Microsystems and Nanotechnology Industry Spotlight:

Microsystems and Nanotechnology Industry Spotlight:
Unified release of ANSYS,
ANSYS CFX and ANSYS ICEM
CFD software offers advances
Industry Spotlight:
Microsystems and
Nanotechnology
Century Dynamics software
helps oil and gas industries
develop better systems
Cylinder example demonstrates
nonlinear effects, follower forces
and modeling techniques used
for buckling analysis
Contents
Departments
Industry Spotlight
5
Microsystems and
Nanotechnology
Thinking outside of the box
A continuing series on the value
of engineering simulation in
specific industries
Announcements and Upcoming Events
Editorial
..............................................................
2
Industry News
.........................
3
CAE Community
College Design Engineering Award
What’s New in Computational Fluid Dynamics
....
22
Simulation at Work
Highlights of
ANSYS 10.0
Personal Rapid Transit Vehicles
Unified release of ANSYS,
ANSYS CFX and ANSYS ICEM
CFD software offers advances
Forefront of Product Development
...............................................
30
Guest Commentary
.......................................
32
Tech File
Meshing in Workbench
14 Designing Safe and
.....................................................................
34
Tips and Techniques
Efficient Drill Rigs
Modal Analysis of Models with Friction
Century Dynamics software
helps oil and gas industries
develop better systems
ANSYS CFX Performance and Scaling
........................
36
Hardware Update
17 Comparing ANSYS
.........................
About the cover
Shell Elements for
Buckling Analysis
Industry Spotlight article
beginning on page 5
discusses how engineering
simulation is used in
developing microsystems
and nanotechnology
products.
Cylinder example demonstrates
nonlinear effects, follower forces
and modeling techniques used
for buckling analysis
Want to continue receiving ANSYS Solutions?
Visit www.ansys.com/subscribe to update your information.
Plus, you’ll have the chance to sign up to receive ANSYS eNews
and email alerts when the latest electronic version of ANSYS
Solutions becomes available!
For ANSYS, Inc. sales information, call 1.866.267.9724, or visit www.ansys.com.
Go to www.ansyssolutions.com/subscribe to subscribe to ANSYS Solutions.
Editorial Director
John Krouse
[email protected]
Designers
Miller Creative Group
[email protected]
Ad Sales Manager
Beth Mazurak
[email protected]
Editorial Advisor
Kelly Wall
[email protected]
Managing Editor
Jennifer Hucko
[email protected]
Art Director
Dan Hart
[email protected]
Circulation Manager
Elaine Travers
[email protected]
CFD Update Advisor
Chris Reeves
[email protected]
ANSYS Solutions is published for ANSYS, Inc. customers, partners and others interested in the field of design and analysis applications.
Neither ANSYS, Inc. nor Miller Creative Group guarantees or warrants accuracy or completeness of the material contained in this publication. ANSYS, ANSYS Workbench, CFX,
AUTODYN, and any and all ANSYS, Inc. product and service names are registered trademarks or trademarks of ANSYS, Inc. or its subsidiaries located in the United States or other
countries. ICEM CFD is a trademark licensed by ANSYS, Inc. All other trademarks or registered trademarks are the property of their respective owners. POSTMASTER: Send
change of address to ANSYS, Inc., Southpointe, 275 Technology Drive, Canonsburg, PA 15317 USA.
©2005 ANSYS, Inc. All rights reserved.
www.ansys.com
20
CFD Update
Features
10
.....................................
ANSYS Solutions
|
Summer 2005
38
Editorial
2
Innovation: The New Competitive Necessity
Thinking outside of the box to fully leverage simulation technology
in product development
Time-to-market, cost and
quality traditionally have
been the major competitive
issues for manufacturers.
Companies that launched
products faster than competitors, made them better and
had the lowest prices were
generally the most successful
in the market. These remain
critical requirements, of
By John Krouse
course. But in today’s comEditorial Director
petitive business climate,
ANSYS Solutions
[email protected] faster–better–cheaper isn’t
good enough anymore.
Now the competitive
edge goes to firms that meet these criteria while
delivering killer products with innovative designs that
customers flock to buy. Design innovation differentiates manufacturers, builds brand value and drives
top-line revenue growth. Innovative products stand
out from the crowd in terms of value, performance,
size or capacity. In some cases, a company may come
out with an entirely new class of never-before-seen
product such as the SkyWeb Express personal rapid
transit vehicle, where critical joints connecting the
cabin to the chassis were evaluated with ANSYS tools
as covered in the Simulation at Work department
beginning on page 30.
Analysis can play a key role in developing these
types of innovative designs by leveraging the expertise
of a company’s technical professionals, from whom
creative ideas and concepts originate. The tools can
be applied in a simulation-driven development
approach by identifying and correcting problems up
front, optimizing designs early, evaluating alternatives,
studying tradeoffs and analyzing product performance
not practical by building and testing physical
prototypes.
As described in the article “Highlights of ANSYS
10.0” on page 10, the latest release of ANSYS software offers major advances in analysis core
technology within an integrated simulation
environment. The ANSYS Workbench platform is the
backbone for unprecedented levels of CAE integration
as well as process automation features for capturing,
streamlining and standardizing simulation operations.
Product and process innovation go hand in hand, and
ANSYS, Inc. is at the forefront of providing tools
for both.
www.ansys.com
In a growing number of industries, such
advanced technologies are becoming indispensable
tools in developing innovative products. State-of-theart multiphysics software, for example, is a must-have
in determining the effect of stress, deformation and
temperature distribution on tiny micron-size structures
of Micro-Electro-Mechanical Systems (MEMS)
discussed in the fascinating Industry Spotlight article
“Microsystems and Nanotechnology” beginning
on page 5.
Often product development teams use simulation
in navigating through uncharted territory in their drive
toward innovation. Take a look at the CAE Community
department on page 20 to learn how a student design
team from the University of Washington boosted
output of a micropump design threefold using ANSYS
tools. These students are probably the first to use CFD
coupled to an optimizer to determine the best design
of the pump’s microvalve concept and also the first to
stack micropumps in parallel to increase flow rate.
In striving for innovation, evaluating and changing
well-entrenched corporate practices to fully leverage
simulation technology often present enormous
challenges. As Charles Foundyller points out in his
guest commentary “CAE Moves to the Forefront of
Product Development” on page 32, technical issues
pale in comparison to confronting cultural, organizational and procedural issues. His cow path concept
hits the mark exactly in showing how process chains
often bend and wander in a convoluted trail that
companies follow because things have been done that
way for decades.
The difficulties in confronting these organizational
issues are daunting, of course, which is why many
timid companies may never get off square one in
making necessary changes. Rather than merely plug
the technology into existing procedures, forwardlooking companies closely examine their ways of
operating and think outside the box in terms of
when simulation is performed, what tools are used,
where refinements are made and how analysis fits into
a company’s overall business strategy. These smart
enterprises know full well the critical value of innovation, have the fortitude to break new ground in product
development and will likely be among the world’s
manufacturing superstars in the coming years. ■
ANSYS Solutions
|
Summer 2005
Industry News
Recent Announcements
and Upcoming Events
Now Available! ANSYS Parametric Design
Language Customization Guide
Phoenix Analysis and Design Technologies (PADT) is
pleased to announce the release of its Guide to
ANSYS Customization with APDL. This guide is a
compilation of course notes from PADT’s very popular
“ANSYS Customization with APDL” class. By popular
demand, PADT has turned the notes into a 288-page
guide that steps new and experienced ANSYS users
through all of the details of APDL scripting. Its 12
chapters include reference information, examples, tips
and hints and eight workshops. This guide, available
in hardcopy only, is an invaluable resource to anyone
who wants to start using APDL or become an ANSYS
“power user.” Priced at $75, it will quickly pay for itself
by saving you hours of research and trial-and-error.
For more information, visit www.padtinc.com/
support/techguides or call 1-800-293-PADT.
ence materials and code-check various structures
against international standards. For example, engineers in the nuclear industry can analyze stress on a
structure design and then conduct a code check of a
welding to ensure it meets the code of the particular
country in which it is being built — all before creating
a sketch in any CAD program.
The integration also offers a convenient way for engineers to manage their projects, since they are able to
compile all information and track their progress on a
project in one location. Because ESOP is an Excelbased program, the technology is also easy to use.
To learn more, visit the news section at www.ansys.com.
Save the Date for 2006 International
ANSYS Conference
ANSYS, Inc. recently announced a partnership with
RoboBAT, a leading supplier of analytical and CAD
software solutions for the structural engineer, to integrate their solutions in order to streamline customers’
product development processes.
Continuing what avid ANSYS users consider a CAE
tradition, preparations are under way for the 2006
International ANSYS Conference to be held May 2 – 4
at a new venue this year — the David L. Lawrence
Convention Center — with training April 30 – May 1
at the Westin Convention Center, in Pittsburgh,
Pennsylvania, USA. The event brings together
engineers and analysts worldwide from all disciplines
of computer-aided engineering.
The integration of RoboBAT’s Engineering System
Open Platform (ESOP) software with ANSYS Workbench expands users’ access to structural engineering
applications and increases their productivity. Users
have access to more than 100 engineering applications as well as the ability to write their own modules,
either stand-alone or integrated with ANSYS Workbench models (for example, writing pre- and/or postprocessors for Workbench).
Since 1983, ANSYS, Inc. has hosted the International
ANSYS Conference to showcase the advances in
computer-aided engineering and related technologies.
The three-day conference addresses the complete
spectrum of engineering professionals including
engineers, analysts and engineering managers.
Conference attendees have the opportunity to
network with peers, interact with ANSYS professionals
and participate in a variety of technical sessions.
Through ESOP for Workbench, users can assess
structures quickly and easily. They have a central location where they can conduct calculations and analyses
in the pre-design phase, access databases of refer-
For more information and to sign up for email updates,
visit www.ansys.com/conf2006.
ANSYS and RoboBAT Announce an Integrated
Structural Engineering Solution
www.ansys.com
ANSYS Solutions
|
Summer 2005
3
Industry News
4
Partner Announcements
Matereality v.2.1 Features New Products,
Functionality, Content
Matereality, L.L.C., USA, the developer of the first-ofits-kind complete material data management (MDM)
system in TrueDigital™ format, announces the release
of Matereality® Version 2.1.
The much awaited Version 2.1 release of the Matereality material data management platform brings a number of advances to this rapidly developing field. In
addition to creating the flagship line that provides a
Web-based resource for the storage and selective
deployment of material property data, the company
now has two new products. Material Data Server has
been created to enable manufacturing enterprises to
store all their material property data on all their materials
on a single extensible platform within their intranets
and maintain complete control over who has access to
it. With this, companies can now finally go paperless
with the storage of all their test data, cradle to grave.
The Materials Databases product line provides
material suppliers with a ready-to-deploy technology
to enhance their customer support with one powerful
secure platform for material selection data as well
as design properties, 365/24/7 globally! For more
information, visit www.matereality.com.
Durability & Fatigue Seminars from Safe Technology
Noted fatigue & durability authority, Professor John
Draper of Safe Technology Ltd (UK ),will be visiting
Downers Grove, Chicago, Ill., September 12 - 16, to
present a series offatigue and durability seminars
hosted and sponsored by Belcan Corporation.
A regular speaker at conferences worldwide, Professor Draper has over 25 years' experiencein fatigue
design and life assessment in the aerospace and
ground-vehicle industries.
These courses will provide an introduction to modern
theories of metal fatigue and their practical application
through worked examples and interaction/discussion.
There is a strong emphasis on what is possible and the
pitfalls to avoid. A professional volume of course notes
forms a self-contained reference book.
Upcoming Events
ISABE
September 4 – 9
Munich, Germany
www.isabe2005.com
Nuclear Energy for New Europe
September 5 – 8
Bled, Slovenia
www.drustvo-js.si/bled2005
Offshore Europe 2005 Oil & Gas
Exhibition & Conference
September 6 – 9
Aberdeen, UK
www.offshore-europe.co.uk
German Fuel Cell Congress 2005
September 26 – 28
Stuttgart, Germany
www.f-cell.de
German Aerospace Congress 2005
September 26 – 29
Friedrichshafen, Germany
www.dglr.de
Automobile & Engine Technology
Colloquium
October 4 – 6
Aachen, Germany
www.rwth-aachen.de/ac-kolloquium/index_e.htm
SNAME Maritime Technology Conference
& Expo and Ship Production Symposium
October 19 – 21
Houston, TX, USA
www.sname.org/AM2005/
Visit www.ansys.com for more events.
