamber; the presence of<br />
dead zones points to underutilization of carbon<br />
One carbon canister design from Expert Corp. has<br />
been simulated using FLUENT. The purpose of the<br />
CFD simulation was twofold: first, to compute the<br />
canister capacity and pressure drop at 60 l/min of<br />
air flow, and second, to predict canister breakthrough<br />
for a gaseous mixture of butane and nitrogen<br />
entering the canister at 15 g/hr. The mixture<br />
reacts with packed carbon pellets (adsorbing and<br />
Contours of butane mass fraction on a slice through the canister during loading (left) and once the canister has<br />
reached capacity (right)<br />
<strong>Fluent</strong> News · Spring 2006 17
POWER TOOLS<br />
Scavenging in a Stratified, Charged<br />
Two-Stroke Engine<br />
By Wolfgang Emmerich, SOLO Kleinmotoren GmbH, Sindelfingen, Germany<br />
TWO-STROKE GAS ENGINES are commonly used in hand-held machines,<br />
such as chain saws and grass trimmers. These engines have a tendency to cause<br />
pollution as a result of unburned hydrocarbons being released along with the<br />
exhaust. At SOLO Kleinmotoren, a new by-pass system has been developed that<br />
allows the combustion chamber to be flushed with either a lean fuel mixture or<br />
pure air early in the engine cycle. Fuel is injected as a rich mixture later in the<br />
cycle when the outlet port is closed. The by-pass system prevents the fuel from<br />
escaping from the combustion chamber before combustion is complete.<br />
A numerical investigation of the by-pass system has been carried out using the<br />
dynamic mesh model in FLUENT. Throughout several engine cycles, the mass<br />
fractions of pure air, a rich fuel mixture, and a mixture of exhaust gases were<br />
tracked. Combustion was approximated by resetting the pressure, temperature,<br />
and components of the three gaseous species each time the piston was in the top<br />
dead center (TDC) position.<br />
The operation of the engine is illustrated at right. During the suction phase, clean<br />
air is drawn into the crank case (blue arrow) and a fresh mixture of fuel and air<br />
is drawn into the by-pass channel (green arrow). As the crank case compresses,<br />
the clean air enters the combustion chamber through transfer ports, while the<br />
fuel mixture is pressed into the chamber through an injection port. Due to fluid<br />
dynamic constraints dictated by the local flow, the fresh air in the chamber prevents<br />
the fuel mixture from leaving the combustion chamber directly through<br />
the outlet port. Only exhaust gases (red) and pure air can be released into the<br />
environment.<br />
A schematic of the engine with the by-pass system: pure air (blue) and fuel (green)<br />
enter through ports on the left; following combustion, pure air and exhaust gases<br />
leave through the outlet port on the right<br />
The distributions for the three gaseous species at three different piston positions<br />
illustrate the success of the by-pass system. The blue surface corresponds to fresh<br />
air, the green surface to the fuel mixture, and the red surface to the exhaust gas<br />
mixture. At bottom dead center (BDC), the positions of the surfaces indicate that<br />
the fresh air forms a fluid shield between the fuel mixture and the outlet port.<br />
When the piston has advanced 50° beyond BDC, the fluid shield is not perfect,<br />
so a small portion of the fuel mixture can leak past the shield near the cylinder<br />
walls. When the piston has advanced 100° after BDC, the outlet port is closed by<br />
the rising piston, just before the fuel mixture has a chance to escape.<br />
The results of the simulation show the ability of the new engine to reduce environmental<br />
pollution by limiting the release of unburned hydrocarbons. The<br />
results also suggest that there is potential to improve the fluid dynamics of the<br />
fuel injection as well as the flushing of the combustion chamber to further<br />
enhance the performance of the new by-pass system. <br />
Surfaces of fresh air (blue), fuel (green), and exhaust (red) at three piston positions:<br />
a) bottom dead center (BDC), where the air forms a fluid shield to block the fuel,<br />
b) 50° beyond BDC, where some fuel leaks past the air shield near the cylinder wall,<br />
and c) 100° beyond BDC, at which time the rising piston has closed the outlet port<br />
18 <strong>Fluent</strong> News · Spring 2006
F L U E N T N E W S S U P P L E M E N T<br />
Focus on CFD<br />
For the Process Industries<br />
s2 OVERVIEW<br />
Transporting CFD to<br />
Process Engineers<br />
s3 MIXING<br />
Tracing Homogenization<br />
s5<br />
Finding the Optimum<br />
Blend Time Calculation<br />
s7 MICROREACTORS<br />
Liquid Mixing<br />
in Microreactors<br />
s8<br />
THERMAL RUNAWAY<br />
Preventing Runaway<br />
Reaction Accidents<br />
s10 EMISSIONS<br />
Scrubbers for<br />
Flue Gas Cleanup<br />
s12 FURNACES<br />
Ultra-low NOx Burners<br />
Get Cracking<br />
s13 MULTIPHASE<br />
Controlling Droplet Size<br />
Distribution in Emulsions<br />
s14 Understanding<br />
Fluid-Bed Coating<br />
s15 PUMPS<br />
Pumping out New<br />
Designs More Quickly<br />
s16 EXTRUSION<br />
Polymer Processing<br />
Simulation for Foam<br />
Extrusion
OVERVIEW<br />
PROCESS INDUSTRIES<br />
Transporting CFD<br />
to Process Engineers<br />
By Ahmad Haidari, Chemicals and Process Industry Segment Manager, <strong>Fluent</strong> Inc.<br />
CHEMICAL AND OTHER PROCESS INDUSTRY<br />
companies are driving issues such as green engineering,<br />
energy reduction, chemical sustainability and corporate<br />
profitability through innovation and the use of ingenious<br />
engineering and technology. Simulation software plays a<br />
key role in helping engineers to better understand<br />
processes and is assisting in productivity and efficiency<br />
gains across the industry. CFD in particular is gaining<br />
broad traction with researchers, process engineers, and<br />
equipment designers to help analyze and design the flow<br />
and performance of process equipment such as stirred<br />
tanks, fluidized bed reactors, separators, combustion<br />
systems, heat exchangers, and polymer and material<br />
processing and handling equipment.<br />
formulation accounts for phenomena such as nucleation,<br />
growth, dispersion, dissolution, aggregation, and<br />
breakage, so it is possible to describe and track changes<br />
in the particle population. Applications of this technology<br />
span a wide range of multiphase systems such as<br />
solid-liquid dispersions, crystallization, precipitative reactions,<br />
gas-sparged reactors, liquid-liquid dispersions, and<br />
liquid-liquid separation equipment. Second, for granular<br />
mixtures, the macroscopic particle model (MPM), available<br />
through user-defined functions (UDFs) in FLUENT,<br />
allows for the presence of large particles where particle<br />
collisions, rotation, adhesion, and other forces are<br />
accounted for using a hard sphere approach. Third,<br />
modifications to the volume of fluid (VOF) free surface<br />
model in FLUENT have been developed to simulate capillary-driven<br />
flow in an unsaturated porous region.<br />
The primary modeling goal was to study the spread<br />
of unbound liquid inside the voids formed by fibrous<br />
material, but the model also accounts for sorption (the<br />
physical bonding of moisture to the fiber), since the<br />
performance of many hygiene, food, and consumer<br />
products strongly depends on their sorption capacity.<br />
The flow of macroscopic particles<br />
in a hopper<br />
Technical advances at <strong>Fluent</strong> have progressed in two<br />
distinct directions. The traditional focus on more<br />
reliable and advanced physical models is being complemented<br />
by highly customized interfaces to fit the<br />
modeling needs of a specific application or equipment<br />
design. These advances, combined with the successful<br />
track record of CFD in many companies, have brought<br />
the full power of CFD to process engineers.<br />
The applicability of CFD to process industry applications<br />
is achieved through ongoing focused development<br />
and user-driven projects. For example, three customerdriven<br />
technology developments have been implemented<br />
recently that uniquely expand FLUENT’s capabilities.<br />
First, a population balance modeling effort was driven<br />
by a need to combine detailed fluid mechanics analysis<br />
with a better description of the size distribution of particles,<br />
bubbles, or droplets in a multiphase mixture. The<br />
In this special newsletter supplement, a diverse sampling<br />
of process industry applications is presented.<br />
Traditional applications such as mixing in stirred tanks<br />
are joined by mixing in microchannels. Reacting flow is<br />
covered in the context of thermal runaway (and a<br />
promising antidote), the cleanup of emissions, and<br />
the optimization of an ethylene cracking furnace.<br />
Multiphase flow examples include a fluidized bed and<br />
droplet breakup validations. The use of CFD for designing<br />
special purpose pumps is also reviewed. Finally, simulations<br />
of the extrusion of foam – a gas-impregnated<br />
polymer – are described.<br />
As these articles illustrate, the use of CFD is spreading<br />
to more and more applications within the process<br />
industries. With increases in computing capacity and<br />
the advances in technology currently taking place, this<br />
expansion can only continue in the years to come. <br />
<strong>Fluent</strong> News is published by<br />
10 Cavendish Court • Lebanon, NH 03766 USA<br />
603 643 2600 • www.fluent.com<br />
© 2006 <strong>Fluent</strong> Inc. All rights reserved.<br />
Editor: Liz Marshall<br />
Assistant Editor: Susan Wheeler<br />
Contributing Editors: Erik Ferguson and Keith Hanna<br />
Design: Lufkin Graphic Designs<br />
An iso-surface of liquid at two times, corresponding to a 10% saturation level<br />
predicted by FLUENT’s wicking model<br />
FLUENT, FiDAP, GAMBIT, POLYFLOW, G/Turbo, MixSim,<br />
FlowLab, Icepak, Airpak, and FloWizard are trademarks of<br />
<strong>Fluent</strong> Inc. Icepak and Airpak are joint developments of<br />
<strong>Fluent</strong> Inc. and ICEM-CFD Engineering. All other products<br />
or name brands are trademarks of their respective holders.<br />
S2 <strong>Fluent</strong> News · Spring 2006
MIXING<br />
The two-impeller system with conductivity probes<br />
Contours and vectors of velocity magnitude in a<br />
vertical plane mid-way between two baffles,<br />
computed by the sliding mesh model and the<br />
standard k-ε (top) and LES (bottom) turbulence<br />
models<br />
Tracing<br />
Homogenization<br />
By Michal Mo˘stĕk and Milan Jahoda, Department of Chemical Engineering,<br />
Prague Institute of Chemical Technology, Prague, Czech Republic<br />
THE BLENDING OF MISCIBLE LIQUIDS is a<br />
common mixing operation in the chemical, biochemical,<br />
food-processing, and other industries.<br />
Most frequently, it takes place in mechanically<br />
agitated tanks, where the miscible liquids are<br />
blended to a predetermined degree of homogeneity.<br />
One of the critical parameters of a mixer<br />
is the time of homogenization, or blending time.<br />
In practice, it is widespread to use empirical correlations<br />
derived from experiments to predict<br />
mixing times. However, these correlations have<br />
many limitations, mainly in the design of<br />
non-standard tank geometries and in cases with<br />
multiple impellers.<br />
Using CFD, the study of processes taking place in<br />
mixing tanks becomes easier and more economical<br />
than with the use of experiments. Thanks to<br />
the growing performance of today’s computers<br />
and with the help of other timesaving tools like<br />
the preprocessor MixSim, an elementary simulation<br />
of a mixing system can be performed within<br />
an hour. For more advanced simulations, the possibility<br />
of parallel computations and the availability<br />
of new and faster non-iterative solvers today<br />
enable a detailed study of mixing tanks to be performed<br />
in a range of days and weeks, where it<br />
would have taken months not long ago.<br />
In mixing tanks, where baffles are used to eliminate<br />
the central vortex and prevent air entrainment<br />
in the system, two methods of simulating<br />
the impeller rotation are most often used. The<br />
multiple reference frames (MRF) method is a fast<br />
steady-state method, which together with the<br />
k-ε family of turbulence models, provides timeaveraged<br />
information about processes taking<br />
place in the system. The transient sliding mesh<br />
(SM) method is much more accurate and suitable<br />
for the simulation of homogenization. For both<br />
methods, the tank is divided into a cylindrical<br />
part containing the impeller and a stationary part<br />
containing the baffles and tank walls. In the MRF<br />
method the impeller maintains a fixed position<br />
relative to the baffles, and the solution in the<br />
impeller region is solved in the rotating frame of<br />
the impeller. For the SM method, the grid for the<br />
impeller region rotates in time, so that transient<br />
effects, especially those applied to phenomena<br />
such as blending, can be captured. While<br />
the time-averaged k-ε models are often used in<br />
combination with the SM model, a transient<br />
turbulence model can also be used for the most<br />
comprehensive simulation of a transient process.<br />
When this level of detail is of interest, a transient<br />
turbulence model such as large eddy simulation<br />
(LES) can be used. To get the most out of this<br />
kind of modeling effort in terms of accuracy and<br />
precision, care must be taken when setting up<br />
the model, in terms of the size and density of the<br />
computational grid, the time step, and a suitable<br />
choice of sub-grid scale turbulence model.<br />
Two examples of liquid homogenization in a<br />
tank filled with water and equipped with either<br />
one Rushton turbine or two 6-bladed 45°<br />
pitched blade turbines mounted on a common<br />
shaft are presented below. For both cases, the<br />
tank diameter, T, is 0.29m and the shaft rotates<br />
at a constant speed of 300rpm. A second liquid<br />
(a saturated solution of NaCl) is introduced to<br />
the tank and the progress of blending is studied.<br />
Conductivity probes are used for experimental<br />
observations of tracer concentration,<br />
and CFD models are run for comparison. The<br />
measuring volume of the probe is approximately<br />
0.55cm 3 . For the one-impeller system,<br />
the probe is placed between the baffles, with<br />
an off-bottom clearance of T/4 and at a distance<br />
of T/20 from the tank wall. For the twoimpeller<br />
system, one probe is placed near the<br />
tank bottom, with the same off-bottom clearance<br />
and distance from the wall as in the previous<br />
case. A second probe is placed at a distance<br />
of T above the first probe. The tracer is<br />
injected just below the free liquid surface, at a<br />
horizontal distance of T/4 from the tank wall,<br />
opposite the probe. Output signals from the<br />
PROCESS INDUSTRIES<br />
<strong>Fluent</strong> News · Spring 2006<br />
S3
MIXING<br />
PROCESS INDUSTRIES<br />
normalized concentration<br />
normalized concentration<br />
2.0<br />
1.5<br />
1.0<br />
0.5<br />
0.0<br />
0.0 2 4 6 8 10<br />
time (s)<br />
2.0<br />
1.5<br />
1.0<br />
experiments<br />
mrf ske<br />
sm ske<br />
sm les<br />
0.5<br />
experiments<br />
mrf ske<br />
sm ske<br />
sm les<br />
0.0<br />
0.0 5 10 15 20 25<br />
time (s)<br />
The time traces of normalized concentration<br />
recorded (top) at one location in the one-impeller<br />
tank and (bottom) at two locations in the twoimpeller<br />
tank during the homogenization process,<br />
comparing three modeling approaches with five<br />
identical experimental runs; the sliding mesh with<br />
LES approach is the closest fit to the data<br />
probes are processed by a conductivity meter,<br />
digitized by an A/D converter and registered<br />
by a computer for further processing in the<br />
form of dimensionless concentration values.<br />
In the simulations, structured grids composed of<br />
non-uniformly distributed hexahedral cells were<br />
made in GAMBIT. The tank volume was decomposed<br />
into many smaller volumes in order to<br />
make a grid usable for structured meshing. In<br />
the case of the one-impeller system, the whole<br />
vessel was divided into 611,000 hexahedral<br />
cells (99 x 37 x 168 along the axial, radial<br />
and tangential coordinates), in the case of the<br />
two-impeller system, the mesh consisted of<br />
1,200,000 hexahedral cells (198 x 37 x 168 along<br />
the axial, radial and tangential coordinates).<br />
The first group of simulations of the flow field was<br />
performed using the MRF technique. By tracking<br />
the mixing of two fluids with equal properties,<br />
the solution of a single species equation could be<br />
used to follow the mixing progress. Converged<br />
solutions were obtained and then the flow field<br />
was “frozen” and only the unsteady solution of<br />
the passive scalar equation was performed.<br />
In the case of the SM and standard k-ε (SKE)<br />
turbulence model, the MRF solution was used as<br />
a starting point. The simulation was then<br />
switched to unsteady, with a time step set to<br />
0.001s and the non-iterative time advancement<br />
(NITA) option. When the periodic steady-state<br />
was reached (after 25s in the case of the<br />
one-impeller system and about 40s in the case of<br />
the two-impeller system), the homogenization<br />
simulation was initiated in the same way as<br />
