Industry Spotlight: Microsystems and Nanotechnology Spotlight: Cylinder example demonstrates ... ANSYS Customization with APDL. ... welding to ensure it meets the code of the particular

Industry Spotlight:

Cylinder example demonstratesnonlinear effects, follower forcesand modeling techniques usedfor buckling analysis

Unified release of ANSYS,ANSYS CFX and ANSYS ICEMCFD software offers advances

Century Dynamics softwarehelps oil and gas industries develop better systems

Microsystems and Nanotechnology

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.

ANSYS Solutions is published for ANSYS, Inc. customers, partners and others interested in the field of design and analysis applications.

Editorial DirectorJohn [email protected]

Managing EditorJennifer [email protected]

DesignersMiller Creative [email protected]

Art DirectorDan [email protected]

Ad Sales ManagerBeth [email protected]

Circulation ManagerElaine [email protected]

www.ansys.com ANSYS Solutions | Summer 2005

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 othercountries. ICEM CFD is a trademark licensed by ANSYS, Inc. All other trademarks or registered trademarks are the property of their respective owners. POSTMASTER: Sendchange of address to ANSYS, Inc., Southpointe, 275 Technology Drive, Canonsburg, PA 15317 USA.

©2005 ANSYS, Inc. All rights reserved.

Editorial AdvisorKelly [email protected]

CFD Update AdvisorChris [email protected]

Highlights of ANSYS 10.0Unified release of ANSYS, ANSYS CFX and ANSYS ICEMCFD software offers advances

Designing Safe and Efficient Drill RigsCentury Dynamics software helps oil and gas industries develop better systems

Comparing ANSYS Shell Elements for Buckling AnalysisCylinder example demonstratesnonlinear effects, follower forcesand modeling techniques used for buckling analysis

ContentsIndustry Spotlight

Features

Microsystems and NanotechnologyA continuing series on the valueof engineering simulation in specific industries

Industry Spotlight articlebeginning on page 5 discusses how engineering simulation is used in developing microsystemsand nanotechnologyproducts.

About the cover

10

5

14

17

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 eNewsand email alerts when the latest electronic version of ANSYSSolutions becomes available!

Departments

CAE CommunityCollege Design Engineering Award . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

Simulation at WorkPersonal Rapid Transit Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

Guest CommentaryForefront of Product Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

Tech FileMeshing in Workbench . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .34

Industry NewsAnnouncements and Upcoming Events . . . . . . . . . . . . . . . . . . . . . . . . . 3

Tips and TechniquesModal Analysis of Models with Friction . . . . . . . . . . . . . . . . . . . . . . . .36

Hardware UpdateANSYS CFX Performance and Scaling . . . . . . . . . . . . . . . . . . . . . . . . .38

CFD UpdateWhat’s New in Computational Fluid Dynamics . . . . 22

EditorialThinking outside of the box. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

Editorial

2

Thinking outside of the box to fully leverage simulation technology in product development

Innovation: The New Competitive Necessity

Time-to-market, cost andquality traditionally have been the major competitiveissues for manufacturers.Companies that launchedproducts faster than competi-tors, made them better andhad the lowest prices weregenerally the most successfulin the market. These remaincritical requirements, ofcourse. But in today’s com-petitive business climate,faster–better–cheaper isn’tgood enough anymore.

Now the competitiveedge goes to firms that meet these criteria while delivering killer products with innovative designs that customers flock to buy. Design innovation differenti-ates manufacturers, builds brand value and drives top-line revenue growth. Innovative products standout from the crowd in terms of value, performance,size or capacity. In some cases, a company may comeout with an entirely new class of never-before-seenproduct such as the SkyWeb Express personal rapidtransit vehicle, where critical joints connecting thecabin to the chassis were evaluated with ANSYS toolsas covered in the Simulation at Work departmentbeginning on page 30.

Analysis can play a key role in developing thesetypes of innovative designs by leveraging the expertiseof a company’s technical professionals, from whomcreative ideas and concepts originate. The tools canbe applied in a simulation-driven developmentapproach by identifying and correcting problems upfront, optimizing designs early, evaluating alternatives,studying tradeoffs and analyzing product performancenot practical by building and testing physical prototypes.

As described in the article “Highlights of ANSYS10.0” on page 10, the latest release of ANSYS soft-ware offers major advances in analysis core technology within an integrated simulation environment. The ANSYS Workbench platform is thebackbone for unprecedented levels of CAE integrationas well as process automation features for capturing,streamlining and standardizing simulation operations.Product and process innovation go hand in hand, andANSYS, Inc. is at the forefront of providing tools for both.

By John KrouseEditorial DirectorANSYS [email protected]


In a growing number of industries, suchadvanced technologies are becoming indispensabletools in developing innovative products. State-of-the-art multiphysics software, for example, is a must-havein determining the effect of stress, deformation andtemperature distribution on tiny micron-size structuresof Micro-Electro-Mechanical Systems (MEMS) discussed in the fascinating Industry Spotlight article“Microsystems and Nanotechnology” beginning on page 5.

Often product development teams use simulationin navigating through uncharted territory in their drivetoward innovation. Take a look at the CAE Communitydepartment on page 20 to learn how a student designteam from the University of Washington boosted output of a micropump design threefold using ANSYStools. These students are probably the first to use CFDcoupled to an optimizer to determine the best designof the pump’s microvalve concept and also the first tostack micropumps in parallel to increase flow rate.

In striving for innovation, evaluating and changingwell-entrenched corporate practices to fully leveragesimulation technology often present enormous challenges. As Charles Foundyller points out in hisguest commentary “CAE Moves to the Forefront ofProduct Development” on page 32, technical issuespale in comparison to confronting cultural, organiza-tional and procedural issues. His cow path concepthits the mark exactly in showing how process chainsoften bend and wander in a convoluted trail that companies follow because things have been done thatway for decades.

The difficulties in confronting these organizationalissues are daunting, of course, which is why manytimid companies may never get off square one in making necessary changes. Rather than merely plugthe technology into existing procedures, forward-looking companies closely examine their ways ofoperating and think outside the box in terms of when simulation is performed, what tools are used,where refinements are made and how analysis fits intoa company’s overall business strategy. These smartenterprises know full well the critical value of innova-tion, have the fortitude to break new ground in productdevelopment and will likely be among the world’smanufacturing superstars in the coming years. ■

3


Industry News

Recent Announcements and Upcoming Events

Now Available! ANSYS Parametric Design Language Customization Guide

Phoenix Analysis and Design Technologies (PADT) ispleased to announce the release of its Guide toANSYS Customization with APDL. This guide is acompilation of course notes from PADT’s very popular“ANSYS Customization with APDL” class. By populardemand, PADT has turned the notes into a 288-pageguide that steps new and experienced ANSYS usersthrough all of the details of APDL scripting. Its 12chapters include reference information, examples, tipsand hints and eight workshops. This guide, available in hardcopy only, is an invaluable resource to anyonewho wants to start using APDL or become an ANSYS“power user.” Priced at $75, it will quickly pay for itselfby saving you hours of research and trial-and-error.For more information, visit www.padtinc.com/support/techguides or call 1-800-293-PADT.

ANSYS and RoboBAT Announce an IntegratedStructural Engineering Solution

ANSYS, Inc. recently announced a partnership withRoboBAT, a leading supplier of analytical and CADsoftware solutions for the structural engineer, to inte-grate their solutions in order to streamline customers’product development processes.

The integration of RoboBAT’s Engineering SystemOpen Platform (ESOP) software with ANSYS Work-bench expands users’ access to structural engineeringapplications and increases their productivity. Usershave access to more than 100 engineering applica-tions as well as the ability to write their own modules,either stand-alone or integrated with ANSYS Work-bench models (for example, writing pre- and/or post-processors for Workbench).

Through ESOP for Workbench, users can assessstructures quickly and easily. They have a central loca-tion where they can conduct calculations and analysesin the pre-design phase, access databases of refer-

ence materials and code-check various structuresagainst international standards. For example, engi-neers in the nuclear industry can analyze stress on astructure design and then conduct a code check of awelding to ensure it meets the code of the particularcountry in which it is being built — all before creating a sketch in any CAD program.

The integration also offers a convenient way for engi-neers to manage their projects, since they are able tocompile all information and track their progress on aproject in one location. Because ESOP is an Excel-based program, the technology is also easy to use.

To learn more, visit the news section at www.ansys.com.

Save the Date for 2006 International ANSYS Conference

Continuing what avid ANSYS users consider a CAEtradition, preparations are under way for the 2006 International ANSYS Conference to be held May 2 – 4at a new venue this year — the David L. Lawrence Convention Center — with training April 30 – May 1 at the Westin Convention Center, in Pittsburgh, Pennsylvania, USA. The event brings together engineers and analysts worldwide from all disciplinesof computer-aided engineering.

Since 1983, ANSYS, Inc. has hosted the InternationalANSYS Conference to showcase the advances incomputer-aided engineering and related technologies.The three-day conference addresses the completespectrum of engineering professionals including engineers, analysts and engineering managers. Conference attendees have the opportunity to network with peers, interact with ANSYS professionalsand participate in a variety of technical sessions.

For more information and to sign up for email updates,visit www.ansys.com/conf2006.


4

Industry News

Matereality v.2.1 Features New Products, Functionality, Content

Matereality, L.L.C., USA, the developer of the first-of-its-kind complete material data management (MDM) system in TrueDigital™ format, announces the releaseof Matereality® Version 2.1.

The much awaited Version 2.1 release of the Matereal-ity material data management platform brings a num-ber of advances to this rapidly developing field. Inaddition to creating the flagship line that provides aWeb-based resource for the storage and selectivedeployment of material property data, the companynow has two new products. Material Data Server hasbeen created to enable manufacturing enterprises tostore all their material property data on all their materialson a single extensible platform within their intranetsand maintain complete control over who has access toit. With this, companies can now finally go paperlesswith the storage of all their test data, cradle to grave.

The Materials Databases product line provides material suppliers with a ready-to-deploy technologyto enhance their customer support with one powerfulsecure platform for material selection data as well as design properties, 365/24/7 globally! For moreinformation, visit www.matereality.com.

Durability & Fatigue Seminars from Safe Technology

Noted fatigue & durability authority, Professor JohnDraper of Safe Technology Ltd (UK ),will be visitingDowners Grove, Chicago, Ill., September 12 - 16, topresent a series offatigue and durability seminarshosted and sponsored by Belcan Corporation.

A regular speaker at conferences worldwide, Profes-sor Draper has over 25 years' experiencein fatiguedesign and life assessment in the aerospace andground-vehicle industries.

These courses will provide an introduction to moderntheories of metal fatigue and their practical applicationthrough worked examples and interaction/discussion.There is a strong emphasis on what is possible and thepitfalls to avoid. A professional volume of course notesforms a self-contained reference book.

For a complete listing of courses, visit www.safetechnology.com. ■

Upcoming Events

ISABESeptember 4 – 9Munich, Germanywww.isabe2005.com

Nuclear Energy for New EuropeSeptember 5 – 8Bled, Sloveniawww.drustvo-js.si/bled2005

Offshore Europe 2005 Oil & Gas Exhibition & ConferenceSeptember 6 – 9Aberdeen, UKwww.offshore-europe.co.uk

German Fuel Cell Congress 2005September 26 – 28Stuttgart, Germanywww.f-cell.de

German Aerospace Congress 2005September 26 – 29Friedrichshafen, Germanywww.dglr.de

Automobile & Engine Technology ColloquiumOctober 4 – 6Aachen, Germanywww.rwth-aachen.de/ac-kolloquium/index_e.htm

SNAME Maritime Technology Conference & Expo and Ship Production SymposiumOctober 19 – 21Houston, TX, USAwww.sname.org/AM2005/

Visit www.ansys.com for more events.

