Spotlight on Trends and Issues:

Robust Design technology

improves designs by reducing
their sensitivity to real-world
variabilities

Create high-quality shell

models with the ANSYS
Advanced Mesher suite

Simulation provides insight

into critical operating
parameters of this emerging
class of MEMS devices
 Contents
Spotlight on Trends and Issues Departments
6 Meeting the Challenges
Editorial
of Global Product Harnessing the Power of Technical Talent ............. 2
Development
A continuing series on the role of Industry News
engineering simulation in product Recent announcements ................................................................ 4
development processes CFD Update
Computational Fluid Dynamics ............................................ 20
Simulation at Work
Features Fewer prototypes with up-front analysis ................. 27
Managing CAE Processes
10 Developing Robust The value of simulation-driven design ........................... 29
Designs
Improve designs by reducing Tech File
their sensitivity to real-world Restarting ANSYS ................................................................................. 33
variabilities Tips and Techniques
More ways to reduce run times ......................................... 35

13 Midsurfacing Tools for Guest Commentary

The next generation of innovators 37
Meshing Complex ...................................

Geometries Hardware Update

Create high-quality shell models High-performance cluster development ................... 39

16 Finite Element Modeling

About the cover
of Capacitive Micro-
Our cover article describes
machined Ultrasonic how the role of simulation-
Transducers driven design in developing
Analysis provides insight into products on a worldwide
critical operating parameters of scale is expanding.
this emerging class of MEMS
devices

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 Designers Ad Sales Manager Editorial Advisor
John Krouse Miller Creative Group Beth Mazurak Kelly Wall
jkrouse@compuserve.com info@millercreativegroup.com beth.mazurak@ansys.com kelly.wall@ansys.com

Managing Editor Art Director Circulation Manager CFD Update Advisor

Jennifer Hucko Dan Hart Elaine Travers Chris Reeves
jennifer.hucko@ansys.com dan.hart@ansys.com elaine.travers@ansys.com chris.reeves@ansys.com

ANSYS Solutions is published for ANSYS, Inc. customers, partners, and others interested in the field of design and analysis applications.
The content of ANSYS Solutions has been carefully reviewed and is deemed to be accurate and complete. However, 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 ANSYS Solutions | Winter 2005
 Editorial
2
Harnessing the True Power of Technical Talent
Get whopping business gains by leveraging corporate intellectual
capital with simulation-driven design.
Simulation technology continues
to advance, giving engineers
and analysts increasingly
powerful tools for a wide range
of analysis tasks. The ANSYS
Advanced Mesher suite, for
example, has unparalleled
functionality for quickly building
models, as explained in Simon
Pereira’s article “Midsurfacing
By John Krouse Tools for Meshing Complex
Editorial Director Geometries” starting on page
ANSYS Solutions 13 of this issue. The DesignX-
jkrouse@compuserve.com plorerTM software can be used
in efficiently optimizing designs
to minimize failures due to uncertainties such as manufac-
turing variabilities, as Ray Browell discusses in his article
“Developing Robust Designs” on page 10.
These are part of a spectrum of advanced technolo-
gies that serve as the foundation for performing simula-
tion faster and more accurately than was previously
practical. In many cases, they automate and streamline
manual tasks that drag down engineering productivity
when users have to slog through these procedures by
hand. In this respect, analysis tools are extremely
powerful time savers and cost cutters that yield very
respectable returns on investment. Beyond these
productivity benefits, however, lie much broader order-of-
magnitude gains made possible by using the technology
to develop better, more innovative products that enhance
brand value and grow top-line revenue.
Xerox Corp., for example, used up-front analysis
with DesignSpace in the development of their new iGen3,
an innovative digital printing system for penetrating the
commercial printing market and the most complex
product ever developed by the company. As described in
the article “Fewer Prototypes with Up-Front Analysis” on
page 27, the machine is regarded as one of the firm’s
flagship products and is projected to be a major revenue-
generator to lift profits and strengthen the company’s
market leadership position.
Products such as this that are developed through
simulation-driven design aren’t necessarily successful
because engineers worked quickly but because they
worked intelligently, using their engineering know-how
with the aid of simulation to study product performance,
spot potential problems, evaluate alternatives and refine
www.ansys.com
the design. In this manner, companies can use early simu-
lation in achieving product performance targets that incor-
porate “voice of the customer” market requirements into
the design process, as discussed by Dr. Jason Lemon in
his article “The Value of Simulation-Driven Design” begin-
ning on page 29. Lemon is founder and chairman of
ANSYS partner ITI, which has used the approach in
helping numerous companies over the years to develop
innovative winning products. They worked with Murray, Inc.,
for example, in using ANSYS to develop an innovative
tractor mower with all-wheel steering capability that turns
easily without the usual hydraulic power-assist systems
that add complexity and cost to a product.
Organizations around the world are using simulation-
driven design processes to compete more effectively in
global markets, with the role of technology shifting to a
means of product development process change and a
critical part of companies’ business strategies. See the
article “Meeting the Challenges of Global Product
Development” on page 6 to learn how this expanding role
of simulation-driven design is helping companies develop
products on a worldwide scale.
The role of simulation at companies competing in
this global market is addressed in the guest commentary
“The Next Generation of Innovators” by Dr. Michael Lovell
on page 37, which discusses the challenges faced by
engineers when competing with those in countries where
labor costs are considerably lower. Lovell describes how
engineers can bring considerable value to the workplace
beyond their technical skills, most notably in using simula-
tion to develop truly innovative products that enhance the
competitive positions of their companies.
The articles cited above highlight profound corpo-
rate benefits that are possible with simulation-driven
design, which enables engineers to do what they do best:
creatively design innovative, winning products. After all,
the greatest worth of engineers and other technical
professionals is not that they perform individual tasks
faster or cheaper than their counterparts at other
companies. Rather, their real value to the enterprise is the
tremendous inventiveness, knowledge, artistry and
expertise they draw upon in product development. By
harnessing the power of this valuable intellectual capital,
simulation can play a key role in helping companies thrive
amid the challenges and opportunities of today’s global
economy.

ANSYS Solutions | Winter 2005
Take a look at the future of
product development...
...a process that's more automated, more integrated, more
innovative and truer to life. That's where ANSYS is taking
engineering simulation. ANSYS 9.0 combines technologies
like industry-leading meshing, nonlinear analysis and
computational fluid dynamics, reducing costs and
bringing innovative products to market quicker.

Bring your products and processes to life with ANSYS.

Visit www.ansys.com/innovate/sol or call 1.866.267.9724 for
more information about ANSYS 9.0.
AN-Solutions-INDUSRTYNEWS 4/13/05 2:02 PM Page 1
Industry News
Recent Announcements
4 and Upcoming Events
ANSYS Awarded General Services Administration
Contract for CFX Products
The U.S. General Services Administration (GSA) has
awarded ANSYS a contract to sell its CFX® products
and services, joining the other ANSYS tools available
to Federal customers through the GSA program. The
contract augments ANSYS’ ability to provide a full
range of products to Federal customers enabling them
to quickly and effectively purchase the products and
services that they need using pre-established rates.
Federal customers can access information about
ANSYS products and services through the GSA
Schedules e-Library at www.gsalibrary.gsa.gov by
searching the contract number GS-35F-0639N.
ANSYS Northern California User’s Group Meeting
The Northern California ANSYS User’s Group meeting
will be held on May 26 & 27, 2005, at the Santa Clara
Marriott Hotel in Santa Clara, Calif. There will be
workshops on Ball Grid Array, MEMS, Fluid Structure
Interaction, CFD Fundamentals and ANSYS CFX,
Fracture Mechanics and Fatigue simulations
using ANSYS. In addition, users are encouraged
to submit and present papers on biomedical,
semiconductor, high performance computing and
general simulations applications. Exhibition space is
also available. For further information, please visit
www.ozeninc.com/ugmeeting.asp or contact
(408) 732-4665 and info@ozeninc.com.
ANSYS Acquires Century Dynamics
In January, ANSYS, Inc. acquired Century Dynamics,
Inc., a leading provider of sophisticated yet easy-to-
use simulation software for solving linear, nonlinear,
explicit and multi-body hydro-dynamics problems.
Century Dynamics' main product, AUTODYN®, can
solve many types of problems with its extensive solver
library. Solvers include computational structural
dynamics finite element solvers (FE),finite volume
solvers for fluid dynamics (CFD), mesh-free particle
solvers for high velocity, large deformation and frag-
mentation problems (SPH), and multi-solver coupling
for multiphysics solutions including coupling between
FE, CFD and SPH methods.
www.ansys.com
The acquisition of Century Dynamics will support
ANSYS' continued global expansion and will provide
new opportunities for both organizations to further
develop breakthrough and innovative CAD/CAE solu-
tions that accelerate the application of CAE simulation
results to design decisions. Century Dynamics
will continue to operate as a separate subsidiary,
maintaining its products, sales channels and other
partner relationships. For more information visit
www.ansys.com.
Durability & Fatigue Seminars Exploring the
Modern Theories and Practical Application of
Fatigue and Durability
Noted fatigue and durability authority, professor John
Draper of Safe Technology Ltd. (United Kingdom)
will present a series of fatigue and durability seminars.
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. For more infor-
mation, please visit www.safetechnology.com.
Only Pittsburgh CEO to Ring Opening
NASDAQ Bell
In January, ANSYS president and CEO, James E.
Cashman, III, rang the Opening NASDAQ Bell. Of the
currently listed NASDAQ companies, ANSYS is the
only Pittsburgh-based company to participate in this
event twice. Additionally, ANSYS rang the Opening Bell
in January 2004. The invitation to ring the Opening Bell
is further validation of ANSYS' strong performance and
signals the growing importance of computer-aided
simulation for businesses -- ANSYS' sole focus for over
three decades.
ANSYS Solutions | Winter 2005
AN-Solutions-INDUSRTYNEWS 4/13/05 2:02 PM Page 2
HP Technology Gives BMW WilliamsF1 Team
Flying Start to 2005 Season
At the launch of the 2005 BMW WilliamsF1 Team car,
the FW27, HP announced that the WilliamsF1 design
team was able to assess twice as many aerodynamic
models in computational fluid dynamics this year than
ever previously possible, thus allowing the team to
investigate the optimal design for the car.
The aerodynamic and structural characteristics of the
FW27 were modeled on a powerful HP supercomput-
ing infrastructure with flexible capacity provided by a
utility computing system at HP Labs Bristol. This
facility, which provides enterprise customers with a
pay-as-you-go resource, allowed the WilliamsF1 team
to manage peaks in workload.
HP technology enabled the team to perform design
simulation of 1.3 terabytes of aerodynamic data. (One
terabyte is a billion bytes, equivalent to 69,333
volumes of the Encyclopedia Britannica.) The results
of the simulations have allowed the team to optimize
solutions for front and rear wings, brake ducts and
radiator ducts without the expense and time of
fabrication or wind tunnel testing.
The utility computing facility at HP Labs and the
installation of a new HP cluster platform 4000 Super-
computer system as WilliamsF1's primary computing
resource at their Grove, UK, headquarters have
increased the team's computer modeling capabilities
in 2005 by more than a factor of three.
ANSYS Earns Desktop Engineering’s Readers
Choice Award
ANSYS, Inc. announced that its ANSYS CFX-5.7
product received the Desktop Engineering's Readers'
Choice Award for October 2004. Desktop Engineering
selected the winner by determining the number of
reader service inquires received in response to the
product announcements published in the July 2004
issue of Desktop Engineering.
The ANSYS CFX-5.7 announcement received the
most inquiries. As the latest release in the powerful
CFX-5 series, ANSYS CFX-5.7 offers further integra-
tion into the ANSYS family of analysis technologies,
providing better performance, enhanced interoperabil-
ity with other ANSYS tools and more accessible
complex physics and multiphase simulations. In
addition, CFX-5.7 introduces true fluid-structure
interaction (FSI) capability. A simple, one- way transfer
of data from a CFX(R) solution to ANSYS enables
CFX-5.7 to pass information between a fluid and
structural simulation. This capability, as part of ANSYS
9.0, will evolve to full bi-directional dynamic coupling
within the ANSYS Workbench environment.
www.ansys.com
ANSYS Unveils Technology for Integrated Product
Development Environment
ANSYS, Inc. announced ANSYS 9.0, the first software 5
release featuring electromagnetic, computational fluid
dynamics (CFD) and mesh creation technologies
integrated within the ANSYS Workbench™ product
development environment, making simulation more
powerful and more accessible to a broader range of
users.
A major upgrade for the CAE industry's leading
simulation software, ANSYS 9.0 is an important stride
toward the continued evolution of Workbench as a
more productive and collaborative engineering tool
for simulation-driven, integrated product develop-
ment. Workbench's easy-to-use modern architecture
allows companies to be more productive by
eliminating manual file transfer, result translation and
reanalysis, saving time and money.
Ideal for companies utilizing simulation in their
product development processes, ANSYS 9.0 extends
core features and provides powerful robust design
capabilities as well as impressive developments in the
areas of mechanics, computational fluid dynamics,
high-frequency electromagnetics and multiphysics.
In addition, new breakthrough technology for large-
scale parallel computing allows users to analyze
complete systems without unnecessary approxima-
tions to their models.
To learn more about the integrated capabilities within
ANSYS 9.0, visit http://www.ansys.com/product/
newfeatures.
Upcoming Events
Offshore Technology Conference
Houston, Texas, USA
May 2 – 10
NAFEMS World Congress
St. Julians, Malta
May 17 – 20
Electronic Components & Technology
Conference
Lake Buena Vista, Florida, USA
May 31 – June 3
ASME Turbo Expo 2005
Reno, Nevada, USA
June 6 – 9
AIAA Fluid Dynamics Conference & Expo
Toronto, Ontario, Canada
June 6 – 9
ASME Fluids Engineering
Houston, Texas, USA
June 19 – 23
ANSYS Solutions | Winter 2005
 Spotlight on Trends and Issues

6
Meeting the Challenges
of Global Product Development
The role of simulation-driven design in developing
products on a worldwide scale is expanding.

Globalization is a major trend, sweeping across industry and changing forever the way

companies operate. According to a survey by the Foundation for the Malcolm Baldrige

National Quality Award, 95% of CEOs indicated that their top challenge over the next five

years is moving their companies to become more global.

Cost savings is generally the number one driver in the shift toward international suppliers

and outsourcing, with companies in China, India and other regions able to develop and

fabricate products generally at 30% to 50% below that of the United States, for example.