described for the MRF simulation. The only difference<br />
was that the computations of tracer distribution<br />
in the vessels were solved together with<br />
the flow equations as the impeller mesh continued<br />
to rotate.<br />
In the case of the SM and LES approach, the MRF<br />
solution was again used as a starting point. The<br />
dynamic Smagorinsky-Lilly sub-grid scale model<br />
was chosen to capture the effects of turbulent<br />
eddies smaller than a cell size. Special attention<br />
was also given to the discretization schemes used.<br />
In particular, the bounded central differencing<br />
(BCD) scheme was used for discretization of the<br />
momentum and species transport equations, the<br />
body force weighted scheme was used for pressure<br />
discretization, and the fractional step method was<br />
chosen for pressure-velocity coupling.<br />
Comparison of the velocity fields calculated using<br />
the standard k-ε and LES models show the<br />
increased detail provided by the LES approach.<br />
For the two-impeller system, concentration maps<br />
at different times of the LES computation illustrate<br />
the complex dispersion of the tracer during<br />
the blending process. For both vessels, comparisons<br />
of the calculated and experimental concentration<br />
profiles indicate that the sliding mesh<br />
approach using LES provides the closest overall<br />
match to the probe readings throughout the<br />
blending process. <br />
Contours and vectors of velocity magnitude in a<br />
vertical plane mid-way between two baffles,<br />
computed by the sliding mesh model and standard<br />
k-ε (left) and LES (right) turbulence models<br />
1 sec 5 sec 10 sec 15 sec<br />
Contours of normalized tracer concentration in a vertical plane mid-way<br />
between two baffles after the tracer addition<br />
S4 <strong>Fluent</strong> News · Spring 2006
MIXING<br />
Finding the Optimum<br />
Blend Time Calculation<br />
By Benoit Post, <strong>Fluent</strong> France, and Rajiv Lochan Rath, <strong>Fluent</strong> India<br />
THE MIXING OF SINGLE AND MULTIPHASE FLUIDS in stirred tank reactors<br />
is a common operation in many industries, such as chemicals, water treatment,<br />
pharmaceuticals, and petroleum. Understanding the fluid flow in these<br />
tanks is critical for equipment design, scale-up, process control, and economic<br />
factors. CFD is now being used routinely to provide this information,<br />
enabling engineers to select the best agitator design to obtain the desired<br />
process performance. Blending time evaluation is one of the key objectives of<br />
such CFD studies, and there are currently several numerical approaches that<br />
can be used for this purpose [1]. The objective of a project recently performed<br />
at <strong>Fluent</strong> France and <strong>Fluent</strong> India was to compare these different<br />
approaches in terms of quantitative results and CPU time.<br />
PROCESS INDUSTRIES<br />
A cylindrical, flat bottom, stirred tank with 4 baffles and a non-standard, Rushtontype<br />
impeller with 4 radial blades was used for the simulations. Owing to the<br />
rotational symmetry of the geometry, a 90° sector of the tank was modeled.<br />
The different approaches studied made use of either the multiple reference<br />
frames (MRF) or sliding mesh (SM) model. For the MRF runs, a steady-state<br />
flow field was first computed and a transient calculation of a tracer species<br />
was performed using the “frozen” flow field. The first such blending calculation<br />
was done using the default settings, including the use of relative velocities.<br />
A second blending calculation was performed with absolute velocities.<br />
For this case, the paddles on the impeller were changed to interior zones,<br />
since this setting is more appropriate for walls moving normal to the fluid<br />
when absolute velocities are used. The walls exerting shear on the fluid (the<br />
impeller disk) were not changed. For the transient sliding mesh calculations,<br />
the flow field and tracking of the tracer species were computed simultaneously.<br />
Two solver options were used: the iterative time advancement (ITA)<br />
scheme and the non-iterative time advancement (NITA) scheme, which was<br />
introduced in FLUENT 6.2. The sliding mesh results are the most rigorous, so<br />
they served as a standard for comparison with the other methods. A correlation<br />
was also used to compare the results.<br />
Pathlines illustrate the two circulation loops<br />
in the vessel<br />
A steady state, MRF solution for the flow was first obtained. On this frozen<br />
flow field solution, the transient species calculation was performed after initializing<br />
a volume of tracer in the upper part of the vessel. The MRF flow field<br />
was also used as the starting point for the unsteady sliding mesh calculation.<br />
Before introducing the species, the sliding mesh calculation was done for a<br />
few cycles until periodic behavior was obtained. At this point, the species was<br />
introduced in the same location used for the MRF calculations, and the transient<br />
calculation of the flow field and species was performed.<br />
Results were analyzed globally in terms of the blending time, t 99 , or time<br />
required to achieve 99% uniformity in the tank. A published correlation for<br />
t 99 for a six-bladed Rushton turbine was used for comparison [2]. The blend<br />
time predicted by the Rushton correlation was found to be longer that that<br />
predicted by the CFD methods, a result that is due to the fact that the paddles<br />
on the impeller used have more surface area than a standard Rushton.<br />
In addition, the tracer species introduced from one location in the 90° model<br />
was equivalent to it being introduced from four locations in a 360° vessel.<br />
The tracer species initial locations at t=0<br />
<strong>Fluent</strong> News · Spring 2006<br />
S5
MIXING<br />
PROCESS INDUSTRIES<br />
The tracer distribution after 1.5 (left) and 3.0<br />
(right) seconds for the MRF simulation<br />
mass fraction<br />
mass fraction<br />
6E-03<br />
5E-03<br />
4E-03<br />
3E-03<br />
2E-03<br />
1E-03<br />
SM + ITA<br />
SM + NITA<br />
MRF – ABS<br />
MRF – REL<br />
0E-03<br />
0 2 4 6 8 10 12 14<br />
time (s)<br />
Mass fraction of the tracer species computed by the<br />
SM and MRF methods, recorded at points near the<br />
top of the vessel (top) and near the impeller<br />
(bottom)<br />
6E-03<br />
5E-03<br />
4E-03<br />
3E-03<br />
2E-03<br />
SM + ITA<br />
SM + NITA<br />
MRF – ABS<br />
MRF – REL<br />
The correlation assumes that the species is introduced from one location, usually<br />
at the surface of the liquid. Comparison of the CFD methods revealed that the MRF<br />
model predicted longer mixing times than the sliding mesh model. The two<br />
MRF approaches predicted similar blend times, as did the two SM approaches.<br />
Correlation MRF with MRF with Sliding Sliding<br />
for 6-bladed relative absolute Mesh Mesh<br />
Rushton velocities velocities with ITA with NITA<br />
turbine (default) scheme scheme<br />
Blending time(s): t 99 19.0 (+/- 6) 15.8 (+/- 4) 15.3 (+/- 4) 14.8 (+/- 3) 14.9 (+/- 3)<br />
The various approaches were also compared by monitoring the CPU time required<br />
to reach t 99 . It was found that the sliding mesh model using the new NITA scheme<br />
takes about 4 times less CPU time than the sliding mesh model using the ITA<br />
scheme. The steady-state MRF solutions are much quicker, requiring about 2.5<br />
times less CPU time than the sliding mesh model with the NITA scheme.<br />
The results were also analyzed locally by monitoring the mass fraction of the tracer<br />
species at several sensor locations. Two locations in particular are of interest: near<br />
the tip of the blade and near the top of the tank. Near the impeller, the<br />
sliding mesh case picked up the blade passing frequency, and in addition, a low<br />
frequency oscillation of greater amplitude that was also observed for the MRF<br />
approaches. This low frequency oscillation was also picked up at the top of the<br />
tank by the sliding mesh and MRF approaches, although the blade passing<br />
frequency was not (for the former). The low frequency oscillation is indicative of a<br />
macroscopic exchange of the tracer material between the upper and lower regions<br />
of the vessel. This type of result is not uncommon when a radial impeller is used<br />
and distinct circulation patterns develop above and below the impeller.<br />
Overall, the MRF model is an economic approach for delivering sufficient blending<br />
time information. If, however, an accurate representation of the transient dispersion<br />
of a tracer is required – for reacting flows, for example – the sliding mesh approach<br />
is necessary, and the NITA scheme in FLUENT now makes sliding mesh calculations<br />
much more affordable. <br />
References:<br />
1 Marshall, E.M. and Bakker, A.: Computational Fluid Mixing. In: The Handbook of Industrial<br />
Mixing, Ch. 5. Paul, E., Atiemo-Obeng, V. and Kresta, S., Editors, John Wiley, 2004.<br />
2 Fasano, J.B., Bakker, A. and Penney, W.R.: Advanced Impeller Geometry Boosts Liquid<br />
Agitation. Chemical Engineering, August 1994.<br />
1E-03<br />
0E-03<br />
0 2 4 6 8 10 12 14<br />
time (s)<br />
Iso-surfaces of tracer mass<br />
fraction show the polarized<br />
species distribution in the<br />
vessel, even after the liquids<br />
are well mixed<br />
S6 <strong>Fluent</strong> News · Spring 2006
MICROREACTORS<br />
Liquid Mixing<br />
in Microreactors<br />
By M. Hoffmann, M. Schlueter, N. Raebiger; University of Bremen,<br />
Institute of Environmental Process Engineering (IUV), Bremen, Germany<br />
MICROREACTORS ARE BASIC components<br />
of microfluidic systems for chemical and biochemical<br />
applications and are an important area<br />
of research in fields such as analytical chemistry,<br />
chemical engineering, and life science. They are<br />
well suited for the controlled mixing of reactants<br />
and show great potential for optimizing conventional<br />
processes. The large area-to-volume ratio<br />
of microreactors allows for a higher yield and<br />
selectivity than conventionally designed processes.<br />
To take advantage of the full potential of this<br />
ambitious technology, a fundamental understanding<br />
of the transport processes on the relevant<br />
time and length scales is necessary. One<br />
approach is through the application of CFD. An<br />
example that is currently undergoing study using<br />
numerical simulation is that of a T-shaped<br />
micromixer with a rectangular cross section.<br />
High resolution CFD simulations have been<br />
performed using FLUENT 6.1 at the Institute of<br />
Chemical Engineering at the University of<br />
Paderborn, Germany [1]. The computational<br />
domain is meshed by a block structured grid with<br />
about 970,000 cubic cells. A more refined computation<br />
is also being performed on a grid of<br />
about 9 million cells. For the entrance region in<br />
the mixing channel, inside the mixing zone, the<br />
grids have a spatial resolution of 2.5 and 1µm<br />
respectively. The velocity field is given by the stationary<br />
solution of the Navier-Stokes equations<br />
using a velocity profile of fully developed, rectangular<br />
cross section duct flow at both inlets. At the<br />
outlet, the pressure is set to a reference value. All<br />
channels are closed at the top and a no-slip<br />
boundary condition is applied to all walls. The<br />
properties of water at 20°C are used and a tracer<br />
is added at one of the inlets. The species transport<br />
model is used to track the motion of the<br />
tracer through the micromixer.<br />
In addition to using a commercial CFD program<br />
for numerical flow visualization, microscale fluid<br />
flow visualization is an important tool for acquiring<br />
localized flow information within these<br />
devices. By means of two non-invasive measurement<br />
techniques, micro-laser induced fluorescence<br />
(micro-LIF) and micro-particle image<br />
velocimetry (micro-PIV), the concentration and<br />
velocity fields have been measured.<br />
The tracer distribution at different cross sections<br />
enables a quantitative analysis of the mixer to be<br />
performed. To construct the tracer distribution, a<br />
confocal laser scanning microscope (CLSM) is<br />
used. The CLSM can produce image slices with a<br />
minimum thickness of 2 microns. By combining a<br />
series of these slices, a three-dimensional rendering<br />
of the mixing quality in the mixing channel<br />
is possible. The mixing quality can then be quantified<br />
using Danckwerts’ intensity of segregation<br />
[1]. Other observations made in the CFD analysis<br />
and experimental tests include a secondary flow<br />
in the form of a vortex pair in the entrance<br />
region, which results from instabilities caused by<br />
centrifugal forces.<br />
The results are fundamental for improving the<br />
design rules for static micromixers and have<br />
enabled a validation of the CFD results. <br />
Acknowledgements:<br />
The authors gratefully acknowledge the CFD work<br />
done by our collaborators, Dieter Bothe, Center of<br />
Computational Engineering Science, RWTH Aachen,<br />
Germany and Hans-Joachim Warnecke, Institute of<br />
Chemical Engineering, University of Paderborn,<br />
Germany.<br />
References:<br />
1 Bothe, D., Stemich, C. and Warnecke, H.-J.: Fluid<br />
Mixing in a T-shaped Micro-mixer. Chem. Eng. Sci.<br />
61, 2950-2958, 2006.<br />
Visit us at<br />
ACHEMA 2006:<br />
Hall 1.2, Stand E9-F10<br />
Tracer profile in the y-z plane (top) on a crosssectional<br />
slice through the mixing channel<br />
(bottom) at x=100 µm computed using CFD; the<br />
Reynolds number for the mixing channel is 186<br />
(corresponding to an inlet velocity of 1.4 m/s for<br />
each inlet)<br />
Courtesy of the Institute of Chemical Engineering, University of Paderborn<br />
Velocity field along a mixing channel with a width<br />
of 400 µm and height of 200 µm using micro-PIV;<br />
the measurement plane (x-y plane) is at z=143 µm,<br />
and flow is from left to right<br />
Concentration field in the mixing channel of a<br />
T-shaped micromixer with a width of 400 µm and<br />
height of 200 µm, constructed using image slices<br />
from a confocal laser scanning microscope and<br />
micro-LIF in the y-z plane, with an x range of 270<br />
to 782 µm; the Reynolds number of mixing<br />
channel is 207 and the flow is from back to front<br />
PROCESS INDUSTRIES<br />
<strong>Fluent</strong> News · Spring 2006<br />
S7
THERMAL RUNAWAY<br />
PROCESS INDUSTRIES<br />
Preventing<br />
Runaway<br />
Reaction<br />
Accidents<br />
By D. Dakshinamoorthy and J.F. Louvar, Chemical Engineering Department,<br />
Wayne State University, Detroit, Michigan, USA<br />
RUNAWAY REACTIONS ARE AN ONGOING PROBLEM in the chemical<br />
industry, where they account for 26% of major accidents. Runaway<br />
reactions generate a sudden excess amount of heat, which can lead to<br />
an explosion. They can be stopped in two ways: by the addition of cold<br />
diluents and by the addition of an inhibitor, a chemical that acts to suppress<br />
the runaway reaction. The technology that involves the use of<br />
inhibitors is called shortstopping. Power failures are one of the main<br />
reasons for runaways, and after a power failure, the process of adding<br />
an inhibiting agent and mixing it with the reactor contents becomes a<br />
major problem in the shortstopping process. Jets or impellers, driven by<br />
a small generator, can be used for mixing under such circumstances.<br />
From a design standpoint, jet mixing is one of the simplest methods to<br />
achieve mixing. In jet mixing, a part of the liquid in the tank is drawn<br />
out through a pump and returned to the tank as a high-velocity jet<br />
through a nozzle, resulting in fluid mixing. In a recent project [1], CFD<br />
was used to compare the efficiency of jet mixers with impeller stirred<br />
vessels in shortstopping runaway reactions. On the basis of equal power<br />
consumption, this comparative study showed that jet mixers were ineffective<br />
when used for shortstopping, unless certain factors could be<br />
optimized. Due to the hazardous nature of runaway reactions, these<br />
factors cannot be determined with lab scale or pilot plant scale experiments,<br />
but CFD can be used to carry out virtual experiments instead.<br />
Using FLUENT, mixing with a jet mixer was first investigated for different<br />
nozzle diameters and angles of injection. To account for the external<br />
circulating pump, user-defined functions (UDFs) were used to<br />
ensure that the mass fractions of all species entering the vessel through<br />
the inlet were equal to those leaving through the outlet at each<br />
timestep. Before simulating the reacting flow of the shortstopping<br />
process, the flow model alone was validated. Since overall mixing controls<br />
the process, the predicted mixing times were compared with the<br />
available experimental correlations [2, 3] and found to be in excellent<br />
agreement.<br />
Solution domain and pathlines for the jet<br />
mixer studied<br />
The converged flow results of the best jet configuration were then used<br />
for subsequent simulations of the runaway and inhibition reactions.<br />
Laminar volumetric reactions were modeled using UDFs. Guibert et al. [4]<br />
documented the kinetics of the considered runaway reaction. Before analyzing<br />
a specific runaway scenario, reactor conditions were varied and<br />
runaway behavior was studied. From all of the cases considered, the one<br />
with the fastest temperature increase was selected for further study, since<br />