Partner Announcements

Industry Spotlight


5

A rapidly growing number of products in many

industries contain miniature sensors, actuators and

other Micro-Electro-Mechanical System (MEMS)

devices made of small parts measured in microns

(millionths of a meter). Complete devices typically

are on the order of a few millimeters across, with

individual components less than 10 µm thick, and

frequently are combined with integrated circuits on

a single chip to provide built-in intelligence and

signal processing.

Microsystemsand

NanotechnologyA continuing series on the value

of engineering simulation in specific industries


Industry Spotlight

6 MEMS devices must perform accurately and reliably, often in hostile environments. Therefore, engineers rely extensively on engineering simulationsoftware to study these microstructures in determining stress, deformation, resonance, temperature distribution, piezoelectric response, electromagnetic interference and electrical properties.Multiphysics analysis tools, in particular, are essential in accounting for these effects, which can have such huge impact on delicate microstructures.

Most of the commercially deployed devices areused in automotive applications such as accelerome-ters for deploying airbags and sensors for monitoringmanifold pressures, roll-over and fuel injection systems, for example. Biomedical is another largemarket, with MEMS in disposable blood pressure sensors, as another example. The technology also is used for ink-jet printer heads, display devices, hard disk drives and other electronic equipment applications.

Clearly, MEMS technology is past the buzzwordstage of the late 1990s and now represents a growingindustry producing millions of products each year.Industry analysts estimate the total worldwide MEMSmarket at nearly $7 billion and expect it to undergo acompound annual growth rate in the range of 20 percent for the next several years as implementationof the devices expands. The use of MEMS in automo-biles is predicted to double by 2007.

Small, Accurate and Economical

The attraction of MEMS for such applications is theirrelatively low cost and high performance in a compactpackage. Produced through the same semiconductorfabrication methods as integrated circuits (ICs), thou-sands of MEMS can be mass-produced on a singlesilicon wafer along with associated electronic circuits.

Using well-proven semiconductor fabricationtechnology, including bulk micromachining or surfacedeposition, MEMS can be produced and sold for a fraction of the cost of conventional sensing or actuation devices. Conventional blood pressure transducers costing $600, for example, can bereplaced by MEMS intravenous sensors selling forabout $10.

The ability to easily integrate MEMS transducerswith signal processing electronics and the precisemanufacturing tolerances inherent in the semi-conductor fabrication process allow manufacturers tomeet (and in many cases exceed) the performance ofequivalent macroscopic transducers.

Design Challenges

One of the most formidable tasks in MEMS development is designing the microscopic parts tobest fit together and operate properly, often in environments hostile to delicate mechanical components and sensitive electronic circuitry.

Developing internal interconnects, circuitry andpackaging so that MEMS devices operate flawlesslyfor years in these demanding applications is a toughengineering challenge. Semiconductors must resistdamaging internal heat build-up while withstanding a wide range of structural loads and ambient temperature swings. Parts such as diaphragms,valves, membranes, beams and other microstructureson the same silicon chip also must survive severeshock and vibration to adequately perform theirmechanical functions.

Traditional electromechanical products havethese same requirements. But designing MEMS is particularly challenging because of the tremendousdifference in size and overall sensitivity of the devices’internal components compared to their surroundings. A MEMS sensor that measures gas pressures in the range of 0.15 psi by detecting a few micronsdeflection of a microbeam, for example, often mustundergo several Gs of shock and vibration while inservice on a piece of factory-floor equipment. The taskof the MEMS sensor in this application is equivalent todetecting a sneeze in the middle of an earthquake!

Electrostatic-structural analysis of a comb drive MEMS resonator commonly used in electromechanical bandpass filters shows voltage iso-surfaces in the device.


7

SilMach uses ANSYS Multiphysics in the development of electromagnetic actuators and research into a variety of MEMS-based actuatorsand systems such as these unmanned airborne vehicles with flapping wing propulsion systems.

A MEMS designer has further challenges in thepackaging of the device. Packaging, typically a form ofplastic encapsulation similar to ICs, often impactsdevice operation, reliability and accuracy.

Because many of these effects are inter-dependent, predicting output and performance ofMEMS devices is generally a complex problem thatoften defies intuitive approaches used in developing traditional transducer technology.

Developers of MEMS also have greater obstaclesto overcome in the area of prototype testing. Whereasphysical mock-ups of conventional electromechanicaldevices may undergo several test and re-designcycles where parts are modified and switched around,the initial semiconductor fabrication setup for MEMSis so time-intensive that prototype testing simply verifies a design, and physical design interactions arelargely replaced by virtual prototyping.

Multiphysics Solutions

To meet these many design challenges, MEMS engineers universally rely on engineering simulation tocreate and performance-test virtual prototypes of thedevices. Multiphysics simulation is used extensively inthe development of MEMS because of the inherentmultiple interrelated physical phenomena at play, such as stress, temperature, electrostatics, piezo-electrics, fluidic damping, thermoelastic effects and electromagnetism.

Previously, solving such coupled applicationsmeant enduring numerous manual file transfers and problem setups for each physics analysis.

Such cycles were cumbersome, error-prone and time-consuming and required users to learn and maintainseveral different software codes. These issues arebeing addressed by multiphysics solutions that automatically combine the effects of two or more interrelated physics within one unified environment.These solutions automatically manage data exchangebetween the different physics to perform coupledanalysis without requiring users to spend time manually performing these tasks. As a result, coupledanalyses can be performed in a fraction of the timeotherwise required.

Possibly the original and most comprehensivemultiphysics solution is provided by ANSYS, Inc.ANSYS Multiphysics offers the widest range of multiphysics disciplines in a single unified environment: structural, thermal, fluid and electro-magnetic. Another major strength is that ANSYS Multiphysics is fully integrated with many ANSYSanalysis tools such as parametric modeling capabilities, design optimization and probabilisticdesign functionality, and both ECAD and MCADimport features. ANSYS Multiphysics also has anadvanced fluid structure interaction capability realizedthrough bi-directional interface with the ANSYS CFXcomputational fluid dynamics code.

Unlike many other commercial FEA codes thatprovide either direct or iterative multi-field approaches,ANSYS Multiphysics provides both. Direct coupledfield analysis solves all of the physics field’s degrees offreedom in one solution phase. In iterative coupling,results of a single FEA solver iteration of one physics

8


field are passed as loads to the next physics field, iterating between all active physics fields until convergence criteria in the transferred loads are met.The ANSYS Multi-field solver automatically handlesthe data exchanged in these iterations. Users onlyhave to set up the problem initially.

Analysis in Action

A growing number of organizations use ANSYS Multiphysics for analyzing MEMS devices. Today, hundreds of companies and research institutions usethe software in developing leading-edge MEMS technologies globally.

SilMach, for example, develops highly integratedsilicon-based actuators and systems. By usingANSYS Multiphysics, the company has the ability tosolve complex coupled physics to create sensors andactuators within arrays and predict their performancebefore committing to manufacture. Coupled physicssuch as mechanical deformation and nonlinear contact effects with acoustic, electrostatic, thermaland fluid damping can be achieved using the software.

ANSYS Multiphysics has enabled SilMach to create more efficient MEMS devices such as an electromagnetic actuator producing 100 Watts pergram compared to 1 Watt per gram for standarddevices. The company also continues to researchother advanced technologies, including MEMS-basedgas micro-turbines, MEMS-based flap-arrays foractive control of turbulence and flapping-wing propul-

sion systems for artificial insects and nanometer-scaleunmanned airborne vehicles (UAVs).

The software is also used by RTI International inthe development of advanced imaging systems. Aparticular challenge is simulating the effect of 25-100THz infrared waves interacting with a periodicarray reflective structure. The problem was consideredunsolvable using traditional FEA methods, whichwould have required a model size approaching 100million degrees of freedom.

ANSYS Multiphysics was used to perform a fullwave electromagnetics harmonic scattering analysison the device. Reflection coefficients were computedfor the entire frequency range. The analysis took intoaccount the skin depth and loss of RF energy throughjoule heating of the materials. A fully parametric modelenabled rapid changes in materials, geometry andexcitation. The numerical problem size was reducedconsiderably through the use of the periodic boundarycondition.

This approach provided RTI with the ability to validate experimental results and quickly improve

device performance by investigating various structureparameter changes. The software contributed to abetter scientific understanding of experimental resultsbecause researchers could actually visualize the electric field within and around the structure. ANSYSMultiphysics also allowed RTl to analyze results at dis-crete frequency points, which will help plan futureequipment purchases for their experimental work. ■

Industry Spotlight

RTI modeled a unit cell (left) and obtained results of the scattered plane wave (right) using ANSYS Multiphysics in the development ofa periodic array reflective structure.

9


ANSYS Multiphysics provides insight into the trajectories ofcharged particles from a conical field emitter with a tip radiusof 10-20 nm.

This electrostatic field benchmark of an array of eight nano-scalespheres was performed by Dr. Andreas Hieke, an internationallyrecognized authority in modeling and analysis of microsystems andnano-scale systems.

Nanotechnology: The Next Small FrontierThe success of MEMS is generating considerablemomentum in emerging research and developmentsurrounding nanotechnology, where the device scaleis at the atomic or molecular level with dimensions of100 nanometers and smaller.

The most recent developments in nanotechnologyare direct atomic manipulation techniques, single-electron transistors and carbon-60 nanotubes as current-carrying conductors in nano-integrated-circuits. Microsystem technology can be used to build and interact with nano-scale devices and systems, so MEMS and nano remain closely linked.There are many MEMS fabrication technologies beingscaled down to produce nanoscale devices. Perhapsthe best example is the micro-tip field emitter: essen-tially a microscopic electron gun consisting of a conical field emitter with a tip radius of 10-20 nm.

Currently, ANSYS Multiphysics addresses nanotechnology simulation requirements covering ionoptics and electrostatic calculations associated withfield emission tips and carbon nanotube structures.The ion optics capability is a post-processing featurewhere charged particles can be introduced into computed electromagnetic fields and their path tracedwith “streamline” graphics. The user can control the“history” of the charged particle, changing its massand/or charge to simulate fragmentation, for example.

At the nanometer scale, the bulk material modelsused by conventional finite element tools generallybreak down as quantum mechanical effects becomedominant. The availability of customizable, user-programmable material models in ANSYS Multi-physics is helping to address the analysis of somenano-systems. Some users have made reasonableapproximations to polycrystalline grains in surfacemicromachined parts using this approach.

ANSYS structural analysis determined deformation and residualstress levels resulting from fabrication of this MEMS optical gratingdevice used in spectroscopy instruments.

10


The latest release of ANSYS v10.0 offers the first unified release of products under the ANSYS productportfolio: ANSYS, ANSYS CFX and ANSYS ICEMCFD. This milestone provides an integrated simulationenvironment with some of the best technologies formeshing, pre-post processing and multiphysics.

Integration of this diverse software portfolio offersleading-edge new technology and unparalleled productivity advances to the end users across a wideindustry spectrum and physics areas including analysis of composites, rotating machinery, metalforming, interface separation and delamination,MEMS devices, CFD transition turbulence and mesh-ing of shell models for underhood and seam welding.

In addition, ANSYS 10.0 offers further enhance-ments in its Workbench product development environment, integrating more technologies and takingfurther steps toward a more collaborative, integratedand customizable range of virtual prototyping software.

Advanced Fluid Structure Interaction

The new release provides the industry’s most flexibleand advanced fluid structure interaction (FSI) analysissolution required for many industry applications suchas MEMS, including microfluidics, biomedical, auto-motive, aerospace and civil engineering. The ANSYScoupled physics multi-field solver offers enhance-ments to support coupling of ANSYS structuralphysics with CFX CFD for such FSI applications.