Cost alone may not be the only driver in the shift toward globalization, however. Compa-

nies often work with international strategic partners and joint ventures in co-developing

products. Moreover, manufacturers increasingly have worldwide divisions and business

units that often must collaborate in the development process. Developing products inter-

nationally also can take advantage of specialized

technical expertise elsewhere in customizing

designs to meet specific regional requirements,

and in locating design centers close to international

customers. Beyond manufacturing, engineering

software companies also may leverage program-

ming skills at operations in other countries for

efficiently developing highly effective and

innovative packages that professionals around the

world can use productively in their work.

www.ansys.com ANSYS Solutions | Winter 2005

A recent survey of manufacturers indi- 7
cated that more than 24% have shift-
ed to foreign suppliers or moved into
global sourcing this past year,
representing a 40% increase over just
two years ago. In particular, this
rapidly growing reliance on worldwide
supply chains leads to increased
levels of global product development,
with OEMs delegating growing levels
of design responsibility to international
subcontractors. In the automotive
industry, for example, designers for
the overall vehicle might be based in
Detroit, while suppliers with design
responsibility for their individual parts
and assemblies are located around the
The ANSYS user interface is available in Japanese and various other languages, making simulation
world, developing and building the technology easier to use for analysts and engineers around the world.
braking system in Japan, the steering
assembly in Germany, the trim in
China and the engine in Mexico.
“At a growing number of companies, the days of example, engineers in the field need ready access to
products designed entirely by a small group of sophisticated simulation performed at a central
engineers working side-by-side in the same facility are location.
drawing to a close. The shift is now toward globally This was a challenge faced by Fläkt Woods
dispersed product development teams, with members Group, a ventilation system manufacturer based in
collaborating around the clock and around the world Finland with representatives in 95 countries around the
on various parts of an overall product design,” world. Their new Elea air treatment system is a break-
explains Charles Foundyller, president of market through solution to provide improved air quality in a
research and technology assessment firm Daratech wide range of facilities, including hospitals, airports,
Inc. office buildings and educational institutions. To
“In a rapidly growing number of enterprises, optimize the performance of the Elea system for every
product development is by necessity becoming more customer installation, the company has implemented
collaborative, and this collaboration transcends a unique Web-enabled design and analysis tool with
national boundaries. It brings together departments, ANSYS CFX simulation software for performing
groups, suppliers, customers and strategic partners challenging HVAC calculations, including mixed
from around the world,” says Foundyller. “In many convection and transitional turbulence. Submissions
respects, CAE simulation-based technologies coupled of the simulations and management of results are
with the Internet overcome differences in time, managed by the EASA Web-enabled user interface
distance, culture and process and, in many cases, this from AEA Technology plc.
potent combination leverages these differences to “Our sales and application engineers securely
create a winning competitive advantage.” access the system via the Internet from anywhere in
the world, often while they are together with potential
Analysis Anytime, Anywhere customers,” explains Reijo Kohonen, vice president of
technology at Fläkt Woods Group. “They set up the
The Internet is becoming an indispensable tool
customer’s individual design case, including room
for companies operating on a global scale. The
layout and heat sources, device selection, etc. They
technology enables communication to take place
submit the analysis, via the Web, directly to our central
almost instantaneously and provides the capability for
CFD simulation environment, and the results are
individuals in remote sites to perform tasks as if they
returned to them later over the Internet in a clear and
were all co-located. Using Internet technology to its
concise report. The system provides control,
full advantage in these global activities entails
visualization, help and feedback for engineers in the
considerably more than merely bouncing e-mails back
field to readily perform simulations.” He notes that
and forth, however, particularly in cases when, for
CFD experts at headquarters in Finland are available
to provide support as needed but are not required to
process every case end-to-end.

www.ansys.com ANSYS Solutions | Winter 2005

 Spotlight on Trends and Issues
8 “The ability to employ powerful CFD analysis at
the proposal and design stages of every new
Elea project around the world allows engineers to
demonstrate clearly the performance and value of the
Elea product, and to optimize its deployment,” says
Kohonen. “This drives increased customer satisfaction
and profitability for Fläkt Woods.”
Taking Advantage of Time Zones
Just about every company working on a global scale
uses some form of electronic communication network
such as the Internet to overcome the challenges of
long distances separating individual sites. And a
growing number of organizations find that these
time-zone differences can actually be leveraged to
their advantage in performing simulation as part of a
collaborative effort.
Syncrude Canada Ltd., for example, the world’s
largest producer of crude oil from oil sands and the
largest single-source producer in Canada, is working
to improve the operation of its petroleum processing
operations based on ANSYS CFX simulations being
performed overnight at Australia’s CSIRO (Common-
wealth Scientific and Industrial Research Organization,
one of the world’s largest and most diverse global
research organizations).
The software is used extensively by CSIRO for
complex CFD modeling of multiphase, combustion
and reacting processes in the mineral processing,
chemical and petrochemical industries. By using such
simulations to gain a better understanding of the fluid
coker stripper operation, it is anticipated that design
changes will be identified to improve efficiency, reduce
deposits and optimize stripper operation.
“To most efficiently perform the simulations and
utilize the results, the two companies are leveraging
the distance separating their facilities,” says Dr. Peter
Witt, research scientist at CSIRO. “When it is night in
Edmonton, Alberta, Canada, where Syncrude
Research is located, CSIRO staff are hard at work in
Australia performing analyses and posting results,
including pictures and animations, on their extranet.
The next morning, the group in Canada can view
progress of the modeling work and provide feedback
for a quick turnaround.”
When Oceans Separate Facilities
Even when operations are performed within the same
four walls, design problems found late in product
development are expensive and time-consuming to
correct, and hurried late-stage corrections often do
not result in optimal product performance. These
www.ansys.com
problems are greatly magnified when groups are
widely dispersed in different countries around the
world. The challenge is further complicated when
manufacturing, as well as some product development
activities, is moved overseas.
“This global footprint can lead to situations where
a product is conceived and its performance
requirements specified in country A, and it is then
designed and tested in country B and mass-produced
in country C. Therefore, development centers have to
be flexible enough to respond to the needs of their
local market as well as be able to develop products for
different, distant markets,” explains Fereydoon
Dadkhah, mechanical analysis and simulation
engineer at Delphi Electronics & Safety.
In addressing this challenge, Delphi has equipped
engineers at design centers in different countries
with DesignSpace® software as a first-order analysis
tool, with engineers involved in the design process
from the moment contracts are awarded. Even while
some of the system features are being finalized, com-
ponent designs can be analyzed and evaluated, and
up-front simulation is used extensively to evaluate
component performance. In many cases, the early
analysis indicates that modifications are necessary.
The modifications are made and assessed until all
problems are eliminated.
“Shortened design schedules in globally
dispersed product development makes the use of
CAE simulation mandatory, especially in the early
stages,” says Dadkhah. “Because of the distributed
product development process, it is important that all
the engineers and designers use the same processes
and techniques. Using analysis as an integrated part of
product development enables engineers from around
the world to collaborate in unprecedented ways.”
Standardizing Worldwide Processes
When different groups and departments around
the world are trying to work together on projects,
differences in processes, procedures, terminology and
convention can cause significant problems that
hamper effective collaboration, productivity and
accuracy. Facilities in China, Europe and the
United States, for example, may have operated
independently for years, and in many cases might
have been separate companies until a merger or
acquisition brought them together. Ways of modeling
parts, performing simulation, displaying data and
evaluating results can vary widely among these
diverse groups and generally are deeply engrained in
their structure and culture.
ANSYS Solutions | Winter 2005
 A major U.S.-based multi-national agricultural
equipment manufacturer is meeting this challenge by
standardizing its unique simulation-based processes
through the use of templates in the ANSYS Work-
bench™ environment. Process automation capabili-
ties are provided by the software through wizards and
templates that automate repetitive operations to han-
dle simulation problems faster and make the develop-
ment process consistent from project to project and
group to group across the company.
The firm’s project manager for computer-aided
engineering explains that this template approach
enables the company to capture the core
competencies of and invaluable specialized expertise
of particular skilled individuals in these different
groups, while at the same time providing a consistent
methodology for performing these tasks and
collaborating across multiple groups on projects.
Templates also provide a way to ensure that company
processes follow industry-standard best practices.
Automatic checkpoints in the simulation workflow
ensure that the standardized processes are being
followed. Also, automatic report generation
capabilities within Workbench provide valuable
documentation needed by the company for project
histories and accountability for compliance with
relevant standards.
“In the past, engineers here could perform
simulations however they wanted so long as they
obtained accurate results,” notes the manager. “Now
the process is standardized company-wide, allowing
engineers to collaborate more effectively.” The
manager also points out that capturing and
standardizing the process enables the company to
better assess and evaluate their overall product
development process for shifting operations from
time-intensive and costly multiple prototype testing
cycles to higher levels of up-front analysis early in
design. “Old habits are difficult to break,” he says.
“The template approach gives us the tools we need to
effectively institute these changes, provide consistent
ways of working and enable us to move to
simulation-based product development that
companies must adopt to compete on a global scale.”
Leveraging Global Intellectual Capital
In these and other innumerable examples of
simulation-driven design implemented on a global
scale, companies are gaining a competitive edge
with innovative products and processes by tapping
www.ansys.com
9
ANSYS, Inc. supports customers around the world with more than
25 sales offices on three continents and a network of 150 channel
partners in more than 40 countries.
into their global intellectual capital — the collective
knowledge, expertise and insight of workers in
multiple disciplines around the world. Through the
ability to quickly perform what-if studies and evaluate
alternative configurations, simulation provides insight
into product behavior and gives free reign to the
imagination of product team members. They can all
see the way a proposed product would function if it
existed in hardware and have the freedom to
investigate alternative ideas.
In these implementations, the Internet becomes
the conduit for rapidly exchanging critical data
while simulation guides the design process and serves
as the knowledge driver that channels everyone’s
ideas and insights into the problem. In this sense,
simulation is a tremendous collaborative tool, allowing
engineering to demonstrate to others — no matter
where they are located or in what discipline they work
— how various designs perform, and enabling
cross-functional team members at dispersed facilities
to provide valuable input into product design.
In the early stages of development, when
changes are most easily made, a marketing manager
in Chicago could suggest a slimmer case for a
consumer product, for example, or a manufacturing
planner in Tokyo might see ways to reduce the parts
count with a single injection-molded assembly.
Analysis results can quickly show the entire team the
impact of such ideas on stress, deformation, vibration
and other aspects of product behavior.
Through these capabilities, simulation-driven
design enables multiple disciplines at sites around the
world to work together in product development and
facilitates the creativity that comes from such synergy.
In that respect, the approach leverages valuable global
intellectual capital at progressive companies that will
likely be among the world’s superstars in the coming
years.

ANSYS Solutions | Winter 2005

10
ANSYS technology improves designs by reducing their
sensitivity to real-world variabilities. By Raymond Browell, P.E., Product Manager, ANSYS, Inc.
Designs are traditionally developed based on exact
specifications that define structural loads, temperature
distributions or other conditions that products must
withstand. Unfortunately, production processes used
to manufacture the product, materials used in the
product and the environment in which the product
operates are seldom that exact. Rather, real-world
variables fall within a range of possibilities that can
cause rejections during quality inspection or failures in
the field.
The Robust Design capability in ANSYS mini-
mizes such failures by allowing engineers to develop
the design to perform its intended function regardless
of these variations. In this way, users can quantify
projected failure rates and therefore compensate
in the design for uncertainties, including material varia-
tions, environmental changes during the product’s
usage, manufacturing variations and component
deterioration.
The Robust Design method is central to
improving engineering productivity. Pioneered by Dr.
Genichi Taguchi (see the sidebar “Understanding the
Taguchi Method”) after the end of the Second World
War, the method has evolved over the last five
decades. Many companies around the world have
saved hundreds of millions of dollars by using the
method in diverse industries, including automotive,
aerospace, business equipment, telecommunications,
electronics and software.
Robust Design can be applied to nearly any
process or product, even financial processes. For
ANSYS, the capability operates around the simulation
solution to determine the extent to which uncertainties
in the model affect the results of an analysis in terms of
stress, deflection and so on. In this context, the
approach is based on a probabilistic characterization
(see the sidebar “Characterizing Product Behavior”)
that quantifies the reliability or quality of the product by
means of statistical analysis of uncertainties.
Robust Design goes one step further than a prob-
abilistic characterization by allowing you to optimize
design variables to achieve a particular probabilistic
www.ansys.com
Robust Design allows you to
define both Design Variables
and Uncertainty Variables and
then optimize to a set of goals
level, such as Six Sigma, which corresponds to 3.4
failures in one million parts. Six Sigma initiatives are
focused primarily on the production cycle and there-
fore try to optimize the manufacturing process such
that it automatically produces parts conforming to Six
Sigma quality.
Design for Six Sigma
Design directly impacts a huge portion of product life-
cycle costs beyond manufacturing, however, so a
broader initiative called Design for Six Sigma (DFSS)
optimizes the design itself such that the part conforms
to Six Sigma quality even with variations in manufac-
turing. A popular set of methodologies addressing
these issues is based on early work at GE. For Robust
Design and Design for Six Sigma, quality is an explicit
goal of the optimization, and Robust Design may
therefore be a valuable tool at companies
implementing DFSS initiatives.
The Design for Six Sigma approach is focused on
1) increasing engineering productivity so that new
products can be developed rapidly and at low cost,
and 2) value-based management techniques that
minimize product failure rates.
DFSS methods impact the final product quality
from two opposite perspectives. First, variation
is studied and reduced in manufacturing processes
with a goal of having +/-6 standard deviations of
dimensional, material and performance characteristics
controlled within the design tolerance. This method
requires extra continuous effort to identify and resolve
variations resulting from process randomness
and trends.
The second perspective involves designing prod-
ucts that can meet the operating requirements and be
subjected to greater variation at the same time. In
other words, there is value in knowing what variation is
important to control and in maximizing variational
allowance for the sensitive design characteristics. This
is the solution focus of the ANSYS Robust Design
initiative and the DesignXplorer family of products.
ANSYS Solutions | Winter 2005