it would be the most difficult to shortstop. For this scenario, as soon<br />
as the reactor temperature reached 450K, the inhibitor was added and<br />
the species transport equations for the inhibition reaction were solved<br />
simultaneously with the runaway reaction.<br />
The results were used to identify the major and minor factors that contribute<br />
to effective shortstopping when using a jet mixer. These factors<br />
include the location for adding the inhibitor; the amount of the<br />
inhibitor; the rate of the inhibition reaction; the power input; the use of<br />
a cold diluent; and the use of multiple nozzles. The temperature distribution<br />
in the reactor after the shortstopping process and the decrease<br />
in the average reactor temperature were used to assess the importance<br />
of each factor.<br />
S8 <strong>Fluent</strong> News · Spring 2006<br />
Grid near the inlet<br />
For example, the temperature distribution on the mid-plane and an<br />
iso-surface of temperature equal to 500K (temperatures over 500K were<br />
considered hazardous) for two power levels was used to illustrate the<br />
importance of power input. The hotspots, or high temperature regions<br />
over 500K, decrease in size as the power is increased, which indicates<br />
improved shortstopping. In addition to the power input, the decrease in
THERMAL RUNAWAY<br />
mixing time (sec)<br />
10 3<br />
10 2<br />
10 1 Jet Reynolds number<br />
10 2 10 3<br />
10 4<br />
experiment<br />
FLUENT<br />
Comparison of the predicted mixing time and experimental data for a<br />
range of jet Reynolds numbers [2]<br />
the final reactor temperature for the other factors was computed as well. When<br />
compared, the major factors that were found to contribute most to effective<br />
shortstopping were the use of cold diluents and the use of multiple nozzles.<br />
The factors found to be of less importance were the rate of the inhibition reaction,<br />
the location of the inhibitor injection, and the amount of inhibitor. The<br />
study clearly demonstrated the value of using CFD simulations in situations that<br />
are experimentally prohibitive. <br />
Kinetic Details of<br />
Runaway and<br />
Inhibition<br />
Runaway Reaction<br />
Guibert et al. [4] studied the kinetics of the propylene oxide polymerization<br />
reaction. The rate of the reaction is defined as a function of temperature,<br />
propylene oxide concentration, and catalyst concentration.<br />
Monomer polymerizes to polymer in the presence of a basic catalyst. The<br />
kinetic expressions for the monomer concentration and temperature are<br />
as follows:<br />
1.<br />
2.<br />
PROCESS INDUSTRIES<br />
Acknowledgements:<br />
The authors acknowledge NSF for their support of this particular project, and Dr. Vivek.<br />
V. Ranade and his student Dr. Avinash. R. Khopkar of IFMG, National Chemical<br />
Laboratories, Pune, India, for their technical contributions and inspiration.<br />
References:<br />
1 Dakshinamoorthy, D., Khopkar, A.R., Louvar, J.F. and Ranade. V.V.: CFD Simulations<br />
of Shortstopping Runaway Reactions in Vessels Agitated with Impellers and Jets.<br />
Accepted in the Journal of Loss Prevention in the Process Industries, 2005.<br />
2 Lane, A.G.C. and Rice, P.: An Investigation of Liquid Jet Mixing Employing an<br />
Inclined Side Entry Jet. Transactions of Institute of Chemical Engineers, 60, 171-176,<br />
1982a.<br />
3 Lane, A.G.C. and Rice, P.: Comparative Assessment of the Performance of Three<br />
Designs for Liquid Jet Mixing. Industrial & Engineering Chemistry Process Design<br />
and Development, 21, 650-653, 1982b.<br />
4 Guibert, M.R., Plank, A.C. and Gerhard, R.E.: Kinetics of the Propylene Oxide –<br />
Oxypropylated Glycerol Reaction. Industrial & Engineering Chemistry Process<br />
Design and Development, 10 (4), 497-500, 1971.<br />
Inhibition Reaction<br />
The added inhibitor neutralizes the basic catalyst. The kinetic expression<br />
for the inhibition reaction is as follows:<br />
3.<br />
Kinetic Data for<br />
Inhibition and<br />
Runaway Reactions<br />
Activation Energy<br />
Heat of Reaction<br />
Pre-Exponential Factor<br />
Specific Heat Capacity<br />
Pre-Exponential Factor<br />
Gas Constant<br />
E = 6.96E+08 J/kgmol<br />
H = -1.63E+06 J/kg(monomer)<br />
k o = 9.5E+11 kg(total)/(kg(cat)-hr)<br />
c p = 2930 J/kg(total)-K<br />
k 1 = 9.5E+11 kg(total)/(kg(inh)-hr)<br />
R = 8314.34 J/kgmol-K<br />
< 450K > 550K<br />
Contours of temperature on the mid-plane and an iso-surface of tempetature at<br />
500k illustrate the temperature distribution after shortstopping with less (left)<br />
and more (right) power input<br />
<strong>Fluent</strong> News · Spring 2006<br />
S9
EMISSIONS<br />
PROCESS INDUSTRIES<br />
Scrubbers for Flue Gas Cleanup<br />
By Christoph Hochenauer, Austrian Energy & Environment AG, Raaba/Graz, Austria,<br />
Martin Demuth, Institute for Process Technology & Industrial Environmental Protection,<br />
University of Leoben, Austria, and Wolfgang Timm, <strong>Fluent</strong> Germany<br />
The mesh structure<br />
WITH TODAY’S INCREASINGLY RESTRICTIVE environmental<br />
regulations, there is an ongoing need for state-of-theart<br />
modeling of pollutant formation and removal in industrial<br />
processing units. A particularly important issue is the sulphur<br />
dioxide (SO 2 ) content in the flue gases of power plants. A<br />
typical approach for removing this hazardous component is<br />
through absorption by limestone slurry droplets in so-called<br />
flue gas scrubbers. Scrubbers consist of large towers in which<br />
up to several thousands of nozzles provide a homogenous<br />
distribution of droplets within the gas flow. This leads to<br />
the highest possible absorption rate for a given slurry mass<br />
flow. To optimize the nozzle configurations for scrubbers,<br />
CFD simulation is one of the best tools to use.<br />
Mesh in the spray bank zone<br />
The absorption of SO 2 is a complex process with several<br />
mechanisms to be considered. The continuous phase (gas) and<br />
discrete phase (droplets) have to be calculated as well as the<br />
interaction between them. The momentum and temperature<br />
of the droplets affect the gas flow and vice versa.<br />
Additionally, mass transfer between the droplet and gas phases<br />
has to be taken into account. Depending on the difference<br />
in the partial pressure of water vapor in the two phases, there<br />
might be evaporation from the droplets or condensation onto<br />
the droplet surfaces. Rapid changes in temperature cause both<br />
evaporation and condensation to have a significant influence<br />
on the SO 2 absorption rate, which itself is governed by a complex<br />
ion chemistry mechanism within the droplets. The heat<br />
and mass transfer rates are dependent on the droplet surface<br />
area, which changes during evaporation or condensation, or<br />
whenever there is splashing at the walls. To account for the<br />
latter, an appropriate wall interaction model also has to be part<br />
of the CFD simulation. With FLUENT’s user-defined function<br />
(UDF) capability, a set of models addressing these issues has<br />
been implemented, and a test case with typical scrubber<br />
conditions has been created to illustrate the capabilities.<br />
A cylindrical scrubber with two sets of eight radial spray bars<br />
was considered. The grid was built using the polyhedral mesh<br />
S10 <strong>Fluent</strong> News · Spring 2006
EMISSIONS<br />
capability of FLUENT 6.3. The original mesh consisted of a total of 1.25 million<br />
cells, which were reduced to 250,000 cells after the tetrahedral elements<br />
were converted to polyhedra. Using sizing functions, a sufficient grid resolution<br />
in the near wall regions of the spray bank was ensured. A flue gas mass<br />
flow rate of 65kg/s was set at an inlet at the bottom of the unit with SO 2 and<br />
water vapor mass fractions of 0.0024 and 0.12 respectively, and the remainder<br />
consisting of air. Sixty-six full cone nozzles were distributed in a circular<br />
pattern on the sixteen radial pipes for a total slurry mass flow rate of 330kg/s.<br />
Both the gas flow and the droplets started with an initial temperature of<br />
316K. For the droplets, the dispersed phase model (DPM) was used with a<br />
large number of trajectories, whose interaction with the gas phase was evaluated<br />
separately. The RNG k-ε model and the stochastic tracking method<br />
were used to incorporate the effects of turbulent fluctuations on both<br />
phases in the simulation.<br />
In a surrounding air flow at ambient temperature, a water droplet is subject<br />
to a coupled heat and mass transfer process. With increasing velocity<br />
difference between the two phases the transfer is enhanced and can be<br />
calculated with the Ranz-Marshall law, which can be used for SO 2 absorption<br />
as well as vaporization and condensation [1]. Since the latent heat of water<br />
Contours of water vapor content (left), SO 2 mass fraction (middle), and vertical<br />
velocity (right) on four slices within the scrubber; the flow in the unit is from<br />
bottom to top<br />
is quite high, the droplet temperature change along its trajectory cannot be<br />
neglected, since a temperature change of 10K can significantly affect the SO 2<br />
partial pressure at the droplet surface.<br />
When droplets hit the wall there are several possibilities of what can occur.<br />
Under certain conditions a droplet can become part of the liquid film on the<br />
wall and in other instances there may be splashing and breakup of the<br />
droplets. The latter results in an increased number of droplets with a larger<br />
total surface area, a change that again significantly affects the momentum,<br />
heat and mass transfer rates. For this application, it was assumed that the wall<br />
interaction behavior was governed by dimensionless Weber and Reynolds<br />
numbers [2] and a model was developed accordingly.<br />
PROCESS INDUSTRIES<br />
The removal of SO 2 in the droplets is governed by a reaction with limestone,<br />
which can be described by the net reaction:<br />
Droplet temperature evolution along some of the droplet trajectories<br />
Additionally there are several dissolution reactions, since the formation of sulphurous<br />
and carbonic acid results in an ion chemistry mechanism that<br />
depends on the pH-value and the limestone and SO 2 concentrations of the<br />
droplets [3]. This was implemented by means of an externally evaluated<br />
lookup table that was coupled with the discrete phase using FLUENT’s particle<br />
scalar UDF, which allows additional droplet properties to be defined.<br />
Once the total sulphur content in a droplet has been evaluated, the new concentration<br />
of each ion is obtained from the table and stored in the respective<br />
DPM scalar. With this information the calculation of the SO 2 partial pressure<br />
at the droplet-gas interface and the respective sink terms is possible to ensure<br />
SO 2 mass conservation in both phases of the domain.<br />
Droplet splashing and breakup at the walls; the droplet trajectories are colored by<br />
particle diameter, so the red trajectories occur before splashing and breakup occur<br />
The results obtained from the simulation are in very good agreement with<br />
real-life scrubbers. Austrian Energy & Environment AG now includes FLUENT<br />
with these special models in their scrubber design workflow. The polyhedral<br />
cell technology has proved to be a significant improvement in terms of meshing<br />
flexibility and simulation time, enabling engineers to perform a complete<br />
scrubber simulation within a few days. The reliability of the results has meant<br />
that they need to rely less on field tests as well. <br />
References:<br />
1 Aguayo, P. and Weiss, C.: Enthalpy Two Way Coupling for Near Vapor Saturated<br />
Polydispersed Spray Flows. 5th International Conference on Multiphase Flow,<br />
Yokohama, May 30 - June 4, 2004.<br />
2 Mundo, C. and Tropea, C.: Numerical and Experimental Investigation of Spray<br />
Characteristics in the Vicinity of a Rigid Wall. Experimental Thermal and Fluid<br />
Science, 15:228-237, 1997.<br />
3 Weiss, C., Maier, H. and Bärnthaler, K.: Modelling of the Mass Transfer Processes in<br />
the Chemisorption of Flue-Gas Components by Sprays. Proceedings of the ILASS-<br />
Europe 2001; Zürich, 2-6 September 2001.<br />
<strong>Fluent</strong> News · Spring 2006<br />
S11
FURNACES<br />
PROCESS INDUSTRIES<br />
Ultra-low NO x Burners<br />
By Robert J. Gartside and Peter R. Ponzi, ABB Lummus Global, Bloomfield, New Jersey, USA, and David G. Schowalter, <strong>Fluent</strong> Inc.<br />
wall burners<br />
hearth burners<br />
convection<br />
section<br />
The geometry of the ethylene cracking furnace<br />
cracking coils<br />
Get<br />
ENGINEERS AND MANAGERS involved in chemical and<br />
hydrocarbon processing are keenly aware of equipment efficiency<br />
and plant productivity, and the ongoing need for improvement<br />
in these areas. In ethylene production furnaces, light hydrocarbons<br />
are cracked in tubes that are suspended in a combustion<br />
chamber. The tubes have very short residence times. The critical<br />
parameters in the operation of this furnace are the transfer of the<br />
heat of cracking to the tube side hydrocarbons, control of the<br />
tube metal temperatures for prolonged run length, and the<br />
reduction of pollutants such as NO x to meet regulatory statutes.<br />
In the radiant section of a typical furnace, combustion heat is<br />
provided by burning fuel in hearth burners (located on the<br />
floor of the furnace and firing vertically) and in wall burners<br />
(positioned along the wall and firing radially along the wall).<br />
Temperature contours in the original burner<br />
configuration (left) and the first ultra-low NO x<br />
design (right)<br />
The John Zink Company has developed an ultra-low NO x burner<br />
(20 to 30 vppm) in which a portion of the fuel is premixed<br />
with all of the air in both the hearth and wall burners. The<br />
remaining fuel needed to achieve the firing rate is introduced in<br />
a staged manner to control the temperature of the combustion<br />
gases and thereby minimize the production of NO x . This is done<br />
through staged ports located in front of the hearth tile at the<br />
furnace floor. When the introduction of this burner technology<br />
was considered for a client’s ethylene production furnace, it was<br />
suspected that, when compared to the original conventional<br />
burners that generally produce straight vertical flames attached<br />
to the furnace wall, these lean premixed burners might produce<br />
shorter flames that could “roll over” at the bottom of the firebox<br />
and impinge on the tubes. The negative impact of rollover would<br />
be the creation of hot spots on the process tubes and subsequent<br />
increase in coking within the tubes and shortening of tube life.<br />
To investigate this possibility, a realistic CFD study was conducted<br />
to compare the performance of the low NO x and original<br />
burners. The detailed CFD models – one for each type of burner<br />
– included locally refined meshes around the burner inlet ports<br />
and process tubes. Combustion was simulated on the firebox<br />
side to provide the heat generation. A reacting model for the<br />
hydrocarbon cracking that occurs inside the tubes was also<br />
included, as were the effects of turbulence and radiation. Overall,<br />
there was sufficient detail to accurately reflect the heat absorption<br />
and thus provide realistic tube heat fluxes and coil metal<br />
temperatures. The models contained between two and five<br />
Temperatures on three cross-sectional planes in the<br />
original burner configuration (left) and the improved<br />
ultra-low NO x design (right)<br />
S12 <strong>Fluent</strong> News · Spring 2006
MULTIPHASE<br />
Cracking<br />
million computational cells depending on the<br />
type of burner modeled, and they required up<br />
to a week’s worth of CPU time on a high speed<br />
computer cluster.<br />
The initial simulations compared the heater with<br />
the original burners to the heater with the new<br />
burners where the new burners were simply<br />
located at the same position as the original burners.<br />
The results of this study indicated that, while<br />
the original burner design provided straight<br />
flames, the new low NO x configuration resulted<br />
in flames that impinged on the tube surfaces.<br />
Based on these results, an optimization study<br />
using CFD was done to develop design modifications<br />
that would allow the ultra-low NO x burner<br />
to perform acceptably in the heater. These modifications<br />
included redirecting the angle of the<br />
fuel that was being injected through the ports in<br />
front of the new burners and relocating the burners<br />
on the furnace floor to provide lateral spacing<br />
into which the combusting gas could expand,<br />
thereby reducing the tendency for the flames to<br />
roll into the tubes. The modifications served to<br />
straighten the flames, making them more in line<br />
with those of the original burners. While the new<br />
design produces more diffuse flames than the<br />
original burner design, it is much improved from<br />
the first ultra-low NO x design.<br />
The new burner configuration, optimized by<br />
CFD, has been installed in the client’s furnace.<br />
Flame quality and run length have been excellent<br />
and low NO x operation has been experienced.<br />
The start-up was smooth and downtime was<br />
minimal. <br />
Acknowledgement:<br />
The authors wish to acknowledge the cooperation<br />
of the John Zink Company in the low NO x burner<br />
modification CFD studies.<br />
Controlling Droplet<br />