Enhanced solver technology allows structuraland fluid solutions to run simultaneously on the sameor different machines, thus accommodating muchlarger models more efficiently than a multi-field solverusing a single machine environment. Code coupling isbased on high-speed inter-process communicationtechnology without the need for third-party applica-tions. The technology ensures that ANSYS CFX canbe run concurrently, thus allowing fluid and structuralcomputations to be solved on different machines whilealso communicating across a local area network, awide area network, or even via an Internet connection.

Biomedical fluid structure interaction analysis of bloodflowing in an elastic artery using the new two-way FSIcapability of the multi-field solver

Highlights of

ANSYS10.0Unified release of ANSYS, ANSYS CFX and ANSYS ICEM CFDsoftware offers advances in core technology within an integratedenvironment for ease of use and productivity enhancements.

Extending the Realm of CFD Analysis

Enhancements in this latest release deliver advancedCFD technology for a wide range of industrial applications. Software improvements target a numberof industries and sectors including power generation,chemical process industries, external aerodynamics,rotating machinery, automotive engine, external andinternal flows, fire and safety applications, marine andfree surface applications.

Power generation industries benefit fromadvances in a number of areas including advancedcoal combustion analysis, state-of-the-art volumetricporosity abilities, advanced boundary conditions that simplify the setup of high-fidelity simulations and novel R&D advances in non-equilibriumsteam modeling. These improvements are furtherleveraged when combined with the latest advances inFSI capabilities.

11


State-of-the-art advances in turbulence modelingbenefits most sectors, permitting more accurate modeling of true flow physics. For example, release10.0 includes the world’s first commercially availablepredictive transitional turbulence model for flows thatinclude both laminar and turbulent flow regimens.For analysts, this translates into more accurate predic-tions of wall forces, flow separation, heat transfercoefficients, and more. For those applications thatdemand detailed resolution of the turbulent struc-tures, there are more options than ever, includingDetached Eddy (DES) models and introducing thegroundbreaking Scale Adaptive (SAS) R&D models.

In many industries, transient flow simulation interest continues to grow in widely diverse applications such as fuel droplet injection in an internal engine, flow-induced vibrations of a structure,water droplet spray in a fire simulation, noise source estimation and fluidized bed performanceassessment. A number of key transient flow-relatedimprovements have been made for ANSYS CFX 10.0to improve the accuracy of transient simulations, as well as reduce the CPU time needed for such simulations. Further reductions in the time needed to perform such large simulations are obtained byimprovements to the scalable parallel performance.ANSYS CFX 10.0 now supports high-speed interconnects on most hardware platforms, for example the Myrinet, Infiniband and Quadrics interconnects. There is also a full release of all ANSYSCFX components as native 64-bit executables for thenew Intel EM64T/AMD Opteron chips. This all translates into more accurate simulations in less time.

Extending its leadership in providing solutions forturbomachinery design and analysis, ANSYS CFX10.0 includes BladeModeler, a highly customizedrotating machinery design tool for bladed components, and TurboGrid, a high-quality near-automatic hexahedral meshing tool for bladed components. These tools combine with the specialized Turbo-Pre and Turbo-Post processingCFD capabilities to create a comprehensive turbomachinery design and analysis system. Simulation models for stress analysis, computationalfluid dynamics and/or fluid-structure interaction canbe created from one centralized geometry definition. It may include upstream and downstream compo-nents and even whole-machine analysis. Applicationsinclude gas turbine design and analysis, pumps, turbines, compressors, fans, blowers, expanders, turbochargers and inducers.

ANSYS CFX 10.0 also contains a number of newfeatures that represent significant advancements inthe interactivity and usability of working within the

post-processor. These are seen primarily throughViewer Shortcuts and new quick access tools. Othernew capabilities include quantitative tables, pointcloud and viewer object transformations.

Core Technology Enhancements in ANSYS

Design and stability of rotating parts such as those inturbo-machinery, cooling systems, biomedical andautomotive systems require accurate specializedmodeling of the physics of these structures. The newrelease offers capabilities for Coriolis damping andforces for beams, shells and 2-D/3-D solids withinANSYS for static, transient and modal analysis ofrotating equipment. Accurate representation of rotating equipment is possible when using solid elements for analysis, with the need for reducing a 3-DCAD representation into beam models. This furthers capabilities to analyze dynamics of rotating machinerythrough modal, harmonic and transient analysis within ANSYS.

Composites find application in aerospace andautomotive for designing structures with high strengthand less weight. Failure and delamination modeling ofcomposite structures are critical, taking into accountthe accurate loading and material behavior of such composites. Release 10.0 offers enhancements inthree main areas:

• Composites enhancements focus on interfacing with Vistagy’s FiberSim software, a composites pre-processing tool for drapinganalysis. Information from FiberSim is importedinto ANSYS, thereby providing accurate information of the manufacturing process foranalysis.

• New solid-shell elements are extended for layered applications like laminated shells andsandwich structures. Solid-shell elements arepart of the new generation of ANSYS elementsfor modeling thin to moderately thick structuresand allow modeling of variable thickness parts.

• New cohesive zone capability in ANSYS allows modeling of interface separation anddelamination, typical of layered compositestructures. This enhancement also finds application in modeling glue behavior of materials with interface separation.

Imag

e co

urte

sy o

f BM

W A

G.

Automotive intake valve modeled using ANSYS ICEM CFDand the moving mesh capability of ANSYS CFX.


Underhood meshing capability in ANSYS ICEMCFD takes care of “dirty” CAD model geometry(left) for robust meshing (right).

12

Enhancing the rezoning feature in ANSYS 9.0, thecurrent release extends the capability to metal plasticity with an ability to re-mesh the deformeddomain and continue solution toward convergence.This enhancement, with support of contact capabilities, finds application in areas such as metalforming. Completing and further advancing the offering of contact technology, ANSYS 10.0 offers new enhancements in areas of beam–beam contact andorthotropic contact friction. This enhancement, forexample, finds application in areas of water/oil supplylines, nuclear power plant piping, cable wires andcoils, woven fabrics and tennis racquets.

For coupled physics analysis, ANSYS 10.0 allowsthermoelastic damping (TED) and direct structural-thermal-electric coupling as part of the new advancedcoupled-field elements. Thermoelastic damping is animportant internal loss mechanism in metals andMEMS (resonator beams) and the capability providesmore accurate, real-world analysis results for theMEMS industry. This functionality allows ANSYS to bethe leader in solving a physical phenomenon that few,if any, other analysis codes can address.

All Distributed ANSYS (DANSYS) users will benefit from the dramatic improvements made to theinstallation and configuration process. Made possibleby HP-MPI on Linux systems, DANSYS no longerrequires consistent working directory structures, now supports NFS mounting for slave nodes, auto-matically picks the fastest interconnect available andallows both rsh and ssh logins. By adopting HP-MPI, Distributed ANSYS now supports the following interconnects:

• InfiniBand (recommended)

• Myrinet (recommended)

• GigE

• Ethernet (not recommended)

DANSYS now supports static and full transientelectromagnetic analysis, mixed U-P formulation, linking in of User Programmable Features (UPF) onLinux and UNIX, and allowing beam and shell contact,including shell and beam thicknesses, by automaticallygrouping all contact elements together into onedomain. Also, DANSYS now allows serial solution ofthe Radiosity Solver while performing a distributedthermal analysis.

Tools for Efficient Meshing

ANSYS ICEM CFD 10.0 provides a complete set oftools to help reduce the time for meshing underhoodmodels in various applications such as automobiles,aircraft and heavy machinery. New additions includeshrink-wrapping, Delaunay and/or hexa-core meshingand pure Cartesian meshing to allow customers tostandardize on ICEM CFD for underhood meshing.The combination of powerful shrink-wrapping com-bined with feature-rich geometry and mesh editingtools allows customers to selectively capture or ignoredetails in their complicated models.

The representation of welds for sheet metal models is very important for accurate results of durability models. With AI*Environment 10.0, a user can represent welds between sheet metal components in a variety of ways to improve solutionaccuracy or solver speed of the analysis. Welds canbe created automatically between single edges of one part and neighboring parts. These welds can bedefined as rigid beam elements, quad element welds or tent welds, to produce more or less rigidwelds in a weld region. The combination of powerfulmidsurfacing and automation of welding allows the user to create assemblies of midsurface shellmodels quickly.

This unification of ANSYS, ICEM CFD and CFXhas been an essential step in the development process,but more importantly, has allowed strengths of the different solutions to be leveraged across products.With ANSYS 10.0 core meshing technology, ANSYSICEM CFD has been exposed within the Simulation tabof Workbench to provide uniform quad/tri mesh. This

Nonlinear contact analysis of a spring using the newbeam–beam contact capability


meshing option provides more uniform, orthogonalquad and/or tri elements to reduce the mesh size ofthe model and to provide better accuracy in the solu-tion. Behind the scenes it also has helped to provide aframework to help integrate these tools as well asother meshing tools in an easier fashion.

Greater Capabilities for Workbench

Integrating more advanced features within the Workbench environment, the current release allowstransient thermal analysis within the Workbench Simulation environment. Users can now perform transient thermal analysis with complex loading patterns and temperature-dependent boundary conditions. Temperature results from various timepoints can then be transferred as boundary conditionsfor static structural analysis within the same simulationenvironment. The feature extends the power of Workbench to a widely used ANSYS functionality oftransient thermal/thermo-mechanical analysis withapplications in electronics, gas turbine and automotiveindustries.

Further extending the advanced capability offering in the Workbench environment, release 10.0offers new functionality of strain-based fatigue inANSYS Fatigue Module. New to the ANSYS FatigueModule is the ability to analyze Low Cycle Fatigue(LCF) by Strain Life methods for constant amplitudeloading with non-constant under development (Beta).The Fatigue Module adds the capability to simulate performance under anticipated cyclic loading conditions over a product’s anticipated life span. Incorporating both Stress Life and Strain Life analyses

In the world of design simulation, today the goal is to integrateand “close-couple” all of the engineering analysis tools necessary to take a product to market.

We at ANSYS call it CAE Collaboration, and it takes manyforms within a design organization. Here are some:

• results sharing between peers, managers, suppliers,and customers

• process and procedure collaboration to establish andcommunicate best practices to the entire engineeringorganization

• hardware resource sharing as more realistic modelsdrive into increasingly larger problem size

• design data sharing, in the form of geometry, materialproperties and design parameters to establish consis-tent data sources and eliminate recreating data work

The key aspects of the ANSYS Workbench customization capa-bilities are all about sharing data, results, hardware and bestpractices with many people in the products development group.

The value of these assets being accessible to everyone is muchgreater than if limited to only a few people.

ANSYS Workbench with its associated customization toolsallows you:

• to manage the overall engineering simulation processincluding simulation methods and processes, commonservices and data and configuration management

• to readily integrate your own applications or third-party-developed engineering applications into a manageablecommon end-user environment

• to establish unique process flows and controls based onspecific product or process requirements

If your goals are to improve productivity and quality of both your engineering tools and your products, you need to furtherexamine the capabilities of Workbench by visiting the ANSYSWorkbench Community Portal accessible from the ANSYS Customer Portal and contacting an ANSYS representative.

Transient thermal analysis results of circuit board with probedresults using new transient analysis capability in the WorkbenchSimulation Environment

ANSYS Workbench Customization Capabilities

13

with a variety of mean stress correction methods includ-ing Marrow and SMT, the Fatigue Module provides contour plots of fatigue life, factor of safety and stressbiaxiality. Additional results include rainflow matrix,damage matrix, fatigue sensitivity and hysteresis.