For a complete listing of courses, visit
www.safetechnology.com. ■
www.ansys.com
ANSYS Solutions
|
Summer 2005
Industry Spotlight
5
Microsystems
and
Nanotechnology
A continuing series on the value
of engineering simulation in
specific industries
A rapidly growing number of products in many
industries contain miniature sensors, actuators and
other Micro-Electro-Mechanical System (MEMS)
devices made of small parts measured in microns
(millionths of a meter). Complete devices typically
are on the order of a few millimeters across, with
individual components less than 10 µm thick, and
frequently are combined with integrated circuits on
a single chip to provide built-in intelligence and
signal processing.
www.ansys.com
ANSYS Solutions
|
Summer 2005
Industry Spotlight
6
MEMS devices must perform accurately and
reliably, often in hostile environments. Therefore,
engineers rely extensively on engineering simulation
software to study these microstructures in
determining stress, deformation, resonance,
temperature distribution, piezoelectric response,
electromagnetic interference and electrical properties.
Multiphysics analysis tools, in particular, are
essential in accounting for these effects, which can
have such huge impact on delicate microstructures.
Most of the commercially deployed devices are
used in automotive applications such as accelerometers for deploying airbags and sensors for monitoring
manifold pressures, roll-over and fuel injection
systems, for example. Biomedical is another large
market, with MEMS in disposable blood pressure
sensors, as another example. The technology also
is used for ink-jet printer heads, display devices,
hard disk drives and other electronic equipment
applications.
Clearly, MEMS technology is past the buzzword
stage of the late 1990s and now represents a growing
industry producing millions of products each year.
Industry analysts estimate the total worldwide MEMS
market at nearly $7 billion and expect it to undergo a
compound annual growth rate in the range of 20
percent for the next several years as implementation
of the devices expands. The use of MEMS in automobiles is predicted to double by 2007.
Small, Accurate and Economical
The attraction of MEMS for such applications is their
relatively low cost and high performance in a compact
package. Produced through the same semiconductor
fabrication methods as integrated circuits (ICs), thousands of MEMS can be mass-produced on a single
silicon wafer along with associated electronic circuits.
Using well-proven semiconductor fabrication
technology, including bulk micromachining or surface
deposition, MEMS can be produced and sold for
a fraction of the cost of conventional sensing or
actuation devices. Conventional blood pressure
transducers costing $600, for example, can be
replaced by MEMS intravenous sensors selling for
about $10.
The ability to easily integrate MEMS transducers
with signal processing electronics and the precise
manufacturing tolerances inherent in the semiconductor fabrication process allow manufacturers to
meet (and in many cases exceed) the performance of
equivalent macroscopic transducers.
www.ansys.com
Electrostatic-structural analysis of a comb drive MEMS
resonator commonly used in electromechanical bandpass
filters shows voltage iso-surfaces in the device.
Design Challenges
One of the most formidable tasks in MEMS
development is designing the microscopic parts to
best fit together and operate properly, often in
environments hostile to delicate mechanical
components and sensitive electronic circuitry.
Developing internal interconnects, circuitry and
packaging so that MEMS devices operate flawlessly
for years in these demanding applications is a tough
engineering challenge. Semiconductors must resist
damaging internal heat build-up while withstanding
a wide range of structural loads and ambient
temperature swings. Parts such as diaphragms,
valves, membranes, beams and other microstructures
on the same silicon chip also must survive severe
shock and vibration to adequately perform their
mechanical functions.
Traditional electromechanical products have
these same requirements. But designing MEMS is
particularly challenging because of the tremendous
difference in size and overall sensitivity of the devices’
internal components compared to their surroundings.
A MEMS sensor that measures gas pressures in
the range of 0.15 psi by detecting a few microns
deflection of a microbeam, for example, often must
undergo several Gs of shock and vibration while in
service on a piece of factory-floor equipment. The task
of the MEMS sensor in this application is equivalent to
detecting a sneeze in the middle of an earthquake!
ANSYS Solutions
|
Summer 2005
A MEMS designer has further challenges in the
packaging of the device. Packaging, typically a form of
plastic encapsulation similar to ICs, often impacts
device operation, reliability and accuracy.
Because many of these effects are interdependent, predicting output and performance of
MEMS devices is generally a complex problem that
often defies intuitive approaches used in developing
traditional transducer technology.
Developers of MEMS also have greater obstacles
to overcome in the area of prototype testing. Whereas
physical mock-ups of conventional electromechanical
devices may undergo several test and re-design
cycles where parts are modified and switched around,
the initial semiconductor fabrication setup for MEMS
is so time-intensive that prototype testing simply
verifies a design, and physical design interactions are
largely replaced by virtual prototyping.
Multiphysics Solutions
To meet these many design challenges, MEMS
engineers universally rely on engineering simulation to
create and performance-test virtual prototypes of the
devices. Multiphysics simulation is used extensively in
the development of MEMS because of the inherent
multiple interrelated physical phenomena at play,
such as stress, temperature, electrostatics, piezoelectrics, fluidic damping, thermoelastic effects
and electromagnetism.
Previously, solving such coupled applications
meant enduring numerous manual file transfers
and problem setups for each physics analysis.
Such cycles were cumbersome, error-prone and timeconsuming and required users to learn and maintain
several different software codes. These issues are
being addressed by multiphysics solutions that
automatically combine the effects of two or more
interrelated physics within one unified environment.
These solutions automatically manage data exchange
between the different physics to perform coupled
analysis without requiring users to spend time
manually performing these tasks. As a result, coupled
analyses can be performed in a fraction of the time
otherwise required.
Possibly the original and most comprehensive
multiphysics solution is provided by ANSYS, Inc.
ANSYS Multiphysics offers the widest range of
multiphysics disciplines in a single unified
environment: structural, thermal, fluid and electromagnetic. Another major strength is that ANSYS
Multiphysics is fully integrated with many ANSYS
analysis tools such as parametric modeling
capabilities, design optimization and probabilistic
design functionality, and both ECAD and MCAD
import features. ANSYS Multiphysics also has an
advanced fluid structure interaction capability realized
through bi-directional interface with the ANSYS CFX
computational fluid dynamics code.
Unlike many other commercial FEA codes that
provide either direct or iterative multi-field approaches,
ANSYS Multiphysics provides both. Direct coupled
field analysis solves all of the physics field’s degrees of
freedom in one solution phase. In iterative coupling,
results of a single FEA solver iteration of one physics
SilMach uses ANSYS Multiphysics in the development of electromagnetic actuators and research into a variety of MEMS-based actuators
and systems such as these unmanned airborne vehicles with flapping wing propulsion systems.
www.ansys.com
ANSYS Solutions
|
Summer 2005
7
Industry Spotlight
8
field are passed as loads to the next physics field,
iterating between all active physics fields until
convergence criteria in the transferred loads are met.
The ANSYS Multi-field solver automatically handles
the data exchanged in these iterations. Users only
have to set up the problem initially.
Analysis in Action
A growing number of organizations use ANSYS
Multiphysics for analyzing MEMS devices. Today,
hundreds of companies and research institutions use
the software in developing leading-edge MEMS
technologies globally.
SilMach, for example, develops highly integrated
silicon-based actuators and systems. By using
ANSYS Multiphysics, the company has the ability to
solve complex coupled physics to create sensors and
actuators within arrays and predict their performance
before committing to manufacture. Coupled physics
such as mechanical deformation and nonlinear
contact effects with acoustic, electrostatic, thermal
and fluid damping can be achieved using the software.
sion systems for artificial insects and nanometer-scale
unmanned airborne vehicles (UAVs).
The software is also used by RTI International in
the development of advanced imaging systems. A
particular challenge is simulating the effect of
25-100THz infrared waves interacting with a periodic
array reflective structure. The problem was considered
unsolvable using traditional FEA methods, which
would have required a model size approaching 100
million degrees of freedom.
ANSYS Multiphysics was used to perform a full
wave electromagnetics harmonic scattering analysis
on the device. Reflection coefficients were computed
for the entire frequency range. The analysis took into
account the skin depth and loss of RF energy through
joule heating of the materials. A fully parametric model
enabled rapid changes in materials, geometry and
excitation. The numerical problem size was reduced
considerably through the use of the periodic boundary
condition.
This approach provided RTI with the ability
to validate experimental results and quickly improve
RTI modeled a unit cell (left) and obtained results of the scattered plane wave (right) using ANSYS Multiphysics in the development of
a periodic array reflective structure.
ANSYS Multiphysics has enabled SilMach to
create more efficient MEMS devices such as an
electromagnetic actuator producing 100 Watts per
gram compared to 1 Watt per gram for standard
devices. The company also continues to research
other advanced technologies, including MEMS-based
gas micro-turbines, MEMS-based flap-arrays for
active control of turbulence and flapping-wing propul-
www.ansys.com
device performance by investigating various structure
parameter changes. The software contributed to a
better scientific understanding of experimental results
because researchers could actually visualize the
electric field within and around the structure. ANSYS
Multiphysics also allowed RTl to analyze results at discrete frequency points, which will help plan future
equipment purchases for their experimental work. ■
ANSYS Solutions
|
Summer 2005
9
Nanotechnology:
The Next Small Frontier
The success of MEMS is generating considerable
momentum in emerging research and development
surrounding nanotechnology, where the device scale
is at the atomic or molecular level with dimensions of
100 nanometers and smaller.
The most recent developments in nanotechnology
are direct atomic manipulation techniques, singleelectron transistors and carbon-60 nanotubes as
current-carrying conductors in nano-integratedcircuits. Microsystem technology can be used to
build and interact with nano-scale devices and
systems, so MEMS and nano remain closely linked.
There are many MEMS fabrication technologies being
scaled down to produce nanoscale devices. Perhaps
the best example is the micro-tip field emitter: essentially a microscopic electron gun consisting of a
conical field emitter with a tip radius of 10-20 nm.
Currently, ANSYS Multiphysics addresses
nanotechnology simulation requirements covering ion
optics and electrostatic calculations associated with
field emission tips and carbon nanotube structures.
The ion optics capability is a post-processing feature
where charged particles can be introduced into
computed electromagnetic fields and their path traced
with “streamline” graphics. The user can control the
“history” of the charged particle, changing its mass
and/or charge to simulate fragmentation, for example.
At the nanometer scale, the bulk material models
used by conventional finite element tools generally
break down as quantum mechanical effects become
dominant. The availability of customizable, userprogrammable material models in ANSYS Multiphysics is helping to address the analysis of some
nano-systems. Some users have made reasonable
approximations to polycrystalline grains in surface
micromachined parts using this approach.
ANSYS Multiphysics provides insight into the trajectories of
charged particles from a conical field emitter with a tip radius
of 10-20 nm.
This electrostatic field benchmark of an array of eight nano-scale
spheres was performed by Dr. Andreas Hieke, an internationally
recognized authority in modeling and analysis of microsystems and
nano-scale systems.
ANSYS structural analysis determined deformation and residual
stress levels resulting from fabrication of this MEMS optical grating
device used in spectroscopy instruments.
www.ansys.com
ANSYS Solutions
|
Summer 2005
Highlights of
10
ANSYS 10.0
Unified release of ANSYS, ANSYS CFX and ANSYS ICEM CFD
software offers advances in core technology within an integrated
environment for ease of use and productivity enhancements.
The latest release of ANSYS v10.0 offers the first
unified release of products under the ANSYS product
portfolio: ANSYS, ANSYS CFX and ANSYS ICEM
CFD. This milestone provides an integrated simulation
environment with some of the best technologies for
meshing, pre-post processing and multiphysics.
Integration of this diverse software portfolio offers
leading-edge new technology and unparalleled
productivity advances to the end users across a wide
industry spectrum and physics areas including
analysis of composites, rotating machinery, metal
forming, interface separation and delamination,
MEMS devices, CFD transition turbulence and meshing of shell models for underhood and seam welding.