DesignXplorer can work with all types of input parameters such as CAD
parameters, APDL parameters, ParaMesh parameters, and Design Simulation
parameters. Here we see the scoped Equivalent Stresses for a given set of
geometry input parameters. This scoped Equivalent Stress will be used for
determining the fatigue life of a turbine blade under cyclic rotational loads.
DesignXplorer contains both Goal Driven Optimization and Six Sigma Analy-
sis capabilities. Goal Driven Optimization and Six Sigma Analysis combined
together results in Robust Design. A design optimized to provide higher level
of reliability (or conversely, a minimum failure probability). In this image,
DesignXplorer has calculated the histogram and cumulative distribution func-
tion for the turbine blade's fatigue life that result from the natural variations
created in the manufacturing process. Robust Design within DesignXplorer
allows use to then optimize the input parameters to provide for both a higher
fatigue life and the probability of the fatigue life being achieved.
Understanding the Taguchi Method
Pioneered by the work of Dr. Genichi Taguchi after the end of
the Second World War, the Robust Design method has
evolved over the last five decades. The Taguchi Method is a
design-of-experiments technique based on the assumption
that the input-output relationship is only linear and does not
have any interactions between the input parameters. It is also
based on the assumption of discrete distributions for the input
parameters, i.e., every input variable can have values only at
the design-of-experiment levels, but nothing in between.
www.ansys.com
11
Response surfaces generated with DesignXplorer allow users to readily
see the influence of input variables on design performance: the way stress
or deflection is impacted as design geometry or material properties vary,
for example. The technology provides insight into product behavior by quickly
arriving at a range of results that would otherwise be impractical to generate
using individual single analysis runs. In this example, our response surface
shows the variation of fatigue life with respect to 2 geometry input parame-
ters. As we can see the function is rather sharp, so significant gains in
the fatigue life can be made by having DesignXplorer optimize the geometry
input parameters.
DesignXplorer can also work with all types of output parameters such as
deflection, stress, mass, frequency, or fatigue life. The scoped Equivalent
Stresses from the preceding image were used to determine the fatigue
life of this turbine blade under cyclic rotational loads.
In general, it is now an outdated method for the most part, but
was a “better-than-nothing” solution in the days when the
computational effort was prohibitively expensive. It still
remains a popular method in cases where the effort is still
highly expensive, such as in crash testing. With modern and
efficient computer tools available, however, the gross simplifi-
cations and assumptions of the method for most applications
are no longer adequate and no longer needed.
ANSYS Solutions | Winter 2005
 Characterizing Product Behavior
Characterizing the behavior of a product (a part or
an assembly of parts) under operational conditions
can be done in three ways. Whereas empirical and
deterministic characterizations are pass/fail in nature,
probabilistic characterization on which Robust Design
is based quantifies the reliability or quality of the
12 product by means of statistical analysis of
uncertainties influencing product behavior.
Empirical Characterization refers to creating a
prototype (or prototypes) of the product, hopefully
replicating the manufacturing steps that are to be
used in a production run of the product. The product
is then tested to determine its behavior and to make
final judgment on whether the product will perform
successfully in the field. Typically, this judgment is of a
yes/no nature. In other words, the product will fail or
not, hence its design is acceptable or not.
Deterministic Characterization typically refers to an
analysis of the product without testing it. This analysis
could range from simple engineering handbook
calculations to elaborate finite element analysis (FEA).
Once again, the typical judgment is of a yes/no nature.
ANSYS Tools for Robust Design
DesignXplorer VT and DesignXplorer provide you with
the ability to create a Robust Design by allowing you
to define both Design Variables and Uncertainty Vari-
ables and then optimize to a set of goals (such as min-
imizing failure probability or maximizing reliability or
quality of a product) where each of these goals could
be related to stress, deflection, fatigue life and so on.
Robust Design combines Probabilistic Character-
ization with optimization. Additionally, due to the
power of our optimization techniques, a variety of
objectives may be combined, even deterministic
objectives with probabilistic objectives. A common
desire would be to have minimum weight to reduce
manufacturing costs, while also maintaining a low
probability of failure rate to minimize warranty costs.
www.ansys.com
Probabilistic Characterization quantifies the reliability
or quality of the product by means of a statistical
analysis. Probabilistic characterization combines the
deterministic characterization, either handbook or FEA
analysis, with statistical analysis tools to address the
effect of statistical variability and uncertainty influenc-
ing the products behavior. Probabilistic analysis
typically involves four areas of statistical variability: the
geometric shape, the material properties, the loading
and the boundary conditions. For example, the statis-
tical variability of the geometry of a product would try
to capture the product-to-product differences due to
manufacturing imperfections quantified by the manu-
facturing tolerances. Because the statistical analysis
typically requires many data points, a combination of
computer-based FEA analysis with statistical analysis
is the most time- and cost-efficient method in
practice. Unlike the first two methods, probabilistic
characterization provides a probability of success or
failure and not just a simple yes/no evaluation. For
instance, a probabilistic analysis could determine that
1 part in 1,000,000 would fail or what the probability is
of a product surviving its expected useful life.
Robust Design allows for and provides this type of
deterministic as well as probabilistic optimization
quantities. In keeping with industry nomenclature, we
would describe the method of achieving the optimal
solution as Multi-Objective Optimization.
Robust Design is a new capability of the ANSYS
DesignXplorer Family at 9.0. Robust Design can be
used with CAD parameters or any Design Simulation
parameter from within the Workbench Environment.
Robust Design can be used with parameters
contained in ANSYS Parametric Design Language
(APDL)–based files, which may contain ParaMesh
parameters as well, to perform Robust Design on
existing or new ANSYS analyses.
Just like DesignSpace made it possible for
companies to expand their use of simulation,
DesignXplorer will allow companies to take extensive
advantage of Robust Design techniques by allowing
them to expand deployment of DFSS initiatives.

What’s Six Sigma?
Robust Design optimizes design variables to achieve a
particular probabilistic level such as Six Sigma, which
translates into 3.4 failures in one million parts. This
probably is achieved when the performance target is
4.5 standard deviations away from the mean value.
The additional 1.5 standard deviations leading to a
total of 6 standard deviations are used as a safety
margin to allow for “drift of the mean value” in the
properties and environment which the product can
experience over its lifetime.
ANSYS Solutions | Winter 2005
Midsurfacing Tools
13
for Meshing Complex Geometries
Create high-quality shell models with the
ANSYS Advanced Mesher suite.
By Simon Pereira
Senior Technical Support Engineer
ANSYS, Inc.

F-1
Using shell elements to model solid parts is a popular
technique in finite element analysis. Shells lower the
number of nodes in the model, solve times are relatively
fast and results are generally very reliable — provided
the model does not oversimplify actual part geometry.
So, as shell mesh solvers improve, users are continually
trying to apply the widening scope of the technology to
more complicated models while maintaining as much
reality as possible in the models.
The bottleneck for most analysts using shell
elements for solids has been in simplifying the geometry
to a reasonable midsurface. Sometimes the geometry is
“simplified” to the point that it barely represents the
original geometry, and in many cases this is a long and
tedious process that takes up valuable time.
ANSYS can support real-world geometry with
t-sections, variable thickness and other complexities
that engineers know are important design considera-
F-1 Zoom
tions. CAD interfaces are available to import the geometry
directly from the design system, saving the analysts from
duplicating this effort. The challenge is to simplify and
midsurface the geometry as accurately and as quickly as
possible. The ANSYS ICEM CFD™ Advanced Mesher is
at the forefront of this effort.
The majority of midsurfacing is done on stamped
sheet metal parts. Some of these parts may have variable
thickness, but the physics of stamping limits the
complexity, so there are no t-sections or other surprises.
The advanced mesher powers through these parts with
many useful features.
The standard process involves minimal user inter-
action and uses two primary functions: “build diagnostic
topology” and “midsurface.” The “build diagnostic
topology” function can be run on a part-by-part basis. Stamped sheet metal parts such as this floor panel can be
This uses a tolerance to establish connectivity informa- midsurfaced in an automatic mode with minimal user input.
tion between the surfaces of a part. The user does not
first need to sew any poorly fitting edges together; these
gaps and overlaps are allowed for by the tolerance.The
midsurface tool can also be run on a part-by-part basis
and uses the diagnostic topology information to
effectively match opposite surfaces and determine the

www.ansys.com ANSYS Solutions | Winter 2005

 A-14
14
(Image A-15) To create a midsurface of an extended
geometry (see A-14), use original geometry as
construction lines to draw a mid-curve (see A-15).
midsurface location. The user enters a maximum part
thickness to be midsurfaced and selects the parts to
be midsurfaced. There are also advanced options that
can be set to further control the midsurfacing process.
It is usually run as an automatic process that creates
the new midsurfaces and automatically removes the
original surfaces and side-surfaces. Connectivity
between the new midsurfaces is maintained. The
thickness of the original thin solid, even variable
thickness, is automatically transferred to the new mid-
surfaces. Once the shell mesh has been generated,
the (variable) thickness shell element properties can
be calculated directly from the midsurface. This is a
highly automated process, and, once begun, it can
work through a stamped sheet metal assembly, one
part at a time, with no further user interaction.
The assembly process can be handled with auto-
matic seam-weld detection or automatic non-confor-
mal contact setup. Manual or scripted connector
setup is also available and easy to use. The Advanced
Mesher produces a high- quality mesh, with the ability
to ignore slivers and internal surface edges. The
exceptional and easy-to-use mesh editing tools allow
for element-level control. After the model is prepared,
the mesh, loads, constraints and properties, including
variable shell thickness, can be written out to a
number of advanced FEA solvers, including ANSYS,
LS-DYNA, ABAQUS and NASTRAN.
Not all midsurfacing is so simple. More and more
analysts are using shell mesh to model cast,
www.ansys.com
(Image A-18) Original surfaces and midsurface
with variable thickness.
(Image A-19) Shells displayed with variable
thickness.
machined, extruded or injection-molded parts. These
parts are more difficult and may contain T-sections,
webs, cylinders, routered edges, grooves, mis-
matched surfaces, etc. For these more complicated
models, automatic “quiet” midsurfacing methods are
insufficient. However, ANSYS ICEM CFD advanced
mesher’s geometry tab, including other midsurfacing
options, provides the tools to get the job done.
The midsurface geometry can be created using
the original geometry as construction lines. There are
a number of point, curve and surface creation tools
that can be used to create a representative midsur-
face. For instance, if a portion of your midsurface
geometry could be created with an extrusion, create a
“mid-curve” along one side of the thin solid and then
use one of the curves in the perpendicular direction to
create a curve-driven midsurface.
The “modify surface thickness” tool has an auto-
matic option to calculate the midsurface thickness or
variable thickness from the original solid geometry.
This tool also has manual options to set the variable
surface thickness. Again, when the mesh is generated
on these midsurfaces, the thickness can easily be
transferred to the elements and nodes. The shells can
be written out to a number of solvers, including
ANSYS, LS-DYNA, ABAQUS and NASTRAN, with the
thickness calculated at each node or averaged over
each shell. Most other advanced shell properties are
also available.

ANSYS Solutions | Winter 2005
Meshing T-Sections
A common complication in many geometries is a T-section.
This could be a traditional “T” shape, or any situation that
results in a multiple edge between two or more mesh planes.
The usual approach is to generate as much of the midsurface
as possible. The final step of connecting the midsurfaces is
done by extending one of the surfaces to the intersection. In
ICEM CFD, the intersection curve is colored blue to indicate a
multiple edge. When the mesh is generated, a row of nodes
will form along this multiple edge to connect the shells
through the T-section. The accompanying images demonstrate
how this approach is used in modeling a support bracket
having two T-sections.
Modeling Complex Thin Solids
For complicated thin solids, it may be easier to create the midsur-
faces using the geometry creation and repair tools. Feature
detection tools such as button and fillet detection make it easy to
locate and remove these features from the model. Other tools
15
such as “remove hole,” “re-approximate surface” and “un-trim
surface” also help to simplify a model. Often, the geometry may
have two sides that are so different that a true midsurface may
be undefined or undesirable. If at least one side of the geometry
is suitable, the “offset surface” tool can be used.
A_02

(Image A_02)
Die-cast ashtray is
a complicated thin
solid not easily
midsurfaced without
ICEM CFD.

(Image T-1) 1. Initial geometry with two T-sections.

A_3

(Images A_3 and A_6)

Various tools help to
simplify the model.

(Image T-2) 2. Midsurface as much as possible.

A_6

(Image T-3) 3. Extend T-surfaces to base plate.

(Image T-4) 4. Mesh midsurface geometry.

www.ansys.com ANSYS Solutions | Winter 2005

 Finite Element Modeling of
22
16 Capacitive Micromachined
Analysis provides insight into critical operating
parameters of this emerging class of MEMS devices.
By Bradley J. Kirchmayer, Walied A. Moussa and M. David Checkel
Department of Mechanical Engineering
University of Alberta
Ultrasonic sensors have been developed for many
applications, including flow sensing, non-destructive
evaluation and medical imaging. Presently, the most
popular ultrasonic transducers are piezoelectric
crystals, which, although successful in these
applications, still have problems associated with their
design, including impedance mismatch, high
temperature instability and manufacturing difficulties.
The evolution of the MEMS (Micro Electrical
Mechanical Systems) industry has led to the emerging
of Capacitive Micromachined Ultrasonic Transducers
(CMUTs) as an alternative to overcome many of the
piezoelectric problems. A CMUT is an array of
elements that either transmit or receive ultrasonic
waves. Each element consists of a metal electrode
embedded atop a thin silicon nitride membrane. The
membrane is supported by silicon nitride walls, which
suspend the membrane above the bottom electrode
(silicon substrate). Transmitter CMUTs use a DC bias
voltage and a driving AC voltage to induce capacitor
forces between the two electrodes. This causes the
membrane to vibrate, thus emitting ultrasonic waves
www.ansys.com
into the medium. Conversely, receiver CMUTs detect the
change in voltage when the membrane is appropriately
deflected by incoming ultrasonic waves.
One advantage of CMUTs over piezoelectric
transducers is that the membrane’s small impedance
accommodates ultrasonic sensing in gases without the
use of matching layers. Other advantages include greater
dynamic range, higher temperature thresholds, lower
manufacturing costs and higher transmitting frequencies.
Like all of MEMS development, finite element
analysis is proving to be a virtually indispensable tool in
CMUT design. The following study shows the value of
various types of ANSYS FEA models in determining two
critical operating factors relating to the performance of a
CMUT: the collapse voltage and the resonant frequency.
Understanding the Operating Parameters
Both collapse voltage and resonant frequency rely upon
the CMUT’s geometric configuration and material
properties. The importance of these parameters will vary
depending upon the sensing application. The electric
potential between the two electrodes can be increased
to the point where the electrostatic force between the
electrodes overcomes the membrane’s stiffness,
causing the membrane to collapse onto the bottom
electrode. This electric potential is called the collapse
voltage or pull-in voltage, and it governs the maximum
applied voltage for a working CMUT.
For a CMUT to produce ultrasound waves in gases,
the impedance caused by the CMUT’s membrane
must be properly matched to the impedance of the
surrounding medium. To produce the required
membrane amplitudes for accurate impedance matching
and receiver detection, the membrane must generally be
operated at its resonant frequency.
Because of its ability to couple these many
interrelated electrical and structural factors, ANSYS
software is a powerful tool in studying these complex
phenomena.
ANSYS Solutions | Winter 2005
Ultrasonic Transducers 17

Building the Finite Element Models
Three separate models were created using ANSYS to
illustrate CMUT behavior: a 2-D axisymmetric model was
used to investigate the collapse voltage, a 3-D solid
model coupled with a TRANS 126 element demonstrated
the resonant frequency of a pre-stressed membrane, and
a reduced-order model developed with the ROM144
element also depicted the resonant frequency. All three
models utilized identical materials, listed in Table 1.
Table 1. Material Properties for Finite Element Models.
Component Membrane Air Substrate
(Material) (Si3N4) Gap (Si)
The model was created using PLANE 82 and
PLANE 121 elements. To simulate the loading due to a
DC bias voltage, a potential difference was applied
between the substrate and the metal electrode. Physics
environments were used to model the static membrane
deflection caused by the capacitive force, as shown in
Figure 2a and b. The solution using a built-in iterative
solver (ESSOLV) was obtained when the membrane
displacement or electrical field strength converged to
within 5%. The collapse voltage was found when the
solution diverged, causing the displacement to exceed
the model’s structural limitations.
Electrode
(Al)