Size Distribution in<br />
Emulsions<br />
By L. Srinivasa Mohan, <strong>Fluent</strong> India, and Ahmad Haidari and Aniruddha Mukopadhyay, <strong>Fluent</strong> Inc.<br />
Emulsions are an important class of materials produced<br />
and handled by the chemical, food, pharmaceutical,<br />
and cosmetic industries. They consist<br />
of two immiscible liquids, one dispersed in the<br />
other. The properties of an emulsion are based<br />
on the droplet size distribution (DSD) of the<br />
dispersed phase. Because they are often thermodynamically<br />
metastable, there is a persistent<br />
threat that the texture of the emulsion will be<br />
altered during the course of preparation or packaging,<br />
or during the subsequent shelf life. Many<br />
processes over widely varying length scales could<br />
cause the DSD to change, and it is important that<br />
they be well understood so that the emulsion<br />
quality can be maintained.<br />
For an ongoing project at <strong>Fluent</strong>, several aspects<br />
of droplet behavior have been studied using the<br />
volume of fluid (VOF) model. Some of the results<br />
have been compared to experiments that were<br />
carried out on microfluidic devices, where a precise<br />
droplet size distribution could be generated.<br />
For example, the generation of droplets at a<br />
T-junction using two streams of immiscible liquids<br />
has been simulated. Droplets of uniform size<br />
were rapidly and reproducibly produced at the<br />
junction as a result of the surface tension and the<br />
shearing motion of the fluid in the main channel,<br />
in agreement with measurements.<br />
In another study, the geometrically mediated<br />
breakup of droplets in a microfluidic device [2]<br />
was simulated. By changing the length of the<br />
arms of the T-junctions, the droplet can be split<br />
into daughter droplets of unequal size. By using a<br />
network of asymmetric T-junctions, emulsions of<br />
a given DSD can be produced. Both 2D and 3D<br />
simulations matched the qualitative and quantitative<br />
aspects of the experiments, such as the size<br />
of the daughter droplets for a given T-junction<br />
and the critical parameters required for droplet<br />
breakup as a function of capillary number. <br />
References<br />
1 Nisisako, T., Torri, T. and Higuchi, T.: Lab Chip. Vol<br />
2, no 1. pp 24-26, 2002.<br />
2 Link, D.R., Anna, S.L., Weitz, D.A. and Stone, H.A.:<br />
Phy. Rev. Lett. Vol. 92, pp 1178-1180, 2004.<br />
1E-07<br />
8E-08<br />
6E-08<br />
4E-08<br />
2E-08<br />
volume of droplet (m 3 )1.2E-7<br />
0<br />
0.0<br />
continuous phase flow<br />
dispersed phase flow<br />
Schematic of the experimental setup [1]<br />
0.1 0.2 0.3 0.4 0.5 0.6 0.7<br />
time (s)<br />
Volume of droplets produced at the T as a<br />
function of time<br />
Droplet breakup at a symmetric (left) and<br />
an asymmetric (right) T-junction<br />
PROCESS INDUSTRIES<br />
<strong>Fluent</strong> News · Spring 2006<br />
S13
MULTIPHASE<br />
PROCESS INDUSTRIES<br />
Understanding<br />
Fluid-Bed<br />
Coating<br />
Rod Ray, Lisa Graham, Rick Falk, Josh Shockey, and Leah Appel,<br />
Bend Research Inc., Bend, Oregon, USA, and Kumar Dhanasekharan,<br />
<strong>Fluent</strong> Inc. and L. Srinivasa Mohan, <strong>Fluent</strong> India<br />
DURING THE PAST DECADE, the use of fluidized bed systems for<br />
coating has developed to a point where this equipment is now used<br />
for applications in industries ranging from chemical to pharmaceutical to<br />
agriculture. The appeal of fluidized beds for coating is due to the high energy<br />
and mass transfer rates that are available in such a system. There are<br />
three types of fluidized beds that are typically used for this purpose:<br />
top-spray, tangential-spray, and bottom-spray. In a recent joint project<br />
involving engineers at Bend Research and <strong>Fluent</strong>, a fluidized bed with a<br />
bottom-spray of the Wurster type was studied.<br />
Fluid bed coaters are used to obtain controlled release or delayed release<br />
particles in pharmaceutical and agricultural applications. For these types of<br />
coatings, it is important to have control of both coating quality and coating<br />
uniformity. Typically, this process is run in batch mode with particles<br />
circulating through the coating zone multiple times in order to achieve the<br />
desired coating level. In this process, the particles are entrained in a high<br />
velocity gas, they pass through the Wurster column where coating is applied<br />
via a two-fluid atomizer, they then dry in the expansion chamber and<br />
fall back down to the bed to repeat the process. FLUENT has been used to<br />
help understand the process, design process conditions that yield products<br />
within an acceptable variation, and understand the effect of various<br />
parameters on product variability and quality.<br />
coating solution<br />
atomization air<br />
fluidization air<br />
Wurster gap<br />
Schematic illustration of a process stream in a<br />
bottom-spray Wurster fluidized bed coater<br />
Accounting for the presence of a secondary phase within CFD software is<br />
well established. Some of the more recent advances enable simulations of<br />
flows with very high solid loading, such as hoppers and chutes; flows with<br />
large particles, such as particle milling machines; gas-solid flows with particle<br />
size variation, such as polymerization reactors; and flows where the effects<br />
of particle shapes and particle mechanics are important. For the Wurster bed<br />
simulation, the Eulerian granular model in FLUENT was used. The results<br />
have led to a better understanding of the uniformity of coating, spray<br />
patterns, and fluidization behavior. Parameters that control the process<br />
efficiency and coating uniformity were studied. These included the<br />
equipment geometry, air flow rate, and other spray-related characteristics.<br />
Of particular importance for the equipment geometry is the Wurster gap, a<br />
small passage at the base of the unit between the inner and outer (annular)<br />
columns. During operation, the particles in the bed are fluidized and<br />
coated primarily in the inner column, where the fluidization velocity is<br />
highest. As a result, a steady fountain of solid particles is produced, and after<br />
reaching the upper region of the device, these particles stream down the<br />
walls of the outer section prior to being entrained through the gap and<br />
re-entering the primary fluidization region.<br />
The results indicated that there is a steady solids flux through the larger gap,<br />
but in the case of the smaller gap, the flow of solids through the gap is<br />
choked periodically and the bed does not maintain a steady fountain. This<br />
creates a non-uniform solids flux in the spray region, which could lead to<br />
problems with coating uniformity. <br />
Contours of particle volume fraction for two different Wurster gap designs:<br />
wide (left) and narrow (right), for a similar air flow rate<br />
S14 <strong>Fluent</strong> News · Spring 2006
PUMPS<br />
Pathlines through one of<br />
Tuchenhagen’s pump designs<br />
Pumping out<br />
New Designs<br />
More Quickly<br />
By Matthias Südel, Tuchenhagen GmbH, Büchen, Germany<br />
TUCHENHAGEN has been producing process components<br />
for the brewery, beverage, dairy and food industries for more<br />
than seventy years. In addition, the valves, pumps, and cleaning<br />
devices they produce are also used in the cosmetics,<br />
healthcare, pharmacy, and fine chemicals segments, where<br />
hygiene and sterility play a major role. The company, which<br />
has been a member of the GEA group since 1999, introduced<br />
the first twin seat valve with a mixproof technology in 1967.<br />
Customers’ expectations about the quality of process technology<br />
used have been increasing ever since. Many products<br />
must meet the stringent requirements outlined in the<br />
European Hygienic Equipment Design Group (EHEDG) and 3A<br />
sanitary standards. All devices must be capable of cleaning<br />
and flushing and must be free from dead space (clearance<br />
volume).<br />
In the past, designers made predictions about the cleaning,<br />
purification, and purging capabilities of the new components<br />
only after expensive technological trials. Today, CFD is capable<br />
of mapping these prototypes, allowing designers to examine<br />
every corner of the valve, pump, or cleaning device.<br />
FLUENT is now used to accelerate process safety and the<br />
durability of Tuchenhagen’s components. The software makes<br />
flow patterns visible and tracks even the smallest of problem<br />
areas in the course of computations. As a result, the development<br />
costs associated with the recent generation of pumps<br />
have decreased by seven percent compared to the conventional<br />
development process. The time required to reach the<br />
mass production phase has been reduced by six to eight<br />
weeks for each new design.<br />
For many years, engineers at Tuchenhagen could design and<br />
build a new pump based solely on their years of experience.<br />
With CFD, the company now has at its disposal a fast and costeffective<br />
method to support this experience with numerical<br />
data. Using this principle, Tuchenhagen developed three new<br />
pumps last year. Their approach is to carry out a computation,<br />
then built the prototype and make measurements.<br />
To improve upon this process, they have defined special<br />
configuration attributes in the CFD software. In two cases, the<br />
pumps were modeled using CFD, then tested at the operating<br />
point, and quickly put into mass production. A typical pump<br />
simulation makes use of a hybrid mesh of about 1 million cells.<br />
The steady-state MRF model is used to simulate the rotation of<br />
the moving parts. The standard k-ε model with enhanced wall<br />
treatment is used for turbulence.<br />
PROCESS INDUSTRIES<br />
A VARIVENT ® mixproof<br />
valve Type R<br />
Velocity magnitude on different<br />
planes of a VARIVENT ® valve<br />
Type R<br />
The numerical simulations have reduced the time required<br />
to go from the first prototype to the actual mass-produced<br />
product. Additionally, material costs for high-grade steel<br />
during the prototyping phase do not need to be considered.<br />
Typically, making the first sample takes up to two weeks, and<br />
measurements take up another two to four weeks depending<br />
on the components. This is partly due to the fact that the<br />
cleanability of the products, which is so important to<br />
Tuchenhagen, must be repeatedly demonstrated in product<br />
trials. While it is not possible to entirely eliminate test runs, the<br />
company can now examine predictions of the wall shear<br />
stress, which is the deciding quantity for cleanability, since it<br />
is related to the ability of the fluid to dislodge dirt particles. <br />
<strong>Fluent</strong> News · Spring 2006<br />
S15
EXTRUSION<br />
PROCESS INDUSTRIES<br />
Polymer Processing<br />
Simulation for Foam Extrusion<br />
By Hirohisa Shiode and Takeharu Isaki, Mitsui Chemicals, Inc., Material Science Laboratory, Chiba, Japan<br />
MITSUI CHEMICALS and Mitsui Chemicals<br />
Group produce a wide variety of chemical products<br />
from catalysts to polymers and fabricated<br />
products like films and fibers. The Computational<br />
Science Department at the Material Science<br />
Laboratory investigates simulation technologies<br />
and applications to research and development in<br />
the areas of chemistry, engineering, polymeric<br />
material design and polymer processing.<br />
Since 1993, POLYFLOW has been used for polymer<br />
processing applications involving various<br />
rheology models, boundary conditions, and free<br />
surface flows, such as extrusions through<br />
complex and multi-layered dies and profile<br />
extrusions. One area of interest is foam extrusion,<br />
which includes gas dissolution into the polymer<br />
melt, viscosity reduction, nucleation with pressure<br />
drop, and bubble growth. The foam properties<br />
are closely related to cell structures, which are<br />
determined by the process conditions. A simulation<br />
method for gas-dissolved polymer flow has<br />
been developed for predicting these cell<br />
structures and the optimal die design. It makes<br />
use of a viscosity model, which is based on free<br />
volume theory [1,2] and which takes into<br />
account the temperature, pressure, and gas<br />
concentration in the melt. The model is<br />
implemented in POLYFLOW by means of userdefined<br />
functions (UDFs), and is used for 3D<br />
simulations of the flow through the die. To<br />
account for the shear-thinning behavior of the<br />
material, the Cross-Carreau model is adopted.<br />
The pressure distribution in a coat hanger die with<br />
a gas concentration of 1%<br />
pressure drop (MPa)<br />
35<br />
30<br />
25<br />
20<br />
15<br />
10<br />
5<br />
0<br />
0.0<br />
The geometry of a coat hanger die<br />
0% gas<br />
1% gas<br />
2% gas<br />
3% gas<br />
0.2 0.4 0.6 0.8 1.0<br />
dimensionsless length<br />
The pressure drop<br />
through a coat<br />
hanger die along the<br />
centerline for gas<br />
concentrations<br />
ranging from 0 to 3%<br />
As an example, the customized model has been<br />
applied to the flow through a coat hanger style<br />
die. The results indicate that the pressure drop<br />
through the die decreases with increasing gas<br />
concentration as a result of the associated drop in<br />
viscosity. This result is consistent with experimental<br />
observations. Further work is now being conducted<br />
to couple the flow simulations with nucleation<br />
and cell growth. <br />
References<br />
1 Shiode, H. and Isaki, T.: Polymer-Supercritical Fluid<br />
Systems and Foams. Dec., Tokyo P-22 188, 2003.<br />
2 Willams, M.L., Landel, R.F. and Ferry, J.D.: J. Am.<br />
Chem. Soc., 77, 3701, 1955.<br />
S16 <strong>Fluent</strong> News · Spring 2006
AEROSPACE<br />
The impact of a rear-mounted nacelle on the top surface<br />
of a wing is illustrated by static pressure contours<br />
Nacelle Impact<br />
on Aircraft Wing & Fuselage<br />
By Zhu Jie, AVIC 1 (Aviation Industries of China) Commercial Aircraft Company, Shanghai, China<br />
THE GENERAL CONFIGURATION and aerodynamic<br />
design integration of the nacelle on a<br />
commercial aircraft have an important influence on<br />
economy, safety, comfort, and performance. It is a<br />
complex yet important task to study the aerodynamic<br />
interference between the wing/body and<br />
engine nacelle and develop techniques for integrating<br />
these components during the design phase.<br />
These considerations have played a role in the<br />
development of the advanced regional jet ARJ21<br />
from China’s AVIC 1 Commercial Aircraft Company<br />
(ACAC). At ACAC, FLUENT was introduced during<br />
the pre-development phase, and it continues to<br />
play a major role in design, engine selection, and to<br />
save time spent on wind tunnel tests.<br />
Before applying the turbulent Navier-Stokes<br />
equations to the complete aircraft, validations were<br />
performed on a simplified geometry. Using a 3D<br />
wing-body model, predictions of pressure and lift<br />
coefficient were compared to experimental data,<br />
and very good agreement was obtained. These<br />
results gave the engineers confidence in CFD, and<br />
its use has expanded to include many components<br />
and aspects of the full-scale aircraft.<br />
As an example, the flow above the joint that is<br />
formed between the wing and body of an aircraft has<br />
been examined. The flow in this region can separate,<br />
and if it is possible to reduce or eliminate the size of<br />
the separation region by design modifications, the<br />
flight performance of the plane can be improved.<br />
Engineers at ACAC have used FLUENT to study this<br />
phenomenon and improve the flow characteristics.<br />
Pressure (Form) Drag Friction Drag Total Drag<br />
Clean wing 0.01547 0.00412 0.01959<br />
Wing with a rear-mounted nacelle 0.01367 0.00397 0.01764<br />
As another example, a rear-mounted nacelle configuration<br />
has been studied, in which the aircraft’s<br />
engines are installed behind the wings. In this configuration,<br />
the position of the nacelle or the design<br />
of the wing can cause the wake of the wing to be<br />
entrained by the engine, compromising its performance.<br />
Using FLUENT, the pressure distribution on<br />
the wing for several different rear-mounted nacelle<br />
positions has been computed for a flight Mach<br />
number of 0.78 at an altitude of 35,000 feet. The<br />
results showed that the lift coefficient of the wing<br />
with this configuration is lower than that of a clean<br />
wing (without the interference of the nacelle). The<br />
drag coefficient is also lower, however, so that the<br />
lift/drag ratio is increased. Furthermore, the interference<br />
caused by rear-mounted nacelles makes the<br />
nose-down pitching moment decrease and it can<br />
reduce the trim drag produced by the elevators. The<br />
overall analysis of drag indicates that the existence<br />
of a rear-mounted nacelle will have little influence<br />
on friction drag, but can have the advantage of<br />
reducing the pressure drag. <br />
The influence of a rear-mounted engine nacelle on components of drag for a case with a wing lift coefficient of 0.46<br />
CP<br />
-1.5<br />
-1.0<br />
-0.5<br />
experiment<br />
FLUENT<br />
Z = 0.75<br />
Z = 0.37<br />
0.0<br />
0.5<br />
Static pressure on the<br />
surface of the ARJ21<br />
0.0<br />
0.2 0.4 0.6 0.8 1.0<br />
x<br />
A comparison of the pressure coefficient computed<br />
by FLUENT and measured in the wind tunnel for<br />