In addition, ANSYS 10.0 offers further enhancements in its Workbench product development
environment, integrating more technologies and taking
further steps toward a more collaborative, integrated
and customizable range of virtual prototyping
software.
Advanced Fluid Structure Interaction
The new release provides the industry’s most flexible
and advanced fluid structure interaction (FSI) analysis
solution required for many industry applications such
as MEMS, including microfluidics, biomedical, automotive, aerospace and civil engineering. The ANSYS
coupled physics multi-field solver offers enhancements to support coupling of ANSYS structural
physics with CFX CFD for such FSI applications.
Enhanced solver technology allows structural
and fluid solutions to run simultaneously on the same
or different machines, thus accommodating much
larger models more efficiently than a multi-field solver
using a single machine environment. Code coupling is
based on high-speed inter-process communication
technology without the need for third-party applications. The technology ensures that ANSYS CFX can
be run concurrently, thus allowing fluid and structural
computations to be solved on different machines while
also communicating across a local area network, a
wide area network, or even via an Internet connection.
www.ansys.com
Biomedical fluid structure interaction analysis of blood
flowing in an elastic artery using the new two-way FSI
capability of the multi-field solver
Extending the Realm of CFD Analysis
Enhancements in this latest release deliver advanced
CFD technology for a wide range of industrial
applications. Software improvements target a number
of industries and sectors including power generation,
chemical process industries, external aerodynamics,
rotating machinery, automotive engine, external and
internal flows, fire and safety applications, marine and
free surface applications.
Power generation industries benefit from
advances in a number of areas including advanced
coal combustion analysis, state-of-the-art volumetric
porosity abilities, advanced boundary conditions
that simplify the setup of high-fidelity simulations
and novel R&D advances in non-equilibrium
steam modeling. These improvements are further
leveraged when combined with the latest advances in
FSI capabilities.
ANSYS Solutions
|
Summer 2005
www.ansys.com
Image courtesy of BMW AG.
State-of-the-art advances in turbulence modeling
benefits most sectors, permitting more accurate
modeling of true flow physics. For example, release
10.0 includes the world’s first commercially available
predictive transitional turbulence model for flows that
include both laminar and turbulent flow regimens.
For analysts, this translates into more accurate predictions of wall forces, flow separation, heat transfer
coefficients, and more. For those applications that
demand detailed resolution of the turbulent structures, there are more options than ever, including
Detached Eddy (DES) models and introducing the
groundbreaking Scale Adaptive (SAS) R&D models.
In many industries, transient flow simulation
interest continues to grow in widely diverse
applications such as fuel droplet injection in an
internal engine, flow-induced vibrations of a structure,
water droplet spray in a fire simulation, noise
source estimation and fluidized bed performance
assessment. A number of key transient flow-related
improvements have been made for ANSYS CFX 10.0
to improve the accuracy of transient simulations,
as well as reduce the CPU time needed for such
simulations. Further reductions in the time needed
to perform such large simulations are obtained by
improvements to the scalable parallel performance.
ANSYS CFX 10.0 now supports high-speed
interconnects on most hardware platforms, for
example the Myrinet, Infiniband and Quadrics
interconnects. There is also a full release of all ANSYS
CFX components as native 64-bit executables for the
new Intel EM64T/AMD Opteron chips. This all
translates into more accurate simulations in less time.
Extending its leadership in providing solutions for
turbomachinery design and analysis, ANSYS CFX
10.0 includes BladeModeler, a highly customized
rotating machinery design tool for bladed
components, and TurboGrid, a high-quality nearautomatic hexahedral meshing tool for bladed
components. These tools combine with the
specialized Turbo-Pre and Turbo-Post processing
CFD capabilities to create a comprehensive
turbomachinery design and analysis system.
Simulation models for stress analysis, computational
fluid dynamics and/or fluid-structure interaction can
be created from one centralized geometry definition.
It may include upstream and downstream components and even whole-machine analysis. Applications
include gas turbine design and analysis, pumps,
turbines, compressors, fans, blowers, expanders,
turbochargers and inducers.
ANSYS CFX 10.0 also contains a number of new
features that represent significant advancements in
the interactivity and usability of working within the
Automotive intake valve modeled using ANSYS ICEM CFD
and the moving mesh capability of ANSYS CFX.
post-processor. These are seen primarily through
Viewer Shortcuts and new quick access tools. Other
new capabilities include quantitative tables, point
cloud and viewer object transformations.
Core Technology Enhancements in ANSYS
Design and stability of rotating parts such as those in
turbo-machinery, cooling systems, biomedical and
automotive systems require accurate specialized
modeling of the physics of these structures. The new
release offers capabilities for Coriolis damping and
forces for beams, shells and 2-D/3-D solids within
ANSYS for static, transient and modal analysis of
rotating equipment. Accurate representation of
rotating equipment is possible when using solid
elements for analysis, with the need for reducing a 3-D
CAD representation into beam models. This furthers
capabilities to analyze dynamics of rotating machinery
through modal, harmonic and transient analysis
within ANSYS.
Composites find application in aerospace and
automotive for designing structures with high strength
and less weight. Failure and delamination modeling of
composite structures are critical, taking into account
the accurate loading and material behavior of such
composites. Release 10.0 offers enhancements in
three main areas:
• Composites enhancements focus on
interfacing with Vistagy’s FiberSim software, a
composites pre-processing tool for draping
analysis. Information from FiberSim is imported
into ANSYS, thereby providing accurate
information of the manufacturing process for
analysis.
• New solid-shell elements are extended for
layered applications like laminated shells and
sandwich structures. Solid-shell elements are
part of the new generation of ANSYS elements
for modeling thin to moderately thick structures
and allow modeling of variable thickness parts.
• New cohesive zone capability in ANSYS
allows modeling of interface separation and
delamination, typical of layered composite
structures. This enhancement also finds
application in modeling glue behavior of
materials with interface separation.
ANSYS Solutions
|
Summer 2005
11
Underhood meshing capability in ANSYS ICEM
CFD takes care of “dirty” CAD model geometry
(left) for robust meshing (right).
12
Enhancing the rezoning feature in ANSYS 9.0, the
current release extends the capability to metal
plasticity with an ability to re-mesh the deformed
domain and continue solution toward convergence.
This enhancement, with support of contact
capabilities, finds application in areas such as metal
forming. Completing and further advancing the
offering of contact technology, ANSYS 10.0 offers new
enhancements in areas of beam–beam contact and
orthotropic contact friction. This enhancement, for
example, finds application in areas of water/oil supply
lines, nuclear power plant piping, cable wires and
coils, woven fabrics and tennis racquets.
For coupled physics analysis, ANSYS 10.0 allows
thermoelastic damping (TED) and direct structuralthermal-electric coupling as part of the new advanced
coupled-field elements. Thermoelastic damping is an
important internal loss mechanism in metals and
MEMS (resonator beams) and the capability provides
more accurate, real-world analysis results for the
MEMS industry. This functionality allows ANSYS to be
the leader in solving a physical phenomenon that few,
if any, other analysis codes can address.
All Distributed ANSYS (DANSYS) users will
benefit from the dramatic improvements made to the
installation and configuration process. Made possible
by HP-MPI on Linux systems, DANSYS no longer
requires consistent working directory structures,
now supports NFS mounting for slave nodes, automatically picks the fastest interconnect available and
allows both rsh and ssh logins. By adopting HP-MPI,
Distributed ANSYS now supports the following
interconnects:
Tools for Efficient Meshing
ANSYS ICEM CFD 10.0 provides a complete set of
tools to help reduce the time for meshing underhood
models in various applications such as automobiles,
aircraft and heavy machinery. New additions include
shrink-wrapping, Delaunay and/or hexa-core meshing
and pure Cartesian meshing to allow customers to
standardize on ICEM CFD for underhood meshing.
The combination of powerful shrink-wrapping combined with feature-rich geometry and mesh editing
tools allows customers to selectively capture or ignore
details in their complicated models.
The representation of welds for sheet metal
models is very important for accurate results
of durability models. With AI*Environment 10.0,
a user can represent welds between sheet metal
components in a variety of ways to improve solution
accuracy or solver speed of the analysis. Welds can
be created automatically between single edges of
one part and neighboring parts. These welds can be
defined as rigid beam elements, quad element
welds or tent welds, to produce more or less rigid
welds in a weld region. The combination of powerful
midsurfacing and automation of welding allows
the user to create assemblies of midsurface shell
models quickly.
This unification of ANSYS, ICEM CFD and CFX
has been an essential step in the development process,
but more importantly, has allowed strengths of the
different solutions to be leveraged across products.
With ANSYS 10.0 core meshing technology, ANSYS
ICEM CFD has been exposed within the Simulation tab
of Workbench to provide uniform quad/tri mesh. This
• InfiniBand (recommended)
• Myrinet (recommended)
• GigE
• Ethernet (not recommended)
DANSYS now supports static and full transient
electromagnetic analysis, mixed U-P formulation,
linking in of User Programmable Features (UPF) on
Linux and UNIX, and allowing beam and shell contact,
including shell and beam thicknesses, by automatically
grouping all contact elements together into one
domain. Also, DANSYS now allows serial solution of
the Radiosity Solver while performing a distributed
thermal analysis.
Nonlinear contact analysis of a spring using the new
beam–beam contact capability
www.ansys.com
ANSYS Solutions
|
Summer 2005
Transient thermal analysis results of circuit board with probed
results using new transient analysis capability in the Workbench
Simulation Environment
meshing option provides more uniform, orthogonal
quad and/or tri elements to reduce the mesh size of
the model and to provide better accuracy in the solution. Behind the scenes it also has helped to provide a
framework to help integrate these tools as well as
other meshing tools in an easier fashion.
13
Greater Capabilities for Workbench
Integrating more advanced features within the
Workbench environment, the current release allows
transient thermal analysis within the Workbench
Simulation environment. Users can now perform
transient thermal analysis with complex loading
patterns and temperature-dependent boundary
conditions. Temperature results from various time
points can then be transferred as boundary conditions
for static structural analysis within the same simulation
environment. The feature extends the power of
Workbench to a widely used ANSYS functionality of
transient thermal/thermo-mechanical analysis with
applications in electronics, gas turbine and automotive
industries.
Further extending the advanced capability
offering in the Workbench environment, release 10.0
offers new functionality of strain-based fatigue in
ANSYS Fatigue Module. New to the ANSYS Fatigue
Module is the ability to analyze Low Cycle Fatigue
(LCF) by Strain Life methods for constant amplitude
loading with non-constant under development (Beta).
The Fatigue Module adds the capability to simulate
performance under anticipated cyclic loading
conditions over a product’s anticipated life span.
Incorporating both Stress Life and Strain Life analyses
with a variety of mean stress correction methods including Marrow and SMT, the Fatigue Module provides
contour plots of fatigue life, factor of safety and stress
biaxiality. Additional results include rainflow matrix,
damage matrix, fatigue sensitivity and hysteresis.
Extending its capabilities, ANSYS DesignXplorer
allows users to perform optimization and Design for Six
Sigma with any application or sequence of applications.
The new third-party plug-in allows parameter transfer to
and from other applications such as in-house codes.
DesignXplorer will process a sequence of instructions
defined by XML file and use this information to interact
with any parameter source. The new Auto Defined
sampling design option will automatically pick the most
accurate sampling method based upon the number
of input parameters. Auto Defined will select from
between the new VIF-Optimal and G-Optimal designs.
DesignXplorer now has a NLPQL (Nonlinear Programming by Quadratic approximation of the Lagrangian)
algorithm, which provides much faster convergence
for smooth continuous responses that do not have local
optima. With the new memory management implemented at 10.0, Tradeoff studies can now be performed
for very large datasets (>100,000). ■
ANSYS Workbench Customization Capabilities
In the world of design simulation, today the goal is to integrate
and “close-couple” all of the engineering analysis tools
necessary to take a product to market.