Young’s Modulus 3.20E+11 - 1.69E+11 6.76E+1

(Pa) 0

Density (kg/m3) 3270 - 2332 2700

Poisson’s Ratio 0.263 - - 0.3555

Permittivity 7.6 1.04 11.8 -

The two-dimensional model is the electrostatic- Figure 2a. Two-dimensional axisymmetric static
structural model first presented in [4] to model an deflection. ( 1/2 expansion)
unstressed circular CMUT element. It was evaluated
against the thermal-structural analogy found in [3]. Figure
1 depicts the geometry of the axisymmetric model.
Figure 2b. Two-dimensional axisymmetric static
Figure 1. Cross-Sectional View of 2-D CMUT Model.
deflection. ( 3/4 expansion)
2D Model Axis
Aluminum Top Electrode of Symmetry
Silicon Nitride Membrane

A three-dimensional solid model was created to

R
E

TE
investigate the relationship between membrane
T
M
geometry and resonant frequency. The model’s
G AirGap R
M
membrane was loaded with a 60 MPa (tension)
pre-stress and a 30VDC bias voltage. Table 3 lists the
model’s geometry.
Silicon Substrate (Electrode)

The purpose of this analysis was to verify the

Table 3. 3-D Model Geometry.
relationship between varying electrode radius (RE ) and
collapse voltage. All other geometric parameters were Parameter Dimension (µm)
held constant and are listed in Table 2.
Membrane Radius 50
Table 2. Geometric Constants for 2-D Model. Membrane Thickness 1

Parameter TM RM G TE RE Electrode Radius 50

Electrode Thickness 0.2
Dimension (µm) 0.6 25 1 0.1 Variable
Air Gap 1.4

www.ansys.com ANSYS Solutions | Winter 2005


18 Figure 3. Three-dimensional Trans126 model static deflection.
The 3-D model reduced the CMUT to the
membrane, electrodes and air gap. SOLID45 and
SOLID95 elements were used to model the membrane
and top electrode. The air gap was modeled with
TRANS126 elements that coupled the electrostatic and
structural domains. Since the bottom electrode has no
displacement, it was simply modeled by the bottom
nodes of the TRANS126 elements. The membrane’s
residual stress, σthermal, was induced by constraining its
edges and subjecting it to a uniform temperature
change, Tuniform. The required temperature change was
calculated using the following equation, where YO, α, and
v were the membrane’s Young’s modulus, coefficient of
expansion and Poisson’s ratio, respectively.
σthermal (1-v)
Tuniform =
YO α
An electric potential of 30VDC was then applied
across the TRANS126 element, and the membrane’s
static deflection was observed, as shown in Figure 3.
From the static solution, a partial block lanczos solver
calculated the first five modal frequencies. Finally, a
harmonic analysis was solved for a 1 Volt AC load
applied over a range that encompassed the first resonant
(modal) frequency.
The reduced order model using ROM144 ele-
ments simplifies 2-D and 3-D models by developing gov-
erning equations for predetermined master nodes, and
uses these relationships to create reduced order ele-
ments. The reduced order elements have a fraction of the
nodes of the full model, greatly reducing computational
time. This model was created using the same constraints
as the 3-D model. The ROM144 element was generated
from a 3-D model defining the membrane, electrodes
and air. The electrode and membrane were modeled
similar to the 3-D solid model while the air was modeled
using SOLID122 elements. Physics environments
coupled the system as in the 2-D model.
Figure 4. ROM model-mapped in 3-D with 60MPa
Membrane prestress.
www.ansys.com
Next, a generation-pass defined the master
nodes, applied static loads and calculated the modal
frequencies of the CMUT. Governing equations were
developed from static force-displacement solutions
and the first modal frequency. When the structural
response had been properly defined, ANSYS created
the ROM144 element. A use-pass was then utilized to
perform an electrostatic-structural analysis on the
ROM144 model. The use-pass consisted of a
harmonic analysis for a 60 MPa pre-stressed
membrane with a 30VDC bias and a 1V AC driving
voltage, as shown in Figure 4.
Looking at the Results
The three ANSYS FE models were validated using
results previously reported in the literature. The
analysis for the 2-D axisymmetric model was verified
with the thermal-structural analogous solutions
obtained in [3]. The resonant frequencies modeled by
the 3-D solid model and the ROM144 element were
verified using the impedance model of [6].
The 2-D axisymmetric model had the advan-
tage in its ability to calculate static deflection and col-
lapse voltage. The axisymmetric nature also reduces
computational time. The disadvantages of the 2-D
model, however, include the weak coupling forces and
the inability to perform a harmonic or modal analysis.
The 2-D model was used to calculate the collapse
voltage for varying electrode radius with the results
shown in Figure 5.
Figure 5. Collapse voltage for varying.
Comparing the results with those obtained in [3], it is
seen that both models demonstrate similar behavior for
electrode radii greater than or equal to half the mem-
brane radius. As the electrode radius decreases, the two
results diverge. At an electrode radius of 2µm, the results
from the electrostatic-structural modal are double those
of the thermal-structural analogy. Due to inaccuracies
in both models, the actual collapse voltage for small
electrode radius is likely to fall between these two values.
ANSYS Solutions | Winter 2005
 For instance, ESSOLV uses weak coupling and may
not have included the full capacitor force. Conversely,
the thermal-structural analysis did not include the
stiffness of the top electrode. Bozkurt et al. [3] defined
the optimal electrode radius to be half of the membrane
radius (12.5µm in this case), so the 2-D model can be
used to calculate the collapse voltage for an optimized
CMUT design.
The 3-D solid model coupled with the TRANS126
element was used to calculate the resonant frequency for
a pre-stressed membrane. By varying the radius by 2%,
a shift in resonant frequency was observed, as shown in
Figure 6. While the resonant frequencies are in a
reasonable range, the displacements are high because
there is no damping in the model. By comparing the
results obtained in the current study with those in [6], the
3-D model was verified to accurately represent the
resonant frequency of a CMUT. Other advantages of this
model include the strong coupling nature of the
TRANS126 element, and the ability to model pre-stress.
The main disadvantage of this 3-D model is the large
amount of computational time required for the harmonic
analysis.
Figure 6. Resonant frequency shift with 2% change in membrane radius.
References and Resources for Further Reading
[1] W. P. Mason, Electromechanical Transducers and Wave
Filters. New York, NY: Van Nostrand, 1942.
[2] I. Ladabaum, X. C. Jin, H. T. Soh, A. Atalar, and B. T.
Khuri-Yakub, “Surface micromachined capacitive
ultrasonic transducers,” IEEE Trans. Ultrason., Ferroelect.,
Freq. Contr., vol. 45, no. 3, pp. 678-690, 1998.
[3] A. Bozkurt, I. Ladabaum, A. Atalar, and B. T. Khuri-Yakub,
“Theory and analysis of electrode size optimization for
capacitive microfabricated ultrasonic transducers,” IEEE
Trans. Ultrason., Ferroelect., Freq. Contr., vol. 46, no. 6, pp.
1364-1374, 1999.
[4] B. Bayram, G. G. Yaralioglu, A. S. Ergun, and B. T. Khuri-
Yakub, “Influence of the electrode size and location on the
www.ansys.com
19
Figure 7. Resonant frequency of ROM144 element.
The ROM144 element was also used to calculate the
resonant frequency for a pre-stressed CMUT membrane.
Results are shown in Figure 7. The results appear to be in
good agreement with those published by X. Jin et al. [6].
While the ROM144 element is inherently less accurate and
more complex to generate than the 3-D model,
computational savings allow a refined mesh, which
enhances accuracy. Also, the air elements provide realistic
damping, so the displacements are considered correct.
Overall, the ROM144 model can include a residual stress;
can calculate static, modal, harmonic and transient
solutions; and converges within reasonable computational
time, providing a significant advantage over the other
models.
Conclusions
Previously reported findings validated the results of the
ANSYS models. The relationship between collapse
voltage and electrode radius for a CMUT membrane with
a radius of 25µm and a thickness of 0.6µm was
investigated using a 2-D axisymmetric model. The
collapse voltage agreement was within 15% for electrode
radius greater than half the membrane radius. For smaller
electrode radii, the collapse voltage was shown to
increase greatly. A reduced order model and a 3-D solid
model were used separately to calculate the resonant
frequency of a CMUT with 1µm thick, 50µm membrane
radius. For a membrane with a 60 MPa pre-stress and
biased at 30 V DC, the ROM144 and the 3-D model
produced resonant frequencies of 2.29 MHz and 2.32
MHz, respectively.
This correlation of ANSYS simulation with empirical
data confirms that finite element analysis is a valuable tool
in this range of leading-edge MEMS applications.

performance of a CMUT,” Edward L. Ginzton Laboratory,
Stanford University, Stanford, CA 94305-4085.
[5] I. Ladabaum and D. Spoliansky, “Micromachined ultrasonic
transducers: 11.4 MHz transmission in air and more,”
American Institute of Physics, Appl. Phys. Lett. 68 (1),1996.
[6] X. Jin, I. Ladabaum, and B. T. Khuri-Yakub, “The microfabri-
cation of capacitive ultrasonic transducers,” IEEE Journal of
Microelectromechanical Systems, vol. 7, no. 3, pp. 295-302,
1998.
[7] X. Jin, I. Ladabaum, F. L. Degertekin, S. Calmes, and B. T.
Khuri-Yakub, “Fabrication and characterization of surface
micromachined capacitive ultrasonic immersion
transducers,” IEEE Journal of Microelectromechanical
Systems, vol. 8, no. 1, pp. 100-114, 1999.
ANSYS Solutions | Winter 2005
 CFD Update: What’s New in Computational Fluid Dynamics
Bosch Common Rail Injection System.

20

Simulation of Cavitating Flow

in Automotive Injection Systems
Bosch uses CFX in analyzing its high-pressure Common Rail Injection
system for fuel efficiency, sporty vehicle performance and low emissions
in increasingly popular diesel engines.
By Uwe Iben and Martin Voß, Robert Bosch GmbH
and Wolfgang Bauer, ANSYS Germany GmbH

Injection system components are required to operate Predicting Local Flow Field
properly at highly different operational conditions over
Various test configurations have been investigated at
their lifetime. A modern injection system provides a rail
Robert Bosch GmbH. Besides global integral quanti-
pressure up to 1600 bar. Because of the design and
ties such as the mass flow rate as a function of a given
high-speed valves, the pressure drops locally below
pressure drop, the prediction of the local flow field is
the vapor pressure and the fluid cavitates. Vapor
necessary for the prevention of cavitation erosion.
subsequently condenses in regions of higher pressure
Taking the small geometric size and the high fluid
values, which may lead to cavitation erosion.
velocities (up to 300 m/s) into account, a highly
Cavitation erosion may cause fuel injection systems to
transient flow field can be analyzed by high-speed
fail, so the geometry and operating conditions have to
visualization technique. In addition to the well-known
be carefully designed. An experimental check of all
cavitation at the throttle inlet, a strong interaction of
operating conditions is extremely time-consuming
turbulence and cavitation inside the shear layers down-
and, due to the material and the small geometric size,
stream of the constriction was observed. The highly
is often impractical. Simulation has to support the
transient dynamic caused erosion on the sapphire
complex engineering process from the beginning, so
glass windows as well as on the impingement body
reliable simulation methods are essential for the
are eroded due to cavitation within 45 minutes opera-
design process.

Test section geometry of fuel injector. Cavitation erosion with vapor is shown in black. Erosion of impingement body is illustrated in circle.

www.ansys.com ANSYS Solutions | Winter 2005

tional time. The initially sharp corners of the aluminum
impingement body are eroded due to cavitation within
45 minutes operational time.
Although the standard engineering practice in
CFD computations is to use statistical turbulence
models (such as the Reynolds Averaged Navier-
Stokes equations, or “RANS”), it was expected that
the transient large-scale turbulent eddies predicted by
a Detached Eddy Simulation (“DES”) would improve
the accuracy of cavitation simulations.
Validating Circular Throttle Flow
As a first step, the basic cavitation simulation was
tested for a circular throttle flow. The outlet pressure of
the circular throttle is varied with respect to the fixed
inlet conditions. Once the outlet pressure drops below
a certain value, the mass flow rate is limited due to
cavitation inside the throttle. A further decrease of the
outlet pressure does not increase the mass flow rate.
This choked flow is purposely designed into fuel
injection systems in order to guarantee a nearly
independent mass rate of the servo valve independent
of the pressure level of the return passage.
The steady state flow was simulated by using the
steady RANS SST turbulence model. The Rayleigh
Plesset model was used within ANSYS CFX-5
software’s homogeneous multiphase framework. The
chart compares the computed mass flow rate
to measured data. The flow was computed on a
series of successively refined grids to distinguish
between numerical and model errors. The choking
characteristic was predicted well.
Simulating Cavitation Erosion
Next, a series of simulations was set up to model the
“planar” throttle flow, still with a RANS approach. A
two-dimensional approach based on the equations
missed the physics completely, because it could not
resolve the two counter-rotating vortices that form on
either side along the impinging body. A three-
dimensional simulation predicted the global flow
behavior well, but failed to mimic the cavitation inside
the free shear layers and above the impingement
body. With an unsteady RANS approach (URANS), the
solution converged towards the steady state result
and could not improve simulation results. These
results show that even when the mass flow rate is in
line with the experimental data, a simulation based
on the RANS form — even with advanced statistical
turbulence models — could not represent the
cavitation zones and erosion areas.
Using the DES Model
Finally, a simulation was conducted with the DES
model. The concept of the DES approach is to switch
between a statistical turbulence model and a Large
Eddy Simulation (LES) model in regions where the
turbulent length scale predicted by the URANS model
is larger than the local grid spacing. The shear stress
www.ansys.com
21
Comparison between experimental data and simulation of circular throttle flow.
transport (SST) turbulence model is employed inside
the boundary layer, while the large-scale turbulent
structures inside the free shear layer downstream of
the constriction are resolved by the LES method.
The DES simulation predicted the complex
transient cavitation in good agreement with the
experiments, as shown in the figure with the vapor
volume fractions. The momentum loss inside of the
free shear layers downstream of the throttle was much
smaller with the DES model than with the RANS
method. Due to the smaller losses, the total pressure
of the impinging jet was higher, and the pressure fell
below the vapor pressure inside the two counter-
rotating vortices in the impingement region upstream
of the stagnation point. The cavities inside the free
shear layers were forced to collapse and cause the
observed erosion on the sapphire glass windows.
Conclusion
CFX-5 is extensively used for the design of fuel
injection systems in order to guarantee proper
behavior at very different operation points over the
lifetime of the device. Advanced statistical turbulence
models provide an efficient approach to simulate
the global dependency of the mass flow rate on the
pressure difference. The DES method, however, is
needed to successfully predict cavitation erosion in
injection systems because the complex transient
interaction of large-scale turbulent structures and
cavitation is resolved.