two wing span locations for the simplified wing-body<br />
geometry<br />
Flow separation near the wing-body joint (lower left)<br />
was improved by using CFD to alter the design<br />
(lower right)<br />
<strong>Fluent</strong> News · Spring 2006 19
AEROSPACE<br />
Compressors Benefit from<br />
the<br />
NASA Rotor 37<br />
By Răzvan Mahu and Cristina Oprea, TENSOR, Bucharest, Romania<br />
percent span<br />
100<br />
80<br />
60<br />
40<br />
20<br />
0<br />
1.24<br />
experiment<br />
FLUENT<br />
1.28 1.32 1.90 2.10 2.30<br />
total pressure ratio<br />
The total temperature ratio (left) and total<br />
pressure ratio (right) along the wing span,<br />
ranging from the blade root (0%) to the blade<br />
tip (100%), approximately 10cm from the<br />
leading edge at 98.7% of the maximum<br />
(experimentally measured) flow rate (20.93kg/s)<br />
2.2<br />
2.1<br />
2.0<br />
1.9<br />
19.0<br />
experiment<br />
FLUENT<br />
19.5 20.0 20.5 21.0<br />
mass flow (kg/sec)<br />
The predicted compression<br />
ratio is within 1% of the<br />
experimental values<br />
throughout the range of<br />
mass flow rates studied<br />
THE NASA ROTOR 37 is among the most popular rotors that have been manufactured<br />
to date. It was designed and tested in the early 1980s by Reid and<br />
Moore at NASA Glenn Research Center [1]. It consists of 36 blades with multiple<br />
arc profiles, and was designed for compressors and turbines with a compression<br />
ratio of 2.05 at a mass flow rate of 20.19 kg/s.<br />
Steady-state numerical simulations of the NASA Rotor 37 have been performed<br />
using FLUENT for one operating curve. Using a constant rotational speed (17,188<br />
rpm), the simulations covered a range that extends from the maximum flow rate<br />
to the pumping regime. A fully structured mesh of 420,000 cells was built using<br />
GTurbo, <strong>Fluent</strong>’s turbomachinery preprocessor. During the preliminary runs, the<br />
shock wave accuracy was found to strongly depend on the mesh resolution<br />
around the blade. Consequently, most of the cells were concentrated near and<br />
between the blades. A comparison of turbulence models was also conducted<br />
early on, and the most suitable one for this case was found to be the realizable<br />
k-ε model. Since the mean y + value was around 33, non-equilibrium wall<br />
functions were used. An axial flow direction was set at the inlet, and rotationally<br />
periodic lateral boundaries were used. An inlet temperature of 288.16K was<br />
assumed, and the exit pressure was altered for each of seven points on the<br />
operating curve that were simulated.<br />
Using data from published experimental results [2] a comparison of the radial<br />
variation of the pitch-averaged total pressure ratio was made, 10.19 cm from the<br />
blade leading edge. The results indicate that the total pressure is lower at the<br />
blade tip (100% span) resulting in losses due to the tip gap. A comparison of the<br />
total temperature ratio along the blade span, measured at the same location,<br />
shows an increase in the total temperature at the blade tip due to the friction<br />
between the fluid and the shroud, and to the detaching of the boundary layer.<br />
In both cases, the FLUENT results were in very good agreement with the data.<br />
Contours of Mach number at 70% of the blade height<br />
at 98.7% of the maximum experimental flow rate<br />
Predictions of the efficiency of the compression process along the operating<br />
curve were found to be good at low flow rates, but to deviate from the data at<br />
large flow rates. The total temperature ratio along the operating curve was found<br />
to be in good agreement throughout the range, even though the computed<br />
values were higher. A comparison of the compression ratio as a function of the<br />
flow rate was found to be within 1% throughout the range studied.<br />
Contours of Mach number on a plane through the blades illustrate some of the<br />
characteristics of flow in transonic compressors. The detached bow shock wave<br />
at the leading edge of each blade generates a normal shock wave in the flow<br />
channel that is the main compression mechanism for this type of compressor.<br />
There is a low momentum fluid region behind the shock wave, which is the<br />
second main source of total pressure loss and compression efficiency. <br />
References<br />
1 Reid, L. and Moore, D.: Performance of Single-Stage, Axial-Flow Transonic Compressor<br />
With Rotor and Stator Aspect Ratios of 1.19 and 1.26 Respectively, and with Design<br />
Pressure Ratio of 2.05. Tech Rep. TP-1338, NASA.<br />
2 NASA/TM-2003-212457 Report, http://gltrs.grc.nasa.gov<br />
Pathlines colored by relative Mach number at 98.7%<br />
of the maximum experimental flow rate<br />
20 <strong>Fluent</strong> News · Spring 2006
FOOD<br />
Image-based<br />
Easy as Pie<br />
Meshing:<br />
By Emma Johnson, Roderick Ross and Philippe Young, Simpleware Ltd., Exeter, UK; Gavin Tabor, University of Exeter, Exeter, UK<br />
IMAGED-BASED MESHING is opening up exciting new possibilities for the<br />
application of computational continuum mechanics (numerical methods such as CFD<br />
and finite element analysis (FEA)) to problems in biomechanics, biofluid dynamics and<br />
materials characterization. Software solutions that can rapidly generate robust, high<br />
quality meshes from complex 3D image data, as can be obtained from magnetic resonance<br />
imaging (MRI), computed tomography (CT) or ultrasound for example, are<br />
increasingly in demand. Such software improves productivity, leads to significantly<br />
more accurate results and enables engineers to focus on the analysis and generation<br />
of results rather than on the geometry definition and mesh creation.<br />
The University of Exeter and Simpleware Ltd. have collaborated to illustrate the<br />
potential of new meshing techniques by generating a CFD model of a hot mince pie,<br />
cooling in a light breeze. The transient study uses Simpleware software to create a<br />
mesh from medical scan data and FLUENT to analyze the coupled heat transfer and<br />
fluid flow of the system. The example illustrates a number of more general points<br />
pertaining to multi-physics simulations on complex image-based domains. First, the<br />
method allows watertight surface meshes and fully volumetric meshes of complex<br />
structures to be generated. Second, the accuracy of the geometric reconstruction is<br />
only dependent on the image quality, which is a function of the scan resolution, noise,<br />
and contrast between the volumes of interest. In fact, because structures are defined<br />
by iso-surfaces that are interpolated across boundaries, the geometric reconstruction<br />
can be of sub-voxel accuracy. (A voxel is a volume pixel, or smallest volume for which<br />
image data can be stored.) Third, the method allows multiple structures to be<br />
identified and meshed, with perfectly conforming/non-overlapping boundaries, and<br />
with the option of defining contact surfaces between them. Finally material properties<br />
within a given structure can be assigned based on signal strength. In the pie example,<br />
thermal properties could be varied throughout the mesh of the crust based on fat<br />
content, for example, and the different regions could then be modeled as separate<br />
solid zones.<br />
The segmentation of structures (crust, filling, raisins and air) in<br />
ScanIP software<br />
A 60g mince pie was scanned using MRI. Four segmentation masks, representing the<br />
crust, filling, raisins, and air, were generated using Simpleware’s ScanIP segmentation<br />
software. A volumetric mesh was then generated in the +ScanFE mesh generation<br />
module and exported directly to FLUENT. The time required to go from the import of<br />
the 3D data through to the creation of the multi-part input mesh for FLUENT was less<br />
than 10 minutes on a PC.<br />
In FLUENT the physical properties were assigned and the filler temperature was set to<br />
450K. A rectangular space was created around the pie with boundaries 30cm downstream<br />
and 4cm on either side of the pie. Upstream of the pie, an inlet condition was<br />
set with a velocity of 2m/s and temperature of 300K, akin to a short term cooling of<br />
the pie in a breeze. These boundary conditions were used for a transient calculation<br />
during which the filler was allowed to cool naturally for a time of 300s (5 minutes).<br />
The results show features such as the temperature gradients inside the pie and the<br />
recirculation zone downstream of the pie and inlet. <br />
Temperature contours and surfaces inside the pie, and<br />
velocity vectors showing the cooling breeze at the start of<br />
the calculation (top) and after several minutes (bottom)<br />
More.info@<br />
www.simpleware.com<br />
<strong>Fluent</strong> News · Spring 2006 21
FOOD<br />
Looking Inside<br />
Dough Mixers<br />
By Robin Connelly, Departments of Food Science and Biological Systems Engineering, University of<br />
Wisconsin, Madison, Wisconsin, USA, and Jozef Kokini, Department of Food Science, Center for Advanced<br />
Food Technology, Rutgers University, New Brunswick, New Jersey, USA<br />
View of the C.W. Brabender, Inc. Farinograph ® twin<br />
sigma blade mixer<br />
Mixing Index for Newtonian corn syrup for two<br />
mixer positions, where a value of 0 (blue) indicates<br />
pure rotation, 0.5 indicates simple shear, and 1 (red)<br />
indicates pure elongation<br />
22 <strong>Fluent</strong> News · Spring 2006<br />
WHEAT FLOUR DOUGH is a rheologically<br />
complex viscoelastic material, whose unique timedependent<br />
properties are governed by the rate,<br />
amount, and type of deformation applied. The<br />
structure and morphology of dough also depend on<br />
the available moisture and the extent to which the<br />
dough is mixed. As a result, dough is a very dynamic<br />
and unstable material, and dough mixers have<br />
evolved into highly complex geometries that shear,<br />
stretch, and fold. In addition, they often have close<br />
wall clearances to ensure that there are no regions<br />
of ineffective mixing. Changing between mixer<br />
types, especially between batch and continuous<br />
mixers, is difficult because of the very different flow,<br />
shear, extension, and mixing profiles that characterize<br />
each one. In industry, determining mixing times<br />
and designing mixer configurations is largely done<br />
on a trial and error basis.<br />
A more predictive approach is to calculate measures<br />
of mixing effectiveness through the use of mathematical<br />
modeling and numerical simulation. An<br />
ongoing project at the Center for Advanced Food<br />
Technology at Rutgers, The State University of New<br />
Jersey, supported by the National Research Initiative<br />
of the USDA Cooperative State Research, Education<br />
and Extension Service*, is using POLYFLOW to study<br />
the flow and mixing in typical dough mixer geometries<br />
with fluid models that range from simple<br />
Newtonian to non-linear viscoelastic [1,2]. The work<br />
featured here involves the numerical simulation of<br />
the flow and mixing in a fully filled Farinograph ®<br />
mixer using the mesh superposition technique and<br />
particle tracking with a Newtonian fluid model<br />
based on a high viscosity corn syrup [3]. Future<br />
work with this geometry will focus on extending the<br />
simulations to non-Newtonian fluid models. The<br />
Farinograph is a low shear rate batch mixer with<br />
two non-intermeshing, asymmetrical sigma blades,<br />
where the fast (right) blade turns at 93 rpm counterclockwise<br />
and the slow (left) blade turns at 62<br />
rpm clockwise. CAD STEP representations of the<br />
blade geometries were provided by C.W. Brabender<br />
Instruments, Inc., South Hackensack, New Jersey,<br />
producer of the Farinograph Mixer.<br />
Because the blades turn at different speeds, two<br />
revolutions of the slow blade and three revolutions<br />
of the fast blade are required before there is repetition<br />
of the relative blade positions. The left (slow)<br />
blade mesh of 6232 tetrahedral elements and the<br />
right (fast) blade mesh of 6166 elements are superimposed<br />
on the bowl (41860 hexahedral elements)<br />
every 0.027 seconds, giving a total of 72 positions<br />
per blade cycle with 10° between positions for the<br />
slow blade and 15° between positions for the fast<br />
blade. The time marching flow simulation results are<br />
then used to generate particle tracking data for<br />
10,000 massless material points initially randomly<br />
distributed throughout the flow domain, or a set of<br />
1000 massless material points initially randomly<br />
distributed in a 1 cm 3 box. As these abstract points<br />
are tracked throughout the flow domain, the associated<br />
local flow characteristics are recorded, thus<br />
providing a spatial and temporal history of phenomena<br />
such as stretching and deformation. Random<br />
distribution is a requirement of the statistical mixing<br />
measures that are used to evaluate the particle<br />
tracking results.<br />
The simulation results show that the differential in<br />
the speed of the two blades in the Farinograph<br />
causes an exchange of material between the blades<br />
to occur. The primary circulation pattern consists of<br />
material moving from the slow blade up toward the<br />
top of the mixer and over toward the fast blade,<br />
while the fast blade pushes material towards the<br />
slow blade near the bottom of the mixer. A slower<br />
mixing pattern is also observed where material<br />
around the blades moves from the center towards<br />
the walls and then up towards the top and back<br />
down in the center of the mixer. The zone in the
FOOD<br />
center of the mixer between the two blades is<br />
shown to have excellent distributive and dispersive<br />
mixing ability with high shear rates and mixing<br />
index values [4]. The mixing index is a measure of<br />
the type of flow, with values that range from 0 for<br />
pure rotational flow to 0.5 for shear flow to 1 for<br />
pure elongational flow. A high value of the mixing<br />
index combined with a high shear stress and shear<br />
rate indicate an area in the mixer with potentially<br />
good dispersive mixing capability. That region also<br />
has fast distribution throughout both sides of the<br />
lower section of the mixer as shown by material<br />
point clusters that travel through it. In contrast, very<br />
slow mixing is seen in the area away from the region<br />
swept by the blades that is generally not filled<br />
during normal use of this mixer.<br />
The mean of the normalized length of stretch [5]<br />
calculated for material points in the Newtonian fluid<br />
case increased exponentially over time, indicating<br />
effective mixing for the majority of material points.<br />
In the area swept by the blades, the material points<br />
with the highest values of the length of stretch are<br />
generally located near the blade edges or in the area<br />
swept by the blade edges. However, material points<br />
with high length of stretch values are also found<br />
outside these zones in a more random distribution.<br />
The instantaneous efficiency, which can be thought<br />
of as the fraction of energy dissipated locally that is<br />
used to stretch a fluid element at a given instant in<br />
a purely viscous fluid [5], gives a picture of the most<br />
and least effective blade positions for applying energy<br />
to stretch rather than displace material points.<br />
The least effective blade positions are when the flattened<br />
central sections of both blades are horizontal,<br />
while the most effective mixing occurs when the<br />
flattened section of the fast blade is vertical. The<br />
mean time-averaged-efficiency [5] stays above<br />
zero while its standard deviation reduces over time,<br />
indicating that the majority of the points are<br />
continuously experiencing stretching at equivalent<br />
levels over time.<br />
The overall result of this project has been to<br />
demonstrate the effectiveness of numerical simulation<br />
as a means to non-intrusively study the<br />
flow and mixing in a given mixer of materials<br />
with different rheological properties. The results<br />
also have established a basis of comparison<br />
between mixers with very different geometries.<br />
The simulations are providing guidance for the<br />
design of experiments that will be used to<br />
validate the findings of the simulations. Once the<br />
simulations are validated, they will be able to<br />
provide a much higher level of detail than the<br />
experimental results. The insight gained by this<br />
research has already stimulated the development<br />
of ideas by those involved on how the flow and<br />
mixing in a mixer affects the development of<br />
material structure, leading to profitable new lines<br />
of research. <br />
References:<br />
1 Connelly, R.K. and Kokini, J.L.: Analysis of Mixing in a<br />
Model Mixer using 2-D Numerical Simulation of<br />
Differential Viscoelastic Fluids with Particle Tracking.<br />
J. Non-Newt. Fluid Mech., 123:1-17, 2004.<br />
2 Connelly, R.K. and Kokini, J.L.: 2-D Numerical<br />
Simulation of Differential Viscoelastic Fluids in a Single-<br />
Screw Continuous Mixer: Application of Viscoelastic<br />
FEM Methods. Adv. Poly. Tech., 22(1):22-41, 2003.<br />
3 Connelly, R.K. and Kokini, J.L.: Mixing Simulation of a<br />
Viscous Newtonian Liquid in a Twin Sigma Blade Mixer.<br />
AIChE J., In Review, 2006.<br />
4 Yang, H.-H., Wong, T.H. and Manas-Zloczower, I.:<br />
Flow Field Analysis of a Banbury Mixer. In: Mixing and<br />
Compounding of Polymers: Theory and Practice, Ch. 7,<br />