We at ANSYS call it CAE Collaboration, and it takes many
forms within a design organization. Here are some:
• results sharing between peers, managers, suppliers,
and customers
• process and procedure collaboration to establish and
communicate best practices to the entire engineering
organization
• hardware resource sharing as more realistic models
drive into increasingly larger problem size
• design data sharing, in the form of geometry, material
properties and design parameters to establish consistent data sources and eliminate recreating data work
The key aspects of the ANSYS Workbench customization capabilities are all about sharing data, results, hardware and best
practices with many people in the products development group.
www.ansys.com
The value of these assets being accessible to everyone is much
greater than if limited to only a few people.
ANSYS Workbench with its associated customization tools
allows you:
• to manage the overall engineering simulation process
including simulation methods and processes, common
services and data and configuration management
• to readily integrate your own applications or third-partydeveloped engineering applications into a manageable
common end-user environment
• to establish unique process flows and controls based on
specific product or process requirements
If your goals are to improve productivity and quality of both
your engineering tools and your products, you need to further
examine the capabilities of Workbench by visiting the ANSYS
Workbench Community Portal accessible from the ANSYS
Customer Portal and contacting an ANSYS representative.
ANSYS Solutions
|
Summer 2005
14
Simulation Tools for
Designing Safe and
Efficient Drilling Rigs
Part 2 of 2: Century Dynamics software helps oil and gas
industries develop better systems for hostile environments.
By Steve Pilz
Product Manager, ANSYS, Inc.
Naury Birnbaum
President, Century Dynamics
The first installment of this series presented an
introductory overview of Century Dynamics Inc. (CDI)
and explored the AUTODYN explicit dynamics
product. This article investigates CDI software specific
to the oil and gas industry, where pressure is
enormous to explore and exploit new petroleum
sources for satisfying growing world demand.
With most of the easy fields already tapped,
petroleum companies must focus on those where
weather and other environmental conditions
are extremely hostile. To tackle these hostile
conditions, significantly more engineering efforts
are now required before the first drilling rig arrives,
and engineers are increasingly relying on advanced
simulation tools to design safe and financially viable
systems. Since 1993, CDI has provided oil and gas
engineering simulation software to meet these
challenging applications: AutoReaGas for studying
gas cloud explosions, ASAS for designing marine
structures and AQWA for hydrodynamic assessment
of structures.
Studying Gas Cloud Explosions
Designed to aid in modeling and understanding
the effects of gas cloud ignition and explosion
phenomena, AutoReaGas has been under continuous
development by CDI in conjunction with TNO (The
Netherlands Research Organization) since 1994. The
software is particularly helpful at understanding how
structures such as piping networks affect flame front
propagation and the resulting explosive pressures that
occur when an invisible gas cloud forms around the
structure and is accidentally ignited. The purpose of
the software is to allow users of AutoReaGas to
economically conduct appropriate safety studies and
ultimately design safer facilities for oilrigs as well as
onshore plants such as refineries.
AutoReaGas was written in response to
numerous industrial accidents where oilrigs exploded
and many people were killed because of gas cloud
deflagrations and the resulting high blast pressures.
AutoReaGas is useful in studying flame front propagation through offshore platform piping networks.
15
The software can be used to predict the severity of
simulated accidents as well as to aid in the design of
systems that reduce the health and safety risks in the
oil and gas industry. Where there is an interest in the
blast response of equipment or structures, AutoReaGas results can be readily used with structural
analysis software such as CDI’s AUTODYN for
combined blast-structure interaction studies.
AutoReaGas is useful in studying flame front
propagation through offshore platform piping networks.
Modeling Marine Structures
A finite element program designed to specifically
address the needs of marine structural simulation,
ASAS has been used by offshore and marine
engineers for over 30 years. CDI acquired ASAS in
Ship hull design was modeled with ASAS.
2001 and has continued to develop and enhance the
capabilities of the program, which can now be used to
model jack-up platforms, compliant structures, manifold installation and riser analysis as well as most
other floating structures. Features include:
• wave, current and wind load
• regular and random waves
• Airy, Stokes 5th, Cnoidal, stream function and
shell new wave
• tube and beam elements
• flooded or sealed members
• frame, panel and concrete structure code
checks and its visualization
• drag and inertia force
• marine growth
• wave loading within the API code
16
Multibody Hydrodynamic Software
The third program in the CDI oil and gas simulation
suite is AQWA, a multibody hydrodynamic software
tool used around the world by engineers to simulate
offshore jacket launching, calculate response of
ship hulls to wave loading, analyze hydrodynamic
and mechanical interaction effects on floating
structures, design loading and unloading devices for
cargo transport, optimize floating structure mooring
systems such as cables and anchors, simulate lifting
operations between vessels and aid in the design of
port installations.
AQWA provides unique simulation capabilities for
the hydrodynamic assessment of all types of structures
including spars, FPSOs (Floating Production Storage
and Offloading vessels), mooring systems, buoys, TLPs
(Tension Leg Platforms), semi-submersibles, ships,
cable dynamics and structure interaction. ■
Concrete gravity platform was modeled with ASAS.
AQWA is used for the hydrodynamic assessment of offshore structures
such as this mammoth floating airport concept shown with C-17 aircraft.
www.ansys.com
ANSYS Solutions
|
Summer 2005
17
Comparing ANSYS Shell Elements
for Buckling Analysis
An externally pressurized cylinder example
demonstrates nonlinear effects, follower forces and
modeling techniques used for buckling analysis.
Finite element codes consider the linear elastic solution for buckling. Eigenvalue buckling analysis predicts
the theoretical buckling strength (the bifurcation point)
of an ideal linear elastic structure.
This method corresponds to the textbook
approach to elastic buckling analysis: For example, an
eigenvalue buckling analysis of a column will match
the classical Euler solution. However, unlike long, slender columns, many real-world structures such as pressurized cylinders can achieve their theoretical elastic
buckling capabilities.
A nonlinear buckling analysis is a static analysis
with large deflection and/or plasticity to determine
where and when the structure reaches its limit load
or maximum load. In the nonlinear analysis that
follows, the path dependence is shown, and thus it
represents a more accurate solution to find the lowest
buckling mode.
By Dr. Charles H. Roche
Pratt & Whitney
Division of United Technologies
Theoretical Solution and Euler Formula
For a straight slender column, the Euler formula is
represented by:
And the equation for hoop stress in a cylinder is:
Considering the quarter symmetry of two-lobe
buckling and neglecting the arc of a 90˚ segment,
applying Euler’s equation to the shell yields:
Solving for the pressure load, we get:
The equation above shows the important effect of
shell thickness and diameter on buckling capability.
For finite length cylinders, an eigenvalue solution
for the critical pressure for buckling is given by
Flügge:
,
where
www.ansys.com
ANSYS Solutions
|
Summer 2005
22
18
and m is the number of lobes in the buckled shape;
for our example, the lowest mode is two-lobe, so m is
2. Also, E and v are the elastic constants, and r is the
radius of the cylinder.
Finite Element Analyses
In the example below, we consider the linear and nonlinear buckling analysis of a shell model using various
shell elements in ANSYS. The results are compared to
the theoretical solution given by Flügge. Consider a
cylindrical tube geometry of (l = 52 in., O.D. = 6.5 in.,
t = 0.035 in.) made up of stainless steel of material
properties (E = 28.5 msi, r = 0.3). The cylinder
easily meets thin shell requirements as well as
elastic buckling.
The methodology for eigenvalue buckling
analysis is given in ANSYS Documentation. The
accompanying table below shows the comparison of
buckling analysis using various shell elements in
ANSYS with the theoretical solution of 19.58 as given
by Flügge.
Surface effects were resolved with the ESURF
command and the SURF154 elements for the
following legacy elements: SHELL43, SHELL63,
SHELL91, SHELL99 and SHELL143. The technique is
unnecessary and not applicable for the newer
SHELL181 and SOLSH190. With the exception of the
axiharmonic element SHELL61, using the surface
effect elements shows dramatic improvement with the
theoretical solution of 19.58 psi.
The use of SURF153 elements with SHELL61
may be inappropriate, as the results suggest. Unique
to the table would be the axiharmonic shell element
SHELL61 and the new SOLSH190 element available
only in ANSYS 9.0. (Note this new solid/shell element
is very attractive for thin shell analysis and may
quickly surpass the traditional thin shell elements.)
To perform nonlinear buckling analysis, one must
perturbate the model, i.e., take away the perfect
geometry or loading. In this example, small nodal
forces were added to make the loading nonsymmetrical. No material nonlinear was added; only large displacement (NLGEOM, ON) was used. For all models,
the mesh density was adequate and all other spurious
effects were eliminated to make clear the need for
using the ESURF command and the need for running
nonlinear static models. Only the first buckling mode
is presented.
Experimental Verification
A test rig was set up for the stainless steel cylinder
with 52 in. representing the span between the simple
supports. Pressure was slowly increased while deflection probes recorded displacement as a function of
pressure. The test specimen buckled at approximately
18.6 psi. (The test was cut short when permanent
deformation was present.) An initial uploading to 18
psi and back to zero showed no permanent deformation but review of the test displacement plot shows
significant nonlinearity at 18 psi. Note that nonlinearity
in deflection does not mean inelastic behavior.
The comparison between the test and the
SHELL63 nonlinear results is remarkable. There was
no deviation up to 18.6 psi. The small deviation
between the two curves is due to curve fitting
between converged finite element loads. Deflection
measurements were accurate only to about 0.030 in.
of deflection, the limit of the probes. The test datum
matched the nonlinear ANSYS run until the specimen
deflected enough to show an asymptotic approach to
the bifurcation point, adequate for most structural
analyses.
Comparison of Buckling Analysis Using ANSYS Shell Elements
Element type
ANSYS
version
Linear
eigenvalue
Linear eigenvalue
with ESURF
Nonlinear static Solution
(linear elastic material)
SHELL43
8.1
25.84
19.45
19.00
SHELL61
8.1
25.68
25.76
Not applicable
SHELL63
8.1
25.55
19.27
18.87
SHELL91
8.1
25.54
19.26
18.78
SHELL93
8.1
25.54
19.26
19.00
SHELL99
8.1
25.54
19.26
18.30
SHELL143
8.1
25.84
19.45
19.00
SHELL181
8.1
19.36
Not applicable
19.01
SOLSH190
9.0
19.38
Not applicable
18.99
www.ansys.com
ANSYS Solutions
|
Summer 2005
Large increments in solver load steps did not
cause any deviation from the test data. The accuracy
of the simple linear material model would be suspect
as the cylinder underwent plastic stress, but it was
all that was necessary to achieve accuracy within
3 percent.
19
A command mode example highlights the
use of surface effect elements:
Finite element solution showing two lobes
ET,1,SHELL63,,,,,,,,, ! A COMMON SHELL
ELEMENT EXAMPLE
R,1,0.035
ET,2,SURF154,,,, ! CREATE SURFACE EFFECT
ELEMENTS
Test specimen run to fully buckled solution
TYPE,2 ! SELECT THE 3-D SURFACE EFFECT
ELEMENTS
R,2,,,,,,,
Comparison of Nonlinear ANSYS Solutions
ESURF ! GENERATE ELEMENTS ON THE
EXISTING SELECTED ELEMENTS
ESEL,S,TYPE,,2 ! SELECT SURF154 TO APPLY
PRESSURE
Pressure (psi)
EMODIF,ALL,REAL,2
SFE,ALL,1,PRES,,10 ! APPLY 10 PSI TO THE
STRUCTURE
Inward Radial Displacement (in.)
Conclusion
Most of the legacy shell elements, when used alone,
will not be accurate in resolving eigenvalue buckling
for curved surfaces under pressure. It is recommended to use surface effect elements for pressure
application on elements not having pressure load stiffness capability according to the ANSYS Element
Reference manual. It is recommended to consider
using the linear eigenvalue solution only in a preliminary design phase and the nonlinear static solution for
a final design analysis.
All of the new shell elements will converge to
the eigenvector without the need for surface
effects elements. The nonlinear static solution
remains attractive, for it shows the path
dependence of force and displacement and
should confirm a proper eigenvalue solution.
Proper use of shell elements should produce
accurate eigenvalue results. And for any new
geometry, testing may be necessary for FE
verification, especially since some structures can
achieve full theoretical load-carrying capability,
while others may never approach the theoretical
solution. ■
Chuck Roche is a Structures Lead at Pratt & Whitney,
teaches ANSYS composites courses at Pratt & Whitney
Engineering Technical University and is an adjunct
professor at the University of Hartford and the University
of Connecticut.