Extending its capabilities, ANSYS DesignXplorerallows users to perform optimization and Design for SixSigma with any application or sequence of applications.The new third-party plug-in allows parameter transfer toand from other applications such as in-house codes.DesignXplorer will process a sequence of instructionsdefined by XML file and use this information to interactwith any parameter source. The new Auto Defined sampling design option will automatically pick the most accurate sampling method based upon the number of input parameters. Auto Defined will select frombetween the new VIF-Optimal and G-Optimal designs.DesignXplorer now has a NLPQL (Nonlinear Program-ming by Quadratic approximation of the Lagrangian)algorithm, which provides much faster convergence for smooth continuous responses that do not have localoptima. With the new memory management imple-mented at 10.0, Tradeoff studies can now be performedfor very large datasets (>100,000). ■

The first installment of this series presented an introductory overview of Century Dynamics Inc. (CDI)and explored the AUTODYN explicit dynamics product. This article investigates CDI software specificto the oil and gas industry, where pressure is enormous to explore and exploit new petroleumsources for satisfying growing world demand.

With most of the easy fields already tapped,petroleum companies must focus on those whereweather and other environmental conditions are extremely hostile. To tackle these hostile conditions, significantly more engineering efforts are now required before the first drilling rig arrives, and engineers are increasingly relying on advancedsimulation tools to design safe and financially viable systems. Since 1993, CDI has provided oil and gasengineering simulation software to meet these challenging applications: AutoReaGas for studyinggas cloud explosions, ASAS for designing marinestructures and AQWA for hydrodynamic assessmentof structures.

Simulation Tools for Designing Safe and

Efficient Drilling Rigs

Part 2 of 2: Century Dynamics software helps oil and gas

industries develop better systems for hostile environments.By Steve PilzProduct Manager, ANSYS, Inc.

Naury BirnbaumPresident, Century Dynamics

Studying Gas Cloud Explosions

Designed to aid in modeling and understanding the effects of gas cloud ignition and explosion phenomena, AutoReaGas has been under continuousdevelopment by CDI in conjunction with TNO (TheNetherlands Research Organization) since 1994. Thesoftware is particularly helpful at understanding howstructures such as piping networks affect flame frontpropagation and the resulting explosive pressures thatoccur when an invisible gas cloud forms around thestructure and is accidentally ignited. The purpose ofthe software is to allow users of AutoReaGas to economically conduct appropriate safety studies andultimately design safer facilities for oilrigs as well asonshore plants such as refineries.

AutoReaGas was written in response to numerous industrial accidents where oilrigs explodedand many people were killed because of gas clouddeflagrations and the resulting high blast pressures.

14

www ansys com ANSYS Solutions | Summer 2005

The software can be used to predict the severity ofsimulated accidents as well as to aid in the design ofsystems that reduce the health and safety risks in theoil and gas industry. Where there is an interest in theblast response of equipment or structures, AutoRea-Gas results can be readily used with structural analysis software such as CDI’s AUTODYN for combined blast-structure interaction studies.

AutoReaGas is useful in studying flame front propagation through offshore platform piping networks.

Modeling Marine Structures

A finite element program designed to specificallyaddress the needs of marine structural simulation,ASAS has been used by offshore and marine engineers for over 30 years. CDI acquired ASAS in

2001 and has continued to develop and enhance thecapabilities of the program, which can now be used tomodel jack-up platforms, compliant structures, mani-fold installation and riser analysis as well as mostother floating structures. Features include:

• wave, current and wind load• regular and random waves• Airy, Stokes 5th, Cnoidal, stream function and

shell new wave• tube and beam elements• flooded or sealed members• frame, panel and concrete structure code

checks and its visualization• drag and inertia force• marine growth• wave loading within the API code

Ship hull design was modeled with ASAS.

AutoReaGas is useful in studying flame front propagation through offshore platform piping networks.

15

www ansys com ANSYS Solutions | Summer 2005

Multibody Hydrodynamic Software

The third program in the CDI oil and gas simulationsuite is AQWA, a multibody hydrodynamic softwaretool used around the world by engineers to simulateoffshore jacket launching, calculate response of ship hulls to wave loading, analyze hydrodynamic and mechanical interaction effects on floating structures, design loading and unloading devices forcargo transport, optimize floating structure mooringsystems such as cables and anchors, simulate liftingoperations between vessels and aid in the design ofport installations.

AQWA provides unique simulation capabilities forthe hydrodynamic assessment of all types of structuresincluding spars, FPSOs (Floating Production Storageand Offloading vessels), mooring systems, buoys, TLPs(Tension Leg Platforms), semi-submersibles, ships,cable dynamics and structure interaction. ■

AQWA is used for the hydrodynamic assessment of offshore structuressuch as this mammoth floating airport concept shown with C-17 aircraft.

Concrete gravity platform was modeled with ASAS.

16


Theoretical Solution and Euler Formula

For a straight slender column, the Euler formula is represented by:

And the equation for hoop stress in a cylinder is:

Considering the quarter symmetry of two-lobebuckling and neglecting the arc of a 90˚ segment,applying Euler’s equation to the shell yields:

Solving for the pressure load, we get:

The equation above shows the important effect ofshell thickness and diameter on buckling capability.

For finite length cylinders, an eigenvalue solutionfor the critical pressure for buckling is given byFlügge:

,

where

17

Finite element codes consider the linear elastic solu-tion for buckling. Eigenvalue buckling analysis predictsthe theoretical buckling strength (the bifurcation point)of an ideal linear elastic structure.

This method corresponds to the textbookapproach to elastic buckling analysis: For example, aneigenvalue buckling analysis of a column will matchthe classical Euler solution. However, unlike long, slen-der columns, many real-world structures such as pres-surized cylinders can achieve their theoretical elasticbuckling capabilities.

A nonlinear buckling analysis is a static analysiswith large deflection and/or plasticity to determinewhere and when the structure reaches its limit load or maximum load. In the nonlinear analysis that follows, the path dependence is shown, and thus itrepresents a more accurate solution to find the lowestbuckling mode.

Comparing ANSYS Shell Elementsfor Buckling AnalysisAn externally pressurized cylinder example demonstrates nonlinear effects, follower forces andmodeling techniques used for buckling analysis.

By Dr. Charles H. RochePratt & WhitneyDivision of United Technologies


2218 and m is the number of lobes in the buckled shape;for our example, the lowest mode is two-lobe, so m is2. Also, E and v are the elastic constants, and r is theradius of the cylinder.

Finite Element Analyses

In the example below, we consider the linear and non-linear buckling analysis of a shell model using variousshell elements in ANSYS. The results are compared tothe theoretical solution given by Flügge. Consider acylindrical tube geometry of (l = 52 in., O.D. = 6.5 in., t = 0.035 in.) made up of stainless steel of materialproperties (E = 28.5 msi, r = 0.3). The cylinder easily meets thin shell requirements as well as elastic buckling.

The methodology for eigenvalue buckling analysis is given in ANSYS Documentation. Theaccompanying table below shows the comparison ofbuckling analysis using various shell elements inANSYS with the theoretical solution of 19.58 as givenby Flügge.

Surface effects were resolved with the ESURFcommand and the SURF154 elements for the following legacy elements: SHELL43, SHELL63,SHELL91, SHELL99 and SHELL143. The technique isunnecessary and not applicable for the newerSHELL181 and SOLSH190. With the exception of theaxiharmonic element SHELL61, using the surfaceeffect elements shows dramatic improvement with thetheoretical solution of 19.58 psi.

The use of SURF153 elements with SHELL61may be inappropriate, as the results suggest. Uniqueto the table would be the axiharmonic shell elementSHELL61 and the new SOLSH190 element availableonly in ANSYS 9.0. (Note this new solid/shell elementis very attractive for thin shell analysis and may quickly surpass the traditional thin shell elements.)

To perform nonlinear buckling analysis, one mustperturbate the model, i.e., take away the perfectgeometry or loading. In this example, small nodalforces were added to make the loading nonsymmetri-cal. No material nonlinear was added; only large dis-placement (NLGEOM, ON) was used. For all models,the mesh density was adequate and all other spuriouseffects were eliminated to make clear the need forusing the ESURF command and the need for runningnonlinear static models. Only the first buckling modeis presented.

Experimental Verification

A test rig was set up for the stainless steel cylinderwith 52 in. representing the span between the simplesupports. Pressure was slowly increased while deflec-tion probes recorded displacement as a function ofpressure. The test specimen buckled at approximately18.6 psi. (The test was cut short when permanentdeformation was present.) An initial uploading to 18psi and back to zero showed no permanent deforma-tion but review of the test displacement plot showssignificant nonlinearity at 18 psi. Note that nonlinearityin deflection does not mean inelastic behavior.

The comparison between the test and theSHELL63 nonlinear results is remarkable. There wasno deviation up to 18.6 psi. The small deviationbetween the two curves is due to curve fittingbetween converged finite element loads. Deflectionmeasurements were accurate only to about 0.030 in.of deflection, the limit of the probes. The test datummatched the nonlinear ANSYS run until the specimendeflected enough to show an asymptotic approach tothe bifurcation point, adequate for most structuralanalyses.

Comparison of Buckling Analysis Using ANSYS Shell Elements

Element type ANSYS Linear Linear eigenvalue Nonlinear static Solutionversion eigenvalue with ESURF (linear elastic material)

SHELL43 8.1 25.84 19.45 19.00

SHELL61 8.1 25.68 25.76 Not applicable

SHELL63 8.1 25.55 19.27 18.87

SHELL91 8.1 25.54 19.26 18.78

SHELL93 8.1 25.54 19.26 19.00

SHELL99 8.1 25.54 19.26 18.30

SHELL143 8.1 25.84 19.45 19.00

SHELL181 8.1 19.36 Not applicable 19.01

SOLSH190 9.0 19.38 Not applicable 18.99


19

A command mode example highlights theuse of surface effect elements:

ET,1,SHELL63,,,,,,,,, ! A COMMON SHELL ELEMENT EXAMPLE

R,1,0.035

ET,2,SURF154,,,, ! CREATE SURFACE EFFECTELEMENTS

TYPE,2 ! SELECT THE 3-D SURFACE EFFECTELEMENTS

R,2,,,,,,,

ESURF ! GENERATE ELEMENTS ON THE EXISTING SELECTED ELEMENTS

ESEL,S,TYPE,,2 ! SELECT SURF154 TO APPLYPRESSURE

EMODIF,ALL,REAL,2

SFE,ALL,1,PRES,,10 ! APPLY 10 PSI TO THESTRUCTURE

All of the new shell elements will converge tothe eigenvector without the need for surfaceeffects elements. The nonlinear static solutionremains attractive, for it shows the path dependence of force and displacement andshould confirm a proper eigenvalue solution.

Proper use of shell elements should produceaccurate eigenvalue results. And for any newgeometry, testing may be necessary for FE verification, especially since some structures canachieve full theoretical load-carrying capability,while others may never approach the theoreticalsolution. ■

Chuck Roche is a Structures Lead at Pratt & Whitney,teaches ANSYS composites courses at Pratt & WhitneyEngineering Technical University and is an adjunct professor at the University of Hartford and the Universityof Connecticut.

Large increments in solver load steps did notcause any deviation from the test data. The accuracyof the simple linear material model would be suspectas the cylinder underwent plastic stress, but it was all that was necessary to achieve accuracy within 3 percent.

References and Further Reading

F.B. Seely, J.O. Smith, “Advanced Mechanics of Materials,” 2nd ed., Wiley & Sons, 1952

C.H. Roche, C.A. Whitney, “Robust Design of Load BearingStructures,” NAFEMS World Congress 2003, Orlando, FL

W. Flügge, “Stresses in Shells,” Springer-Verlag, Berlin, 1960


Conclusion

Most of the legacy shell elements, when used alone,will not be accurate in resolving eigenvalue bucklingfor curved surfaces under pressure. It is recom-mended to use surface effect elements for pressureapplication on elements not having pressure load stiff-ness capability according to the ANSYS Element Reference manual. It is recommended to considerusing the linear eigenvalue solution only in a prelimi-nary design phase and the nonlinear static solution fora final design analysis.