This article was excerpted from Bauer, Iben, Voss: Simulation
of cavitating flow in injection systems, Conference
Proceedings: Berechnung und Simulation im Fahrzeugbau,
VDI-Berichte 1846, VDI-Verlag, 2004.
Data 3D-RANS DES
Vapor volume fraction
ANSYS Solutions | Winter 2005
 CFD Update: What’s New in Computational Fluid Dynamics
22
Pollen Flow Study
Improves Crop Production
Predicting pollen capture efficiency helps grow
larger kiwifruit in New Zealand’s most important
horticultural export.
By Michael Hii, John Abrahamson and Patrick Jordan,
University of Canterbury, New Zealand
Pollen collection efficiency (collected/fed by jet) by a full-open
green kiwifruit flower targeted by a pollen-laden jet from the front.
The jet exits a φ30-mm nozzle placed 200 mm away from the flower.
The experimental range refers to mean ± one standard deviation.
(a)
(a), (b) and (c) are the CFX predicted trajectories of a cloud of pollen around a single half-open kiwifruit flower under 1 m/s air current from the front,
side and back respectively (as shown by the arrows). The separation between pollen along a trajectory is 0.01 s.
www.ansys.com
Kiwifruit is the most important horticultural export
from New Zealand, with crop value relying heavily on
the size of the fruit to sell at premium prices in
international markets. Growers recognize that the
single limiting factor for a good-sized fruit is the
number of pollen grains fertilizing its flower. Sufficient
fertilization must be completed in the annual eight-day
flowering period, when the flowers are receptive. Often
bees are used to transfer pollen from male flowers to
female flowers, but bees are unreliable. Wind also
assists in this transfer, but this natural pollination is
found by growers to be insufficient for commercial
purposes.
Growers can now blow pre-collected pollen onto
kiwifruit flowers with an air jet and choose the time for
optimal flowering. However, before this study, no one
had measured or estimated the pollen capture
efficiency by kiwifruit flowers from airflow. CFX-5,
incorporating particle trajectories (calculated using the
Lagrangian framework) around a full-scale rigid flower,
has provided a sound prediction of the efficiency of
pollen capture onto the critical area for fertilization —
(b) (c)
ANSYS Solutions | Winter 2005
 (a) (b)

23

(a) Smoke visualization of flow field around a single full-open green kiwifruit flower under a uniform 1.2 ± 0.1 m/s flow from
the front in a glass wind tunnel. The CFX simulated flow field is shown in (b). The arrows indicate the flow direction.

the stigma. This prediction is adequate for air

velocities below those where petal flutter begins
(3 m/s). For example, the predicted pollen collection
efficiency indicated that wind pollination alone is
insufficient, in agreement with growers’ experience.
The use of CFX has enabled a more fundamental
understanding of the capture mechanism of airborne
pollen onto the stigma. The predicted flow fields
exhibit at least two recirculation regions, a larger
recirculation downstream of the flower and some small
eddies in the airspace between the stigma and the
petals. The spiralling large downstream vortices direct Front and side views of a full-open female kiwifruit flower (cv Hay-
the pollen backwards to the stigma area for a second ward). More than 200 female kiwifruit flowers were photographed
and studied to extract the detailed geometry for a model flower in
pollen collection after the first collection by the
CFX simulations.
approaching flow. The eddies near the filament and
stigma bushes trap the pollen, increasing their
chances of being captured by the stigma. Using the
smoke visualization around a real kiwifruit flower in a
wind tunnel confirmed the airflow patterns.
This flower model enables the recommendation
of optimum ways of spraying pre-collected pollen,
which is expensive and sometimes in short supply.
This includes the effect of jet direction onto the flower,
nozzle-to-flower distance, the diameter of the nozzle
and initial jet velocity. Selected CFX simulations were
confirmed with experiment and found to be in good
agreement. Furthermore, the use of user-Fortran in
CFX also allowed the assessment of the enhancement
of pollen collection by electrostatically charging the
pollen. The results here provide valuable insight for
manufacturers to improve operating conditions and
improve design of air jet sprayers for effective and
economic delivery of pollen onto the flowers.

A close-up look of the meshed flower model.

www.ansys.com ANSYS Solutions | Winter 2005

 CFD Update: What’s New in Computational Fluid Dynamics

24
Simulating Combustion Processes
Software tools help meet often-conflicting
legislative and market requirements
in power generation, aviation,
automotive and other industries.
By Dr. Jorge Carregal Ferreira,
ANSYS Germany GmbH

Combustion remains one of the most important

energy conversion processes worldwide. In fact,
more than 80% of the global primary energy
demand is based on combustion. It is used in
many important industrial applications, such as gas
and steam turbine combustors, internal combustion
(reciprocating piston) engines, furnaces, boilers and
gasifiers. A gas turbine combustor is an industrial
example where combustion plays a key
These devices are key components in power role in the energy conversion process.
generation, aviation and automotive industries. In Courtesy of MTU Aero Engines GmbH.
addition, in the oil, gas, glass, minerals, chemical and
process industries, chemically reacting flows are
critical to many large production processes. Due to has been achieved by optimizing designs through
limited resources of fuels and ever-tougher simulation and by reducing the number of experiments
environmental constraints, there is a strong financial through virtual prototyping. In addition, CFD can
incentive to improve combustion devices in order to provide detailed in-flame information, which, in many
enhance the overall efficiency of fuel consumption and cases, is difficult or even impossible to measure. For
to reduce pollutant formation. example, the heat release in a combustor, which is
An increasing number of industries have found determined by the magnitude of chemical source
that Computational Fluid Dynamics (CFD) is a cost- terms, is a property that cannot be measured directly
effective tool for accelerating the design and but is immediately available from a CFD calculation.
development process in order to meet these often-
conflicting legislative and market requirements. This Analyzing Complex Interactions
Combustion is a complex interaction of flow and
mixing comprising a large number of elementary
chemical reactions and species, heat release and heat
Comparison of Models with Real-World Data transfer, turbulent flow and occasionally multiple
phases with mass transfer, for example when the fuel
is a liquid like oil or a solid like coal.
As part of the ongoing model testing process at ANSYS, predictions of the
CFD models must address all these physical
implemented combustion models are compared with laboratory and
processes and their interactions. With ANSYS CFX
industrial-scale flames for which extensive experimental data exists. One
software, ANSYS Inc. offers an advanced general-
example is the Sandia Flame D, which is one of the standard benchmark
purpose CFD package that is well suited to the
flames of the internationally recognized TNF workshop. ANSYS, Inc.
analysis of combustion devices and processes. At the
sponsors the activities of the TNF workshop and therefore benefits from
core of CFX is leading and unmatched solver
the latest model developments of leading groups in the combustion
technology which is used to solve accurately all
research community. A large amount of experimental data that was
required transport equations for mass, momentum,
obtained for specific laboratory flames can be used to verify whether the
energy, turbulence quantities, and chemical species.
applied combustion models can predict, for instance, the local concentra-
tion of minor species or temperature fluctuations.

www.ansys.com ANSYS Solutions | Winter 2005

Advanced turbulence models in CFX allow an
excellent prediction of the flow and the mixing
between fuel and oxidizer, which is a prerequisite of
any combustion process. In addition, flexible handling
and generation of unstructured meshes provides a
high efficiency in discretizing and modeling of often
complex geometries.
Combustion Modeling Strategies
In most industrial applications, combustion and
chemical reactions are strongly coupled to turbulent
fluctuations. This represents a modeling challenge that
is addressed in CFX by the implementation of two
complementary turbulent combustion modeling
strategies. In the first of these, the user is allowed to
specify single or multiple-step chemistry and the
chemical reaction rates are calculated by an Arrhenius
expression or by the Eddy Dissipation Model (EDM),
which sets the reaction rate proportional to a turbulent
mixing time scale, or by a combination of both. The
graphical user interface of CFX allows very intuitive
and fast selection of chemical species and reaction
steps, which can be taken from the library provided or
defined by the user. The EDM represents a quick and
efficient way of predicting overall heat release, and the
distribution of major species and temperature.
This method is commonly applied to gas turbine
combustors to predict temperature distributions and
associated heat loads, which can then be used to
calculate and optimize the required cooling flow. If the
fuel is liquid or solid, one can use one of the Eulerian
or Lagrangian-based models, which are available in
CFX to simulate the multiphase mass and energy
Particle Temperature [K] Residence Time [s]
www.ansys.com
25
Modern laser measuring techniques are used to provide a reliable
experimental data set which covers mean and fluctuating values of
velocity, temperature, and chemical species. This data is used to validate
combustion models. Visit http://www.ca.sandia.gov/TNF/abstract.html
for more information on the TNF workshop. Picture shows the piloted
Sandia Jet Flame, one of the standard benchmark flames of the
internationally recognized TNF workshop.
transfer in combination with the combustion process.
For example, pulverized coal can be used as a fuel in
large power generation furnaces where, as part of the
simulation, the release of volatiles, their subsequent
gas-phase oxidation, and heterogeneous oxidation of
the residual solid char are taken into account.
The second combustion modeling strategy can
be classified as a presumed PDF method combined
with tabulated chemistry. PDF stands for “Probability
Density Function” and is a powerful procedure to
incorporate the statistical nature of turbulent
fluctuations of all variables, in particular, of species
and temperature on the chemical source terms.
As part of this methodology, mixing of the fuel and
oxidant is represented by the mixture fraction, which
can be seen as a normalized equivalence ratio.
The location of the premixed flame front is determined
by a reaction progress variable, where the turbulent
burning velocity is used as an input parameter. Finally,
This CFX analysis shows paths of coal
particles in a coal-fired furnace. The
particle paths can be colored with the
variable of choice. Residence time can
be used to determine if coal particles
have sufficient residence time to react
before they leave the reactor chamber.
ANSYS Solutions | Winter 2005
 CFD Update: What’s New in Computational Fluid Dynamics
26
Axial distribution of mean temperature and mean mass fraction of CO is shown on these charts. The flamelet
libraries include data for radicals such as OH and O which are needed for the prediction of NO.
the local thermochemical state is obtained from
flamelet libraries, which are pre-calculated by CFX-RIF,
an easy-to-use flamelet library pre-processor. This
modeling strategy can be applied to all types of
flames, including non-premixed, partially premixed,
and fully premixed flames. The first advantage of this
approach is that it takes account of detailed chemistry
more rigorously than the EDM and with less
computational effort because the chemistry is
tabulated in advance. The second advantage is that it
models the complex interaction between flame
This ANSYS CFX analysis demonstrates the ratio between turbulent
and molecular viscosity on a surface of constant vorticity of the
swirl flow entering a generic combustor. This information can be
used to identify the local and instantaneous intensity of turbulence
which has a strong impact on the mixing process between fuel and
oxidizer and on the propagation and position of the flame front.
www.ansys.com
propagation and turbulence in premixed or partially
premixed combustion more realistically than the EDM
model.
Accurate Predictions
Both the previously mentioned combustion modeling
strategies and their sub-models are continuously
improved. For instance, the ability to use the
combustion models with new turbulence models such
as Large Eddy Simulation (LES), Detached Eddy
Simulation (DES), and Scale Adaptive Simulation
(SAS). SAS is a new method currently under
development, which dynamically switches between
LES and RANS depending on the required mesh
size that is needed to resolve the energy containing
turbulence length scales. The advantage of these
computationally more demanding simulation methods
is that large-scale turbulence can be predicted more
accurately, which in turn, gives better predictions of
heat release and temperature distribution.
The use of CFX provides an improved under-
standing of the overall behavior of combustion
devices, the impact of operating parameter settings
and of details of the combustion processes in
industrial applications. Specifically, CFX can be used
to analyze the flow, mixing, heat release and heat
transfer, required cooling flows, and formation and
production of chemical species including pollutants. In
combination with its superior models for multiphase
flow, CFX can be used to analyze liquid and solid fuels,
such as gasoline, diesel, coal and biomass. ANSYS
CFX can help to accelerate the development and
design cycle by optimizing the performance of
combustion equipment of all types.

ANSYS Solutions | Winter 2005
Simulation at Work
Fewer Prototypes
with Up-Front Analysis
Xerox refines the design of their new iGen3 digital printing
system early in the design cycle.
Overview
Xerox Corporation is a $15.7 billion technology
and services enterprise that helps businesses
deploy smart document management strategies
and find better ways to work. Its intent is to
constantly lead with innovative technologies,
products and solutions that customers can
depend upon to improve business results.
The company provides the document industry’s
broadest portfolio of offerings. Digital systems
include color and black-and-white printing and
publishing systems, digital presses and “book
factories,” multifunction devices, laser and solid
ink network printers, copiers and fax machines.
Xerox services include helping businesses
develop online document archives, analyzing
how employees can most efficiently share
documents and knowledge in the office,
operating in-house print shops or mailrooms, and
building Web-based processes for personalizing
direct mail, invoices, brochures and more.
Headquartered in Stamford, CN, Xerox is number
130 among the Fortune 500 and has 60,600
employees worldwide, including 35,100 in the
United States.
www.ansys.com
27
Challenges
To expand its business beyond conventional office
copiers, Xerox embarked on a strategy to penetrate
the commercial printing market currently dominated
by traditional offset presses. The company set out to
develop the iGen3: a toner-based digital printing
system with image-quality “look and feel” comparable
to offset presses - yet with a faster speed of 100
pages per minute, greater economy for short-run
press jobs, and the ability to customize each page
with variable information to create personalized
brochures, tailored catalogs, on-demand books,
newsletters and direct-mail pieces.
Over $1 billion was poured into the R&D project,
which resulted in more than 400 patents on a product
that was the most complex system ever developed by
Xerox. The payoff, hoped company executives, would
be a major revenue-generator in the years to come
that would lift Xerox’s profits and strengthen the
company’s leadership position in the fiercely
competitive printing and copying market.
The challenges for the engineers developing the
iGen3 were immense. The product pushed the limits of
the technology for speed and performance, and every
one of the thousands of interconnected parts and
assemblies had to operate flawlessly for years of
reliable service that is the hallmark of Xerox’s brand
value. Moreover, design work had to be completed
quickly so that the product launch could take
advantage of the window of opportunity in the rapidly
evolving market.
ANSYS Solutions | Winter 2005
 Solution
Xerox met these challenges with a simulation-based
design approach in which analysis is performed up
28 front in development. This process enables engineers
to find and correct potential problems earlier in the
cycle as compared with the traditional method of
building and testing numerous physical prototypes
near the end of development.
The company performs much of the structural
analysis on parts and assemblies during conceptual
design with ANSYS DesignSpace. Full CAD
associativity with CAD allows the system to build
analysis models automatically and enables users to
easily modify designs based on simulation results.
Also, wizards and a range of other automated features
for meshing control, convergence checking and
multi-body contact permit engineers to analyze their
own designs on Sun Blade workstations. For more
detailed simulation, models can be re-used in full
ANSYS by a central group of dedicated analysts for
fluid flow, vibration, heat transfer, multiphysics and
other advanced analyses.
In development of the iGen3, DesignSpace was
particularly useful in quickly simulating the many
interconnected parts and assemblies that are welded,
glued, press-fit or otherwise joined. In the analysis of
welds in a support pin, for example, DesignSpace
imported the geometry directly from I-DEAS,
automatically recognized bonded contact between the
welds joining the individual parts (the pin, baseplate
and a reinforcing gusset), and allowed for different
material properties and dissimilar meshes of
contacting parts.
The analysis allowed engineers to determine
structural deflection and weld stresses much faster
than would have been possible with other simulation
packages, which would have required the user to
manually define bonded contact and build separate
models for each of the parts — including the six welds
holding the assembly together.
Baseplate Displacement
Benefits
Simulation-based design has enabled Xerox to
significantly reduce the number of physical prototyping
cycles late in development, when making design
changes is costlier and more time-consuming. By shifting
engineering efforts up front in development with a heavy
emphasis on early analysis, engineers can more readily
study product performance, spot potential problems,
evaluate alternatives and refine the design in the
conceptual stage of development.
By using this approach, Xerox was able to bring the
iGen3 to market on time while keeping costs in line and
maintaining quality and reliability. The machine is now
regarded as one of the premier flagship products at Xerox
and sales have been strong. A growing number of
commercial printers and in-house company printing
departments are now taking advantage of the machine’s
high speed in producing impressions almost 50% faster
than other competitive toner-based digital presses.
Since its introduction, the iGen3 has won
numerous prestigious awards, including the industry’s
Gold Ink Award, the Gallery of Superb Printing Award
from the International Association of Printing House
Craftsmen and the InterTech Technology Award
from the Graphic Arts Technical Foundation for design
innovation and production efficiencies.