References and Further Reading
F.B. Seely, J.O. Smith, “Advanced Mechanics of
Materials,” 2nd ed., Wiley & Sons, 1952
C.H. Roche, C.A. Whitney, “Robust Design of Load Bearing
Structures,” NAFEMS World Congress 2003, Orlando, FL
W. Flügge, “Stresses in Shells,” Springer-Verlag, Berlin, 1960
www.ansys.com
ANSYS Solutions
|
Summer 2005
CAE Community
20
Students boost output of micropump
design threefold using ANSYS structural,
CFD and optimization tools.
A team of engineering students at the University of
Washington in Seattle, Wash., won the fourth annual
College Design Engineering Award for their work in
refining the design of a micropump that may one day
circulate cooling fluid for electronic circuitry in
satellites and spacecraft as well as move small
volumes of fluids in chemical analyzers, medical
equipment and other applications where space is
limited and reliability is essential.
Sponsored by ANSYS, Inc. as part of its continuing support of engineering education, the award
was made based on project scope, engineering
problems encountered and solved, uniqueness of
solutions, potential for commercialization and impact
on the engineering community. Entries were judged
by editors of Design News magazine as part of the
publication’s Excellence in Design Achievement
Program. Funded by ANSYS, Inc, the award for the
winning project is $10,000 in cash to the student team
and a $10,000 scholarship grant for the school’s
engineering department.
Under the supervision of their advisor, Dr. Fred
Forster at the University of Washington, engineering
students Adrian Gamboa, Jone Chung and Chris
Morris worked on the project to maximize the output
of the micropump design incorporating unique
no-moving-part valves based on a concept originated
by famed inventor Nicola Tesla in 1920.
Instead of relying on conventional flap-type
check valves, the micropump uses an oscillating
membrane actuated by a piezoelectric driving element
to move fluid through inlet and outlet fixed-geometry
passive check valves. The valves are essentially a set
of two channels: a straight-line channel connecting
the pump chamber to the inlet/output and a curved
channel that loops around from the chamber to
www.ansys.com
intersect the straight channel. Fluid moves through the
valves based on the position of the membrane as it
vibrates. The shape of the channels and the angle at
which they intersect produce turbulence and restrict
the backflow. In effect, a differential pressure drop is
developed in one direction to create net flow in the
path of least resistance.
Because the valves have no moving parts, the
micropump offers distinct advantages over other
designs. The device is highly reliable, requiring no
repairs or replacement parts. The pumping system
handles any fluid, even air, which may make it useful
as a sampling device in medical and security systems.
Also, the pump is clog-resistant, even though valve
channels measure only 300 microns wide and 750
microns deep. Probably the greatest advantage of the
micropump is its small one-square-inch footprint,
providing fluid-moving capability where space is at
a premium.
A major challenge in developing a micropump for
practical applications has been increasing flow rate
and pressure to useful levels. Using water as a coolant,
the system was capable of dissipating 35 Watts of
heat with pump average power usage of only 100 mW
and a flow rate of 11 mL/minute.
The University of Washington students were able
to maximize pressure and flow rate using ANSYS
software. A finite element model of the membrane was
solved with ANSYS structural software to determine
the component’s deflection and resonant behavior.
Pump performance was simulated with ANSYS
FLOTRAN computational fluid dynamics (CFD)
software coupled to the Subproblem Approximation
Optimization Method included in ANSYS, which
iteratively computed pressure and flow for multiple
value shapes until the solution converged on an
ANSYS Solutions
|
Summer 2005
Previous Winners
The University of Washington team is the latest winner of
the annual College Design Engineering Award sponsored by
ANSYS, Inc. to recognize outstanding achievement in design
by engineering students.
2003
An engineering student team from Texas Christian University
designed an automated quality inspection system for checking
vacuum strength of saline bottles in the pharmaceutical industry.
2004
Students from the Massachusetts Institute of Technology
developed a projector that produces large, clear images from
inexpensive microfilm so people without ready access to
teaching materials can share information.
2002
The first award went to students from the University of
Evansville (Indiana) for designing an adaptive tricycle for a
disabled child unable to use a standard tricycle.
optimum. Using this CFD and optimization approach,
the students were able to increase pump output
threefold for both pressure and flow rate, thus taking
the device much closer to practical applications.
“Other people built and tested configurations to
try to optimize the general design of the Tesla valve.
I think we were first to use computational fluid
dynamics with an optimizer to determine the best
design,” said Adrian Gamboa. “Automation of the
process revealed a lot about what makes this valve
work and how we could maximize pump output.”
According to Gamboa, the team also may be the
first to stack micropumps connected in parallel to
increase flow rate. Made of flat plastic panels using
Micro-Electro-Mechanical System (MEMS) machining
fabrication techniques, the micropumps are readily
stackable, making the plastic devices well suited for
high-density cooling.
The team’s advisor, Dr. Forster, thinks Gamboa,
Morris and Chung did outstanding work in refining the
micropump design. “Beyond learning basic theory and
fundamentals, today’s engineering students absolutely
must know what analysis tools are available and how
to use them on a practical basis,” said Forster. “I’m
encouraged to see students confronting difficult
design challenges never before tackled. The talent,
know-how and innovative thinking of this bright new
generation represent the future of the engineering
profession and the key to successful designs in the
coming decades.” ■
21
A stacked array of four micropumps increases cooling flow to the
heat sink without adding to the footprint of the assembly.
CFD simulation with ANSYS FLOTRAN
shows high-velocity flow around the
sharp corner of the value channel.
www.ansys.com
Finite element analysis of one-quarter of
the membrane shows transverse displacement of the component.
The ANSYS optimization
routine enabled the student
team to develop an optimal
valve configuration for the
micropump.
Probably the greatest advantage of the
micropump is its small one-square-inch
footprint.
ANSYS Solutions
|
Summer 2005
CFD Update: What’s New in Computational Fluid Dynamics
22
Incoming flow from the pipeline is distributed by
a header into the six separation fingers of the slug
catcher. The lighter gas passes upward into a
gas header leading to an outlet while the heavier
liquid flows through the downcomers into the
liquid fingers.
Debottlenecking the
Hannibal Slug Catcher
ANSYS CFX simulates transient two-phase
flow in a complex pipe network.
When petroleum company BG Group commissioned
Genesis Oil and Gas Consultants to study the
feasibility of increasing gas production from their
Miskar concession off Tunisia, it became apparent that
a major bottleneck could be the slug catcher at the
end of the Miskar pipeline in the on-shore Hannibal
terminal.
The slug catcher receives a mixture of production
gas and liquid condensate from the pipeline into six
separation pipes or “fingers,” where gravity acts to
separate the two phases: the gas passing upward to
the gas outlet with the heavier liquid falling through
short downcomers into long liquid fingers, where it is
stored. If the surge of gas and liquid in the 20-minute
period after daily cleaning (“sphering”) of the pipeline
is too great, then liquid could overflow into the gas
outlet causing problems downstream. So Genesis
needed to know if the existing slug catcher has the
www.ansys.com
By Justin Penrose and Phil Stopford
ANSYS CFX Technical Services
ANSYS Europe
capacity to cope with higher flows of gas and liquid. If
not, then a new slug catcher would have to be built at
an estimated capital cost of $25 million.
To assess the capacity of the slug catcher,
Genesis asked ANSYS to simulate the timedependent two-phase fluid flow in the system using
the Eulerian multiphase flow model in ANSYS CFX
software. To forecast liquid overflow reliably, the CFD
model needed to track the motion of the gas-liquid
interface accurately throughout the pipe system.
Normally, the size of mesh required to do this would
make a long transient simulation intractable but, by
using transient mesh adaption to concentrate cells
into the region around the interface, we were able to
reduce the total mesh size by an order of magnitude.
The peak liquid levels were predicted for the
original pipeline gas/liquid flow rate and for a flow rate
increased by 45 percent. It was found that the
ANSYS Solutions
|
Summer 2005
23
Example of transient adaptive meshing to resolve the gas–liquid interface, where blue represents gas and red represents liquid.
Comparison of maximum liquid height for current and 45 percent higher flow rate. In both cases, no liquid overflow to the gas outlet
is predicted.
maximum height of the liquid in the separation fingers
increased significantly when the flow rate was
increased but no catastrophic overflow of liquid into
the gas outlet was found. As a result, it was concluded
that the existing slug catcher will be able to cope with
the increased pipeline capacity.
The innovative nature of this work was
recognized when the BG Tunisia team, including the
present authors, received an award under the BG
Group Chief Executive Innovation Awards Scheme for
2004. The submission was the winner in the “Alliance
with an External Party” category, one of only three
winning projects from over 100 entries submitted. ■
www.ansys.com
This view of the slug catcher at the BG Hannibal terminal shows the
inlet pipe and header as well as the short separation fingers above
the much longer liquid fingers.
ANSYS Solutions
|
Summer 2005
CFD Update: What’s New in Computational Fluid Dynamics
24
The frozen rotor CFD
model shown in red
utilizes interface area
matching (not to scale).
ANSYS CFX Advances
Centrifugal Stage Technology
Development
Software predicts overall performance of a centrifugal compressor
as well as individual components.
By Dr. Basuki N. Srivastava, Lead Engineer
General Electric Company, Aircraft Engines
Typical CFD model of the full geometry of a compressor (not to scale)
www.ansys.com
Secondary flow development in a diffuser
ANSYS Solutions
|
Summer 2005
Engineers at General Electric Company in Lynn, Mass.,
have made significant progress in demonstrating that
a steady-state model available in ANSYS CFX computational fluid dynamics software can predict not only
the overall performance but also individual component
performance of a centrifugal compressor. This can
be done using a mixing plane formulation of an
impeller-diffuser-deswirler centrifugal geometry. Many
investigators have computationally studied the aerodynamics of a centrifugal compressor stage and some
have concluded that unsteady-stage CFD is necessary
for centrifugal-stage predictions. However, GE has
determined that a cost-effective design method tuned
to provide component performance and consistency
with design cycle requirements is needed for the design
application engineers. Advanced CFD validation on
several challenging geometries has successfully been
completed, while validation on others is in progress.
This capability is currently being used to design an
advanced new generation of centrifugal stages.
Mixing plane formulation of the stage interface in
a centrifugal compressor offers an added advantage
of a computationally tractable CFD model (as compared to a frozen rotor model) that is very well-suited
to rapid turnaround of results for design iterations.
A typical speedline map comparison of the test
data with the CFX-TASCflow CFD model, based on
mixing plane concept for a similar centrifugal stage,
predicts the performance and loss coefficient.
Success of the steady CFD analysis using
ANSYS CFX for stage promises to be a considerable
savings in the cost of development for future generations of centrifugal stages by reducing the number of
rig tests and enhancing the current technology base.
This GE effort is a result of a team technical effort
by GE engineers coupled with a support team from
both ANSYS and GE Global companies. ■
Vanes
Deswirler
Diffuser
Transition
zone with
bleed flow
extraction
Impeller
Inlet
Mixing plane CFD model using one passage
Centrifugal stage performance and loss are predicted by ANSYS CFX software.
www.ansys.com
ANSYS Solutions
|
Summer 2005
25
CFD Update: What’s New in Computational Fluid Dynamics
26
Designing a Better Sports Car
Through Engineering Simulation
Aerodynamic analysis with ANSYS CFX fine-tunes the vehicle shape.
NUMA sports car
By Nuno Ricardo Rosmaninho
and António Gameiro Lopes
Department of Mechanical Engineering
University of Coimbra, Portugal
The study of aerodynamics in the automotive
industry is important to improve fuel economy
as well as vehicle comfort and safety. Since
the 1980s, automobile industries have relied
more and more on numerical methods for
vehicle design in order to reduce expensive
experimental tests traditionally required for
aerodynamic studies.
In the Department of Mechanical Engineering
at the University of Coimbra, Portugal, engineers have
been working on designing an innovative vehicle
that combines a sporty design with standard
transportation capability, the NUMA car. In order to
fine-tune the shape of the car, an aerodynamic
analysis was performed with ANSYS CFX computational fluid dynamics software.