pp. 187-223, Manas-Zloczower, I. and Tadmor, Z.<br />
Editors, Carl Hanser Verlag, New York, 1994.<br />
5 Ottino, J.M.: The Kinematics of Mixing: Stretching,<br />
Chaos and Transport. Cambridge University Press,<br />
1989.<br />
6 Connelly, R.K. and Kokini, J.L.: 3D Numerical<br />
Simulation of the Flow of Viscous Newtonian and Shear<br />
Thinning Fluids in a Twin Sigma Blade Mixer. Adv. Poly.<br />
Tech., In Review, 2006.<br />
* Grant numbers 2001-35503-10127 & 2003-35503-<br />
13907.<br />
Newtonian corn syrup fluid velocity vectors for blade<br />
positions of 180° and 270° on the z=0 (top), x=0<br />
(middle), and y=4.225 cm (bottom) planes; the vectors<br />
are all the same length and colored by the velocity<br />
magnitude<br />
Initial (left) and final (right) positions after 3 blade cycles (6 revolutions of the slow blade and 9 revolutions of the<br />
fast blade) of 1,000 material points in the Newtonian corn syrup, colored according to the length of stretch [6]<br />
For the Newtonian corn syrup, 3D positions of<br />
10,000 initially randomly distributed material points<br />
(top) and after three cycles (bottom) with<br />
concentrations of 1 (red) and 0 (blue) [6]<br />
<strong>Fluent</strong> News · Spring 2006 23
HEALTHCARE<br />
CFD Assists Neonatal<br />
By Maciej K. Ginalski and Andrzej J. Nowak, Institute of Thermal Technology, Silesian University of Technology, Gliwice, Poland<br />
Pathlines, colored by temperature, illustrated the<br />
circulation inside the left side of the incubator<br />
Postprocessing courtesy of Michal / Nowak<br />
PREMATURE INFANTS generally enter the world with little protection from<br />
the harsh environment. Thermal comfort thus plays a crucial role in their survival<br />
and health. To provide the optimal environmental conditions for these infants,<br />
incubators are widely used. However, due to the complexity of the<br />
physical processes occurring within modern neonatal units during treatment<br />
procedures, comprehensive analysis involving the thermal comfort of premature<br />
infants can be difficult. The application of analytical calculations is currently not<br />
practical, and therefore numerical modeling is required.<br />
At the Silesian University of Technology in Gliwice, Poland, a CFD model of<br />
conjugate fluid flow and heat transfer in an infant incubator has been developed<br />
to support infant healthcare and improve medical equipment design. Accurate<br />
geometrical models representing the human body in a variety of postures are<br />
now easily created with high accuracy. Fluid flow analysis can be performed by<br />
taking into account mechanisms of heat transfer, such as radiation, convection,<br />
conduction, and even the evaporation of sweat and moisture from human skin.<br />
Since all infants, and particularly premature infants, are different from each other,<br />
the complex shape of the human body was generated using the CATIA package<br />
and described by a number of scaleable parameters. By changing the value of<br />
those parameters, a geometrical model can easily be adjusted to the individual<br />
24 <strong>Fluent</strong> News · Spring 2006
HEALTHCARE<br />
Intensive Care<br />
features of each analyzed patient or to his nursing position. The researchers used<br />
TGrid 4.0 to wrap the surface model of the infant’s body and to create the surface<br />
mesh to be exported to GAMBIT. In parallel, they created the geometry of<br />
the incubator using the CATIA package and combined the baby and incubator<br />
geometries using GAMBIT.<br />
Once a fine tetrahedral volume mesh was created, numerical calculations were performed<br />
in FLUENT. The research team enhanced the FLUENT model using several<br />
user-defined functions (UDFs) to control the processes of heat transfer together<br />
with internal heat generation inside the infant’s body, evaporation of moisture from<br />
the infant’s skin, and respiration. This last component of heat balance can be modeled<br />
as a steady-state or transient process. To make the calculation process fully<br />
automatic, managing software called MARCEL was developed to assist the team at<br />
all essential stages of the numerical simulation. MARCEL reads the provided data,<br />
manages the process of geometry and mesh creation, and sets the appropriate<br />
boundary and initial conditions. After receiving the results from the CFD solver,<br />
MARCEL then provides a basic report about the heat balance of the infant being<br />
analyzed. It is anticipated that a modified version of the program could be used by<br />
medical staff to perform heat balance calculations without possessing any knowledge<br />
of how to build and run CFD models.<br />
Contours of velocity magnitude on two planes through the<br />
incubator; red regions correspond to the flow inlets at the<br />
ends and outlets at the sides<br />
Validation of the code has been performed with comparison of the results to a<br />
series of clinical tests described in the medical literature. In all cases numerical<br />
models were created to represent analyzed infants and environmental conditions<br />
measured in the described tests. Results from the numerical simulations were<br />
compared with those obtained from the literature and proved to be very accurate.<br />
Researchers also performed a series of air temperature and velocity measurements<br />
inside an empty prototype of a new design. These measurements were<br />
fairly consistent with the results obtained from the numerical simulations. Based<br />
on this conclusion, modification of the ventilation system has been proposed and<br />
verified numerically.<br />
The incubator modeling project is still in the phase of testing and validation.<br />
However, even at the present stage, it is providing crucial information and is<br />
becoming an important tool in the treatment of premature babies at neonatal<br />
intensive care units. <br />
Suggested Reading:<br />
Ginalski, M., Nowak, A.J. and Wrobel, L.C.: Computational Model of Selected Transport<br />
Processes of the Premature Newborn Baby within an Infant Incubator. XXI International<br />
Congress of Theoretical and Applied Mechanics, Conference proceedings, Warsaw,<br />
Poland, August 2004.<br />
Ginalski, M., Nowak, A.J. and Wrobel, L.C.: Combined Heat and Fluid Flow in a Double<br />
Wall Infant Incubator. International Conference of Computational Methods in Sciences<br />
and Engineering, Conference proceedings, Loutraki, Greece, October 2005.<br />
Ginalski, M., Nowak, A.J. and Brandt, J.: Numerical Optimization of the Infant<br />
Incubator Ventilating System. Conference Proceedings, XIX National Congress of<br />
Thermodynamics, Sopot, Poland, September 2005.<br />
Acknowledgments<br />
The authors wish to express their gratitude to Nicolae Mera from the Center for<br />
Computational Fluid Dynamics at Leeds University, UK; Professor L.C. Wrobel from Brunel<br />
University, UK; Keith Hanna from <strong>Fluent</strong> Europe; and Michal / Nowak for their contribution<br />
and help in the project.<br />
Temperature contours show a uniform field in the region<br />
surrounding the infant<br />
Velocity vectors without (left) and with (right) an overhead plastic screen;<br />
the screen was found to help reduce radiative heat loss from the child<br />
<strong>Fluent</strong> News · Spring 2006 25
ENVIRONMENTAL<br />
Activated Sludge Basins<br />
Get on Track<br />
By Gregory Cartland Glover, Karim Essemiani and Jens Meinhold, Veolia Environment R&D Center, Maisons Laffitte, France,<br />
and Stephanie Vermande, Technical Division of Veolia Water, St. Maurice, France<br />
Hydrodynamic phenomena in<br />
the ps-ASB, showing gas fraction<br />
iso-surfaces (top) and pathlines<br />
of liquid velocity (bottom)<br />
concentration (mg/l)<br />
25<br />
20<br />
15<br />
10<br />
5<br />
0<br />
0<br />
Feed Aerobic Reaction<br />
experiment<br />
FLUENT ASM1 Anoxic Reaction<br />
WEST ASM1<br />
Ammonia<br />
Oxygen<br />
Nitrate<br />
30 60 90 120 150 180 210 240 270<br />
time (min)<br />
Profile plots of the reaction phenomena in the ps-ASB<br />
ACTIVATED SLUDGE BASINS (ASBs)<br />
are a key step in the treatment of municipal<br />
wastewater. ASBs are used to degrade<br />
biochemically reactive pollutants from<br />
water that is discharged into the natural<br />
environment. The bacteria found in the<br />
sludge consume and assimilate nutrients<br />
such as carbon, nitrogen, and phosphorous<br />
under several environmental conditions.<br />
The design of ASBs is difficult,<br />
however, due to factors including:<br />
• the treatment aims (pollutants<br />
to be degraded)<br />
• the influent wastewater composition<br />
(time and location<br />
dependent)<br />
• the application of liquid agitation<br />
and aeration (brushes,<br />
disks, membrane aerators,<br />
impellers, turbines and venturi<br />
tube mixers)<br />
• the location and spatial<br />
requirements (local weather,<br />
above or below ground,<br />
sparsely or densely populated<br />
regions)<br />
• the daily treatment loading<br />
(from a few hundred liters to<br />
thousands of cubic meters)<br />
Each one of these factors influences the<br />
size, form, and mode of operation (continuous<br />
or sequenced batch treatment) that<br />
is used when designing an ASB. Variation<br />
in some of the factors, such as daily<br />
changes in the temperature and in the<br />
influent composition and flow rate, can<br />
impact the efficacy of the treatment<br />
process. Thus, for the design of any ASB,<br />
an appropriate selection of agitators and<br />
aerators is essential to provide optimal<br />
environmental conditions for the treatment<br />
of wastewater.<br />
A study that was co-funded by Anjou<br />
Recherche, the R&D arm of Veolia Water,<br />
and the European Union was recently performed<br />
to develop a computational strategy<br />
for modeling ASBs. The primary focus<br />
of the study was to examine the effect of<br />
the local hydrodynamics on the biochemical<br />
reactions observed in ASBs for both<br />
pilot and real scale processes. The results<br />
provided an improved understanding of<br />
the hydrodynamic impact on the environmental<br />
conditions experienced by the<br />
bacteria.<br />
A pilot-scale ASB (ps-ASB) was used to<br />
establish a database of hydrodynamic data<br />
(circulation velocity and mean bubble<br />
sizes) and interphase mass transfer data<br />
under different operating conditions. The<br />
ps-ASB was then used to study the impact<br />
of the normal and abnormal operating<br />
conditions on the biochemical reaction<br />
phenomena by monitoring the change<br />
in concentration of key wastewater<br />
constituents, such as ammonia, organic<br />
carbon, nitrates, organic nitrogen, and<br />
particulate matter.<br />
The ps-ASB applied different environmental<br />
conditions in a programmed sequence.<br />
The different conditions were affected by<br />
contacting the mixed liquor, a mixture of<br />
wastewater and activated sludge, with air<br />
in the form of fine bubbles, and applying<br />
a circulation velocity with a marine-type<br />
impeller. This condition enabled the aerobic<br />
oxidation of both organic carbon and<br />
nitrogen (in the form of ammonia) to carbon<br />
dioxide and nitrate, respectively. A<br />
second condition only agitated the biomass<br />
in the mixed liquor to reduce the<br />
overall nitrogen content of the wastewater.<br />
Under these conditions a different<br />
type of biomass (anoxic bacteria) consumes<br />
the nitrate producing nitrogen,<br />
which then evolves as a gas.<br />
The operational data was used to create a<br />
26 <strong>Fluent</strong> News · Spring 2006
ENVIRONMENTAL<br />
hydrodynamic and biochemical reaction model of<br />
the ps-ASB. The first step was to model the biochemical<br />
reactions assuming perfectly mixed reactor<br />
conditions with the wastewater modeling tool<br />
WEST ® [1], which uses the Activated Sludge Model<br />
No. 1 (ASM1) protocol [2].<br />
The mixed liquor composition and the calibrated<br />
kinetic and stoichiometric parameters derived from<br />
this global model were then applied to a converged<br />
local hydrodynamic solution constructed in FLUENT.<br />
The Eulerian multiphase model was employed with<br />
the dispersed phase form of the k-ε turbulence<br />
model to resolve the gas-liquid motion under steady<br />
flow conditions.<br />
The application of the ASM1 protocol to FLUENT<br />
was validated by analysis of the component profiles<br />
during the aerobic and anoxic reaction phases. The<br />
rates under the different operating conditions were<br />
accurate to within 10% of the experimentally measured<br />
reaction rates, where the accepted error was<br />
10 to 15%.<br />
Liquid velocity pathlines and volume fraction contours<br />
of the ps-ASB were used to illustrate the<br />
hydrodynamics of the basin. The distribution of dissolved<br />
oxygen after one hour of aeration was also<br />
examined. The difference between the cold and<br />
warm iso-surfaces was found to be 0.4 mg/l over<br />
the range from 2.9 to 3.4 mg/l. At this point in the<br />
simulation, the oxygen concentration gradient does<br />
not influence which biochemical reaction processes<br />
are predominant. However, at a mean concentration<br />
of less than 0.5 mg/l different treatment<br />
regimes could be found in the reactor (informal<br />
aerobic and anoxic zones).<br />
The aeration and agitation regimes applied to the<br />
ps-ASB were considered to be perfectly mixed and at<br />
a comparatively small scale for treatment processes.<br />
Thus, the effects observed with the ps-ASB may be<br />
different from those of real scale Activated Sludge<br />
Basins (rs-ASBs) where volumes of up to 10,000m 3<br />
are widely used. At the larger scale, perfectly mixed<br />
reactor or plug flow regimes are not guaranteed,<br />
and the potential influence of the mixing regimes is<br />
far greater and more difficult to categorize. Thus,<br />
incorrect operating conditions will cause mixing<br />
regimes that would hinder the ability of the ASB to<br />
meet the treatment objectives.<br />
To test this possibility, two rs-ASBs with different<br />
geometries and hydrodynamic regimes were simulated.<br />
The basin rs1-ASB is a carrousel type ASB, whereas<br />
rs2-ASB is a race-track type ASB. The information<br />
derived from these simulations (circulation velocity<br />
and oxygen mass transfer characteristics) indicated<br />
how well each ASB could meet its objectives.<br />
The pathlines for rs1-ASB indicate that the conditions<br />
are near optimal, while pathlines in rs2-ASB<br />
showed that the process was not close to optimal<br />
conditions. The flow field in rs2-ASB was characterized<br />
by low horizontal liquid velocities and large<br />
recirculation zones and short-circuiting upstream of<br />
the bubble plumes. The liquid circulation patterns<br />
were disrupted by the bubble plumes, causing them<br />
to degrade. In general, poor liquid circulation<br />
reduces the oxygen mass transfer rates and the dispersion<br />
of oxygen throughout the ASB. The ability<br />
of the biomass in this particular basin to aerobically<br />
treat the wastewater could therefore be limited by<br />
the operating conditions applied.<br />
Further simulations were performed to improve the<br />
aeration strategy by analyzing the effect that different<br />
aerator and agitator configurations had on these<br />
two design parameters. The simulations examined<br />
the positions of the aerators and agitators as well as<br />
the agitation rate and size of the impellers, with the<br />
single goal of improving the mass transfer characteristics<br />
of rs2-ASB. The hydrodynamics of the modified<br />
design showed improved liquid circulation<br />
with less disruption of the flow by the bubble<br />
plumes, resulting in a 250% increase of the circulation<br />
velocity.<br />
The computational strategies developed during the<br />
course of this study have enabled Veolia Water<br />
to obtain a numerical tool that can aide decisions<br />
made in selecting aeration and agitation configurations<br />
for Activated Sludge Basins, for the design of new<br />
technologies and retrofitting of existing installations. <br />
References:<br />
1 Hemmis, N.V., WEST ® , Hemmis N.V., Kortrijk, Belgium,<br />
2003.<br />
2 Henze, M., Gujer, W., Mino, T., and van Loosdrecht, M.:<br />
Scientific and Technical Report No. 9. IWA Publishing,<br />
London, UK, 2000.<br />
rs1-ASB rs2-ASB Modified rs2-ASB<br />
Hydrodynamic phenomena in the rs-ASB, showing gas fraction iso-surfaces (top) and pathlines of liquid velocity (bottom)<br />
<strong>Fluent</strong> News · Spring 2006 27
ACADEMIC NEWS<br />
A Dry Passage<br />
to the Afterlife<br />
By Essam E. Khalil and Omar A.A. AbdelAziz,<br />
Cairo University, Egypt<br />
EGYPT WAS ONCE HOME to the Pharaohs,<br />
one of the oldest and most sophisticated civilizations<br />
in the ancient world. Many Egyptian artifacts,<br />
treasures, and buildings are a priceless part of our<br />
world heritage. The Valley of the Kings at Luxor is<br />
a unique and world famous site where thousands<br />
of years ago, many Egyptian monarchs had elaborate<br />
tombs built to ensure their safe passage into<br />
the afterlife. Many of these tombs were robbed<br />
after being discovered, but others have survived<br />
and their greatest remaining treasures are the<br />
beautiful, vivid, yet fragile wall paintings and decorations<br />
that lie within. Each year, tourists flock to<br />
Egypt to see these awe-inspiring tombs, but their<br />
visits cause a major problem for curators. The heat<br />
and humidity given off by visitors in the enclosed<br />
chambers and passageways cause damage to the<br />
plasterwork and paintings. Hence, the ventilation<br />
system and resultant air flow patterns in a tomb<br />
are critical to the preservation of the exhibits so<br />
that they may be kept open to tourists for many<br />
years to come.<br />
The Egyptian government and the Supreme<br />
Council of Antiquities approached ventilation<br />
experts at Cairo University for help in solving this<br />
problem for the archeaological tombs in the Valley<br />
of the Kings. The engineers were asked to devise a<br />
climate control system for these tombs; a pilot<br />
study was conducted on the tomb of Ramses VII.<br />
Using GAMBIT, a model of approximately one<br />
million cells was created that yielded good geometric<br />
representation of the tomb passageway<br />
with the sarcophagus in the main room and a<br />
large number of visitors. They then used custom<br />
models in FLUENT to simulate human breathing<br />
and heat generation, and considered a worst case<br />
scenario when 20 adult visitors are in the tomb at<br />
the same time. Parametric CFD simulations were<br />
performed to predict what the relative humidity, a<br />
key variable, would be like near the wall paintings<br />
for a given air extraction system design and tomb<br />
vent locations. Their modeling work was used to<br />
determine the optimized air flow pattern for the<br />
tomb. Based on their findings, the installation of<br />
raised flow exhausts that were unobtrusive to<br />
visitors was recommended. The tomb now has<br />
minimal adverse flow and humidity gradients<br />
within the chamber, thereby helping to preserve<br />
this priceless piece of history for future generations<br />
of visitors. <br />
Predicted local humidity levels on a plane near the<br />
main wall paintings of the tomb of Ramses VII with a<br />
representative adult male tourist figure; the<br />
sarcophogus is in the foreground<br />
Postprocessing courtesy of Maciej Ginalski<br />
Convective<br />
By Daniele Melideo, Davide Mazzini, Enrico<br />
A NUMBER OF BENEFITS have resulted from<br />
recent reductions in weight and improvements to<br />
the efficiency of aeronautical gearboxes. Reduced<br />
heat generation inside the gearbox means that less<br />
oil is needed to maintain a certain transmission<br />
operating temperature. In addition, reductions in<br />
surge system losses allow the lubrication system<br />
components, such as tanks, high pressure tubes,<br />
filters, and pumps to be smaller in size.<br />
28 <strong>Fluent</strong> News · Spring 2006<br />
Predicted pathlines colored by moisture<br />
in the air in the tomb of Ramses VII;<br />
the flow is from left to right<br />
Postprocessing courtesy of Maciej Ginalski<br />
In the framework of thesis work carried out in the<br />
Department of Mechanics, Nuclear and Production<br />
Engineering (DIMNP) of the University of Pisa, the<br />
convective motion of the air inside a gearbox<br />
induced by the rotation of the gear drive components<br />
has been investigated. FLUENT was used for<br />
the numerical modeling component of the project,<br />
which focused on a single gear pair. The reference<br />
equipment is located at the Research Center on
ACADEMIC NEWS<br />
Student Submariners Peddle<br />
Their Way to Victory<br />
By Keith Hanna, <strong>Fluent</strong> News and Joost Sterenborg, Technical University Delft, Netherlands<br />
MANY UNIQUE CFD applications<br />
come across the desk of <strong>Fluent</strong><br />
News, and one that recently appeared<br />
was of a human powered submarine.<br />
Students from university engineering<br />
teams around the world have come<br />
together for an international humanpowered<br />
submarine competition for<br />
the last eight years. The teams of student<br />
engineers have to design their<br />
own submarines from the ground up<br />
using modern computer-based tools,<br />
and manufacture and test their<br />
designs for performance and safety.<br />
They also need to secure enough<br />
sponsorship funding to cover the cost<br />
of their team, including travel to the<br />
competition venue.<br />
WASUB, an intrepid team from the<br />
Technical University of Delft in the<br />
Netherlands approached <strong>Fluent</strong><br />
Benelux, along with other companies<br />
in the region, for sponsorship last<br />
year. They used Pro/ENGINEER ®<br />
from PTC to design their submarine<br />
in CAD, and FLUENT to hone the<br />
hull and fin shapes for optimal hydrodynamic<br />
drag characteristics. After<br />
extensive testing of the WASUB<br />
in the MARIN towing tank in<br />
Wageningen, the craft was shipped<br />
to America for the races. The team<br />
came away with several prestigious<br />
prizes, including best design and<br />
fastest in two categories. Throughout<br />
the process, the young engineers<br />
gained modern product design engineering<br />
experience in a competitive<br />
situation. <strong>Fluent</strong> salutes these brave<br />
Dutch submariners and their revolutionary<br />
WASUB design! <br />
More.info@<br />
http://wasub.oli.tudelft.nl<br />
The WASUB Team in Maryland, June 2005<br />
CAD layout of WASUB for a single<br />
human-powered propulsion system<br />
Axisymmetric FLUENT prediction of static<br />
pressure around the WASUB hull shape<br />
Motions Inside a Gearbox<br />
Manfredi, and Maria Vittoria Salvetti, The University of Pisa, Italy<br />
Advanced Technology Mechanical Transmissions<br />
(CRTM) of DIMNP, as part of a collaboration with<br />
the Italian aerospace company Avio. The test section<br />
consists of two rotating shafts that support the test<br />
gears. Wheel lubrication is ensured by two spray<br />
bars injecting oil upstream and downstream of the<br />
mesh point, where the gear wheels come together.<br />
The first activity conducted was to test different<br />
turbulence models (Spalart-Allmaras, k-ε, k-ω, and<br />
Reynolds stress). Their performance was evaluated<br />
by studying 2D cases of Couette motion, consisting<br />
of two coaxial cylinders rotating at different velocities.<br />
The k-ω model was chosen as the most costeffective<br />
turbulence model, since it was able to<br />
capture the swirling velocity component better than<br />
the other RANS models, and did not require the<br />
additional computational resources of RSM. This<br />
model was used for subsequent numerical studies.<br />
The 3D simulation of the gearbox focused on the air<br />
motion induced by the wheel rotation, and included<br />
the effects of protruding hardware, such as bolt<br />
heads. Owing to a plane of symmetry, half of the<br />
gearbox was modeled. The oil presence was neglected<br />
in the first round of simulations. The wheel<br />
rotation was simulated using the multiple reference<br />
frames (MRF) model. The gear teeth, small on the<br />
scale of the wheels, were neglected. The bolt heads<br />
were found to entrain the air creating a venting<br />
effect. While the wheel rotation produced mostly<br />
circumferential flow, a significant axial velocity was<br />
observed between the spray bars and the mesh<br />
point, suggesting a possible deflection of the oil jet.<br />
In the next stage of the work, a more refined grid<br />
and the use of the sliding mesh model will be<br />
adopted, and the oil presence will be taken into<br />
account. <br />
Axial velocity on the symmetry<br />
plane, with vectors showing the<br />
axial flow near the mesh point<br />
<strong>Fluent</strong> News · Spring 2006 29
PRODUCT NEWS<br />
<strong>Fluent</strong> & Microsoft Team to Deliver<br />
64-bit FLUENT on Windows Clusters<br />
By Diana Collier, Barbara Hutchings, and Rongguang Jia, <strong>Fluent</strong> Inc.<br />
ing the way for cluster-based FLUENT simulations for Windows<br />
users. With this 64-bit support, Windows users will now be able to run<br />
much larger simulations than currently feasible and they will also see<br />
performance improvements due to the enhanced memory management<br />
and wider memory bandwidth available with 64-bit processors.<br />
This good news for Windows users is the result of close collaboration<br />
between <strong>Fluent</strong> and Microsoft. With excellent technical support from<br />
Microsoft, <strong>Fluent</strong>’s development team has optimized FLUENT 6.3 on<br />
CCS. The resulting cluster solution uses the Microsoft MPI (Message<br />
Passing Interface) software layer for data communication between<br />
processors on the cluster, and supports a variety of interconnect<br />
options including Gigabit Ethernet (GigE), Infiniband, and Myrinet.<br />
FLUENT 6.3 also takes advantage of the Microsoft job scheduler that<br />
ships with CCS, providing an off-the-shelf solution for launching and<br />
controlling jobs on the cluster.<br />
Water ski jumper simulation showcased on the 64-bit Windows cluster at SC05<br />
Courtesy of Sports Engineering Group, Sheffield Hallam University; postprocessed using EnSight from CEI<br />
speedup<br />
9<br />
8<br />
7<br />
6<br />
5<br />
4<br />
3<br />
2<br />
1<br />
0<br />
0<br />
FL5L1<br />
FL5L2<br />
FL5L3<br />
Linear<br />
2 4 6 8 10<br />
number of processors<br />
Parallel performance of 64-bit FLUENT running the<br />
FL5L1, FL5L2, and FL5L3 large benchmark cases,<br />
using Windows CCS 2003 on 3.4 GHz Intel EM64T<br />
dual-CPU nodes connected with gigE<br />
30%<br />
28.5%<br />
25%<br />
22.0%<br />
20%<br />
16.7%<br />
14.3%<br />
15%<br />
11.8%<br />
12.5%<br />
15.3%<br />
15.7%<br />
10%<br />
5%<br />
0%<br />
15s1 15s2 fl5s3 fl5m1 15m2 15m3 fl5l1 fl5l2<br />
performance boost<br />
Serial performance boost with the 64-bit solution,<br />
relative to 32-bit, for the small (FL5S1 and FL5S2),<br />
medium (FL5M1, FL5M2, and FL5M3), and large<br />
(FL5L1 and FL5L2) benchmark cases<br />
THE AVAILABILITY OF 64-BIT<br />
off-the-shelf computing from<br />
Intel and AMD has provided<br />
great value to <strong>Fluent</strong> customers<br />
seeking to run larger, more<br />
memory intensive CFD simulations.<br />
With the upcoming<br />
release of FLUENT 6.3, 64-bit<br />
will become an option to customers<br />
running the Microsoft ®<br />
Windows ® operating system.<br />
Many FLUENT customers are<br />
already running a 32-bit version<br />
of FLUENT on the 64-bit<br />
Windows XP operating system,<br />
but in this configuration FLUENT<br />
does not take advantage of<br />
extended memory addressing.<br />
With FLUENT 6.3, full 64-bit<br />
capability will be supported on<br />
both desktop and server systems<br />
running Windows. On the server<br />
side, FLUENT 6.3 has been ported<br />
and optimized to run under<br />
the new Microsoft ® Windows ®<br />
Compute Cluster Server 2003<br />
(CCS) operating system, open-<br />
Benchmarking<br />
The performance of 64-bit FLUENT under CCS is excellent. Speed<br />
improvements for serial performance, relative to the 32-bit FLUENT<br />
version, are in the range of 10-30% for many of the examples in<br />
<strong>Fluent</strong>’s standard benchmark problem set. Parallel scaling is also quite<br />
good. Initial scaling studies using the larger benchmarks on a Windows<br />
CCS cluster are on par with what is observed on similarly configured<br />
clusters running Linux. For example, on an 8-CPU cluster connected<br />
with GigE, a speed-up of 6.4 was obtained using the FL5L2 benchmark.<br />
Supercomputing 2005<br />
Microsoft showcased their new high-performance computing platform<br />
at Supercomputing 2005 (SC05), held in Seattle November 14-18,<br />
2005. FLUENT was proud to be part of the Microsoft exhibit, demonstrating<br />
FLUENT 6.3 running as a 64-bit application under Windows<br />
Compute Cluster Server 2003 on a 16 CPU cluster provided by<br />
IBM and using a Myrinet interconnect. The showcased simulation was<br />
a multi-million cell model of a water ski jumper, courtesy of Sports<br />
Engineering Group, Sheffield Hallam University. In fact, several <strong>Fluent</strong><br />
partners at SC05 demonstrated FLUENT simulations running on CCS<br />
clusters, including Dell, Mellanox, Voltaire, and Broadcom. The widespread<br />
success of these demonstrations, using pre-release versions of<br />
both the operating system and FLUENT, was a confirmation of<br />
Microsoft’s stated goal to provide HPC solutions that are easy to<br />
deploy, operate, and integrate with existing infrastructure and tools. <br />
For information on obtaining FLUENT 6.3 beta for Windows Compute Cluster<br />
Server 2003, contact your local <strong>Fluent</strong> account manager.<br />
More.info@<br />
www.microsoft.com/windowsserver2003/<br />
ccs/overview.mspx<br />
30 <strong>Fluent</strong> News · Spring 2006
PRODUCT NEWS<br />
Quick Turnaround with<br />
Rapid Flow Modeling<br />
By André Bakker, FloWizard Product Manager, and Laurent Collonge, FLUENT for CATIA V5 Product Manager<br />
THE OBJECTIVE OF RAPID FLOW MODELING is<br />
to bring CFD to the broader engineering design<br />
market, by providing fluid flow analysis tools that let<br />
engineers quickly turn their CAD models into CFD<br />
results. There are now two full-fledged rapid flow<br />
modeling products available in <strong>Fluent</strong>’s product<br />
line: FloWizard and FLUENT for CATIA V5. With<br />
these products, <strong>Fluent</strong> is taking the lead in what is<br />
expected to be a fast growing segment of the CFD<br />
software market.<br />
Many recent technology advances have made rapid<br />
flow modeling possible. The increased speed of<br />
computer hardware has reduced the calculation<br />
time for many CFD applications to fit within engineering<br />
design timeframes. A simulation that previously<br />
might have taken tens of hours on a workstation<br />
can now be run in well under an hour on a<br />
standard PC. On the software side, it is now possible<br />
to embed the CFD experience in highly automated<br />
products that significantly reduce the turnaround<br />
time for a fluid flow analysis.<br />
Modern companies have an increasing need for engineering<br />
analysis to be tightly integrated into their<br />
design and PLM (product lifecycle management)<br />
processes. This requires software that interfaces well<br />
with CAD and other engineering software tools. It<br />
also means that the fluids analysis software should be<br />
accessible to engineering designers, who may have<br />
only intermittent needs to run simulations. FloWizard<br />
and FLUENT for CATIA V5 share characteristics that<br />
allow them to meet these needs. They each have a<br />
high level of connectivity with CAD or PLM products.<br />
This is important, since most of their users are engineers<br />
involved in product design. FloWizard accepts<br />
a wide variety of CAD and mesh file formats, and<br />
has tight CAD connections with SolidWorks ® ,<br />
Pro/ENGINEER ® and UGS’ NX. FLUENT for CATIA V5<br />
embeds the same rapid flow modeling technology<br />
deep into the CATIA V5 PLM system.<br />
Both packages focus on modeling well understood<br />
physics, including compressible and incompressible<br />
fluid flow and heat transfer, both laminar and turbulent.<br />
Stationary equipment can be modeled as well<br />
as rotating machinery. Tasks that often require a lot<br />
of user interaction, such as geometry cleanup,<br />
meshing, and solving are fully automated. Not only<br />
does this save time, it also leads to a significantly<br />
reduced learning curve.<br />
Both products make use of the fast, accurate, and<br />
well-validated FLUENT 6 solver in the background.<br />
As an additional benefit, the files generated by<br />
FloWizard and FLUENT for CATIA V5 can be shared<br />
with FLUENT users as needed. This also helps<br />
improve the level of collaboration between the<br />
analysis and design teams. Furthermore, FloWizard<br />
Starting with a pump model in SolidWorks, engineers can quickly send their<br />
model to FloWizard. There it will be meshed and solved using FloWizard’s<br />
automated tools. The final results can be used by the engineer to decide<br />
upon possible design changes, thus closing the full design cycle.<br />
offers built-in collaboration tools that let multiple<br />
users simultaneously connect to the same session<br />
for design reviews. When there is a need to model<br />
large problems, both products offer the ability<br />
to run calculations in parallel on a local or remote<br />
network.<br />
In many companies, there is also a need to efficiently<br />
model recurring problems related to their product<br />
lines. For this purpose, custom tools can be built<br />
upon the rapid flow modeling platform. FloWizard is<br />
fully customizable using the Python programming<br />
language which has been used by several companies<br />
to build custom, organization specific analysis<br />
tools. <br />
<strong>Fluent</strong> News · Spring 2006 31
PRODUCT NEWS<br />
High quality Cooper and tet meshes in a redesigned adaptorconnector<br />
originating from a Pro/E part file<br />
Impressive Prepro<br />
By Erling Eklund, GAMBIT and TGrid Product Manager<br />
THE SPRING RELEASES of several preprocessing products from <strong>Fluent</strong><br />
promise many exciting new features and capabilities, ranging from CAD<br />
integration to wrapping technology.<br />
GAMBIT<br />
In GAMBIT 2.3 the focus continues to be on CAD import, geometry manipulation,<br />
and advanced meshing. A new CATIA V5 translator has been added, as well<br />
as new CAD Connections for SolidWorks ® , Pro/ENGINEER ® , and UGS’ NX.<br />
These Connections use native surface, connectivity, and topology information,<br />
and eliminate the need to use a third party generic CAD format to translate your<br />
geometry model. With the CAD Connections, tools developed by <strong>Fluent</strong> reside<br />
in the CAD packages to perform a number of integrity checks prior to export,<br />
which guarantees that clean CAD integration is achieved.<br />
All edge and face construction tools in GAMBIT 2.3 can be utilized on these<br />
CAD geometries including the new closest point projection. In addition,<br />
design changes and general mesh-to-CAD conversion can be made using a<br />
new Advanced Covering algorithm. The cleanup tools continue to expand<br />
with a new sliver face removal capability.<br />
A surface mesh of occupants<br />
in a truck cabin, created<br />
using the new surface<br />
wrapper in TGrid<br />
In the area of meshing, the size function capability has been improved<br />
through speed-ups and added functionality, such as individual minimum<br />
size assignments, 2D proximity, size function blending, and a new, exciting<br />
meshed size function, starting on a boundary layer cap.<br />
For boundary layers, a new first height, last-ratio-based cell growth function<br />
has been added, and the full 3D boundary layer can be examined prior to<br />
volume meshing. A new quad split option in tetrahedral meshing allows for<br />
combined hex and tet meshing, without the need for pyramids, which in<br />
many cases leads to significant cell quality improvements.<br />
TGrid<br />
The most important new feature in TGrid 4.0 is the surface wrapper, which<br />
allows connectivity, defeaturing, and surface meshing to be handled in a<br />
single operation. Once the large holes have been covered, the tool wraps<br />
the geometry with a high-quality surface mesh from which a volume mesh<br />
can then be created. Gaps, slivers, and overlaps in the geometry are automatically<br />
cleaned-up in the process. The advanced wrapping procedure<br />
provides a full spectrum of size controls, including local and global<br />
32 <strong>Fluent</strong> News · Spring 2006
PARTNERSHIPS<br />
Discrete Element<br />
Modeling of<br />
Particles for FLUENT<br />
By Peter Weitzman and John Favier, DEM Solutions, Edinburgh, UK<br />
The 3Matic-for-<strong>Fluent</strong> interface, showing all three levels of operations:<br />
Manual (buttons on top), Autofix (right panel), and Batch Fix (left panel)<br />
cessing<br />
proximity, curvature, and zone-based size functions. It also supports<br />
controlled size growth, and unique coarsening and feature capturing capabilities.<br />
Other functionalities include the ability to fully examine the region<br />
to be wrapped, prior to the wrapping procedure, with a guaranteed hole<br />
finder. There are local and global feature extraction capabilities, automatic<br />
fixing of incorrect topology, and automatic mesh quality improvement.<br />
Using a strategy based on an “over-refined” initial mesh, imprinting and<br />
coarsening together produce the highest-quality surface mesh on the<br />
market today that is based on wrapper technology.<br />
In addition to the introduction of the surface wrapper, the prism layer<br />
technology in TGrid 4.0 has been expanded. The initial normal growth direction<br />
has been improved, leading to overall speed and quality improvements.<br />
Two automated functions have been added: one to detect prism layer proximity/collision,<br />
which includes automatic prism height adjustment, and another<br />
to detect and peal off prism layers in sharp corners, which includes automatic<br />
non-conformal re-meshing of prism sides. The ability to handle prism<br />
layer growth on double-sided, zero-thickness walls has also been added.<br />
DEM SOLUTIONS is the developer of EDEM, discrete element method<br />
(DEM) software for particle flow simulation. EDEM is used to simulate the dynamics<br />
of solid particles including inter-particle and particle-wall collisions.<br />
<strong>Fluent</strong> and DEM Solutions have jointly developed a new plug-in for FLUENT that<br />
can be used to simulate the dynamics of multiphase fluid-particle systems. The<br />
EDEM plug-in complements the multiphase models already available in FLUENT<br />
by adding the ability to include particle attributes such as shape and size distribution,<br />
material properties, and initial placement, and to obtain detailed information<br />
on collisions, clumping, adhesion to boundary surfaces, and exchange<br />
with the surrounding fluid. In addition to the fluid drag, other forces such as<br />
gravity, electrostatic, electromagnetic, bonding, and cohesion can be applied to<br />
the particles. An open interface allows additional user-defined forces and<br />
coupling with other physical models.<br />
The EDEM plug-in can provide important information on multiphase flows across<br />
a broad range of industries and applications. It is useful in applications where the<br />
particles must be transported efficiently, without damage or excessive buildup,<br />
or when particle-equipment interactions are critical. Some examples are the<br />
optimization of pneumatic transport systems, fluidization, filtration, prevention<br />
of sand damage, and reactor transport.<br />
The EDEM software can be used by itself to model single-phase systems of solid<br />
particles. Some examples of single phase applications include bulk material handling<br />
in conveyors, chutes and hoppers, rock drilling, excavation, agricultural<br />
machinery, and various pharmaceutical manufacturing processes. <br />
A large number of enhancements have also been added in the areas of<br />
surface mesh intersection, re-meshing, local manipulation, and more.<br />
3Matic-for-<strong>Fluent</strong><br />
The wrapping of highly complex industrial models often involves dealing<br />
with hundreds of files, all of different quality and origin. In collaboration<br />
with Materialise, a new product has been developed to meet this type of<br />
challenge. 3Matic-for-<strong>Fluent</strong> 1.0 is designed to handle all types of CAD<br />
formats (native, standard, and faceted) and produce a coherent, optimal,<br />
and clean faceted mesh that is ready to wrap in TGrid 4.0. This customized<br />
tool allows data to be handled in three ways. First, connectivity, cleanup,<br />
hole filling, and faceting of a single CAD file can be investigated, step-bystep,<br />
so that the impact of parameter settings can be better understood.<br />
Second, some of the files can be run through an auto-fix sequence to evaluate<br />
consistent behavior through different CAD formats. Third, hundreds<br />
of mixed CAD files can be run through the batch converter using the<br />
selected parameters, all producing TGrid mesh files. <br />
3Matic-for-<strong>Fluent</strong> is sold and distributed by <strong>Fluent</strong>. Please contact your local <strong>Fluent</strong><br />
office for more information.<br />
Entrainment of particles in an air-stream<br />
using FLUENT-EDEM co-simulation<br />
More.info@<br />
www.dem-solutions.com<br />
<strong>Fluent</strong> News · Spring 2006 33
SUPPORT CORNER<br />
Mapping Thermal Data<br />
from FLUENT to Structural<br />
Codes Quickly<br />
By Aleksandra Egelja-Maruszewski, <strong>Fluent</strong> Inc.<br />
“<strong>Fluent</strong>’s add-on UDF for temperature<br />
mapping is an excellent tool. It allows<br />
quick and accurate interpolation of<br />
temperatures between a CFD model<br />
and a structural mesh. It has cut the<br />
time and effort it takes us to map<br />
temperature data from FLUENT to<br />
FEA drastically. It is simple to use<br />
and flexible. If you need to map<br />
temperatures from a CFD model<br />
to a structural mesh, this is the tool<br />
to use.”<br />
– William Towne<br />
CAE Engineer<br />
Performance Prediction<br />
Mercury Marine<br />
IN MANY MODERN PROCESSES TODAY there is<br />
an increased need to solve multiphysics problems,<br />
such as thermal coupling. Some of these problems<br />
can be solved with one simulation software only;<br />
others require the combination of multiple software<br />
packages. In extreme thermal environments, there<br />
are various types of thermal loads on components<br />
and structures. Applying these thermal loads correctly<br />
is one of the first steps required for achieving<br />
a reliable thermal stress analysis of the component.<br />
Thermal loads can be applied as a temperature<br />
distribution within the solid, or as heat transfer<br />
coefficients on the skin of the solid where the heat<br />
transfer takes place.<br />
Most CFD tools have advanced fluid dynamics and<br />
heat transfer algorithms, but do not have built-in<br />
advanced solid mechanics analysis capabilities such<br />
as thermal stress analysis and fatigue prediction. On<br />
the other hand, most of the advanced finite element<br />
analysis (FEA) tools have advanced solid mechanics<br />
algorithms, but do not offer the necessary advanced<br />
fluid dynamics and heat transfer capabilities to<br />
determine the correct thermal loads for the thermal<br />
stress analysis. For scenarios involving fluid flow and<br />
thermal loading, the two simulation tools become<br />
dependent on each other.<br />
To make it easier to perform combined simulations<br />
such as these, engineers at <strong>Fluent</strong> have developed a<br />
user-defined function (UDF) for coupling CFD and<br />
FEA for thermal stress analysis. With this new capability,<br />
FLUENT is used to solve for the flow and thermal<br />
analysis in the fluid and solid zones of a part or<br />
component. The temperature data of the solid<br />
region is then supplied to the FE solver as the thermal<br />
load profile for the thermal stress analysis.<br />
Because the CFD mesh is typically much finer than<br />
the FE mesh, it is necessary to interpolate, or map,<br />
the CFD results to the FE mesh. The mapping is<br />
accomplished through a Scheme file (read into<br />
FLUENT) and the UDF.<br />
For the case of exporting data from FLUENT to<br />
ABAQUS, for example, the coupled solution would<br />
proceed as follows.<br />
1. The user solves the case in FLUENT.<br />
2. The user reads the Scheme file and<br />
compiled UDF into FLUENT.<br />
3. The user reads the ABAQUS ASCII format<br />
mesh into FLUENT. The FLUENT UDF<br />
uses import filters to read the nodes,<br />
elements, and connectivity from the<br />
ABAQUS mesh.<br />
34 <strong>Fluent</strong> News · Spring 2006
SUPPORT CORNER<br />
The panel used for input to the mesh interpolator<br />
4. For each FE node/element, the UDF finds<br />
the closest solution location in the CFD<br />
mesh. A binary space partitioning (BSP)<br />
tree is used to find the closest node<br />
by recursively subdividing the space<br />
containing the geometry, making fast<br />
geometric searching possible.<br />
5. The UDF interpolates from the closest<br />
node.<br />
6. The UDF instructs FLUENT to export the<br />
results in a format suitable for ABAQUS.<br />
There are two methods for performing this<br />
mapping procedure:<br />
• The temperature data inside the solid<br />
region is exported. For each node in the<br />
FE mesh of the solid region, temperature<br />
data will be written to an output file.<br />
• The temperature and heat transfer coefficient<br />
at the walls that are solid/fluid<br />
interfaces are exported. For each surface<br />
element of the FE mesh, the temperature<br />
and heat transfer coefficient will be<br />
written to an output file.<br />
The UDF is not designed to manage strongly coupled<br />
CFD and FEA calculations that require many<br />
iterations back and forth between the solvers.<br />
Instead, it is designed to make individual data transfers<br />
easier to do. The UDF, which has panels for<br />
input, is currently available for ABAQUS, ANSYS,<br />
and PATRAN formats, but it can easily be<br />
customized for other structural analysis codes. The<br />
mesh can be built in millimeters, centimeters, inches,<br />
or meters. Temperatures can be in Kelvin,<br />
Fahrenheit, or Celsius. The UDF performs in serial<br />
and parallel versions of FLUENT. Data for different<br />
solid zones or wall zones can be exported separately<br />
by reading the individual mesh files of the zones.<br />
With additional customization, other properties can<br />
be exported as well.<br />
The compiled libraries are freely available for<br />
exporting steady-state data. Simply download the<br />
CFD/FEA Thermal Coupling UDF at no charge from<br />
the UDF examples section of the UDF archive on the<br />
User Services Center at www.fluentusers.com. (You<br />
will need your username and password to log on<br />
to the site.)<br />
To customize the UDF or obtain libraries for the<br />
export of transient or time-averaged data, or if you<br />
have questions regarding the use of this UDF, please<br />
contact your local <strong>Fluent</strong> office or log a help<br />
request through the Online Technical Support<br />
Portal, accessible from the User Services Center:<br />
http://clarify.fluent.com/eSupport<br />
The geometry and mesh of the powerhead of a<br />
marine outboard is used to illustrate this capability.<br />
Using the UDF, a 2.8 million cell FLUENT mesh<br />
was mapped to a 430,000 element ABAQUS mesh<br />
in 14 seconds. Contour plots of static temperature<br />
computed by FLUENT (left) and ABAQUS (right) are<br />
shown for several views of the powerhead assembly<br />
Courtesy of Mercury Marine<br />
<strong>Fluent</strong> News · Spring 2006 35
AROUND FLUENT<br />
<strong>Fluent</strong> Opens Larger<br />
Office in Ann Arbor, MI<br />
FLUENT HAS OPENED a new, larger office in<br />
Ann Arbor, MI. This expanded office location allows<br />
for the addition of more customer service staff,<br />
an expanded training facility, and a much larger<br />
computer resource center. The new training facility<br />
can accommodate more people and allows for the<br />
addition of new educational sessions on advanced<br />
automotive application areas such as underhood<br />
thermal management, noise prediction, and HVAC<br />
system analysis.<br />
Upcoming User<br />
Group Meetings<br />
36 <strong>Fluent</strong> News · Spring 2006<br />
<strong>Fluent</strong> invites you to take full advantage of the<br />
facility. Meet with the US customer services team,<br />
get trained on new product innovations, and<br />
discuss how the Ann Arbor team can provide consulting<br />
services to meet your engineering needs.<br />
<strong>Fluent</strong><br />
3005 Boardwalk, Suite 100<br />
Ann Arbor, MI 48108<br />
734 213 6821<br />
US CFD Summit May 22 – 24<br />
US Automotive CFD Summit Sept. 11 – 12<br />
<strong>Fluent</strong> Sweden Sept. 18 – 19<br />
<strong>Fluent</strong> Europe Sept. 20 – 22<br />
<strong>Fluent</strong> Benelux Oct. 4 – 6<br />
<strong>Fluent</strong> China Nov. 6 – 7<br />
ATES – Korea Nov. 8 – 10<br />
<strong>Fluent</strong> France Nov. 9<br />
<strong>Fluent</strong> Germany Nov. 14<br />
<strong>Fluent</strong> Asia Pacific Nov. 16 – 17<br />
<strong>Fluent</strong> Italy Nov. 21<br />
Australia Nov. 21 – 22<br />
<strong>Fluent</strong> India Nov. 21 – 22<br />
<strong>Fluent</strong> in Spain Nov. 24<br />
International Oil & Gas CFD Conference Nov. 30 – Dec. 1<br />
Contact your local<br />
<strong>Fluent</strong> office for details.<br />
<strong>Fluent</strong> Worldwide<br />
Corporate Headquarters<br />
<strong>Fluent</strong> Inc.<br />
10 Cavendish Court<br />
Lebanon, NH 03766, USA<br />
Tel: 603 643 2600<br />
800 445 4454<br />
Fax: 603 643 3967<br />
Email: info@fluent.com<br />
USA Regional Offices<br />
Ann Arbor, MI 48108 Austin, TX 78746<br />
Tel: 734 213 6821 Tel: 512 306 9299<br />
Santa Clara, CA 95051 Evanston, IL 60201<br />
Tel: 408 522 8734 Tel: 847 491 0200<br />
Morgantown, WV 26505<br />
Tel: 304 598 3770<br />
European Regional Offices<br />
<strong>Fluent</strong> Benelux<br />
Wavre, Belgium<br />
Tel: 32 1045 2861<br />
Email: info@fluent.be<br />
<strong>Fluent</strong> Deutschland GmbH<br />
Darmstadt, Germany<br />
Tel: 49 6151 36440<br />
Email: info@fluent.de<br />
<strong>Fluent</strong> Europe Ltd.<br />
Sheffield, England<br />
Tel: 44 114 281 8888<br />
Email: info@fluent.co.uk<br />
<strong>Fluent</strong> France SA<br />
Montigny le Bretonneux, France<br />
Tel: 33 1 3060 9897<br />
Email: info@fluent.fr<br />
<strong>Fluent</strong> Italia<br />
Milano, Italy<br />
Tel: 39 02 8901 3378<br />
Email: info@fluent.it<br />
<strong>Fluent</strong> Sweden AB<br />
Goteborg, Sweden<br />
Tel: 46 31 771 8780<br />
Email: info@fluent.se<br />
Asian Regional Offices<br />
<strong>Fluent</strong> Asia Pacific Co., Ltd.<br />
Tokyo, Japan<br />
Tel: 81 3 5324 7301<br />
Email: info@fluent.co.jp<br />
Osaka, Japan<br />
Tel: 81 6 6359 7371<br />
<strong>Fluent</strong> Software (Shanghai) Co., Ltd.<br />
Shanghai, China<br />
Tel: 86 21 53855180<br />
Email: info_china@fluent.com<br />
<strong>Fluent</strong> India Pvt. Ltd.<br />
Pune, India<br />
Tel: 91 20 2293770<br />
Email: info@fluent.co.in<br />
Distributors<br />
ANOVA Ltd. – Turkey<br />
ATES – Korea<br />
Beijing Hi-Key Technology Corp. Ltd. – China<br />
Beijing Tianyuan – China (Icepack products only)<br />
Cavendish Instruments de Mexico, S.A. de<br />
C.V.(CIM) – Mexico, Venezuela, Argentina,<br />
Chile, Colombia<br />
CFD.HU Kft – Hungary<br />
Flowmen Technology Co., Ltd. – Taiwan<br />
Fluid Codes Ltd. – UK (sole distributor for the Middle<br />
East except for Hungary)<br />
Fluvius Pty. Ltd. – Australia & New Zealand<br />
J-ROM Ltd. – Israel<br />
Plasma Venture Ltd. – Russia (serving the Auto, Aero<br />
and Oil/Gas industries only)<br />
Process Flow Ltd. – Finland & Baltics; Russia<br />
(serving Chemical, Power and Electronics industries only)<br />
Qfinsoft – South Africa<br />
Simcon Intl (Pvt.) Ltd. – Pakistan<br />
SimTec Ltd. – Greece (serving Southeastern Europe)<br />
SMARTtech Fluidos Services & Systems, Ltd. –<br />
Brazil<br />
SymKom – Poland<br />
Techsoft Engineering s.r.o. – Czech Republic &<br />
Slovak Republic<br />
TENSOR SRL – Romania