Comparison of Nonlinear ANSYS Solutions

Finite element solution showing two lobes

Test specimen run to fully buckled solution

Pre

ssur

e (p

si)

Inward Radial Displacement (in.)

A team of engineering students at the University ofWashington in Seattle, Wash., won the fourth annualCollege Design Engineering Award for their work inrefining the design of a micropump that may one daycirculate cooling fluid for electronic circuitry in satellites and spacecraft as well as move small volumes of fluids in chemical analyzers, medicalequipment and other applications where space is limited and reliability is essential.

Sponsored by ANSYS, Inc. as part of its cont-inuing support of engineering education, the awardwas made based on project scope, engineering problems encountered and solved, uniqueness ofsolutions, potential for commercialization and impacton the engineering community. Entries were judged by editors of Design News magazine as part of the publication’s Excellence in Design Achievement Program. Funded by ANSYS, Inc, the award for thewinning project is $10,000 in cash to the student teamand a $10,000 scholarship grant for the school’s engineering department.

Under the supervision of their advisor, Dr. FredForster at the University of Washington, engineeringstudents Adrian Gamboa, Jone Chung and Chris Morris worked on the project to maximize the outputof the micropump design incorporating unique no-moving-part valves based on a concept originatedby famed inventor Nicola Tesla in 1920.

Instead of relying on conventional flap-typecheck valves, the micropump uses an oscillatingmembrane actuated by a piezoelectric driving elementto move fluid through inlet and outlet fixed-geometrypassive check valves. The valves are essentially a setof two channels: a straight-line channel connectingthe pump chamber to the inlet/output and a curvedchannel that loops around from the chamber to

Students boost output of micropumpdesign threefold using ANSYS structural,CFD and optimization tools.

intersect the straight channel. Fluid moves through thevalves based on the position of the membrane as itvibrates. The shape of the channels and the angle atwhich they intersect produce turbulence and restrictthe backflow. In effect, a differential pressure drop isdeveloped in one direction to create net flow in thepath of least resistance.

Because the valves have no moving parts, themicropump offers distinct advantages over otherdesigns. The device is highly reliable, requiring norepairs or replacement parts. The pumping systemhandles any fluid, even air, which may make it usefulas a sampling device in medical and security systems.Also, the pump is clog-resistant, even though valvechannels measure only 300 microns wide and 750microns deep. Probably the greatest advantage of themicropump is its small one-square-inch footprint, providing fluid-moving capability where space is at a premium.

A major challenge in developing a micropump forpractical applications has been increasing flow rateand pressure to useful levels. Using water as a coolant,the system was capable of dissipating 35 Watts ofheat with pump average power usage of only 100 mWand a flow rate of 11 mL/minute.

The University of Washington students were ableto maximize pressure and flow rate using ANSYS software. A finite element model of the membrane wassolved with ANSYS structural software to determinethe component’s deflection and resonant behavior.Pump performance was simulated with ANSYS FLOTRAN computational fluid dynamics (CFD) software coupled to the Subproblem ApproximationOptimization Method included in ANSYS, which iteratively computed pressure and flow for multiplevalue shapes until the solution converged on an

CAE Community

20


optimum. Using this CFD and optimization approach,the students were able to increase pump output threefold for both pressure and flow rate, thus takingthe device much closer to practical applications.

“Other people built and tested configurations totry to optimize the general design of the Tesla valve. I think we were first to use computational fluid dynamics with an optimizer to determine the best

design,” said Adrian Gamboa. “Automation of theprocess revealed a lot about what makes this valvework and how we could maximize pump output.”

According to Gamboa, the team also may be thefirst to stack micropumps connected in parallel toincrease flow rate. Made of flat plastic panels usingMicro-Electro-Mechanical System (MEMS) machiningfabrication techniques, the micropumps are readilystackable, making the plastic devices well suited forhigh-density cooling.

The team’s advisor, Dr. Forster, thinks Gamboa,Morris and Chung did outstanding work in refining themicropump design. “Beyond learning basic theory andfundamentals, today’s engineering students absolutelymust know what analysis tools are available and howto use them on a practical basis,” said Forster. “I’mencouraged to see students confronting difficultdesign challenges never before tackled. The talent,know-how and innovative thinking of this bright newgeneration represent the future of the engineering profession and the key to successful designs in thecoming decades.” ■

The University of Washington team is the latest winner of the annual College Design Engineering Award sponsored byANSYS, Inc. to recognize outstanding achievement in design by engineering students.

2004Students from the Massachusetts Institute of Technology developed a projector that produces large, clear images frominexpensive microfilm so people without ready access to teaching materials can share information.

Previous Winners

2003An engineering student team from Texas Christian Universitydesigned an automated quality inspection system for checkingvacuum strength of saline bottles in the pharmaceutical industry.

2002The first award went to students from the University of Evansville (Indiana) for designing an adaptive tricycle for a disabled child unable to use a standard tricycle.

Probably the greatest advantage of themicropump is its small one-square-inchfootprint.

CFD simulation with ANSYS FLOTRANshows high-velocity flow around thesharp corner of the value channel.

A stacked array of four micropumps increases cooling flow to theheat sink without adding to the footprint of the assembly.

The ANSYS optimizationroutine enabled the studentteam to develop an optimalvalve configuration for themicropump.

Finite element analysis of one-quarter ofthe membrane shows transverse displace-ment of the component.

21


22

When petroleum company BG Group commissionedGenesis Oil and Gas Consultants to study the feasibility of increasing gas production from theirMiskar concession off Tunisia, it became apparent thata major bottleneck could be the slug catcher at theend of the Miskar pipeline in the on-shore Hannibalterminal.

The slug catcher receives a mixture of productiongas and liquid condensate from the pipeline into sixseparation pipes or “fingers,” where gravity acts toseparate the two phases: the gas passing upward tothe gas outlet with the heavier liquid falling throughshort downcomers into long liquid fingers, where it isstored. If the surge of gas and liquid in the 20-minuteperiod after daily cleaning (“sphering”) of the pipelineis too great, then liquid could overflow into the gasoutlet causing problems downstream. So Genesisneeded to know if the existing slug catcher has the

Debottlenecking the

Hannibal Slug CatcherANSYS CFX simulates transient two-phaseflow in a complex pipe network.

Incoming flow from the pipeline is distributed by a header into the six separation fingers of the slugcatcher. The lighter gas passes upward into a gas header leading to an outlet while the heavierliquid flows through the downcomers into the liquid fingers.

By Justin Penrose and Phil StopfordANSYS CFX Technical ServicesANSYS Europe

capacity to cope with higher flows of gas and liquid. Ifnot, then a new slug catcher would have to be built atan estimated capital cost of $25 million.

To assess the capacity of the slug catcher, Genesis asked ANSYS to simulate the time-dependent two-phase fluid flow in the system usingthe Eulerian multiphase flow model in ANSYS CFXsoftware. To forecast liquid overflow reliably, the CFDmodel needed to track the motion of the gas-liquidinterface accurately throughout the pipe system. Normally, the size of mesh required to do this wouldmake a long transient simulation intractable but, byusing transient mesh adaption to concentrate cellsinto the region around the interface, we were able toreduce the total mesh size by an order of magnitude.

The peak liquid levels were predicted for the original pipeline gas/liquid flow rate and for a flow rateincreased by 45 percent. It was found that the


CFD Update: What’s New in Computational Fluid Dynamics

23

Example of transient adaptive meshing to resolve the gas–liquid interface, where blue represents gas and red represents liquid.

This view of the slug catcher at the BG Hannibal terminal shows theinlet pipe and header as well as the short separation fingers abovethe much longer liquid fingers.

Comparison of maximum liquid height for current and 45 percent higher flow rate. In both cases, no liquid overflow to the gas outlet is predicted.

maximum height of the liquid in the separation fingersincreased significantly when the flow rate wasincreased but no catastrophic overflow of liquid intothe gas outlet was found. As a result, it was concludedthat the existing slug catcher will be able to cope withthe increased pipeline capacity.

The innovative nature of this work was recognized when the BG Tunisia team, including thepresent authors, received an award under the BGGroup Chief Executive Innovation Awards Scheme for2004. The submission was the winner in the “Alliancewith an External Party” category, one of only threewinning projects from over 100 entries submitted. ■


ANSYS CFX Advances Centrifugal Stage TechnologyDevelopmentSoftware predicts overall performance of a centrifugal compressoras well as individual components. By Dr. Basuki N. Srivastava, Lead Engineer

General Electric Company, Aircraft Engines

Typical CFD model of the full geometry of a compressor (not to scale) Secondary flow development in a diffuser

The frozen rotor CFDmodel shown in red utilizes interface areamatching (not to scale).

24



Centrifugal stage performance and loss are predicted by ANSYS CFX software.

Success of the steady CFD analysis usingANSYS CFX for stage promises to be a considerablesavings in the cost of development for future genera-tions of centrifugal stages by reducing the number ofrig tests and enhancing the current technology base.

This GE effort is a result of a team technical effortby GE engineers coupled with a support team fromboth ANSYS and GE Global companies. ■

Engineers at General Electric Company in Lynn, Mass.,have made significant progress in demonstrating thata steady-state model available in ANSYS CFX compu-tational fluid dynamics software can predict not onlythe overall performance but also individual componentperformance of a centrifugal compressor. This can be done using a mixing plane formulation of animpeller-diffuser-deswirler centrifugal geometry. Manyinvestigators have computationally studied the aero-dynamics of a centrifugal compressor stage and somehave concluded that unsteady-stage CFD is necessaryfor centrifugal-stage predictions. However, GE hasdetermined that a cost-effective design method tunedto provide component performance and consistencywith design cycle requirements is needed for the designapplication engineers. Advanced CFD validation onseveral challenging geometries has successfully beencompleted, while validation on others is in progress.This capability is currently being used to design anadvanced new generation of centrifugal stages.

Mixing plane formulation of the stage interface ina centrifugal compressor offers an added advantageof a computationally tractable CFD model (as com-pared to a frozen rotor model) that is very well-suitedto rapid turnaround of results for design iterations.

A typical speedline map comparison of the testdata with the CFX-TASCflow CFD model, based onmixing plane concept for a similar centrifugal stage,predicts the performance and loss coefficient.

Transition zone withbleed flowextraction

Diffuser

Impeller

Inlet

Vanes

Deswirler

Mixing plane CFD model using one passage

25


26

The study of aerodynamics in the automotive

industry is important to improve fuel economy

as well as vehicle comfort and safety. Since

the 1980s, automobile industries have relied

more and more on numerical methods for

vehicle design in order to reduce expensive

experimental tests traditionally required for

aerodynamic studies.

In the Department of Mechanical Engineering at the University of Coimbra, Portugal, engineers have been working on designing an innovative vehicle that combines a sporty design with standard transportation capability, the NUMA car. In order tofine-tune the shape of the car, an aerodynamic analysis was performed with ANSYS CFX computa-tional fluid dynamics software.

Velocity vector field at the rear of the car

NUMA sports car

Designing a Better Sports CarThrough Engineering SimulationAerodynamic analysis with ANSYS CFX fine-tunes the vehicle shape.

By Nuno Ricardo Rosmaninho and António Gameiro LopesDepartment of Mechanical EngineeringUniversity of Coimbra, Portugal



27

An inflated boundary of prismatic elements wasused near the car surface to improve spatial resolu-tion and gain a better understanding of boundary-layer phenomena. An unstructured mesh with tetrahedral elements was used for volume meshing.Simulations were carried out with the SST turbulencemodel, coupled with a blend factor of 0.5 for theadvection scheme.

The research started with three models with characteristics of a sports car. The models differed inthe design of the rear end. Aerodynamic optimizationwas used to determine the best configuration of airfoils, spoilers and diffusers. Because standard solutions for optimizing aerodynamics have been welldeveloped over the years, we simply studied the bestchoices to apply. And, since the car was to bedesigned for sporty performance, study of both dragand lift were important. Little details on shape and position of spoilers and airfoils played an importantrole in the compromise between lift and drag.

The ANSYS CFX analyses provided informationon flow separation, pressure and velocity fields, vortices and forces interacting with the vehicle. Thisinformation allowed engineers to make modifications

in the car shape, producing better results. In the finaldesign, airfoils were positioned based on the visualiza-tion of the flow field in the rear of vehicle and on theanalysis of the corresponding effect in terms of dragand lift. Pressure coefficient charts helped with thedefinition of the overall geometry and gave insight onthe distribution of the applied forces.

Thanks to ANSYS CFX software, the universitycould reduce the positive lift and still bring the drag to95 percent of the initial value. ANSYS CFX proved tobe an excellent tool for automotive aerodynamics. ■

Pressure coefficient results on the car symmetry plane

Evolution of CD and CL for different configurations of thecar geometry

Airflow past the rear airfoil

ANSYS CFX visualization of flow streamlines


28

In the early years, FLOTRAN was used to carryout CFD solutions. Recently, however, ANSYS CFXwas adopted along with the Workbench environmentto greatly expand the scope of the problems that areaddressed during the semester. Since then, I beganteaching at the University of St. Thomas in St. Paul,Minnesota, and have integrated a similar course intothat curriculum, and it’s consistently filled to capacity.In the meantime, Dr. Sparrow and graduate studentsSandra Sparr and Jimmy Tong have continued todevelop the course, which is now taken by more than150 students each year.

Seven years ago, Dr. Ephraim Sparrow and I

developed a course at the University of Minnesota

called Case Studies in the Thermal Sciences. This

senior-level course utilized ANSYS software to

solve industry-related problems in the thermal

and fluid sciences. Students immediately recog-

nized the value, and class enrollment swelled to

approximately 60 students per year.

Minnesota Moving Ahead withANSYS in the ClassroomTwo engineering schools revolutionize undergraduate instructionwith ANSYS Workbench and ANSYS CFX software.

By Dr. John P. Abraham, ProfessorUniversity of St. Thomas, Minnesota



29

While CFD is typically studied as a graduate-levelcourse, the ease of use and broad capability ofANSYS CFX have enabled it to be brought into theundergraduate classroom. Both Dr. Sparrow and Ibelieve in two prime benefits of exposing undergrad-uates to CFD. First, the experience with ANSYS CFXdeepens the students’ understanding of fluid flowphenomena. In particular, the visualization capabilitygreatly enhances students’ intuition of flow behavior.Second, ANSYS CFX opens a door to a new class ofproblems that can be solved by undergraduates whoare no longer limited by the narrow range of classicalflow solutions.

The decision to use ANSYS CFX at both the University of Minnesota and the University of St. Thomas was made after a careful consideration ofall the leading CFD packages. It was concluded thatthe ongoing integration of ANSYS CFX with the popu-lar ANSYS software suite, the robust solver providedby ANSYS CFX and the CAD capabilities of ANSYSWorkbench gave ANSYS CFX a clear advantage overcompeting software alternatives.

Both schools are now teaching ANSYS CFX and Workbench as part of a one-semester, 15-weekcourse that combines a theoretical basis with thecomputational exercises. The courses are built aroundindustrial case studies that expose students toincreasingly complex flow situations as the semesterprogresses. The incredible popularity of the coursehas alleviated concerns about the difficulty of introducing advanced CFD simulations at the undergraduate level.

The success of ANSYS CFX at the undergraduatelevel has not gone unnoticed. The SupercomputingInstitute for Digital Simulation and Advanced Computation at the University of Minnesota hasrecently acquired the ANSYS CFX program, and anumber of graduate students are currently using thesoftware for their doctoral research. ■

Dr. John Abraham uses ANSYS CFX software inhis undergraduate CFD course.

Using ANSYS CFX CFD software in the Workbench environmentallows students at the University of St. Thomas to solve a widerrange of problems. In this plot, streamlines represent air-flow patterns through an array of heated tube banks. The results indicate regions of separated flow (eddies) on the trailing edges of the tubes. The streamline colors quantify the heating of the airas it passes through the array.


Simulation at Work

Taxi 2000, Inc. was formed to develop and commer-cialize the concept of a personal rapid transit (PRT)system emerging from work done at the University ofMinnesota. In contrast to mass transit systems thatmove many people in large vehicles traveling succes-sively from station to station along an establishedroute, SkyWeb Express saves considerable time withsmaller vehicles that travel directly to a destinationwithout intermediate stops.

Each electrically driven vehicle carries up to 650pounds (typically three or four people) and travels on araised network of single-rail guideways connectingconveniently located stations. The rider swipes a pre-paid card through a stanchion in front of an emptywaiting vehicle, punches in a destination number, andtakes a seat. A computer system chooses the fastestroute to the destination and proceeds there non-stop.Since stations are off the main line, riders go directlyfrom origin to destination, bypassing all the other stations along the way. There are no stop signs, redlights or interferences along the route.

Design Challenges

The company needed to get a working prototype sys-tem built as quickly as possible to demonstrate theoperation and feasibility of SkyWeb Express. Vehicleweight was a key economic issue, so the cabin andchassis were designed to be constructed of tubularaluminum members bonded together with structuraladhesive, much the same as aircraft are fabricated.Safety and reliability were primary concerns, so components were required to be strong enough tosafely withstand operational loads.

Of particular concern was the bonded aluminumT-joint connecting the cabin to the narrow chassis.This part had to provide adequate support for turningand wind-load forces as the vehicle rides along theguideway, as well as repeated unbalanced load cyclesas passengers enter and exit the cabin.

The Simulation Solution

To determine if the bonded aluminum T-joint andcabin structure were strong enough to withstandthese operational loads, the company turned to the consulting firm Summit Analysis Inc. to studycomponent stresses.

ANSYS Tools Help Develop First-Generation Personal Rapid Transit Vehicles

Small computer-controlled cars

take passengers non-stop to

their destination via a network

of single-rail guideways.

30


Working on a Dell Pentium 3 desktop PC, Summit Analysis used ANSYS meshing capabilities toconstruct a parametric finite element model of thetubular vehicle chassis and overlapping gussetsheets. ANSYS surface-to-surface contact elementswere used to represent the adhesive joining the structural members. The bonded contact feature ofthese elements allows both tangential and normalforces to be transferred between adjacent surfaces.

Using these software features, the study enabledengineers to develop a satisfactory design and quicklyverified that joint forces were within acceptable limitsfor the shear and peel strengths for the adhesive.

Key Benefits

Summit Analysis used ANSYS contact elements toquickly and accurately model the T-joint structures.

“ANSYS Structural software was valuable instudying stress levels for this project, especiallythe convenience and flexibility of the surface-to-surface bonded contact feature. This function-ality greatly facilitated modeling the structure andprovided a convenient mechanism for extractingload values on each adhesive joint to verify thatforces were within acceptable limits. The speedof this approach enabled Taxi 2000 to confirm theperformance expectations of the design so aworking prototype of the SkyWeb Express couldbe completed within the time constraints of thedevelopment program.”

Mitchell VoehlCEO and Engineering ConsultantSummit Analysis, Inc.

Traveling on a network of raised single-rail guideways, computer-controlled SkyWeb Express personalrapid transit (PRT) vehicles take people directly to their destination with no intermediate stops.

31

ANSYS Structural was used to determine if stresses were withinacceptable limits for the adhesive-bonded tubular aluminum T-jointconnecting the cabin to the narrow vehicle chassis.

For such analyses, ANSYS contact technology hasuseful capabilities, including automatic assembly contact detection, robust default settings and a rangeof contact behavior and solution options. Other FEApackages may require extensive manual interventionin setting up these contact conditions. ANSYSsoftware also can generate customized reports easilywith critical information summed, listed or plotted foreach joint.

Verification of the safety of the SkyWeb Expressstructure enabled the company to construct a workingprototype of the SkyWeb Express on schedule. Sincethen, more than 5,000 people have ridden the proto-type PRT vehicle, with cities in North America, Europe,the Middle East and Asia investigating the feasibility ofimplementing the transit system. ■


32

Guest Commentary

Computer-aided engineering(CAE) is in the midst of a renaissance. Once a highlyspecialized and somewhatunder-appreciated field, CAEis muscling its way to theforefront of product develop-ment. Slashing costs is stillpriority number one acrossnearly all verticals in the manufacturing industry, andcompanies are scrutinizingtheir processes to uncoverways to save money. Applied

and understood properly, CAE can lead to better models and, subsequently, less design rework. It canallow designers and analysts to explore more designalternatives and free them to spend more time ondesign work and core engineering. CAE can reduce,not replace, the number of physical prototypes, notoriously costly and time exhaustive. And it can helpboost product quality, which builds brand loyalty, andcan help avoid program-killing warranty problems orrecalls down the road.

Digital simulation is not all peaches and cream,however. Challenges abound. For example, integratingsimulation more fully into the product developmentprocess requires overcoming hindrances to inter-operability between the various design and analysisapplication packages. No tool lives in a vacuum. Interoperability and integration are key attributes of any technology solution. Oftentimes, software products aren’t even backward compatible with

themselves, let alone with other systems. The NationalInstitute of Standards and Technology has estimatedthat the automotive industry throws away $1 billionannually on interoperability-related expenses: anastounding number.

Streamlining the Process Chain

The challenge of overcoming such technical issues pales in comparison to confronting cultural, organiza-tional and process issues, however. Companies aiming to reap the tremendous benefits of CAE mustmeet these formidable challenges head-on. Often thisinvolves taking a close look at the company’s productdevelopment cycle from start to finish in evaluatinghow digital simulation can best be integrated into theprocess. In such an examination, companies oftenfind that their process chain resembles a cow paththat bends and wanders in a convoluted trail thatcompanies follow because things have been donethat way for decades.

Obliterating these inefficiencies and streamliningthe process chain to leverage simulation often is donein conjunction with PLM to digitally create, store, manage, share and reuse data. As CAE information isincreasingly relied upon in product development, somust a place for it be carved out within — not around— the PLM framework with capabilities for managingmultiple iterations of analysis data, correlating analysisresults with physical test data and associating analysis and test data with CAD models in ways thatlet digital and physical prototyping have broader anddeeper impact on product development.

CAE Moves to the

Forefrontof Product Development

Companies must meet challenges head on to reapthe tremendous benefits of digital simulation.

By Charles FoundyllerPresidentDaratech, Inc.


33

Bridging the Gap Between Design and Analysis

In scrutinizing their product development processes toimprove efficiency, some manufacturers are findinggains in and around the early product design phase. In many organizations, a process gap exists betweenthe design and analysis teams whereby designers waituntil their work is finished before “throwing it over the wall” to analysis, which in turn makes suggestionsand then throws it back over the wall to the designer to rework. This serial process wastes time, and the separation of design and analysis can create a competitive atmosphere rather than fostering an environment of shared goals.

Once technologically impractical, a new focus onimproved usability of analysis codes through wizardsand more simplified user interfaces, combined withmore powerful desktop computing power, have madeupfront analysis a viable option. Some organizationsare now seizing this opportunity to bridge the designand analysis gap by having their designers performmore of the analysis work and enabling the experts to concentrate on the more complex issues. Mostly cultural considerations and cemented processes get in the way of such initiatives. Many companiesremain unconvinced that putting analysis on everydesigner’s workstation is a good idea, thinking that misinterpretation of the results by people who don’t know how to question the inputs and outputsmay be too great a risk.

While many OEMs acknowledge the merits of anupfront analysis strategy, some remain hesitant tomake this leap. Many Tier 1 and 2 suppliers, on theother hand, have more readily embraced the strategyas a way to survive and are leading the way in developing approaches for best implementing digitalsimulation early in the development process.

Test and Analysis Must Get Along

In many organizations, there is a perception that analysis will replace test. But most agree that we’llnever get to a point where you’ll be able to analyzestraight through to production. Moreover, companiesmust continue to rely heavily on physical testing tostudy competitive products.

The real issue is improving the interface betweenthe analysis and test organizations to create the mostefficient process possible, to get the best and maximum amount of information possible out of every

turn. Clearly, there will always be testing, but undeniably more and more performance verificationwill be carried out using virtual models and simulation.

The resulting reduction in physical prototypes willsave companies billions. Indeed, Daratech’s model ofthe savings digital simulation can yield at automotiveOEMs and their suppliers indicates that, by 2010, theautomotive industry will have saved on the order of $2 billion on a cumulative basis.

Where Do We Go from Here?

With each successful program comes a higher level ofconfidence in digital simulation. While physical testresults are still typically seen as the data of record,many organizations are beginning to second-guessthis long held and once unchallengeable position.

Today it is difficult to keep CAE on the back burner due to the maturity and refinement of the software; the ability to simulate multiple scenariosrather than build multiple prototypes; the speed afforded by new high-performance computing architectures; the ability to better correlate modelswith physical test results; and the expanding body ofevidence before us... the cars we drive, the planes wefly, the tractors we ride, the defense systems that protect us and on and on.

While manufacturers formerly could design products such as these by the seat of their pants without much simulation, today it is virtually impossible to satisfy all of the technical requirementsof product development and remain competitive without a thorough and comprehensive analysis program. Simply put, engineering simulation is nolonger an option but a necessity. And for that reason,the use of analysis tools will undoubtedly accelerate in the coming years, propelling companies that wisely implement the technology to leading positionsin their industries. ■

Charles Foundyller is founder and president of Daratech, Inc.,a market research and technology assessment firm specializing in CAD, CAM, CAE, PLM and related areas. The company hosts a variety of user forums and technologyworkshops throughout the year and publishes a range of studies, newsletters, market reports, sourcebooks and industry statistics. For more information on the company, callDaratech at 617.354.2339 or visit www.daratech.com.

In integrating digital simulation into the product developmentcycle, companies must streamline process chains that oftenresemble a convoluted trail followed because things have beendone that way for decades.

Source: Daratech, Inc.

COW PATH PROCESS CHAIN

According to Daratech, automotive OEMs and their supplierswill save $2 billion on a cumulative basis by 2010 by reducingphysical prototypes through greater use of digital simulation.

CUMULATIVE AUTOMOTIVE INDUSTRY SAVINGS

THROUGH REDUCED PHYSICAL TESTING

Source: Daratech, Inc.


Tech File

As much as we might not wantto think about it, there are errorsin our work. One source of erroris inadequate mesh refinement.A coarse mesh can under-predict deflections as well asstresses, and this is of particular

concern because it is nonconservative.An experienced and conscientious analyst will try

to estimate meshing errors by comparing averagedand unaveraged nodal results, plotting the estimatederror, and using experience and insight to review theresults and determine whether the mesh is adequatelyrefined to provide the required accuracy. If the meshisn’t satisfactory, it would be necessary to remesh themodel, reapply boundary conditions, run the solutionagain and review the results.

All of this can be labor-intensive and time-consuming, especially if several iterations are requiredto get an error that has been reduced to an acceptablelevel. Is there a better way to do this? If you’re usingANSYS Workbench, the answer is “yes.”

Origins of Meshing Errors

Meshing errors can have several origins. Any time youhave a sharp inside corner (often called a re-entrantcorner), you have an infinite stress concentration, orsingularity. While the stress at a singularity is theoreti-cally infinite, finite element analysis will always return anon-infinite stress result at these locations. Dependingon the mesh density at the singularity, the calculatedstress may be too high or too low. But no matter whatthe value is, it cannot be trusted.

Another source of meshing error is a mesh that isinsufficiently refined to give good resolution. To getsome idea of what the meshing error is, compare theaveraged nodal result to the unaveraged nodal result.The greater the differences between them, the greaterboth of them are from the true answer. ANSYS software provides an estimate of the meshing errorwith the PLES,SDSG command. However, even whenwe have some idea where the mesh is unsatisfactory,we don’t have an easy way of remeshing the modeland rerunning the analysis.

By John CrawfordConsulting Analyst

Part 2 of 3:

Using adaptive meshing tools to converge on a solution

34

Meshing in Workbench Simulation


Converge on a Solution with Adaptive Meshing

Workbench introduces the ability to automaticallyremesh and resolve a structural or heat transfer analysis until a predetermined criterion has been met.You can use global adaptivity to refine the wholemodel, or scoped adaptivity to restrict adaptive meshing to regions of particular interest. For structuralproblems, adaptive meshing can be used to convergeon a solution for stress results like von Mises stress, displacements and even natural frequencies, while in heat transfer applications it can converge on temperatures and heat flux.

Using a structural problem as an example, here’show we might go about using adaptive meshing.Begin by creating a model and running an analysis.Review the results and see if you can identify regionswhere you might want to use adaptive meshing torefine the mesh and improve the quality of the results.If you are interested in using a stress value like vonMises stress as the criteria for meshing refinement,insert an error branch beneath it in the Outline Treeand then select the error branch to display an estimateof the error in von Mises stress that is caused by meshing.

After reviewing the results, decide if you want touse global adaptivity to refine the entire model, or usescoping to restrict adaptive meshing to certainregions. You may not want to refine the mesh every-where in the model because only certain areas may be of interest, or you may want to avoid singularities.It is important to avoid using stress-based adaptivemeshing on singularities because it is impossible toachieve convergence in these regions. Scoping isdone by picking specific features like an edge or aface, and then inserting a convergence branchbeneath the result you would like to use for adaptivemeshing. You can insert a convergence branch underseveral result entities, and Workbench will continue torefine the mesh until all the desired branches havebeen meshed satisfactorily.

Setting Controls, Convergence Criteriaand Refinement Loops

Once you have picked a result or results that adaptivemeshing will be based upon, go to the Detail View win-dow and set controls for each of them. It’s necessary todefine whether you want to converge on a minimum ormaximum value because you may want to converge onthe maximum value of maximum principal stress, or theminimum value of minimum principal stress.

Also, set the amount of change between succes-sive solutions that will be your criteria for successfulconvergence. It’s important to remember that meshrefinement is based upon the idea that, as the meshdensity is increased, the result value that convergenceis based upon will asymptotically approach the targetanswer. Workbench compares the results from theprevious solution to the results from the current solu-tion. If they differ by an amount that is less than orequal to the convergence criteria, then convergencehas been achieved and the solution will be stopped.

Finally, you should set the maximum number ofrefinement loops that will be allowed. For a well-behaved analysis, it is common for convergence to be achieved after three or four refinement loops, but if a singularity is present, it will be impossible for convergence to be achieved no matter how manyloops are done.

Getting Results

When you have defined all the regions where adaptivemeshing is to take place and set their convergencecontrols, you right-click on the Solution branch in theOutline Tree window and pick Solve. Workbench willbegin running through adaptive meshing loops andwill continue until mesh refinement has satisfied all ofthe convergence criteria or the maximum number ofrefinement loops has been met.

When the solution has been completed, you simply left-click on the result of interest and view theresults from the last refinement loop. You can learnmore about the convergence solution by left-clickingon the convergence branch. This will display a chartthat shows the convergence history of the solutions, a table that lists the number of nodes and elementsthat were used in each refinement loop, the value ofthe result that adaptive meshing was being applied to and the percent change in this result between thelast solution and the previous solution.

The ability to automatically refine a mesh is a verypowerful capability that has been needed for a very long time. Workbench makes this possible in amanner that is relatively easy once you get the hang ofit. Pretty soon you’ll be using it for all the analyses youdo in Workbench. ■

Workbench users can find more detailed information on adaptive meshing in the Tips and Tricks section of the newWorkbench community area now available from the ANSYSCustomer Portal: www1.ansys.com/customer

35


Tips and Techniques

The last several releases of ANSYS software have significant enhancements to the QR damped eigenvalue extraction method, which now supportsunsymmetric stiffness [K] matrices, output of complexeigenvectors and use of the PSOLVE command. Support of unsymmetric stiffness matrices makesQRDAMP particularly useful in solving complex problems such as brake squeal, where frictionaleffects can introduce an unsymmetric [K] term.

Keep in mind that by simply defining friction,unsymmetric [K] matrices may not be present. First ofall, unsymmetric [K] must be requested with theNROPT,UNSYM command. Second, sliding mustoccur for the matrices to be unsymmetric.

QRDAMP Background

In the QRDAMP eigenvalue extraction method, theundamped eigenvalues and eigenvectors based onthe symmetric [K] matrix are calculated first. Dampingand the unsymmetric [K] portion are then included inthe modal equations.

To perform a modal analysis, including frictionaleffects, follow these steps:

1. Perform a nonlinear, large-deflection analysis(NLGEOM,ON) with the unsymmetric Newton-Raphson method (NROPT,UNSYM) and pre-stress effects on (PSTRES,ON). The firsttwo commands allow calculation of unsym-metric [K] of the contact elements, and the third command informs the program that pre-stress effects will be included in a subsequent modal analysis. Also,EMATWRITE,YES is usually required to forceANSYS to write the .emat file, needed in themodal analysis phase.

2. Post-process the results after the static solution. Specifically, plotting or listing the contact status (PLESOL/PRESOL,CONT,STAT)allows you to determine if there are any areas insliding (contact status = 2). If no contact elements have a contact status of “2” (“Sliding”), then no unsymmetric [K] term will be present.

3. Re-enter the solution processor and specify a modal analysis with the QR damped eigen-value extraction method (MODOPT,QRDAMP).Complex eigenvector output also can berequested at this point. Specify PSTRES,ON toinclude the pre-stress effects calculated in step 1. Then, instead of performing a regular solution, run a partial solution withPSOLVE,EIGQRDA and expand the modes withPSOLVE,EIGEXP. A partial solution is requiredbecause a regular solution will automatically re-create the matrices, but in this situation, the matrices (including the unsymmetric [K])calculated from step 1) need to be used. Thesteps outlined here will allow you to haveANSYS calculate the unsymmetric [K] termsdue to frictional sliding, then include the unsymmetric matrices in the eigensolution.

The eigenvalues (i.e., frequencies) will have real and imaginary parts if damping [C] and/or anunsymmetric [K] matrix are present. The imaginarycomponent reflects the damped frequency. The realcomponent indicates whether or not the mode is stable — unstable modes will have a large, positive real eigenvalue.

The eigenvector also will be complex in either case.The real and imaginary eigenvectors represent the“motion” of the mode shape — if the imaginary eigen-vector is non-zero, this means that a phase difference ispresent, analogous to harmonic analysis output.

Brake squeal and other complex friction problems are readily solvedusing an unsymmetric stiffness matrix supported by QRDAMP.

By ANSYS, Inc. Technical Support

Modal Analyses of Models with Friction

36


Example Problem

As an example for comparing the MATRIX27 andCONTA17x methods, a simple disc model shown inthe accompanying diagram was analyzed. The internalradius is constrained in all DOF, and two contactingareas are present, modeled with surface-to-surfacecontact elements TARGE169 and CONTA171.

A nonlinear static analysis is run first with theunsymmetric option and pre-stress effects on. Angularvelocity and acceleration are used to apply a normaland tangential force component on the contact points.Examination of the contact output shows that thesetwo areas are in contact and sliding. A modal analysisis then performed, using the unsymmetric matricescalculated in the static phase.

With the static coefficient of friction as 0.5 for this model, one finds that the unsymmetric [K] term couples the ninth and tenth modes into an unstable mode.

Disc model was analyzed for comparing the MATRIX27 and CONTA178 methods.

The effect of the coefficient of friction can bestudied by running multiple iterations and varying �in a probabilistic analysis initiated using the coefficient of friction as a uniformly distributed random variable, while the complex eigenvalues were the response variables.

Imaginary vs. real eigenvalues for modes 9 and10 are shown on the left plots, where the coupling ofthe modes and the unstable frequencies are apparent.If the points lie on the axis, these indicate stablemodes with nodes in phase with each other (such as a regular undamped case). Points lying in the nega-tive half of the real eigenvalue axis indicate stable modes (another example of negative real is due todamping). However, points on the positive x-axis showunstable modes.

Mode 9 shows stable response (no damping ispresent in this example) between 453 and 456 Hz.Above that, the modes become unstable. Conversely,for mode 10 the response is stable from about 456.5to 459 Hz. The coupling and instability below thisrange are apparent from the graph.

Real eigenvalues vs. coefficient of friction (variedbetween 0 and 1) are plotted in the right graphs. It isinteresting to note how the stability (real eigenvalue) isaffected by �, as shown on the left near a value of0.45, where both modes 9 and 10 converge andbecome unstable. ■

Contact ANSYS Technical Support engineer Sheldon Imaoka

([email protected]) for the entire paper on the

QRDAMP eigenvalue extraction method from which this

article was excerpted.

37

Left plots show imaginary vs. real eigenvalues for modes 9 and 10. Real eigenvalues vs. coefficientof friction are plotted in the right graphs.



AMD Opteron™ processors provide an easy upgraderoute to 64-bit computing. Memory bottlenecks thatlimited performance in legacy x86 machines are elimi-nated. Low system cost and power requirements nowbring 64-bit processing into the mainstream. In tandem,ANSYS CFX software now exploits the benefits of inex-pensive and powerful hardware, combining increasedsimulation performance and modest compute platformownership costs.

In this article, we compare performance of ANSYSCFX-5.7 and ANSYS CFX 10.0 on Intel Xeon and AMDOpteron processors. CFX now exploits the 64-bit fea-tures of the Opteron and provides excellent scaling onmulti-processor compute nodes and clusters. TheOpteron processor’s Direct Connect Architecture yieldssuper linear speedups in marked contrast to the limita-tions of legacy shared bus SMP nodes. Results areshown for the latest Opteron “Dual-Core” processorsthat fit in the same sockets but provide two completeprocessors on a single die. Results are also providedshowing performance gains from upgrading a two-socket machine to dual core processors.

64-bit Computing with x86 Binary Compatibility

The AMD Opteron processor provides 64-bit exten-sions to the x86 architecture. It can run traditional x86binary images and also codes recompiled to exploit 64-bit extensions, and it allows the ability to address largeamounts of memory. This overcomes the limitations of32-bit addressing, which is restricted to less than 4Gbytes of memory. Both 32- and 64-bit images can runconcurrently on a single Opteron processor. In 64-bitmode, Opteron processors provide more registers toboost performance. ANSYS CFX has a 64-bit versionthat exploits these new features to boost performance.

How Tests Were Run

AMD Opteron processor model numbering uses threedigits: XYY, where X is the number of processors that asingle system can scale up to with that particularOpteron part based on the number of HyperTransportlinks; and YY is the performance metric of the part.Hence, an Opteron processor Model 250 is a 2.4 GHzpart that can scale to two-way. An Opteron processor

Model 248 is a 2.2 GHz part. An AMD Opteron 842 is a 1.6 GHz part, which could be used in up to an eight-way single system configuration. Today, AMDOpteron 800 series parts are generally used in four-waysystems, but eight-way systems should be available inthe near future.

The test machine used in this analysis consisted ofa two-way 1U AMD Opteron Model 250 (2.4 GHz)processor-based server with 8 GB of 400 MHz DDRmemory. The operating system used was Novell’s SLES9 (x86-64 with 2.6 Linux kernel).

The second test machine configuration was a four-way 1U AMD Opteron Model 248 (2.2 GHz) processor-based server with 32GB of 333 MHz DDR memory. Theoperating system was Red Hat AS 3.0 x86-64. The“Smith” cluster has 32 of these four-way servers with atotal of 128 AMD Opteron processors and one Tbyte ofmemory connected with gigabit Ethernet and Myrinetinterconnects.

Each AMD Opteron processor has an on-chipmemory controller and is directly connected to memory.In a multi-processor system design, each processorhas direct attached memory that can be accessed byany of the processors in the node through the inte-grated HyperTransport™ technology, which is part of the AMD64 architecture design.

Table 1 describes the size of each test case. Singlenode results refer to the three different problem sizes:small (100k), medium (200k) and large (400k). Largerproblems (LeMans Car and OHS RC Aircraft) were runfor cluster scaling tests.

AMD brings 64-bit computing into the mainstream of engineering simulation.

By Dan Williams, ANSYS, Inc. David Cownie and Dmitriy Safro, Advanced Micro Devices, Inc.

Table 1: Benchmark Test Cases

38

Test Case Number of Number of Number of Grid Number ofName Nodes Elements Interfaces Domains

Static Mixer 92,993 517,145 0 1100k (small)

Static Mixer 184,450 1,037,239 0 1200k (medium)

Static Mixer 370,162 2,091,423 0 1400k (large)

LeMans Car 1,861,696 10,285,401 0 1

OHS RC Aircraft 2,954,574 14,369,304 3 2

Hardware Update

64-bitChart 3: Comparing Opteron 250 (2.4 GHz) and Opteron875 (2.2 GHz Dual-Core) Performance

Single Processor Performance

Performance is normalized to give speed relative to theP4 Xeon (3.0 GHz) for both ANSYS CFX-5.7 andANSYS CFX 10.0. CFX 10.0 runs on the AMD Opteronprocessor use the 64-bit executable. All CFX-5.7 runsuse the 32-bit executable. In serial, the AMD Opteron250 CPU/memory combo is twice the speed of a 3.0GHz P4 Xeon, and 33 percent faster than a 3.4 GHz P4Xeon, as shown in Chart 1.

More Registers Make a Difference

With AMD Opteron processors, more registers are avail-able in the 64-bit mode. The extra register typicallygives a speed improvement of about 15 percent. Theextra registers more than compensate for longer pointerlengths and increased memory bandwidth needed tohandle these. Results are shown in Chart 2.

Dual-Core Processor Performance

AMD Opteron 250 (2.4 GHz) processors used in the previous tests are fabricated using a 130-nmprocess. AMD recently introduced “Dual-Core” Opteronprocessors fabricated on the new 90-nm process thatallows two entire processors to be placed on a single-die. They are plug-compatible with existing single- core 90-nm AMD Opteron products. Test results inChart 3 show the performance gains possible when theoriginal two-socket machine is upgraded to the newdual-core processors so that the machine then has four processors.

Single Node Scaling

Chart 3 shows the results from several runs on both single- and dual-core, two-socket AMD Opteronprocessor-based systems. In all cases, scaling is excellent, especially if the individual ANSYS CFXprocesses fit inside one CPU/memory package. Theresults clearly indicate one benefit of the on-chip mem-ory controller and dual processor NUMA architecture.Superlinear scaling is achieved in both the single-coreand dual-core two-way benchmarks. There are twopossible causes of superlinear scaling: poor schedulingon the single-processor case between processorsreduces performance, while two-processor cases benefit from explicit decomposition of the problem,which improves the overall cache and local memory hitrates. The net effect is a significant increase in theeffective memory bandwidth and a decrease in thelatency.

When individual processes do not fit in a CPU/memory package, say for the four-way dual-core run(speedup of 3.6), respectable results are still achieved.Non-uniformity in the results is due to individualprocessor memory being scheduled by the operatingsystem such that it is moved away from local memory.In this case, memory access must be done through theHyperTransport link, which has more latency thanaccessing memory on the local CPU/memory package.

Chart 1: Medium Problem Performance

Chart 2: Comparing 32 and 64-bit Performanceon Opteron 250 (2.4 GHz)


39


Hardware Update

To ensure the best possible performance on anAMD Opteron system, you should run a Linux operatingsystem with kernel support for the Linux command taskset. This kernel feature can control locking a process toa CPU (or set of CPUs), which allows optimal processscheduling and memory allocation on multi-processorcompute nodes. For example, SuSE Linux Enterpriseserver release (SLES 9) uses version 2.6 of the Linux kernel, which has support for the Linux commandtask set.

There is scope to optimize the communicationpackage to use these OS hooks to tie a process to apreset CPU/Memory combo. It is possible to use one of several communication APIs with ANSYS CFXincluding PVM, MPICH-1.2.5 and Hewlett-Packard MPI, so to take advantage of such operating system-dependent controls will ultimately require support bythe API authors.

With the latest Linux kernels, scaling can be highlysuperlinear when processes get allocated within oneCPU/Memory package per process. However, evenwhen running Linux versions without the latest Linuxkernel, scaling is still very good: much better than anymulti-processor P4 Xeon offering tested so far by AMD.

Cluster Scaling

Clusters can be configured either with fast, low-latencyinterconnects, or nodes can simply be connected viaGigabit Ethernet and a commodity Ethernet switch.Fast interconnect fabrics (like Myrinet or InfiniBand) offer better scaling but are generally more expensive. TheAMD developer center “Smith” cluster has 32 computenodes, each with 32 Gbytes of memory and four AMDOpteron Model 848 (2.2 GHz) processors. Each node isconnected with both Gigabit Ethernet and a singleMyrinet card. Test results are shown in Chart 4.

Tests with these clusters indicated that Myrinetscales only a bit better than PVM (message passingover gigE) in the equation assembly and linear solveportion of the solution. This is because of low commu-nication overhead in this part of the solution. However,Myrinet speed is far better for communication-intensiveparts of the solution. In ANSYS CFX, this shows up especially during file I/O. Startup/Shutdown times aretwice as fast using the Myrinet interface than one Gigabit Ethernet with PVM.

A cluster configured with a parallel file system (likeLustre or Panasas) was not tested. But in I/O-intensivejobs, these could show a significant speedup over simply cross-mounting a single disc via NFS to everynode. Also, note that Serial ATA disc arrays now are relatively inexpensive, offering speeds closer to SCSIdiscs at prices comparable to IDE drives.

Conclusions

On serial code tests, the two-way AMD Opteron Model250 (2.4 GHz) processors with 400 MHz memory outperform the P4 Xeon (3.0 GHz) by 100 percent.Scaling is excellent with superlinear speedups meas-ured when running parallel code within a computenode. Good Linux distributions are available from bothSuSE and Red Hat that support x86-64.

AMD Opteron dual-core processors can upgradea traditional two-processor machine to four proces-sors. On the single-node test cases, ANSYS CFX runson the dual-core two-socket machine almost as fast asa four-socket four-processor machine. ■

...scaling can be highly

superlinear when

processes get allocated

within one CPU/Memory

package per process.

Chart 4: ANSYS CFX 10.0 (64-bit) Performance on AMD Opteron Model 848 (2.2 GHz)

40


Documents

Industry Spotlight: Microsystems and Nanotechnology Spotlight: Cylinder example demonstrates ... ANSYS Customization with APDL. ... welding to ensure it meets the code of the particular