Velocity vector field at the rear of the car
www.ansys.com
ANSYS Solutions
|
Summer 2005
27
Airflow past the rear airfoil
An inflated boundary of prismatic elements was
used near the car surface to improve spatial resolution and gain a better understanding of boundarylayer phenomena. An unstructured mesh with
tetrahedral elements was used for volume meshing.
Simulations were carried out with the SST turbulence
model, coupled with a blend factor of 0.5 for the
advection scheme.
The research started with three models with
characteristics of a sports car. The models differed in
the design of the rear end. Aerodynamic optimization
was used to determine the best configuration of
airfoils, spoilers and diffusers. Because standard
solutions for optimizing aerodynamics have been well
developed over the years, we simply studied the best
choices to apply. And, since the car was to be
designed for sporty performance, study of both drag
and lift were important. Little details on shape and
position of spoilers and airfoils played an important
role in the compromise between lift and drag.
The ANSYS CFX analyses provided information
on flow separation, pressure and velocity fields,
vortices and forces interacting with the vehicle. This
information allowed engineers to make modifications
Pressure coefficient results on the car symmetry plane
www.ansys.com
Evolution of CD and CL for different configurations of the
car geometry
in the car shape, producing better results. In the final
design, airfoils were positioned based on the visualization of the flow field in the rear of vehicle and on the
analysis of the corresponding effect in terms of drag
and lift. Pressure coefficient charts helped with the
definition of the overall geometry and gave insight on
the distribution of the applied forces.
Thanks to ANSYS CFX software, the university
could reduce the positive lift and still bring the drag to
95 percent of the initial value. ANSYS CFX proved to
be an excellent tool for automotive aerodynamics. ■
ANSYS CFX visualization of flow streamlines
ANSYS Solutions
|
Summer 2005
CFD Update: What’s New in Computational Fluid Dynamics
28
By Dr. John P. Abraham, Professor
University of St. Thomas, Minnesota
Minnesota Moving Ahead with
ANSYS in the Classroom
Two engineering schools revolutionize undergraduate instruction
with ANSYS Workbench and ANSYS CFX software.
Seven years ago, Dr. Ephraim Sparrow and I
developed a course at the University of Minnesota
called Case Studies in the Thermal Sciences. This
senior-level course utilized ANSYS software to
solve industry-related problems in the thermal
and fluid sciences. Students immediately recognized the value, and class enrollment swelled to
approximately 60 students per year.
www.ansys.com
In the early years, FLOTRAN was used to carry
out CFD solutions. Recently, however, ANSYS CFX
was adopted along with the Workbench environment
to greatly expand the scope of the problems that are
addressed during the semester. Since then, I began
teaching at the University of St. Thomas in St. Paul,
Minnesota, and have integrated a similar course into
that curriculum, and it’s consistently filled to capacity.
In the meantime, Dr. Sparrow and graduate students
Sandra Sparr and Jimmy Tong have continued to
develop the course, which is now taken by more than
150 students each year.
ANSYS Solutions
|
Summer 2005
29
While CFD is typically studied as a graduate-level
course, the ease of use and broad capability of
ANSYS CFX have enabled it to be brought into the
undergraduate classroom. Both Dr. Sparrow and I
believe in two prime benefits of exposing undergraduates to CFD. First, the experience with ANSYS CFX
deepens the students’ understanding of fluid flow
phenomena. In particular, the visualization capability
greatly enhances students’ intuition of flow behavior.
Second, ANSYS CFX opens a door to a new class of
problems that can be solved by undergraduates who
are no longer limited by the narrow range of classical
flow solutions.
The decision to use ANSYS CFX at both the
University of Minnesota and the University of
St. Thomas was made after a careful consideration of
all the leading CFD packages. It was concluded that
the ongoing integration of ANSYS CFX with the popular ANSYS software suite, the robust solver provided
by ANSYS CFX and the CAD capabilities of ANSYS
Workbench gave ANSYS CFX a clear advantage over
competing software alternatives.
Both schools are now teaching ANSYS CFX and
Workbench as part of a one-semester, 15-week
course that combines a theoretical basis with the
computational exercises. The courses are built around
industrial case studies that expose students to
increasingly complex flow situations as the semester
progresses. The incredible popularity of the course
has alleviated concerns about the difficulty of
introducing advanced CFD simulations at the
undergraduate level.
The success of ANSYS CFX at the undergraduate
level has not gone unnoticed. The Supercomputing
Institute for Digital Simulation and Advanced
Computation at the University of Minnesota has
recently acquired the ANSYS CFX program, and a
number of graduate students are currently using the
software for their doctoral research. ■
Using ANSYS CFX CFD software in the Workbench environment
allows students at the University of St. Thomas to solve a wider
range of problems. In this plot, streamlines represent air-flow
patterns through an array of heated tube banks. The results
indicate regions of separated flow (eddies) on the trailing edges
of the tubes. The streamline colors quantify the heating of the air
as it passes through the array.
Dr. John Abraham uses ANSYS CFX software in
his undergraduate CFD course.
www.ansys.com
ANSYS Solutions
|
Summer 2005
Simulation at Work
30
ANSYS Tools Help Develop
First-Generation Personal Rapid Transit Vehicles
Small computer-controlled cars
take passengers non-stop to
their destination via a network
of single-rail guideways.
Taxi 2000, Inc. was formed to develop and commercialize the concept of a personal rapid transit (PRT)
system emerging from work done at the University of
Minnesota. In contrast to mass transit systems that
move many people in large vehicles traveling successively from station to station along an established
route, SkyWeb Express saves considerable time with
smaller vehicles that travel directly to a destination
without intermediate stops.
Each electrically driven vehicle carries up to 650
pounds (typically three or four people) and travels on a
raised network of single-rail guideways connecting
conveniently located stations. The rider swipes a prepaid card through a stanchion in front of an empty
waiting vehicle, punches in a destination number, and
takes a seat. A computer system chooses the fastest
route to the destination and proceeds there non-stop.
Since stations are off the main line, riders go directly
from origin to destination, bypassing all the other
stations along the way. There are no stop signs, red
lights or interferences along the route.
www.ansys.com
Design Challenges
The company needed to get a working prototype system built as quickly as possible to demonstrate the
operation and feasibility of SkyWeb Express. Vehicle
weight was a key economic issue, so the cabin and
chassis were designed to be constructed of tubular
aluminum members bonded together with structural
adhesive, much the same as aircraft are fabricated.
Safety and reliability were primary concerns, so
components were required to be strong enough to
safely withstand operational loads.
Of particular concern was the bonded aluminum
T-joint connecting the cabin to the narrow chassis.
This part had to provide adequate support for turning
and wind-load forces as the vehicle rides along the
guideway, as well as repeated unbalanced load cycles
as passengers enter and exit the cabin.
The Simulation Solution
To determine if the bonded aluminum T-joint and
cabin structure were strong enough to withstand
these operational loads, the company turned to
the consulting firm Summit Analysis Inc. to study
component stresses.
ANSYS Solutions
|
Summer 2005
“ANSYS Structural software was valuable in
studying stress levels for this project, especially
the convenience and flexibility of the surfaceto-surface bonded contact feature. This functionality greatly facilitated modeling the structure and
provided a convenient mechanism for extracting
load values on each adhesive joint to verify that
forces were within acceptable limits. The speed
of this approach enabled Taxi 2000 to confirm the
performance expectations of the design so a
working prototype of the SkyWeb Express could
be completed within the time constraints of the
development program.”
ANSYS Structural was used to determine if stresses were within
acceptable limits for the adhesive-bonded tubular aluminum T-joint
connecting the cabin to the narrow vehicle chassis.
Working on a Dell Pentium 3 desktop PC,
Summit Analysis used ANSYS meshing capabilities to
construct a parametric finite element model of the
tubular vehicle chassis and overlapping gusset
sheets. ANSYS surface-to-surface contact elements
were used to represent the adhesive joining the
structural members. The bonded contact feature of
these elements allows both tangential and normal
forces to be transferred between adjacent surfaces.
Using these software features, the study enabled
engineers to develop a satisfactory design and quickly
verified that joint forces were within acceptable limits
for the shear and peel strengths for the adhesive.
Key Benefits
Summit Analysis used ANSYS contact elements to
quickly and accurately model the T-joint structures.
Mitchell Voehl
CEO and Engineering Consultant
Summit Analysis, Inc.
For such analyses, ANSYS contact technology has
useful capabilities, including automatic assembly
contact detection, robust default settings and a range
of contact behavior and solution options. Other FEA
packages may require extensive manual intervention
in setting up these contact conditions. ANSYS
software also can generate customized reports easily
with critical information summed, listed or plotted for
each joint.
Verification of the safety of the SkyWeb Express
structure enabled the company to construct a working
prototype of the SkyWeb Express on schedule. Since
then, more than 5,000 people have ridden the prototype PRT vehicle, with cities in North America, Europe,
the Middle East and Asia investigating the feasibility of
implementing the transit system. ■
Traveling on a network of raised single-rail guideways, computer-controlled SkyWeb Express personal
rapid transit (PRT) vehicles take people directly to their destination with no intermediate stops.
www.ansys.com
ANSYS Solutions
|
Summer 2005
31
Guest Commentary
CAE Moves to the
32
Forefront
of Product Development
Companies must meet challenges head on to reap
the tremendous benefits of digital simulation.
By Charles Foundyller
President
Daratech, Inc.
Computer-aided engineering
(CAE) is in the midst of a
renaissance. Once a highly
specialized and somewhat
under-appreciated field, CAE
is muscling its way to the
forefront of product development. Slashing costs is still
priority number one across
nearly all verticals in the
manufacturing industry, and
companies are scrutinizing
their processes to uncover
ways to save money. Applied
and understood properly, CAE can lead to better
models and, subsequently, less design rework. It can
allow designers and analysts to explore more design
alternatives and free them to spend more time on
design work and core engineering. CAE can reduce,
not replace, the number of physical prototypes,
notoriously costly and time exhaustive. And it can help
boost product quality, which builds brand loyalty, and
can help avoid program-killing warranty problems or
recalls down the road.
Digital simulation is not all peaches and cream,
however. Challenges abound. For example, integrating
simulation more fully into the product development
process requires overcoming hindrances to interoperability between the various design and analysis
application packages. No tool lives in a vacuum.
Interoperability and integration are key attributes
of any technology solution. Oftentimes, software
products aren’t even backward compatible with
www.ansys.com
themselves, let alone with other systems. The National
Institute of Standards and Technology has estimated
that the automotive industry throws away $1 billion
annually on interoperability-related expenses: an
astounding number.
Streamlining the Process Chain
The challenge of overcoming such technical issues
pales in comparison to confronting cultural, organizational and process issues, however. Companies
aiming to reap the tremendous benefits of CAE must
meet these formidable challenges head-on. Often this
involves taking a close look at the company’s product
development cycle from start to finish in evaluating
how digital simulation can best be integrated into the
process. In such an examination, companies often
find that their process chain resembles a cow path
that bends and wanders in a convoluted trail that
companies follow because things have been done
that way for decades.
Obliterating these inefficiencies and streamlining
the process chain to leverage simulation often is done
in conjunction with PLM to digitally create, store,
manage, share and reuse data. As CAE information is
increasingly relied upon in product development, so
must a place for it be carved out within — not around
— the PLM framework with capabilities for managing
multiple iterations of analysis data, correlating analysis
results with physical test data and associating
analysis and test data with CAD models in ways that
let digital and physical prototyping have broader and
deeper impact on product development.
ANSYS Solutions
|
Summer 2005
CUMULATIVE AUTOMOTIVE INDUSTRY SAVINGS
THROUGH REDUCED PHYSICAL TESTING
Bridging the Gap Between Design
and Analysis
In scrutinizing their product development processes to
improve efficiency, some manufacturers are finding
gains in and around the early product design phase.
In many organizations, a process gap exists between
the design and analysis teams whereby designers wait
until their work is finished before “throwing it over
the wall” to analysis, which in turn makes suggestions
and then throws it back over the wall to the designer
to rework. This serial process wastes time, and the
separation of design and analysis can create a
competitive atmosphere rather than fostering an
environment of shared goals.