“DesignSpace is an ideal tool in early product development for conceptual

simulation-based design and was instrumental in the success of developing
the iGen3 digital printing system. Also, full ANSYS is one of our primary analy-
sis tools for advanced simulation, particularly in multiphysics applications
where multiple physical factors must be evaluated. In typical product devel-
opment programs at Xerox, simulation-based methods using these types of
predictive tools have definitely helped reduce the number of prototype testing
iterations, each costing tens of thousands of dollars and weeks of time. In the
end, development time and costs are reduced. But more significantly, our
high-quality standards are met and time-to-market is shortened in developing
innovative winning new products such as the iGen3, enabling Xerox to grow
top-line revenue and increase market share.”
Weld Stresses
Dr. Korhan Sevenler
Director, Product Lifecycle Management
Xerox Corp.

www.ansys.com ANSYS Solutions | Winter 2005

Managing CAE Processes

The Value of
Simulation-Driven Design 29

Analysis-led product development is a paradigm shift that

could save a growing number of manufacturers.
By Dr. Jason Lemon
Founder and Chairman
International TechneGroup Incorporated

In a project to develop a new all-wheel–steering mower for Murray, Inc. ITI used ANSYS in the conceptual phase of design to study critical
mower parts such as this inner pivot bracket and cam plate.

Conventional ways of developing new products can be long, arduous and costly. Companies rush to
create designs so they can analyze, build and test physical prototypes to verify that the product per-
forms as required. Problems are detected late in development, when they are expensive and time-con-
suming to correct, and frequent design changes inevitably occur along the way. Consequently, costs
skyrocket, schedules slip and products often fall short of market demands, customer expectations and
business requirements. What started out as a good idea often flops in the market because of outdated
product development methods.

Few companies can continue to operate successfully
in today’s world using these outdated methods.
Indeed, many such manufacturers will have a difficult
time competing in tough markets if they fail to update
their product development processes to take
advantage of the latest technology and best-practice
know-how.
Some companies have improved upon conven-
tional processes by using what can be termed digital
engineering, where CAD models replace paper draw-
ings and some level of computer analysis is performed
after designs are created. This results in incremental
time and cost savings, mostly due to increased engi-
neering productivity. But this savings does not even
come close to offsetting the enormous time and
expense required for multiple physical prototypes in
the traditional design-analyze-build-test approach.
To break this cycle, successful manufacturers are
adopting simulation-driven approaches that represent
a paradigm shift in product development. Simulation is
performed first in the concept stage to explore alterna-
tives, spot flaws and optimize product performance
before the detailed design is created and the first
prototype is built. In this way, important decisions can
be made on functionality, geometry and materials early
in the cycle based on simulation results. Moreover,

www.ansys.com ANSYS Solutions | Winter 2005

 Managing CAE Processes
rather than doing multiple operations serially, as in
conventional development, companies work concur-
rently in performing many phases of product develop-
ment, from initial product planning to completion of
30
detailed design.
Hardware prototypes then are built and tested to
validate these virtual prototypes. Of course, some
changes still occur during physical testing, but the
number of costly and time-consuming changes is
reduced by orders of magnitude. Likewise, risk/recall
programs are substantially less. In this way, designs
satisfy engineering and business objectives much
more closely than those developed using the old
design-analyze-build-test cycle.
Tools and Technologies
A wide range of technologies are used in simulation-
driven product development, including multiphysics
simulations, electromagnetics, fluid dynamics, struc-
tural finite element analyses, failure analyses and
design optimization. Technologies from ANSYS, Inc.
are of significant value in implementing simulation-
driven design, particularly those tools that facilitate
up-front simulation, efficient evaluation of alternative
designs, iterative modification of designs based on
simulation results and collaboration throughout the
product development process.
DesignSpace, for example, enables engineers
with minimal finite element experience to perform first-
pass analysis right on their desktops in studying the
performance of conceptual ideas. DesignXplorer is a
tool for exploring product performance using response
surfaces to graphically depict analysis results over an
entire range of multiple input variables. Workbench
has the ability to control simulation variables as well as
CAD parameters, thus providing bi-directional asso-
ciativity with CAD software. Also, simulation wizards
and templates in Workbench automate repetitive oper-
ations to make the product development process
Analysis in Action
In one recent project utilizing simulation-driven design, ITI
consultants used ANSYS technology in a collaborative product
development program with lawn and garden equipment
manufacturer Murray, Inc. to design a tractor mower with all-
wheel steering capability.
One of the primary goals was to minimize steering forces so the
mower could be turned easily without hydraulic power-assist
systems that add complexity and cost to the design. The
manufacturer also wanted the mower to have a tight turning
radius. High reliability was mandatory, as the Murray brand image
is built on product quality and value. Staying on schedule was
paramount for the mower to be introduced before the peak selling
season.
www.ansys.com
faster and more consistent throughout the company.
In this way, process automation enables companies to
embed best practices into the design cycle and facili-
tates collaboration between separate groups.
Using these and other types of analysis tools in
up-front development, simulation guides the direction
of the design to optimally satisfy performance, struc-
tural integrity, reliability, durability, cost and other
requirements. Most important, simulation guides criti-
cal trade-off decisions to balance competing product
objectives — reliability, cost and weight requirements,
for example. Usually hundreds of concept alternatives
are evaluated before detailed design is begun.
Where the Approach Pays Off
An ANSYS partner, International TechneGroup Incor-
porated (ITI) has helped hundreds of clients for over
two decades implement simulation-driven processes,
which we call more specifically Systems Engineer-
ing/Analysis Leads Design (SE/ALD™).
Using this process and simulation tools such as
ANSYS, ITI has helped numerous companies achieve
significant benefits in the development of a wide range
of products in various industries, including automotive,
aerospace, electronic equipment and consumer prod-
ucts. An agricultural tractor manufacturer developed a
new product in half the time of conventional methods
at a 30% cost reduction, with the tractor setting best-
in-class performance standards in noise reduction and
turning radius. Similarly, an aircraft engine
manufacturer got to market a full year before its
closest competitor by evaluating more than a thou-
sand concept alternatives using simulation-driven
design. A power tool manufacturer produced a cord-
less driver on its initial simulation-driven development
program that increased market share 63% and is now
being implemented throughout the company.
Early in conceptual design, ANSYS Structural™ software was used
to study 12 critical components that were identified as contribut-
ing most to mower performance and reliability. The software
determined stress distributions over a range of operating
conditions for these components and was used as an input for
fatigue-life predictions based on a typical customer duty cycle
measured in the field on an existing two-wheel-steering mower.
Performing simulation up-front in mower development allowed
engineers to study several alternative configurations and optimize
the final design. As a result, the new Murray mower was
introduced on time and has maintained good reliability. Turning
force was minimized so that costly, complex hydraulic power-
assist systems were not required. Moreover, a tight turning radius
of only 14.25 inches was achieved, well below the original goal of
18.5 inches.
ANSYS Solutions | Winter 2005
Paradigm Shift in Product Development

Traditional Process

Design then Evaluate

Create geometry Analyze, build and test

– Layout drawings – Geometry layouts

31
– CAD wireframe and surfaces – Experience, management
– Solids packaging opinion
– Test prototypes

Simulation-Driven Design

Simulate Design
then
Modify and optimize Define and create geometry Traditionally, companies rush to create
validated baseline and material requirements designs so they can analyze, build and test
computer models that achieve targets physical prototypes. Problems thus are
detected late in development, when they are
– Virtual layout/parametrics – Virtual concept layouts that expensive and time-consuming to correct. In
– System engineering and satisfy targets the simulation-driven approach, analysis is
performed in the concept stage to optimize
analysis-led design – Multiple iterations until design product performance, then detailed design is
– Active QFD target setting concepts satisfy targets created and the first prototype is built.
– Tradeoffs and target cascading

In these types of applications, most of the benefit
of simulation-driven development is derived from the
fewer number of engineering changes made after
detailed design begins, as compared with conventional
methods. Building multiple physical prototypes is
rarely practical or cost-effective. Conversely, validated
virtual models representing a number of design alter-
natives can be modified, analyzed and evaluated to
lead design decisions at a fraction of the time and
cost. With simulation-driven methods, overall design
costs are cut because changes in the latter stages are
reduced with successful products assured at the
beginning of detailed design.
Setting Targets for Product Performance
The simulation-driven process relies heavily on setting
targets for overall product performance requirements
deemed necessary to strengthen market share and
brand image. Targets are cascaded from system to
subsystems and assemblies to individual compo-
nents. In automobile design, for example, dynamic
motions that characterize ride and vibration targets are
translated into resulting forces and displacements in
the vehicle suspension, which in turn are used to
establish design targets for shock absorbers and indi-
vidual bushings and connectors in terms of stress and
deformation determined from structural analysis. The
product can then be designed from the component
level up to satisfy these various levels of requirements,
with much greater assurance of overall product
success.
Target setting often involves establishing a base-
line consisting of your own existing product, or that of
one or more competitors. Products are tested to
compile a quantitative knowledge base of information
on product behavior that will be critical in correlating
the analytical and the physical, and in providing empir-
ical data where computer modeling is impractical. In
this case, hybrid models may be created combining
virtual models and test data to accurately represent
the system. Once this baseline is established, various
components and subsystems then can be modified to
evaluate multiple alternatives before beginning the
detailed design process.
Customer Usage Profiling (CUP) and related load
and duty cycle definitions serve as a foundation
for overall product target setting. In a typical CUP
program, products fitted with instrumentation and
data-collection gauges are provided to customers for
use over an extended period of time. This indicates,
often with surprising results, how products are actually
used (and abused) in the real world. For example, the
armrest in a car is designed to support the weight of
an arm, but will it safely support the weight of a child
climbing into the back seat? CUP saves warranty,
liability and material costs by making sure components
are neither under nor over designed.
Stepping Outside the Comfort Zone
Achieving the orders-of-magnitude gains possible
with simulation-driven product development requires a
major shift in priorities at many companies. Because
greater engineering effort is expended up-front,
companies must focus more resources earlier in the
process, with as much as two-thirds of their engineering
budget spent before detailed design begins.

www.ansys.com ANSYS Solutions | Winter 2005

 Managing CAE Processes
Reducing Time to Market and Risk
Traditional “Design/Build/Test” Development Environment
LAUNCH
32 Design
Build
Minimal
development Test (Development)
occurs here Design
Too much
development
occurs here
Fully Implemented “Analysis Leads Design” Environment
LAUNCH Much lower
Simulation-Driven Design risk exists
here
Build
Much more
development Test (Verification)
occurs here
Some development, but
focus is on verification
This can represent a tremendous readjustment of
rewards and objectives for mid-level managers,
whose performance has been traditionally evaluated
on rushing to final design in the least amount of time
and at minimal expense. In this respect, commitment
from top management is critical in providing sufficient
funding early enough in the program to make simula-
tion-driven design as it should. True paradigm shifts
represent drastic, sometimes uncomfortable, change.
It’s not surprising, therefore, that they often are met
with organizational resistance as leaders and workers
step outside their comfort zone.
Key Best Practices
Achieving the greatest gains in moving from traditional design-
build-test development programs to simulation-driven design
processes often requires companies to change the way they
operate. Key best practices that facilitate these changes include
many of the following:
• Concurrent product development processes optimize overall
time-to-market and development productivity.
• Incorporating the “Voice of the Customer” (customer demands
and expectations) into the product content as measurable and
predictable customer-focused targets ensures that the
product meets or exceeds market expectations.
• Full application and integration of systems engineering assure
that the customer-focused targets are methodically managed
from all system levels through component levels.
• The integration of simulation-based analysis and problem
solving within the systems engineering discipline virtually
www.ansys.com
PLANNED ACTUAL
RELEASE RELEASE
Too much risk Companies struggle
exists here to meet market
and business
requirements
Build
Test (Verification
PLANNED Schedules slip and unexpected
& ACTUAL problems arise in traditional product
RELEASE development that relies on numerous
design/build/test cycles with physical
Companies lead in prototypes. Simulation-driven
time-to-market and development yields breakthroughs in
productivity reducing time-to-market and product
risk by considering multiple design
alternatives and refining product
performance early in the cycle.
The key to making these changes is under-
standing the enormous value of simulation-driven
design to a company’s business potential, brand
image and profitability. Companies willing to make
the necessary changes and investments to achieve
these benefits will likely emerge as leaders in the
coming years, while less perceptive competitors will
fade away.

International TechneGroup Incorporated (www.iti-global.com)
consults with a wide range of manufacturing companies on
implementation of simulation-driven development and other
processes related to engineering analysis and concurrent
product and manufacturing process development.
designs out problems and validates new designs before
expensive prototypes and tooling are built. This includes the
use of CAD/CAE models to make design decisions during
concept selection.
• Product testing is performed ahead of and concurrent with
development programs to understand and quantify current
product behavior and applied in simulation-based validation
(such as competitive benchmarking, current product
baselining, and customer usage profiling).
• Effective global collaboration, communication and data
management to support development, including shared use of
information and solutions for analysis, testing, and data
management.
• Disciplined optimization and integration of the above items is
achieved through applying best practices in cross-functional
teams, and product, project and risk management.
ANSYS Solutions | Winter 2005
 Tech File
Restarting ANSYS
LSREAD
/POST1
Quickly recover from a mistake, avoid a
nonconvergence problem or try different
/SOLUloads without re-running the entire analysis.
RESCONTROL
By John Crawford
Consulting Analyst
Here’s an all too familiar scenario.
You look at the results of compli-
cated nonlinear analysis and
discover that you placed a load
on a wrong node. Oops. The
good news is that you detected
the problem and identified the
cause. The bad news is that it will take another three
days to rerun the entire analysis.
Don’t panic. There’s still hope — if you told
ANSYS before you ran the solution to save the results
for each load step so you could do a restart from any
one of them. This means that instead of rerunning the
entire analysis you can restart the analysis from the end
of a load step, apply the proper load, and continue on.
This won’t take nearly as much time. The day is saved.
Restarting an analysis is a great way of
recovering from a mistake or a nonconvergence
problem, or for just trying a different loading scenario.
ANSYS offers two types of restarts — single-frame
and multi-frame —that can be used for all static analy-
ses, 2-D electromagnetic harmonic analyses and tran-
sient analyses that use the full solution method.
Single-Frame Restarts
A single-frame restart is a continuation of a prior
solution, picking up where the last solution left off. It’s
useful for recovering from a nonconverged solution or
continuing an analysis with another set of boundary
conditions. Single-frame restarts are available for all
analyses that support restarts and are the default
restart method for all solutions except nonlinear static
or full transient analyses. To do a restart, return to the
solution processor (/SOLU) and use the ANTYPE
command to tell ANSYS to, instead of doing a new
analysis, do a restart of the current analysis
(ANTYPE,,REST).
www.ansys.com
33
If you have run several load steps and reviewed
them in /POST1, be sure the one read into /POST1
using the SET command is the one with the boundary
conditions you wish to use as the basis for doing a
restart. When ANSYS reads a result set, it also reads in
the loads that correspond to that result set. Once you
are certain that the desired boundary conditions are
active, you can make any changes you wish and
restart the analysis. Alternatively, if you have written
the original loads to load step files before you ran the
solutions, you can use the LSREAD read to retrieve the
boundary conditions you wish to use as the basis for
the restart. As with all analyses, it’s always good prac-
tice to review the boundary conditions before you run
the solution and make sure that they really are what
you think they are.
Multi-frame Restarts
Multi-frame restarts are the default restart method for
nonlinear static and full transient analyses, and offer
the ability to restart from any previous converged solu-
tion. Before running the solution, you can use the
RESCONTROL command to identify how many solu-
tions you would like to save data for to use for
restarting. The default value is to only save the data for
the last solution.
There are some pretty neat applications for multi-
frame restarts. For example, let’s say you are analyzing
an assembly held together with bolts that has contact
elements with friction between the components, has a
temperature applied to it, and has operating loads
imposed by applying forces to some of the nodes. The
load step history for this problem might go like this:
Load Step 1: Use pre-tension elements to “tighten”
the bolts.
Load Step 2: Apply temperatures to the assembly.
Load Step 3: Apply operating forces to nodes.
ANSYS Solutions | Winter 2005
 Tech File
34 If the model is fairly large, it could take several days to
get through all three load steps. It might be wise to use the
RESCONTROL command to save the restart data for each
load step so you can restart the analysis from any of them.
This would save time when you want to apply different
(restart from load step 1).
/SOLU
loads (restart from load step 2) or different temperatures
Another benefit offered by multi-frame restarts is that
you can restart an analysis that has had trouble converging.
RESCONTROL
You can restart from an earlier load step, change some
convergence parameters, and try solving it again.
Keep in mind when using multi-frame restarts that
when you tell ANSYS you want to do a multi-frame restart
from a specific load step, the software will immediately
remove data for subsequent load steps from the working
directory. For example, suppose you ran an analysis that
had 25 load steps and used RESCONTROL to tell ANSYS
to write out data for each load step so you can restart from
any one of them. When the analysis is done running, you
will have restart files in the working directory ranging from
jobname.r001 through jobname.r025.
After reviewing the results in /POST1, you decide that
you want to restart the analysis from load step 20. You
return to /SOLU and use the ANTYPE command to tell
ANSYS you want to do a restart, but you accidentally
type in 2 instead of 20. When ANSYS executes the
ANTYPE command, it will remove the files jobname.r003
through jobname.r025 from the working directory. So if you
accidentally tell ANSYS to restart from load step 2 instead
of load step 20, you can’t go back and correct your mistake.
You’ll only have the option to restart from load step 2 or an
earlier solution for which restart files still exist. (Helpful Hint:
If your operating system has a trash can, you might find the
deleted files there.)
There is a price to be paid for the ability to perform
Restarting an
analysis is a great way of
recovering from a mistake or
a nonconvergence problem
www.ansys.com
LSREAD
/POST1
multi-frame restarts. ANSYS writes a number of files (such
as jobname.rdb and jobname.rnnn) that allow a multi-frame
restart to be done, and these files might consume more disk
space than you have available. You may wish to do away
with the ability to do multi-frame restarts for nonlinear static
and transient structural solutions by using RESCONTROL,
DEFINE, NONE before solving. This will force ANSYS to
revert back to writing the files that are needed for a single-
frame solution.
Other Uses for Restarting
You can use single and multi-frame restarts to make
adjustments to a model to get the behavior you are looking
for. An example of this would be adjusting contact element
stiffness until you are satisfied with the results. You can run
a solution and then view the results in /POST1 to see how
much penetration is taking place and how the stress and
displacement results look in the zones you are particularly
interested in. You then return to /SOLU, change the contact
element stiffness, restart the analysis and solve it again.
When the solution is done, go back to /POST1 and review
the results. If the changes in the stress and displacement
results are small enough to be acceptable, you can consider
the results to be adequate for your needs. If you’re not sure,
you can return to /SOLU, change the contact element
stiffness again, restart the analysis and solve it again.
This is a useful technique for getting contact element
problems to converge because “soft” contact elements
tend to converge more quickly than “hard” contact
elements, but “hard” contact elements usually provide
more accurate answers. By progressing through a series of
analyses, you can start with a soft contact stiffness to get
the solution going, and then increase its stiffness through
subsequent solutions until you are satisfied with the results.
This approach can be used for single-frame restarts as long
as friction, plasticity, creep and other phenomena that
make an analysis path dependent are not present. If the
problem is path- dependent you must use multi-frame
restarting to restart the analysis from a load step that ends
before the path-dependent phenomena occur. As always, it
is the responsibility of the analyst to ensure that the results
are realistic and true to what nature already knows.

Practice Doing Restarts
Learning and using the restart features available in
ANSYS will allow you to work more efficiently. The details
are discussed in the Basic Analysis Procedures Guide
and are worth looking at. Build yourself a simple model
that will allow you to try different features. A macro called
runcontact.mac is available at the ANSYS Solutions Web
site at www.ansyssolutions.com that will allow you to try
different settings and practice doing single-frame and
multi-frame restarts.
ANSYS Solutions | Winter 2005
 Tips and Techniques
SolidWorks Support
Assembly.

Efficient Power
Spectral Density 35

Techniques
Part 2 of 2:
More ways to reduce run times — expand only
significant modes, limit result calculation to one area of
interest, use the less rigorous single point spectrum method.
By ANSYS, Inc. Technical Support

Power Spectral Density (PSD) analysis is a ! Sample APDL to only

statistical dynamic analysis dealing with the effects of ! expand significant
a random excitation: a base excitation (displacement, ! modes (> SIGNIF)
velocity, acceleration) or nodal excitation (force, /solu
pressure.) Details on performing PSD analyses in antype,modal
ANSYS are given in Chapter 6 of the Structural modopt,lanb,MODES
Analysis Guide as well as Section 17.7 of the ANSYS spopt,sprs
Theory Reference. svtyp,TYPE
Two CAD models were used to quantify the freq,FREQB,FREQE
efficiencies of some sample PSD analyses: a Solid- sv,0,1,1
Works support assembly meshed with SOLID92 sed,X,Y,Z
elements and connected with TARGE170 and allsel,all
CONTA174 contact elements, and an Inventor oilpan solve
part meshed with SHELL181 elements. finish
Part one of this article discussed how total run /solu
times in a PSD analysis can be greatly reduced by expass,on
using the covariance significance ratio on the PSD- mxpand,,,,yes,SIGNIF
COM command. Now we’ll take a look at ways to save solve
additional time by expanding only significant modes, finish
limiting result calculation to one area of interest, and
using the less rigorous single point spectrum method. There is no way to quantify the error introduced
Some of the suggested short cuts may be valuable by this technique, but in most structures there are
simply to obtain initial result estimates. many modes that are not excited at all by excitation in
a certain direction. Therefore, this technique is espe-
Reducing Number of Modes with MXPAND cially well suited to single direction base excitation
problems.
Reducing the number of modes that are expanded will
speed up the expansion itself and make the PSD Table 1: Solution Times vs. Expanded Modes.
mode combination and the calculation of a response
Element Types Number of Modal + PSD UX UY UZ
PSD faster as well. A technique for reducing the num- Modes CP-Time/
ber of expanded modes is to run a dummy Single Expanded Total Time (sec.)
Point Response Spectrum (SPRS) with a significance SOLID92 Test Model 77 5570 / 6560 7.74e-03 2.17e-02 1.69e-03
level specified with the MXPAND command.
A sample set of APDL commands is shown. SOLID92 Test Model 22 1870 / 2400 7.73e-03 2.17e-02 2.44e-03
During the modal analysis (first SOLVE), mode SHELL181 Test Model 200 36100 / 42800 9.20e-04 1.16e-02 8.16e-04
expansion is not requested. Instead, a response SHELL181 Test Model 86 14300 / 14700 8.87e-04 1.15e-02 1.74e-03
spectrum analysis is specified with a base excitation
direction (SED) the same as the PSD analysis to be run
As shown in Table 1, a significance level of 1e-3 for MXPAND in the solid92 model
later. The second SOLVE performs a mode expansion
resulted in 22 modes being used instead of 77. In the shell181 model, a significance
for only the modes with a significance level greater level of 1e-3 reduced the number of modes from 200 to 86. The significance level
than what is specified with the SIGNIF parameter of was based on participation factors in the Y direction, so the maximum 1-sigma UX
the MXPAND command. In a response spectrum and UZ values vary with the number of modes, but the UY values compare closely.
analyses, the significance level is defined as the ratio The dummy spectrum could have excited all three directions (SED,1,1,1) to produce
of the mode coefficients, so this process will result in better results, but the savings would have been less.
only modes with significant mode coefficients being
expanded.

www.ansys.com ANSYS Solutions | Winter 2005

 Tips and Techniques

Reducing File Size with OUTRES

The size of the results file can also be reduced by only
calculating the result types of interest in only the regions
of interest. Typical input is
36
OUTRES,ALL,NONE
OUTRES,NSOL,1
OUTRES,STRS,1,Cname ! Cname is the name of an
Inventor Oilpan Part.
element component

Faster PSD with SPRS

A less rigorous, but much faster PSD analysis can be There was no attempt to use any of the techniques
performed using Single Point Response Spectrum previously discussed to make the PSDCOM solution
(SPRS) with SVTYP,4. more efficient, so the memory and computation time is
A type 4 spectrum analysis utilizes simplified PSD larger than necessary. Still, the SPRS-type of PSD
calculations, which may be suitable for single PSD base analysis with SRSS mode combination method will
excitations with a relatively flat spectrum.
The procedure for SVTYP,4 is the same as for a Table 3: Run Times and Memory for Standard vs SPRS PSD.
regular Response Spectrum analysis. The user does Element Method / Memory Total UX UY UZ SX SY SZ SEQV
not specify which of the constrained DOF are to be Types No. of (MB) Time
used for the base excitation — all constrained DOF are Modes (sec.)
assumed to be excited. The PSD curve is input via
FREQ and SV commands, and it must take the form of SOLID92 Modal 380 38,400 8.33e-03 2.34e-02 1.80e-03 3350 4060 2670 10,400
acceleration2/Hz rather than g2/Hz. Lastly, mode com- Model 139 1640 93,300 8.33e-03 2.34e-02 2.06e-03 3780 4430 3110 10,000
PSDCOM 130 970 8.33e-03 2.34e-02 1.82e-03 3470 4160 2760 10,000
bination is done with a method such as SRSS, CQC or SRSS 12 130 2350
GRP. A comparison between the rigorous PSD method CQC 12
as well as the alternative SPRS-type PSD method is
outlined in Table 2. SHELL181 Modal 260 19,500 9.20e-04 1.16e-02 8.16e-04 1830 1190 2040 1940
Model 200 470 29,300 9.36e-04 1.16e-02 7.78e-04 1610 1190 1660 1580
Table 2: Comparing Standard and SPRS PSD. PSDCOM 31 530 9.15e-04 1.16e-02 8.04e-04 1910 1190 2230 2110
SRSS 49 31 11,000
Types of Calculations Standard PSD SPRS PSD CQC 49

Non-Uniform node excitation Yes No

PSD Specification Various Acceleration only always be much faster and use less memory. The
response in the direction of excitation (UY) is captured
Graph PSD Curve Yes (PSDGRAPH) No well with SRSS. The CQC combination method
Damping Specification Various MP,DAMP includes the ‘coupling’ effect of modes at the expense
of increased computation time. It is still much faster
Mode Combination PSDCOM SRSS, CQC, GRP, …
than PSDCOM if the number of included modes is not
Multiple Correlated PSD’s Yes No too high. The case of 12-13 modes is fast, but with 49
Spatial Correlation Yes No modes the extra load case combinations result in a
much slower solution.
Wave Propagation Excitation Yes No In the SHELL181 model, the maximum RMS
1 sigma results Results File (LS 3-5) Load Case Combination SEQV values do not match as well as the displace-
ments. This is due in part to the calculation of SEQV
Relative or Absolute Results Yes Relative Only
using the Segalman-Reese technique versus direct
Segalman-Reese SEQV Yes No combinations from 1-sigma component values in an
Response PSD Yes No SPRS analysis. The introduction of the Segalman-
Reese technique in ANSYS 8.1 for computing statisti-
Covariance Results Yes No cally meaningful values of SEQV was an important
improvement in PSD post processing (Ref: ANSYS
The mode coefficients of an SPRS analysis are Theory Manual, Section 17.7.12.4.)
comparable to the square root of the PSD Covariance Even for an analysis in which the more rigorous
diagonal. Modes can be combined by one of the PSD analysis is required, an SPRS type analysis can be
various mode combination methods, SRSS, GRP or used to establish a “first-pass” preliminary solution.

CQC depending on whether it is desired to combine ________________________________________________
closely spaced modes.
The cross term mode coefficients generated by the Part one of this article appearing in the last issue of ANSYS
Solutions discussed use of the covariance significance ratio
CQC combination method for different modes are on the PSDCOM command to reduce total run times.
comparable to the square root of the off-diagonal terms
of the PSD covariance matrix.
Using the SOLID92 and SHELL181 models dis-
cussed in this article, two cases were run to compare
SRSS and CQC with PSDCOM using default values for
the significance factors. Run times and memory are
listed in Table 3.

www.ansys.com ANSYS Solutions | Winter 2005

Guest Commentary
The Next Generation
of Innovators
Integrating simulation technology into
coursework promotes design innovation
among future engineers.
Engineers and the companies they work for in regions
such as North America and Europe are becoming
hard-pressed to compete with the growing number of
highly skilled technical professionals in countries
where labor costs are considerably lower.
Consider the following facts. The equivalent cost
of a $70,000 U.S. engineer is approximately $15,100
in China, $14,400 in Russia and $13,500 in India.
Moreover, whereas U.S. schools presently grant
approximately 60,000 B.S. engineering degrees per
year, substantially less than the 270,000 in India,
195,000 in China, 103,000 in Japan and 84,000 in
Russia. Only 5% of all post secondary degrees
awarded in the U.S. were granted in engineering fields.
This is in dramatic contrast to China, India, Japan and
Russia with 44%, 31%, 19% and 15%, respectively.
The result has been unprecedented levels of
offshoring of professional services, particularly from
the U.S. Forrester Research recently predicted that 3.3
million U.S. technical service jobs will be relocated
abroad in the next 15 years, and Goldman Sachs
estimates that job loses in technical fields could reach
$6 million over the next 10 years.
www.ansys.com
37
By Dr. Michael Lovell
Associate Dean for Research
Director, Swanson Institute for
Technical Excellence
University of Pittsburgh
School of Engineering
These challenges in the industrial sector are
reflected in higher education. Unlike industry, however,
a university’s primary product is human capital and
intellectual property rather than goods and services.
Clearly, the engineering field has been dramatically
altered over the past decade and traditional education
practices must be modified to address the changing
needs of society. In this light, several fundamental
questions are being asked in the academic
community. What new skills and capabilities do
engineering graduates need to compete in today’s
world? How can we assure that engineering graduates
will bring value to the workplace beyond their technical
skills? How will graduates function in an international,
interdisciplinary team setting?
The Spirit of Design Innovation
At the University of Pittsburgh, we are attempting to
meet these challenges by embedding a spirit of design
innovation and entrepreneurship into our engineering
coursework. These activities are specifically focused
on interdisciplinary engineering activities through
virtual product development and prototyping.
ANSYS Solutions | Winter 2005
 Guest Commentary
We have been very fortunate to be able to focus
our educational efforts within the Swanson Institute for
Technical Excellence (www.engr.pitt.edu/SITE)
established through a generous gift from ANSYS
founder John Swanson. Within the institute is the
38
42 Swanson Center for Product Innovation (SCPI): the
epicenter of the innovative design activities within our
engineering curriculum.
Originally opened in April of 2001, the mission of
the Swanson Center is to provide a mechanism for
training students in the latest product development
techniques while fostering economic growth and
development of partnering industries. Simply stated, we
are trying to nurture an environment that promotes
innovative discovery in an educational experience that
is as close to industry practice as possible. The
foundation of the educational programs of the SCPI lies
within the formation of interdisciplinary teams of
engineering and business students that produce
working prototypes of new products or processes for
industry sponsors.
Virtual Product Development in the
Classroom
Although innovation is easily recognizable, it is often
difficult to quantify and measure, especially in an
academic setting. As instructors of the courses in the
SCPI where new products are being developed, we
have been trying to understand how some groups of
students develop highly innovative new products
while others do not. Do some students have learning
experiences in their academic career that lead them
down a different design path than other students? How
can we as educators synthesize a scientific approach
for learning innovative design?
Since starting the Swanson Center, we have come
to believe that focusing on virtual product development
activities has enabled our students to be more
innovative. This belief has been verified by learning
scientists such as Barbara Tversky at Stanford who
have directly linked student innovation with the ability to
design and analyze concepts in a virtual environment.
www.ansys.com
In the case of our engineering innovation focus, the
virtual environment critical to our educational efforts is
provided by ANSYS, Inc. through their Workbench
product simulation platform that includes an intuitive,
integrated tool set for carrying out performance
assessments rapidly across design alternatives.
ANSYS’ development priorities have allowed us to
move virtual design and analysis into the classroom at
much earlier stages of a student’s career.
The metrics in the capstone Product Realization
course taught in the SCPI indicate that the present
approach has been very successful. Over the past three
years, student projects in the Product Realization
course have resulted in the formation of four
companies, the selection of five student teams for the
prestigious March Madness for the Minds Competition
sponsored by the National Collegiate Inventors and
Innovators Alliance (NCIIA), and the submission of more
than 10 invention disclosures that are moving toward
patents.
Weaving Technology Into Coursework
The most recent release of the ANSYS Workbench
product has allowed the Swanson Center to take its
educational activities to a new level. Each summer, the
School of Engineering conducts a six-week program for
high school sophomore and juniors that are
predominantly from underrepresented groups.
The goal of the program is to promote interest and
discovery in the areas of science, technology and
mathematics.
This past summer program culminated with a
student soap box competition in which the high school
students virtually designed their cars using SolidWorks,
analyzed and optimized their performance within
ANSYS Workbench and fabricated them using the
prototyping facilities of the SCPI. Upon exiting the
summer program, more than 90% of the high school
participants indicated that they planned on pursuing a
college degree in the technical fields of science,
mathematics or engineering, which is more than double
the proportion of a representative group of high school
students planning on attending college. In addition,
nearly all of the students stated that the virtual design
and analysis tools that they learned increased their
interest in technology, which may be the most
important aspect of the summer program.
Traditionally, manufacturing companies that are the
most successful in the long run are those that develop
innovative products with high brand value. In today’s
world, their future success also depends on their ability
to quickly adapt to technological change, a process
that is accelerating at a dizzying pace. By providing
software tools that allow design innovation to be woven
into the fabric of higher education, ANSYS, Inc. is
helping recruit and produce the next generation of
innovators who will generate winning new products,
create employment opportunities, enhance the
competitive position of their companies and ultimately
strengthen the national economy.

ANSYS Solutions | Winter 2005
Hardware Update
Reducing the Pain of cluster
High-Performance Cluster Deployment
Better integration and management of system
components overcomes major problems.
By Brian Lowe
HP ANSYS Alliance Manager
High Performance Technical Computing Division
Hewlett-Packard Company
Clustering technology is firmly entrenched as the
leading computing choice for high-performance users.
According to the IDC white paper “HP’s Strategy for
Delivering Cluster Technology to Technical Computing
Environments” by Dr. Christopher Willard, clusters will
account for more than half of all technical computer
sales by 2008. Although the cluster market is growing
rapidly, several challenges are holding back clustered
computing from reaching its full potential. Unwieldy
complexity, uncertain results, restricted choices,
software integration, storage and visualization issues
are all challenges that customers face when deploying
clusters in an ever-changing technology world.
Clusters are certainly not new, but today’s
technologies and choices are incredibly dynamic and
diverse. Customers want the ability to exploit new
technology. Currently, if you want an integrated
solution from a cluster vendor, they exist (to some
extent), but offerings may be limited and often don’t
keep pace with technology. In other words, there’s not
much choice in terms of processors or interconnects.
Additionally, you’ll find that systems typically feature a
single proprietary cluster management environment
and have incomplete and uncertain application
availability. If you want something different from what
you are offered, you’re on you own — utilizing precious
research and engineering time integrating and testing,
tuning and maintaining a ‘one-off’ solution.
What is being done to resolve these cluster
challenges? One company, Hewlett-Packard (HP), has
Clusters will account for
more than half
of all technical computer sales by 2008.
www.ansys.com
39
taken the approach of offering the Unified Cluster
Portfolio, a tested collection of modular offerings
(hardware, software and support) delivered as a single
solution. This portfolio provides a single environment
for multiple applications and tools from HP, software
partners such as ANSYS and the open source
community. The portfolio utilizes a common
implementation across servers and operating systems
to ensure agile deployment of new technologies.
Improving Integration
One way to advance cluster technology is through
improved integration. Integrating computation, data
management and visualization functions greatly
enhances the productivity and performance of
clusters. This level of integration (combining
computation, data management and visualization) is
new for the high-performance computing market, and
is an HP first. Such comprehensive functional
integration is important because data consumed and
generated is growing as clusters become more
capable. Consequently, the need for one system to
compute, manage and visualize data is becoming
more critical. Clusters must be able to multi-task.
Capabilities for multiple functions (job launch,
resource scheduling, software installation and
monitoring, for example) must be unified into a
common framework for rapid deployment. Issues that
choke productivity when clusters scale and run real-
world applications must be identified and corrected.
ANSYS Solutions | Winter 2005
 Hardware Update
Building Better Clusters
A cluster consists of some number of nodes (systems)
that communicate with each other through an inter-
40 connect. Generally, these nodes are supervised from a
single contact point. A compute cluster usually resides
in a single machine room and is managed as a single
computer system. Let’s examine some of these
components and suggest solutions that help alleviate
and simplify some of the challenges of adopting
clusters for CAE applications.
Platforms. Users want the hardware platform to offer
excellent performance and price/performance, without
locking them into proprietary solutions that are limited
in flexibility. Industry-standard Intel and AMD
processor-based systems can meet or exceed these
requirements and are surrounded by an infrastructure
that satisfies the intense compute and I/O needs of
CAE applications, such as ANSYS CFX. Processor
choices include Intel Itanium 2, Intel Xeon EM64T and
AMD Opteron, all of which are capable of running
32-bit and 64-bit applications. It is important that
customers consider future expansion needs —
selecting, for example, cluster platforms that are
available in configurations from 5 to 512 nodes and
beyond with packaging options for density and
expandability. A modular platform allows for easier
expansion, which means users do not need to
re-architect the enterprise to expand and add new
systems.
Operating Systems. Customers are looking for
price/performance, strong proven robustness,
scalability and flexibility. By offering a choice of
operating systems — HP-UX, Windows and Linux —
customers are free to choose the best environment
depending on their particular design needs. This
flexibility is important because some applications are
High Speed Interconnect
Network Servers
Network I/O
Sys Admin
Management
Compute
A cluster consists of nodes (systems) that communicate with one another through a high-speed interconnect. Integrating and managing computation,
data management and visualization functions enhances the productivity and performance of clusters.
www.ansys.com
only available on certain operating systems, or it may
be more economical to run an application on a
particular operating system. The key here is to get a
“one-stop” portfolio of OS solutions that offers
freedom and flexibility along with single-vendor
support and consulting.
Interconnect. A cluster assumes some kind of logical
and physical connection between nodes. In general,
the higher the performance of the interconnect, the
better the overall performance of the cluster. Industry
standard interconnect technologies such as GigE,
InfiniBand, Myrinet or Quadrics offer a wide range of
options and packages and help maximize price/
performance. Currently, InfiniBand technology is
emerging as the standards-based, high-performance
clustering interconnect with low latency and high
bandwidth and offering increased performance over
Ethernet.
Cluster Manageability. One main benefit of clusters is
that they are composed of a fairly large number of
individual (and mostly independent) platforms. While
this provides excellent price/performance, it also
introduces management and usability issues, in that
each node has its own copy of an operating system,
middleware, and applications and user-support
databases. To make the promise of clustering a
reasonable proposition, these systems must be easy
to use and to manage. Simplification of cluster
management begins with a single point of control, or a
single user interface, that customers can deploy to
support multiple users, mixed workloads, and different
applications. The professional life of the system
manager is simplified through validated software, both
applications and middleware. For example, HP-MPI
provides a common tool or layer that is above the
interconnect so users don’t have to get into lower-level
Users
Service
Visualization
File Servers and File I/O
ANSYS Solutions | Winter 2005
Nimble, turnkey cluster platforms
help CAE customers
achieve productivity
without sacrificing performance.
libraries. HP-MPI is an implementation of the industry-
standard messaging passing interface and is
specifically developed to free software designers
from decisions about interconnect and platform
dependencies. HP-MPI is fully interconnect-aware and
host platform–aware, ensuring optimal performance
regardless of the underlying hardware. HP-MPI
supports 32-bit and 64-bit applications and provides
optimal performance on all popular cluster
interconnects.
Scalable Storage. Clusters are scalable,
computational engines that can deliver immense
computing power to meet the most demanding
compute-intensive research projects. However, recent
technology advances in computational performance
have outpaced capabilities in scalable storage and
I/O. Many Linux clusters use shared I/O techniques,
such as Network File System (NFS), the current
defacto standard for sharing files. On larger
configurations, the limited scalability of NFS can
diminish the speed and throughput of the Linux
cluster. Additionally, programmers must often
implement numerous time-consuming and difficult
techniques to use hundreds of disjointed, distributed
file systems. As a result, applications run slowly and
users waste valuable time and effort on file system
housekeeping. HP StorageWorks Scalable File Share
(SFS) solves the distributed file system I/O challenge
while simplifying the use and administration of Linux
clusters to deliver faster processing and a higher
return on investment. HP StorageWorks SFS is based
on open Lustre™ technology, a high-performance,
highly scalable file share engineered to provide
extremely fast, scalable and reliable I/O for Linux
clusters. Lustre is a major breakthrough for fast I/O on
Linux clusters, delivering up to 100 times more
bandwidth than typical clusters. HP SFS is a
www.ansys.com
41
self-contained fileserver built from multiple, industry-
standard HP ProLiant servers and StorageWorks disk
arrays. The SFS server runs a combination of Lustre
and HP-specific value-added software.
Application Availability. Without application software,
nothing will work! Software partners like ANSYS are
integrating and releasing applications with many of
these cluster requirements and solutions in mind —
such as choice of platforms and choice of operating
systems. It is important to verify that the software
partner has certified a particular configuration to
ensure reliable performance and support. Finally, it is
important to have access to consulting and integration
services to make sure that the cluster solution can be
implemented with minimal start-up time and to get
help tuning the system for maximum performance and
throughout.
As more and more clusters are deployed for high-
performance cluster (HPC) applications, customer
expectations for cluster performance will increase as
well. To that end, clusters must become easier to
manage and deploy while offering tremendous
technical flexibility in terms of functionality
(computation, visualization and storage) and in
processor and interconnect advancements. Nimble,
turnkey cluster platforms help CAE customers achieve
productivity without sacrificing performance.

For More Information
Further details on high-performance cluster
deployment can be found in the book “Building
Clustered Linux Systems” by Robert W. Lucke,
Prentice Hall, PTR 2004. The following Web sites
also have useful information:
www.hp.com/go/clusters
www.hp.com/go/mpi
www.lustre.org
www.infinibandta.org
ANSYS Solutions | Winter 2005
Unparalleled CAE performance and infrastructure
The fastest CAE performance on the planet just got more versatile
Choose between:
■ Intel® Itanium® processor’s fastest floating point in the industry.
■ The CAE software portfolio offered by HP-UX, Linux and
Microsoft® Windows®.
■ The cluster capability of HP Integrity servers.
■ The value offered by HP ProLiant industry-standard servers.
■ The pre/post processing power offered by HP Workstations.

Whatever your choice, HP and our CAE partners deliver.

www.hp.com/go/cae

Screen images courtesy of (left to right): ANSYS® ICEM CFD™ Cabin Modeler™ and
ANSYS® CFX™
Intel and Itanium are registered trademarks of the Intel Corporation
in the United States and other countries.
Microsoft and Windows are U.S. registered trademarks of Microsoft Corporation.
© 2004 Hewlett-Packard Development Company, L.P.