Once technologically impractical, a new focus on
improved usability of analysis codes through wizards
and more simplified user interfaces, combined with
more powerful desktop computing power, have made
upfront analysis a viable option. Some organizations
are now seizing this opportunity to bridge the design
and analysis gap by having their designers perform
more of the analysis work and enabling the experts
to concentrate on the more complex issues. Mostly
cultural considerations and cemented processes
get in the way of such initiatives. Many companies
remain unconvinced that putting analysis on every
designer’s workstation is a good idea, thinking that
misinterpretation of the results by people who
don’t know how to question the inputs and outputs
may be too great a risk.
While many OEMs acknowledge the merits of an
upfront analysis strategy, some remain hesitant to
make this leap. Many Tier 1 and 2 suppliers, on the
other hand, have more readily embraced the strategy
as a way to survive and are leading the way in
developing approaches for best implementing digital
simulation early in the development process.
Test and Analysis Must Get Along
In many organizations, there is a perception that
analysis will replace test. But most agree that we’ll
never get to a point where you’ll be able to analyze
straight through to production. Moreover, companies
must continue to rely heavily on physical testing to
study competitive products.
The real issue is improving the interface between
the analysis and test organizations to create the most
efficient process possible, to get the best and
maximum amount of information possible out of every
COW PATH PROCESS CHAIN
Source: Daratech, Inc.
In integrating digital simulation into the product development
cycle, companies must streamline process chains that often
resemble a convoluted trail followed because things have been
done that way for decades.
www.ansys.com
33
Source: Daratech, Inc.
According to Daratech, automotive OEMs and their suppliers
will save $2 billion on a cumulative basis by 2010 by reducing
physical prototypes through greater use of digital simulation.
turn. Clearly, there will always be testing, but
undeniably more and more performance verification
will be carried out using virtual models and simulation.
The resulting reduction in physical prototypes will
save companies billions. Indeed, Daratech’s model of
the savings digital simulation can yield at automotive
OEMs and their suppliers indicates that, by 2010, the
automotive industry will have saved on the order of
$2 billion on a cumulative basis.
Where Do We Go from Here?
With each successful program comes a higher level of
confidence in digital simulation. While physical test
results are still typically seen as the data of record,
many organizations are beginning to second-guess
this long held and once unchallengeable position.
Today it is difficult to keep CAE on the back
burner due to the maturity and refinement of the
software; the ability to simulate multiple scenarios
rather than build multiple prototypes; the speed
afforded by new high-performance computing
architectures; the ability to better correlate models
with physical test results; and the expanding body of
evidence before us... the cars we drive, the planes we
fly, the tractors we ride, the defense systems that
protect us and on and on.
While manufacturers formerly could design
products such as these by the seat of their pants
without much simulation, today it is virtually
impossible to satisfy all of the technical requirements
of product development and remain competitive
without a thorough and comprehensive analysis
program. Simply put, engineering simulation is no
longer an option but a necessity. And for that reason,
the use of analysis tools will undoubtedly accelerate
in the coming years, propelling companies that
wisely implement the technology to leading positions
in their industries. ■
Charles Foundyller is founder and president of Daratech, Inc.,
a market research and technology assessment firm
specializing in CAD, CAM, CAE, PLM and related areas.
The company hosts a variety of user forums and technology
workshops throughout the year and publishes a range of
studies, newsletters, market reports, sourcebooks and
industry statistics. For more information on the company, call
Daratech at 617.354.2339 or visit www.daratech.com.
ANSYS Solutions
|
Summer 2005
Tech File
34
Meshing in
Workbench Simulation
Part 2 of 3:
Using adaptive meshing tools to converge on a solution
By John Crawford
Consulting Analyst
As much as we might not want
to think about it, there are errors
in our work. One source of error
is inadequate mesh refinement.
A coarse mesh can underpredict deflections as well as
stresses, and this is of particular
concern because it is nonconservative.
An experienced and conscientious analyst will try
to estimate meshing errors by comparing averaged
and unaveraged nodal results, plotting the estimated
error, and using experience and insight to review the
results and determine whether the mesh is adequately
refined to provide the required accuracy. If the mesh
isn’t satisfactory, it would be necessary to remesh the
model, reapply boundary conditions, run the solution
again and review the results.
All of this can be labor-intensive and timeconsuming, especially if several iterations are required
to get an error that has been reduced to an acceptable
level. Is there a better way to do this? If you’re using
ANSYS Workbench, the answer is “yes.”
www.ansys.com
Origins of Meshing Errors
Meshing errors can have several origins. Any time you
have a sharp inside corner (often called a re-entrant
corner), you have an infinite stress concentration, or
singularity. While the stress at a singularity is theoretically infinite, finite element analysis will always return a
non-infinite stress result at these locations. Depending
on the mesh density at the singularity, the calculated
stress may be too high or too low. But no matter what
the value is, it cannot be trusted.
Another source of meshing error is a mesh that is
insufficiently refined to give good resolution. To get
some idea of what the meshing error is, compare the
averaged nodal result to the unaveraged nodal result.
The greater the differences between them, the greater
both of them are from the true answer. ANSYS
software provides an estimate of the meshing error
with the PLES,SDSG command. However, even when
we have some idea where the mesh is unsatisfactory,
we don’t have an easy way of remeshing the model
and rerunning the analysis.
ANSYS Solutions
|
Summer 2005
35
Converge on a Solution with
Adaptive Meshing
Workbench introduces the ability to automatically
remesh and resolve a structural or heat transfer
analysis until a predetermined criterion has been met.
You can use global adaptivity to refine the whole
model, or scoped adaptivity to restrict adaptive
meshing to regions of particular interest. For structural
problems, adaptive meshing can be used to converge
on a solution for stress results like von Mises stress,
displacements and even natural frequencies, while
in heat transfer applications it can converge on
temperatures and heat flux.
Using a structural problem as an example, here’s
how we might go about using adaptive meshing.
Begin by creating a model and running an analysis.
Review the results and see if you can identify regions
where you might want to use adaptive meshing to
refine the mesh and improve the quality of the results.
If you are interested in using a stress value like von
Mises stress as the criteria for meshing refinement,
insert an error branch beneath it in the Outline Tree
and then select the error branch to display an estimate
of the error in von Mises stress that is caused
by meshing.
After reviewing the results, decide if you want to
use global adaptivity to refine the entire model, or use
scoping to restrict adaptive meshing to certain
regions. You may not want to refine the mesh everywhere in the model because only certain areas may
be of interest, or you may want to avoid singularities.
It is important to avoid using stress-based adaptive
meshing on singularities because it is impossible to
achieve convergence in these regions. Scoping is
done by picking specific features like an edge or a
face, and then inserting a convergence branch
beneath the result you would like to use for adaptive
meshing. You can insert a convergence branch under
several result entities, and Workbench will continue to
refine the mesh until all the desired branches have
been meshed satisfactorily.
Setting Controls, Convergence Criteria
and Refinement Loops
Once you have picked a result or results that adaptive
meshing will be based upon, go to the Detail View window and set controls for each of them. It’s necessary to
define whether you want to converge on a minimum or
maximum value because you may want to converge on
the maximum value of maximum principal stress, or the
minimum value of minimum principal stress.
www.ansys.com
Also, set the amount of change between successive solutions that will be your criteria for successful
convergence. It’s important to remember that mesh
refinement is based upon the idea that, as the mesh
density is increased, the result value that convergence
is based upon will asymptotically approach the target
answer. Workbench compares the results from the
previous solution to the results from the current solution. If they differ by an amount that is less than or
equal to the convergence criteria, then convergence
has been achieved and the solution will be stopped.
Finally, you should set the maximum number of
refinement loops that will be allowed. For a wellbehaved analysis, it is common for convergence to
be achieved after three or four refinement loops, but
if a singularity is present, it will be impossible for
convergence to be achieved no matter how many
loops are done.
Getting Results
When you have defined all the regions where adaptive
meshing is to take place and set their convergence
controls, you right-click on the Solution branch in the
Outline Tree window and pick Solve. Workbench will
begin running through adaptive meshing loops and
will continue until mesh refinement has satisfied all of
the convergence criteria or the maximum number of
refinement loops has been met.
When the solution has been completed, you
simply left-click on the result of interest and view the
results from the last refinement loop. You can learn
more about the convergence solution by left-clicking
on the convergence branch. This will display a chart
that shows the convergence history of the solutions,
a table that lists the number of nodes and elements
that were used in each refinement loop, the value of
the result that adaptive meshing was being applied
to and the percent change in this result between the
last solution and the previous solution.
The ability to automatically refine a mesh is a very
powerful capability that has been needed for a
very long time. Workbench makes this possible in a
manner that is relatively easy once you get the hang of
it. Pretty soon you’ll be using it for all the analyses you
do in Workbench. ■
Workbench users can find more detailed information on
adaptive meshing in the Tips and Tricks section of the new
Workbench community area now available from the ANSYS
Customer Portal: www1.ansys.com/customer
ANSYS Solutions
|
Summer 2005
Tips and Techniques
36
Modal Analyses
of Models with Friction
By ANSYS, Inc.
Technical Support
Brake squeal and other complex friction problems are readily solved
using an unsymmetric stiffness matrix supported by QRDAMP.
The last several releases of ANSYS software have
significant enhancements to the QR damped
eigenvalue extraction method, which now supports
unsymmetric stiffness [K] matrices, output of complex
eigenvectors and use of the PSOLVE command.
Support of unsymmetric stiffness matrices makes
QRDAMP particularly useful in solving complex
problems such as brake squeal, where frictional
effects can introduce an unsymmetric [K] term.
Keep in mind that by simply defining friction,
unsymmetric [K] matrices may not be present. First of
all, unsymmetric [K] must be requested with the
NROPT,UNSYM command. Second, sliding must
occur for the matrices to be unsymmetric.
QRDAMP Background
In the QRDAMP eigenvalue extraction method, the
undamped eigenvalues and eigenvectors based on
the symmetric [K] matrix are calculated first. Damping
and the unsymmetric [K] portion are then included in
the modal equations.
To perform a modal analysis, including frictional
effects, follow these steps:
1. Perform a nonlinear, large-deflection analysis
(NLGEOM,ON) with the unsymmetric NewtonRaphson method (NROPT,UNSYM) and
pre-stress effects on (PSTRES,ON). The first
two commands allow calculation of unsymmetric [K] of the contact elements, and
the third command informs the program
that pre-stress effects will be included
in a subsequent modal analysis. Also,
EMATWRITE,YES is usually required to force
ANSYS to write the .emat file, needed in the
modal analysis phase.
www.ansys.com
2. Post-process the results after the static
solution. Specifically, plotting or listing the
contact status (PLESOL/PRESOL,CONT,STAT)
allows you to determine if there are any areas in
sliding (contact status = 2). If no contact
elements have a contact status of “2”
(“Sliding”), then no unsymmetric [K] term will
be present.
3. Re-enter the solution processor and specify
a modal analysis with the QR damped eigenvalue extraction method (MODOPT,QRDAMP).
Complex eigenvector output also can be
requested at this point. Specify PSTRES,ON to
include the pre-stress effects calculated
in step 1. Then, instead of performing a
regular solution, run a partial solution with
PSOLVE,EIGQRDA and expand the modes with
PSOLVE,EIGEXP. A partial solution is required
because a regular solution will automatically
re-create the matrices, but in this situation,
the matrices (including the unsymmetric [K])
calculated from step 1) need to be used. The
steps outlined here will allow you to have
ANSYS calculate the unsymmetric [K] terms
due to frictional sliding, then include the
unsymmetric matrices in the eigensolution.
The eigenvalues (i.e., frequencies) will have
real and imaginary parts if damping [C] and/or an
unsymmetric [K] matrix are present. The imaginary
component reflects the damped frequency. The real
component indicates whether or not the mode is
stable — unstable modes will have a large, positive
real eigenvalue.
The eigenvector also will be complex in either case.
The real and imaginary eigenvectors represent the
“motion” of the mode shape — if the imaginary eigenvector is non-zero, this means that a phase difference is
present, analogous to harmonic analysis output.
ANSYS Solutions
|
Summer 2005
Example Problem
As an example for comparing the MATRIX27 and
CONTA17x methods, a simple disc model shown in
the accompanying diagram was analyzed. The internal
radius is constrained in all DOF, and two contacting
areas are present, modeled with surface-to-surface
contact elements TARGE169 and CONTA171.
Disc model was analyzed for comparing the MATRIX27
and CONTA178 methods.
A nonlinear static analysis is run first with the
unsymmetric option and pre-stress effects on. Angular
velocity and acceleration are used to apply a normal
and tangential force component on the contact points.
Examination of the contact output shows that these
two areas are in contact and sliding. A modal analysis
is then performed, using the unsymmetric matrices
calculated in the static phase.
With the static coefficient of friction as 0.5 for
this model, one finds that the unsymmetric [K]
term couples the ninth and tenth modes into an
unstable mode.
The effect of the coefficient of friction can be
studied by running multiple iterations and varying in a probabilistic analysis initiated using the coefficient
of friction as a uniformly distributed random
variable, while the complex eigenvalues were the
response variables.
Imaginary vs. real eigenvalues for modes 9 and
10 are shown on the left plots, where the coupling of
the modes and the unstable frequencies are apparent.
If the points lie on the axis, these indicate stable
modes with nodes in phase with each other (such
as a regular undamped case). Points lying in the negative half of the real eigenvalue axis indicate stable
modes (another example of negative real is due to
damping). However, points on the positive x-axis show
unstable modes.
Mode 9 shows stable response (no damping is
present in this example) between 453 and 456 Hz.
Above that, the modes become unstable. Conversely,
for mode 10 the response is stable from about 456.5
to 459 Hz. The coupling and instability below this
range are apparent from the graph.
Real eigenvalues vs. coefficient of friction (varied
between 0 and 1) are plotted in the right graphs. It is
interesting to note how the stability (real eigenvalue) is
affected by , as shown on the left near a value of
0.45, where both modes 9 and 10 converge and
become unstable. ■
Contact ANSYS Technical Support engineer Sheldon Imaoka
([email protected]) for the entire paper on the
QRDAMP eigenvalue extraction method from which this
article was excerpted.
Left plots show imaginary vs. real eigenvalues for modes 9 and 10. Real eigenvalues vs. coefficient
of friction are plotted in the right graphs.
www.ansys.com
ANSYS Solutions
|
Summer 2005
37
Hardware Update
38
AMD brings 64-bit computing into the mainstream of engineering simulation.
By Dan Williams, ANSYS, Inc.
David Cownie and Dmitriy Safro, Advanced Micro Devices, Inc.
AMD Opteron™ processors provide an easy upgrade
route to 64-bit computing. Memory bottlenecks that
limited performance in legacy x86 machines are eliminated. Low system cost and power requirements now
bring 64-bit processing into the mainstream. In tandem,
ANSYS CFX software now exploits the benefits of inexpensive and powerful hardware, combining increased
simulation performance and modest compute platform
ownership costs.
In this article, we compare performance of ANSYS
CFX-5.7 and ANSYS CFX 10.0 on Intel Xeon and AMD
Opteron processors. CFX now exploits the 64-bit features of the Opteron and provides excellent scaling on
multi-processor compute nodes and clusters. The
Opteron processor’s Direct Connect Architecture yields
super linear speedups in marked contrast to the limitations of legacy shared bus SMP nodes. Results are
shown for the latest Opteron “Dual-Core” processors
that fit in the same sockets but provide two complete
processors on a single die. Results are also provided
showing performance gains from upgrading a twosocket machine to dual core processors.
64-bit Computing with x86 Binary
Compatibility
The AMD Opteron processor provides 64-bit extensions to the x86 architecture. It can run traditional x86
binary images and also codes recompiled to exploit 64bit extensions, and it allows the ability to address large
amounts of memory. This overcomes the limitations of
32-bit addressing, which is restricted to less than 4
Gbytes of memory. Both 32- and 64-bit images can run
concurrently on a single Opteron processor. In 64-bit
mode, Opteron processors provide more registers to
boost performance. ANSYS CFX has a 64-bit version
that exploits these new features to boost performance.
How Tests Were Run
AMD Opteron processor model numbering uses three
digits: XYY, where X is the number of processors that a
single system can scale up to with that particular
Opteron part based on the number of HyperTransport
links; and YY is the performance metric of the part.
Hence, an Opteron processor Model 250 is a 2.4 GHz
part that can scale to two-way. An Opteron processor
www.ansys.com
Model 248 is a 2.2 GHz part. An AMD Opteron 842 is
a 1.6 GHz part, which could be used in up to an
eight-way single system configuration. Today, AMD
Opteron 800 series parts are generally used in four-way
systems, but eight-way systems should be available in
the near future.
The test machine used in this analysis consisted of
a two-way 1U AMD Opteron Model 250 (2.4 GHz)
processor-based server with 8 GB of 400 MHz DDR
memory. The operating system used was Novell’s SLES
9 (x86-64 with 2.6 Linux kernel).
The second test machine configuration was a fourway 1U AMD Opteron Model 248 (2.2 GHz) processorbased server with 32GB of 333 MHz DDR memory. The
operating system was Red Hat AS 3.0 x86-64. The
“Smith” cluster has 32 of these four-way servers with a
total of 128 AMD Opteron processors and one Tbyte of
memory connected with gigabit Ethernet and Myrinet
interconnects.
Each AMD Opteron processor has an on-chip
memory controller and is directly connected to memory.
In a multi-processor system design, each processor
has direct attached memory that can be accessed by
any of the processors in the node through the integrated HyperTransport™ technology, which is part
of the AMD64 architecture design.
Table 1 describes the size of each test case. Single
node results refer to the three different problem sizes:
small (100k), medium (200k) and large (400k). Larger
problems (LeMans Car and OHS RC Aircraft) were run
for cluster scaling tests.
Table 1: Benchmark Test Cases
Test Case
Name
Number of
Nodes
Number of Number of Grid Number of
Elements
Interfaces
Domains
Static Mixer
100k (small)
92,993
517,145
0
1
Static Mixer
200k (medium)
184,450
1,037,239
0
1
Static Mixer
400k (large)
370,162
2,091,423
0
1
LeMans Car
1,861,696
10,285,401
0
1
OHS RC Aircraft
2,954,574
14,369,304
3
2
ANSYS Solutions
|
Summer 2005
64-bit
Single Processor Performance
Dual-Core Processor Performance
Performance is normalized to give speed relative to the
P4 Xeon (3.0 GHz) for both ANSYS CFX-5.7 and
ANSYS CFX 10.0. CFX 10.0 runs on the AMD Opteron
processor use the 64-bit executable. All CFX-5.7 runs
use the 32-bit executable. In serial, the AMD Opteron
250 CPU/memory combo is twice the speed of a 3.0
GHz P4 Xeon, and 33 percent faster than a 3.4 GHz P4
Xeon, as shown in Chart 1.
AMD Opteron 250 (2.4 GHz) processors used in
the previous tests are fabricated using a 130-nm
process. AMD recently introduced “Dual-Core” Opteron
processors fabricated on the new 90-nm process that
allows two entire processors to be placed on a singledie. They are plug-compatible with existing singlecore 90-nm AMD Opteron products. Test results in
Chart 3 show the performance gains possible when the
original two-socket machine is upgraded to the new
dual-core processors so that the machine then has
four processors.
Chart 1: Medium Problem Performance
Chart 3: Comparing Opteron 250 (2.4 GHz) and Opteron
875 (2.2 GHz Dual-Core) Performance
More Registers Make a Difference
With AMD Opteron processors, more registers are available in the 64-bit mode. The extra register typically
gives a speed improvement of about 15 percent. The
extra registers more than compensate for longer pointer
lengths and increased memory bandwidth needed to
handle these. Results are shown in Chart 2.
Chart 2: Comparing 32 and 64-bit Performance
on Opteron 250 (2.4 GHz)
www.ansys.com
Single Node Scaling
Chart 3 shows the results from several runs on both
single- and dual-core, two-socket AMD Opteron
processor-based systems. In all cases, scaling is
excellent, especially if the individual ANSYS CFX
processes fit inside one CPU/memory package. The
results clearly indicate one benefit of the on-chip memory controller and dual processor NUMA architecture.
Superlinear scaling is achieved in both the single-core
and dual-core two-way benchmarks. There are two
possible causes of superlinear scaling: poor scheduling
on the single-processor case between processors
reduces performance, while two-processor cases
benefit from explicit decomposition of the problem,
which improves the overall cache and local memory hit
rates. The net effect is a significant increase in the
effective memory bandwidth and a decrease in the
latency.
When individual processes do not fit in a CPU/
memory package, say for the four-way dual-core run
(speedup of 3.6), respectable results are still achieved.
Non-uniformity in the results is due to individual
processor memory being scheduled by the operating
system such that it is moved away from local memory.
In this case, memory access must be done through the
HyperTransport link, which has more latency than
accessing memory on the local CPU/memory package.
ANSYS Solutions
|
Summer 2005
39
Hardware Update
40
To ensure the best possible performance on an
AMD Opteron system, you should run a Linux operating
system with kernel support for the Linux command task
set. This kernel feature can control locking a process to
a CPU (or set of CPUs), which allows optimal process
scheduling and memory allocation on multi-processor
compute nodes. For example, SuSE Linux Enterprise
server release (SLES 9) uses version 2.6 of the
Linux kernel, which has support for the Linux command
task set.
There is scope to optimize the communication
package to use these OS hooks to tie a process to a
preset CPU/Memory combo. It is possible to use
one of several communication APIs with ANSYS CFX
including PVM, MPICH-1.2.5 and Hewlett-Packard
MPI, so to take advantage of such operating systemdependent controls will ultimately require support by
the API authors.
With the latest Linux kernels, scaling can be highly
superlinear when processes get allocated within one
CPU/Memory package per process. However, even
when running Linux versions without the latest Linux
kernel, scaling is still very good: much better than any
multi-processor P4 Xeon offering tested so far by AMD.
Cluster Scaling
Clusters can be configured either with fast, low-latency
interconnects, or nodes can simply be connected via
Gigabit Ethernet and a commodity Ethernet switch.
Fast interconnect fabrics (like Myrinet or InfiniBand) offer
better scaling but are generally more expensive. The
AMD developer center “Smith” cluster has 32 compute
nodes, each with 32 Gbytes of memory and four AMD
Opteron Model 848 (2.2 GHz) processors. Each node is
connected with both Gigabit Ethernet and a single
Myrinet card. Test results are shown in Chart 4.
Chart 4: ANSYS CFX 10.0 (64-bit) Performance on
AMD Opteron Model 848 (2.2 GHz)
www.ansys.com
...scaling can be highly
superlinear when
processes get allocated
within one CPU/Memory
package per process.
Tests with these clusters indicated that Myrinet
scales only a bit better than PVM (message passing
over gigE) in the equation assembly and linear solve
portion of the solution. This is because of low communication overhead in this part of the solution. However,
Myrinet speed is far better for communication-intensive
parts of the solution. In ANSYS CFX, this shows up
especially during file I/O. Startup/Shutdown times are
twice as fast using the Myrinet interface than one
Gigabit Ethernet with PVM.
A cluster configured with a parallel file system (like
Lustre or Panasas) was not tested. But in I/O-intensive
jobs, these could show a significant speedup over
simply cross-mounting a single disc via NFS to every
node. Also, note that Serial ATA disc arrays now are
relatively inexpensive, offering speeds closer to SCSI
discs at prices comparable to IDE drives.
Conclusions
On serial code tests, the two-way AMD Opteron Model
250 (2.4 GHz) processors with 400 MHz memory
outperform the P4 Xeon (3.0 GHz) by 100 percent.
Scaling is excellent with superlinear speedups measured when running parallel code within a compute
node. Good Linux distributions are available from both
SuSE and Red Hat that support x86-64.
AMD Opteron dual-core processors can upgrade
a traditional two-processor machine to four processors. On the single-node test cases, ANSYS CFX runs
on the dual-core two-socket machine almost as fast as
a four-socket four-processor machine. ■
ANSYS Solutions
|
Summer 2005
www.ansys.com
ANSYS Solutions
|
Summer 2005
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement