ADVANTAGE - Ansys · ANSYS Advantage • Volume I, Issue 3, 2007 ... ANSYS, ANSYS Workbench, CFX,...

GREEN POWERPAGE s6

RELIABLE WHEELSPAGE 14

OLYMPIC PERFORMANCEPAGE 20

HEATING THINGS UP IN THE GLASS INDUSTRY

PAGE 8

A D V A N T A G EE X C E L L E N C E I N E N G I N E E R I N G S I M U L A T I O N

V O L U M E I I S S U E 3 2 0 0 7

www.ansys.comANSYS Advantage • Volume I, Issue 3, 20072

ANSYS Advantage • Volume I, Issue 3, 2007

EDITORIAL

www.ansys.com 1


old data files and analysis models from past simulationprojects is difficult and often impossible, even for the people who created them.

Manufacturers are implementing product lifecyclemanagement (PLM) solutions in record numbers to dealwith issues such as this for engineering data and docu-ments across the enterprise. According to statistics fromCIMdata Inc., the PLM market grew 10.4 percent in 2006 toreach $20.1 billion; it is forecast to increase at an 8.5 per-cent compound annual growth rate — exceeding anestimated $30 billion by 2011. Handling computer-aideddesign (CAD) files, part lists, technical documents andchange orders is a lot easier than managing simulationdata, however, which is, inherently, a much more demandingtask for PLM because of the huge file sizes and the complexity of capturing the context of the simulation.

A major step forward in closing this gap is the develop-ment of the ANSYS Engineering Knowledge Manager(EKM) tool. Scheduled for release this year, the Web-basedtool will be targeted at managing simulation processes anddata along with capabilities for backup and archival, trace-ability, process automation, collaboration, capture ofengineering expertise and intellectual property protection.By managing simulation data and processes within such aframework, companies can more effectively leverage thefull power of this critical information and the tremendousexpertise of the analysts and engineers who created it. ■

John Krouse, Editorial Director

Where’s the Data?

One unmistakable global trendis the oncoming wave of digitaliza-tion of the product developmentprocess. Manufacturers are makinggreater use of upfront analysis,collaborative tools, digital mock-ups, assembly modeling andcomplex system simulation. Onerecent study, “The Digital ProductDevelopment Benchmark Report”from the Aberdeen Group, quantifies compelling reasons why

companies are moving to a paperless process. Specifically,evaluating and refining designs early in developmentenables top companies to eliminate an entire prototypecycle, in some cases getting products to market a full threemonths faster and saving up to $1.2 million in developmentcosts, depending on product complexity.

The study also notes major challenges to digital productdevelopment. Topping the list is accessibility to digitalproduct information: that is, how the right people can get tothe right data at the right time. And therein lies a big problemthat the engineering simulation community has faced sincethe early days of the industry: Analysis files — includingmodels, results data and the processes that go into the simulation — are not well managed. More often than not,keeping track of this information is left to the individual whogenerated it, so typically it is buried in obscurity somewhereon a hard drive — or possibly deleted — at the end of a project. Also, this valuable intellectual property may be lostforever when individuals leave the company. Tracking down

Companies embracing digital product development must implementtools for better managing simulation information.

For ANSYS, Inc. sales information, call 1.866.267.9724, or visit www.ansys.com.To subscribe to ANSYS Advantage, go to www.ansys.com/subscribe.

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

Editorial DirectorJohn Krouse

Production ManagerChris Reeves

Art DirectorSusan Wheeler

Ad Sales ManagerBeth Bellon

EditorsMarty MundyFran HenslerErik FergusonChris HardeeDave SchowalterTim Roolf

Production AssistantJoan Johnson

Editorial AdvisorKelly Wall

Circulation ManagersElaine TraversSharon Everts

DesignersMiller Creative Group

Neither ANSYS, Inc. nor the editorial director nor Miller Creative Group guarantees or warrants accuracy or completeness of the material contained in this publication.ANSYS, ANSYS Workbench, CFX, AUTODYN, FLUENT, DesignModeler, ANSYS Mechanical, DesignSpace, ANSYS Structural, TGrid, GAMBIT and any and all ANSYS, Inc.brand, product, service, and feature names, logos and slogans are registered trademarks or trademarks of ANSYS, Inc. or its subsidiaries located in the United States or other countries. ICEM CFD is a trademark licensed by ANSYS, Inc. All other brand, product, service and feature names or trademarks are the property of their respective owners.

© 2007 ANSYS, Inc. All rights reserved.

About the CoverGlass has fascinating and uniqueproperties, and producing it canbe a complex undertaking. Readabout how PFG Building Glass inSouth Africa is changing glass-making from an art to a science,on page 8.

Email: [email protected]


CONTENTS

Table of ContentsARTICLES

4 TURBOMACHINERY

Streamlined Flutter AnalysisIntegrated fluid structure interaction enables high-fidelity turbomachinery blade flutter analysis.

6 TRANSPORTATION/HVAC

Ventilating Giant Railway TunnelsHigh-speed trains in Spain cross more than just the plain.

8 GLASS

Glass-Making Goes from Art to ScienceModeling glass furnaces helps improve batch transition time and reduce product defects.

10 THOUGHT LEADERS

Getting It Right the First TimeIn a corporate-wide initiative, Cummins Inc. refines designs early with Analysis Led Design to shorten development time, reduce costs and improve product performance.

13 BIOMEDICAL

Special DeliveryResearchers use simulation and medical imaging to explore new options for managing pain.

14 TRANSPORTATION

More Certainty by Using UncertaintiesEngineers apply probabilistic methods to historically deterministic problems.

16 GOVERNMENT AND DEFENSE

Out of Harm’s WayEngineers used simulation to design an innovative military gun turret.

18 MARINE

Designing Out the Weakest LinkEngineering simulation validates the design of a mooring system component,a critical wheel/chain assembly that holds ships in place during oil and gasoperations in the North Sea.

20 SPORTS

Going for the GoldSimulation helps design low-drag canoes for Olympic-medal performance.

20

16

14

6

4


CONTENTS

www.ansys.com 3

Spotlight on Engineering Simulation for

Environmental Design

DEPARTMENTS22 PARTNERS

Something in the MixResearchers use the Poincaré plane method to obtain quantitative time scale information from CFD simulations.

24 Cluster Computing with Windows CCSNew clustering technology from Microsoft speeds up engineering simulation.

26 TIPS & TRICKS

Component Mode Synthesis in ANSYS Workbench SimulationCMS superelements provide flexibility of simulation models while reducing the number of degrees of freedom for highly efficient solutions.

28 ANALYSIS TOOLS

Accelerating to ConvergenceANSYS VT Accelerator technology can help solve nonlinear transient and static analyses faster.

30 Seeing is BelievingDevelopments in version 11.0 software from ANSYS allow inclusion of solidparts during pre- and post-processing, making for more intuitive problem setupand results visualization.

32 Predicting Liquid AtomizationSimulation can be used to produce sprays with desired characteristics using the FLUENT VOF model.

24

26

s1 It’s Getting Easier to Be GreenFrom air to water to power, industries are using engineering simulation touncover new ways to be environmentally responsible.

s3 In the WorksUsing simulation to model wastewater treatment plants effectively.

s6 Cooling Down Powered-Up Fuel CellsResearchers use probabilistic methods and design optimization to improveheat-transfer characteristics of fuel cell stacks.

s8 Making Electricity through Chemistry Analysis helps power fuel cell design.

s10 The Future of FuelA European research project is developing internal combustion engines powered by hydrogen.

32

s1

s10

www.ansys.com4 ANSYS Advantage • Volume I, Issue 3, 2007

TURBOMACHINERY

Streamlined Flutter AnalysisIntegrated fluid structure interaction enables high-fidelityturbomachinery blade flutter analysis.

The “flutter” of blades within compressors and turbinesis a serious cause of machine failure that is difficult to predict and expensive to correct. This aeromechanical phe-nomenon usually occurs at a blade natural frequency andinvolves sustained blade vibration resulting from the changingpressure field around the blade as it oscillates. For theprocess to occur, it is necessary that, over one cycle, thereis an input of energy from the gas stream to the blade of asufficient magnitude to overcome the mechanical damping.

Clearly, flutter is dependent on both the aerodynamicand structural characteristics of the blade, and, until recently, it has been beyond the design capability to satisfactorily investigate and avoid this phenomenon.Historically, empirical design criteria have been used basedon parameters involving blade natural frequencies and flowtransit times, but these methods fail to take into accountgenerally found vibrational modes or the influence of adjacent blades.

Improvements in unsteady computational fluid dynamics(CFD) capability combined with the ability to easily andaccurately transfer information between CFD and finite element analysis (FEA) has enabled the development of anadvanced yet efficient and cost-effective methodology foranalyzing forced vibration processes.

A key enabling development now provided by ANSYS,Inc. is the ability to deform the CFD computational grid inresponse to deformations at the fluid structure interface and integrate this with unsteady flow computations. Theprocess is straightforward to set up and is facilitated by theintuitive and intrinsic functionality of the user interface andlayout in the ANSYS Workbench platform. PCA EngineersLimited, based in the U.K., has utilized this capability bymapping time-dependent deformations computed from afinite element analysis to the CFD computational grid.

As a rule, blade flutter occurs at a blade natural frequency that is determined together with its correspondingmode of vibration using traditional finite element techniques.

A bladed disc assembly can be classified as a rotation-ally periodic structure, and, therefore, the mode shape ofadjacent blades within a row are fully defined by a phase

By Robin Elder and Ian Woods, PCA Engineers Limited, Lincoln, U.K.

Simon Mathias, ANSYS, Inc.

ANSYS Mechanical analysis tools can predict vibration modes that occur over an entirewheel from a single blade component model. Shown here are exaggerated deformationsfor a four-nodal diameter mode shape, meaning that the mode repeats itself four timesover the entire wheel circumference. Engineers are interested in determining whethervibration modes such as these will be amplified by interaction with the fluid or safelydamped out.

www.ansys.com 5ANSYS Advantage • Volume I, Issue 3, 2007

TURBOMACHINERY

Deformations of a four-nodal diameter which repeat once over each quarter ofthe wheel, were exported from the modal analysis vibration mode to ANSYS CFXsoftware as a boundary profile. The mode shape is used to create a periodicboundary motion in the CFD software and to evaluate the net work input due tothe blade motion.

Damping coefficients can be calculated from the CFD results. Negative net workinput due to blade motion results in a positive damping coefficient. Negative damping coefficients induce sustained blade vibration, or flutter, which couldlead to blade failure. These results show positive damping for all inter-bladephase angles.

difference. This phase difference (the inter-blade phaseangle or IBPA) depends on the number of blades in the rowand the number of patterns repeating around the annulus.This latter parameter is often called the nodal diameter (ND)and can move either in the direction of rotation or againstthe direction of rotation.

The significant development is that this modal displace-ment information now can be applied to the computationalgrid and the resulting time varying flow through a blade rowas well as the dynamic pressure field over each definedblade calculated using ANSYS CFX software. The computeddynamic pressure distribution and the corresponding modaldisplacements then are used to compute the work done onthe blade over one complete cycle. If the net work done onthe blade is positive, then work is being imparted to theblade, creating negative damping, a potentially unstable situation leading to a self-sustained vibration (flutter) likely tocause a material fatigue failure. On the other hand, if theaerodynamic work done on the blade is negative, the blademotion is doing work on the fluid and leads to a stable ordamped vibration.

In the aerodynamic damping case illustrated, the bladeis stable (no flutter) because the damping is always positive.This information is critical to the designer as blades are relatively easy to modify before manufacture but extremelycostly to rectify in an operational plant. By utilizing bladeflutter prediction early in the design cycle, costly damageand repairs can be avoided. This integrated design andanalysis approach in multiphysics technology from ANSYScan lead to improved quality and dependability of thedesign process, realizing further cost benefits to clients.

ANSYS, Inc. and PCA Engineers now are applying suchtechnology to a wide range of applications extending fromlarge steam turbines to small turbochargers. These tech-niques are assisting engineers to design compressor andturbine blading in which both aerodynamic efficiency andstructural integrity are paramount over the operational rangeof the machine. ■

www.pcaeng.co.uk

Finite element (FE) mesh at the fluid–structure interface A typical torsional blade mode, where the relativeamplitude of each node point on the gas swept surface of the blade is known as a function of time

Equivalent stresses

DampingCoef(log)

-40 -30 -20 -10 0 10 20 30 40 50 60 70 80

Inter-Blade Phase Angle (deg)

0.2

0.16

0.12

0.08

0.04

0

Pressure contours on a train with vortices shown by streamlines in a tunnel


TRANSPORTATION/HVAC

High-speed trains in Spain cross more than just the plain.

By José Carlos Arroyo and Pedro Luis Ruiz, INECO-TIFSA, Madrid, Spain

Yannick Ducret and Roberto Garcia, ANSYS, Inc.

Sunny beaches filled with sunbathers may be the firstthing that comes to mind when imagining Spain. However,the reality of its geography is a lot more variable. In fact, theIberian Peninsula sports many mountain ranges that largelyhinder the development of complicated infrastructure, suchas the high-speed rail network that is planned to connectSpanish urban areas. This project has led the Spanish railway industry to bore some of the world’s longest high-speed transit railway tunnels, such as the 28-km Guadarrama tunnel and the 24.5-km Pajares tunnel.INECO-TIFSA, a transport and telecommunication company located in Madrid, Spain, has contributed to theongoing expansion of high-speed railways throughout thecountry by participating in the design of superstructures,such as these giant tunnels.

When designing a railroad tunnel, the ventilation systemis a critical component. The ventilation units themselvesconsist of longitudinal jet fans that are placed at several positions along the tunnel. Their performance is affectedseverely by air disturbances that result from the motion of atrain traveling in the tunnel at speeds of up to 350 km/h (218 mph). Not only does the train movement quickly forcethe air toward both tunnel exits, it also creates a complex system of pressure waves that propagate throughout thespace. A positive-amplitude pressure wave is created whenthe train enters the tunnel. When the train’s rear end passes into the tunnel, another wave, of negative amplitude,originates at the tunnel entrance. Both waves propagatetoward the tunnel exit where they are primarily reflected.

Traditionally, 1-D or 2-D simulations have been satisfac-tory to predict the average pressure correctly inside the

Ventilating GiantRailway TunnelsImage courtesy Eurorail Group.

ANSYS Advantage • Volume I, Issue 3, 2007www.ansys.com 7

tunnel. Only by means of a complete 3-D simulation,though, is it possible to obtain an accurate estimate of thevelocity components, magnitude and their distribution in thetunnel sections in order to allow for accurate fan sizing.

To design the ventilation system for a long (>5 km) tunnel, INECO-TIFSA chose FLUENT software to perform afull 3-D unsteady simulation of a train passing through sucha tunnel. The movement of the train was simulated using theFLUENT sliding mesh capability, in which the train and thedomain that it encompasses slide along a non-conformalinterface. The interface was placed at the tunnel wall, andthe mesh was extruded accordingly. Due to the speed of the train, the ideal gas model was used to account for the effects of compressibility. The computations were performed using the pressure-based solver, which was chosen because the flow is only slightly compressible —that is, there is only a weak coupling between density andvelocity, and, thus, the computation does not require thedensity-based solver. This unsteady simulation was performed using non-iterative time-advancement (NITA) in order to reduce the computational time required. Thisapproach was validated by a series of 2-D computations.Special consideration was given to the determination of theappropriate time step, since it needed to be small enough topredict the wave’s propagation correctly.

The velocity components and the static pressure weremonitored in seven different locations along the tunnellength corresponding to the ventilators’ positions. The flowpatterns also were analyzed using velocity contours in various sections of the tunnel. The results showed theamplitude of the flow created by the train’s passage. When

a train enters a tunnel, air first escapes at the tunnelentrance at the side of the train, both because it is the closest exit and because the mass of air between the frontof the train and the tunnel exit has yet to be put in motion.When the train has passed, the flow then changes direction.At that moment the air is pushed by the train and travelsbackward in the narrow gap between the train and the tunnel. Speeds of up to 35 m/s were observed at the fanpositions. Furthermore, some sudden changes of slightlyhigher amplitude could be seen when the front of the trainreached the jet fans, showing how carefully this equipmentneeds to be selected.

As expected, the pressure waves created by the train’smotion do have a discernible effect on flow patterns withinthe tunnel. Seconds after the train has passed the jet fans atthe entrance of the tunnel, the wave patterns form such thatthey accelerate the air flow by up to 25 m/s. When the air iscompressed by a positive-amplitude wave, the air velocitydiminishes according to conservation of mass, while theinverse (acceleration) occurs if the wave is of negativeamplitude. The train creates both of these types of waves asit passes through the tunnel, thus inducing both accelera-tions and decelerations in the surrounding tunnel airflow.This complex and decaying phenomenon then continueslong after the train has exited the tunnel. Even though thehighest velocities observed are longitudinally oriented, thetransversal velocity profiles revealed the benefits of a 3-Dstudy, since velocities of the same order of magnitude wereobserved. Overall, this modeling approach has shown interesting results and proven beneficial for INECO-TIFSAby the level of detail achieved. ■

Computational fluid dynamics (CFD) contours at three locations within a tunnel show how the longitudinal velocity changes in the tunnelas a train passes through it. The plane cuts represent the position of jet fans where fluctuating velocities have been monitored.

TRANSPORTATION/HVAC

Glass-Making Goes from Art to ScienceModeling glass furnaces helps improve batchtransition time and reduce product defects.By Eddie W. Ferreira, PFG Building Glass, Springs, South Africa

To create glass from its raw materials is to invest in both the art andthe science of the process. Glass is a fascinating engineering material withunique properties; however, producingit can be a complex undertaking and isoften thought of as an art. As a result, commercial glassmakers strive contin-ually to understand the science of itsmanufacture in order to optimize andimprove the process. As one suchmanufacturer, PFG Building Glass inSouth Africa is using FLUENT compu-tational fluid dynamics (CFD) softwareto model the flow inside its glass furnaces, track processing defects andimprove overall production systems.

At the basic level, glass-makingconsists of three steps. The first ismelting a blend of raw materials, whichcan include sand, limestone, soda ash,feldspar and saltcake. The next is refining, in which bubbles containedwithin the molten raw materials areremoved. Finally, there is conditioning,

in which the glass is cooled to a suitableworking temperature. There are variousmethods of accomplishing each stepthat affect the process differently. Differ-ent glass compositions require differentoperating envelopes, due to the changein physical and chemical properties.

Because glass-making requires furnace temperatures of 1500 degrees C(about 2700 degrees F), heat transfer andchemical diffusion dominate the processkinetics, and the reaction tank itself isslowly dissolved by the molten glass.These factors make experimental studiesdifficult. As an alternative, simulation arises as a good way to understand how furnaces behave and how processimprovements can be made.

Using the 3-D version of FLUENTsoftware and the pressure-based solver,PFG Building Glass developed a CFDsolution for steady-state glass process-ing conditions. The company created asimplified initial simulation, one that didnot include any time-dependent events.

Model of a float glass furnace. The most common method for glass production is floating molten glass on top ofmolten tin, thus giving it the name “float glass.” This process results in the formation of plates or ribbons of glass.

Melter

Refiner

Waist

Conditioner

Contours of batch species fraction in the glass domain of a container furnace. The red area represents the introduction of a new species into the glass flow.

Once these simulation results wereacquired, time-dependent events, suchas color transitions, were incorporatedinto the simulation and accounted for byswitching to the transient solver in theFLUENT product. To simulate this morecomplicated type of process, PFGenabled the species transport model inFLUENT software and incorporated itsown batch models via user-definedfunctions (UDFs) for the species proper-ties, boundary conditions and sources.These additions allowed the team toobserve factors such as mixing.

Apart from the glass flow itself, whathappens in the combustion spaceabove the processing glass is veryimportant. Combustion that occurs inthis region of the furnace is a heatsource for melting and heating the glassmixture. In order to include this region in the analysis, PFG incorporated combustion, radiation and turbulencemodeling into the simulation. By includ-ing these factors, the model complexitywas greatly increased.

The combustion space model and the glass flow model then were


GLASS

factors driven by this transition time,glass that does not fall within approvedspecifications is produced with anassociated loss of revenue. Any reduc-tion in transition time is, therefore, ofgreat value. PFG has been able to partially model this transition processusing FLUENT software, leading tomodifications in operating procedures.

The use of CFD modeling has ledto a better understanding of glassflows and combustion conditionsinside glass furnaces. This has allowedPFG Building Glass to achieve itsobjective of producing high-qualityglass at the lowest possible price,

Contours of temperature in the combustion space over the melter region of an oil-fired, six-port float glass furnace. Red areas identify regionsof higher temperature.

Contours of velocity magnitude through the center of a float furnace. The simulation includes both the combustionspace above the glass melt and the melted glass itself.

while also maintaining long furnace life,all without a hit-and-miss approach.Product quality has improved as aresult of defect tracking, and losseshave been reduced as the process hasbecome more of a science than an art. This experience and the models drawn up allowed PFG to sim-ulate planned expansions before theywere installed and, thereby, eliminate problem areas before installation. ■

The authors would like to acknowledge PeetDrotskie and Corne Kritzinger from PFG, wholaid the groundwork for these modeling efforts,as well as Danie de Kock and his support teamat Qfinsoft for their invaluable input.

combined into a coupled simulation,making use of the FLUENT non-premixed combustion model, discreteordinates (DO) radiation model andrealizable k-ε turbulence model. Further UDFs were used to define thematerial properties and source terms.For boundary conditions, it was impor-tant to maintain the glass zone as alaminar zone and the combustion zoneas a reaction zone.

The quality of the final glass product is influenced by the presenceof small bubbles, which manufacturerstry to remove from the batch during arefining phase because bubbles canlead to discrete faults in the final product. There are numerous sourcesthat can lead to an unacceptable rise inthe number of faults. Using simulationfor defect tracking has helped PFG topinpoint the areas that are most proba-bly the origin for the faults. PFGaccomplished this with the reverseparticle tracking capability in the FLUENT product. By defining trackinglocations throughout the glass fluiddomain and using the FLUENT dis-crete phase model (DPM), PFG wasable to examine a particular particle’sflow path history and determine statis-tically probable fault positions in thefinal glass ribbon.

One other complication of a batchprocess is the transition from onebatch to another, which involves moving the complete furnace glassvolume. Usually the transition processtakes a number of days. As a result of


GLASS

www.ansys.com 9


THOUGHT LEADERS

In developing complex mechanical products such asdiesel engines, going through multiple build-and-test hard-ware prototype cycles to verify performance, stress andfatigue life is tremendously expensive and time-consuming.This issue can be addressed by evaluating and refiningdesigns with analysis tools up front in development, sofewer test cycles will be needed later in development.

Five years ago, such a Simulation Driven Product Development approach was started at Cummins Inc., a corporation of complementary business units that design,manufacture, distribute and service engines and related

technologies, including fuel systems, controls, air handling, filtration, emissionsolutions and electrical power generation systems. Applications include trucks,construction and mining equipment, agricultural machinery, electrical generators,fire trucks, recreational vehicles, buses, cars, SUVs and pickup trucks. The Cummins Analysis Led Design (ALD) strategy is a corporate-wide initiative tochange the prevalent test-first culture; it has had a major impact at the company,with significant benefits that include shorter development time, lower costs andimproved products.

ALD can shorten product development time by getting designs right the firsttime. Many Cummins-designed parts have extensive lead times because toolingneeds to be created. Beyond this, traditional hardware testing can take weeks oreven months to validate a design. Leveraging analysis early in the process caneliminate tooling changes and repetition of lengthy endurance testing, thus providing significant reductions in overall development time.

Getting It Rightthe First TimeIn a corporate-wide initiative, Cummins Inc. refines designsearly with Analysis Led Design to shorten development time,reduce costs and improve product performance.

By Bob Tickel, Cummins Inc., Indiana, U.S.A.

Modal analysis identified deformationof the crankshaft.

The Cummins QSK78 engine delivers more power thanany other engine for gigantic haulers in the mining industry. The 18-cylinder, 12-ton super-engine is rated at 3,500 horsepower and stands almost eight feet high.Cummins also provides engines for agricultural andindustrial equipment and heavy-duty pickup trucks.

Bob TickelDirector of Structural and Dynamic Analysis


THOUGHT LEADERS

www.ansys.com 11

Simulation also radically lowers the total cost of productdevelopment through less dependency on hardware testsand a reduced number of long-hour tests, which sometimescan last for days. At Cummins, some of this traditionalendurance testing can cost in the range of $50k to $100kper test, so eliminating even a single cycle can result in sig-nificant savings. The intention is not to eliminate all testingbut, rather, to use targeted component and assembly-leveltesting first to validate analysis models and then to validatethe overall design with only a few long-hour tests of theentire engine. Savings also are achieved by eliminatingredesigns, in which costs are lowered by reducingresources required to manage the design process (engi-neering, drafting, clerical time, etc.) as well as reducingretooling costs.

While shortening development time and lowering costsare important aspects of ALD, it can be argued that themost significant benefit of the approach is the ability to create improved products by considering a broad range ofdesign alternatives. Simulation allows engineers to readilyperform what-if studies and large-scale design of experi-ments in order to understand more fully the design spaceand trade-offs involved. Otherwise, once the first set ofhardware is created, the design space narrows and designsare much harder to modify.

Various measures have been used within Cummins tohelp determine the effectiveness of ALD. In looking at testtime and cost in one example, validation testing for a cylinder block traditionally required $72k of rig testing and$30k to $80k for engine testing for a single block design.Each repetition costs the same amount: in the range of$100k to $150k. Testing usually took about one month,once hardware was available. Lead time for the tooling andpart procurement took about 12 weeks.

Through the ALD initiative, engine testing has beenremoved as a requirement for some cylinder block

validations. Now when a new heavy-duty engine design isbeing developed, a series of repetitions are done throughsimulation until the entire block meets the design limits. Thisrequires the time of one analyst for about a month of work,or approximately $7k. Once the hardware is procured, rigtesting is completed on the initial pass — a first for this typeof design. The result is that a minimum $30k of engine testing is eliminated. Also, redesigns are eliminated that,most likely, would have occurred over many more weeks ormonths and at an additional cost of $100k ($72k of rig and $30k of engine testing), which does not include the significant additional expense of prototype hardware.

There are several reasons why ALD has been successfulat Cummins: It is a top-down initiative that was driven byupper management, appropriate resources were allocated,and an infrastructure was established to support the initiative.

From the beginning of the program, top managementhas been a strong proponent of ALD. Cummins’ chief technical officer coined the acronym ALD, and he has continued to push the initiative. The progress of ALD has

In the development of Cummins diesel engines, engineers use the ANSYS Mechanical software to determine (left to right) cylinder block deformation and stresses, cylinder headassembly stresses and temperature distribution in the cylinder head and valves.

This customized Kenworth tanker truck has a Cummins 565-horsepower engine.


THOUGHT LEADERS

been monitored continually and reported in quarterly messages by Tim Solos, Cummins CEO. At the executivelevels, there has never been a question about whether toreduce testing and increase analysis but rather how to bestaccomplish this objective with limited resources.

Along with driving ALD, Cummins management provided resources to do more analysis. Shortly after theALD initiative was started, a technical center was set up(Cummins Research and Technology, or CRT, in India). Thisanalysis center focuses solely on design, computationalfluid dynamics (CFD) and structural analysis in supportingall Cummins business units.

Infrastructure to support ALD at Cummins has takentwo forms: Engineering Standard Work (ESW) processes

Thermal analysis shows temperature distribution for a diesel engine piston.

Coarse mesh of detailed geometry for an inline six-cylinder head created using theANSYS Workbench platform

and Six Sigma tools. ESW defines the work, tools and limitsrequired to release a part for production. This became a natural focus for ALD as Cummins examined where testingwas being reduced and where analysis was beingincreased. Six Sigma has been an invaluable support forALD in validating new tools and methods to ensure thatanalysis can be used to replace testing. So, ALD is the initiative, ESW is the process to ensure that all necessarywork is completed and Six Sigma is the set of tools used todetermine that the appropriate work is included.

In performing the underlying work for ALD, the Structural and Dynamic Analysis group within the CumminsCorporate Research and Technology organization is responsible primarily for developing tools and methods as well as conducting analyses to ensure that structural components meet both reliability and durability require-ments. The group partners with key software vendors inefforts to develop improved simulation tools, and one of theprimary relationships is with ANSYS, Inc. In fact, the relationship has been the benchmark set for subsequentpartnerships. Technology from ANSYS has become the primary finite element tool within all Cummins business unitsfor conducting static structural, thermal, transient thermal,modal, harmonic and other analyses. This partnership withANSYS has resulted in joint development of advanced features in continuing to meet analysis needs at Cummins.

Any culture shift is difficult, requiring vision, leadership,planning and tangible benefits. The ALD initiative, in particular, has driven considerable change and has provento be of tremendous value at Cummins. While significantprogress has been made, there is room for expansion, andCummins will continue to evaluate new and improved technologies, processes and strategies in using simulation tofurther strengthen its position in the diesel engine industry. ■

ANSYS Advantage • Volume I, Issue 3, 2007www.ansys.com s1

A D V A N T A G E

s1 It’s Getting Easier to Be Green

s3 In the Works

s6 Cooling Down Powered-Up Fuel Cells

s8 Making Electricity through Chemistry

s10 The Future of Fuel

The term “green engineering” has become ubiquitousin recent years, with references even on the covers of tradejournals and magazines. The U.S. Environmental ProtectionAgency defines green engineering as “the design, commer-cialization and use of processes and products that arefeasible and economical while reducing the generation ofpollution at the source and minimizing the risk to humanhealth and the environment.” So while green engineeringencompasses environmental engineering, it also can referto any engineering field in which environmental and humanhealth impacts are minimized. Increasingly, the term hasbecome associated with sustainable development, inwhich processes and products can continue to be producedindefinitely with a minimum of resource depletion or environmental degradation.

Along with increased awareness of environmentalimpact, well-known corporations have launched campaigns

that show how they are developing green technologies. Of particular note are General Electric’s ecomagination™,the BP™ campaign Beyond Petroleum and Chevron Corporation’s willyoujoinus.com advertising promotion. One thing is clear: Major companies believe there is money to be made in developing environmentally friendly technology, which should encourage even the most contrarian environmentalist.

In building a better world, global companies are learning that the right engineering simulation can improveefficiency in the design of real-world systems. Simulationcapabilities from ANSYS, Inc. are particularly visible in theareas of pollution control, architecture, energy and sustain-able technology. This spotlight on the environmentalindustry provides details about how hard-working users ofengineering solutions from ANSYS are improving the environment. Perhaps readers will find themselves inspired.

It’s Getting Easier to Be GreenFrom air to water to power, industries are using engineering simulationto uncover new ways to be environmentally responsible.By Dave Schowalter, ANSYS, Inc.

Image © iStockphoto/Elena Elisseeva

www.ansys.comANSYS Advantage • Volume I, Issue 3, 2007s2

ENVIRONMENTAL DESIGN

Clean AirAir pollution comes primarily from transportation and

point-source industrial processes. In the transportationarena, there is particular emphasis on particulates andnitrous oxides (NOx), with increasing efforts to reduce car-bon dioxide emissions through efficiency improvements.Reducing any type of pollution can involve heavy simulationusage for flow, chemistry, heat transfer and thermal stressminimization.

Industrial sources are concerned with particulates, NOx,and carbon dioxide, as well as sulfur oxides (SOx) and mercury. The low pollutant levels achieved today throughoptimized furnace combustion and optimized flow distribu-tion in downstream pollutant capture systems would not bepossible without virtual prototyping through computationalfluid dynamics (CFD). Additionally, minimization of materialusage requires an understanding of thermal stress loadsthrough structural analysis.

Clean WaterEngineers are using fluid flow modeling — including

solutions from ANSYS — to optimize both municipal andcommercial purification processes, such as tank mixing,ultraviolet disinfection, chlorination and ozone contactors.Modeling also comes into play in wastewater treatment,which involves similar processes, in addition to phase separation.

Protection of fish is another aspect of clean water, andsimulation has been used to design oxygenation systemsand retrofits in hydropower dams, that are aimed at increas-ing downstream oxygen levels. Modeling of water intakestructures at industrial plants also is contributing to reduc-tion of ecosystem impact.

Run-off and drift from commercial and residential pesticide treatments can affect water as well as air; simula-tion is used to optimize chemical dosing, and to model andunderstand dispersion.

Green BuildingGreen building refers to designing commercial and res-

idential buildings that minimize non-renewable energyusage; use materials whose production has a minimal environmental impact; and use heating, ventilation and airconditioning methods that maximize air quality. Safely minimizing material usage and maximizing passive ventila-tion through natural circulation makes this an active andgrowing area for simulation.

Renewable EnergyOf all renewable energy technologies, wind power has

taken the most advantage of simulation capabilities.Today’s large wind turbines require advanced materials,increased efficiency, reduced weight while avoiding fluidstructure interaction, and the ability to withstand seismicvibrations. Because the power that can be extracted scalesas the cube of the wind velocity, placement decisions canhave a major impact on the profitability of a project. Otherrenewable energy technologies that take advantage ofproducts from ANSYS include tidal power systems, solarpower installations, and biomass power and energy.

Sustainable TechnologyDrastic reductions in energy usage and pollution produc-

tion are possible with new technologies such as fuel cells,advanced nuclear power plants (including nuclear fusionresearch), advanced coal power (including gasification) andhybrid automobiles. For these technologies, simulation is in on the ground floor of development, playing an especiallyactive role in next-generation products.

In order to support the ever increasing rate of technologydevelopment that is required for global environmental sus-tainability, computer aided engineering tools themselvesmust be scalable and sustainable, which is why ANSYSgives the highest priority to developing multidisciplinary,multiphysics tools all within a single accessible environment,deployable on the desktops of engineers in the small venturestart-up as well as on the large parallel servers in engineeringdepartments of major multinational corporations. ■

References

[1] U.S. Environmental Protection Agency,http://www.epa.gov/oppt/greenengineering, 2006.

Particle tracks in a wet SO2 scrubber in which simulation was usedto optimize pollutant capture efficiency. Image courtesy URS Corporation.

Finite element analysis was used in the design of a solar car, which hadsevere weight limitations. Image shows the stress analysis on A-Arm clevis for the car’s suspension. Image courtesy University of Toronto Blue Sky Solar Racing.


In the WorksUsing simulation to model wastewater treatment plants effectively.

By David J. BurtMMI EngineeringBristol, U.K.

In response to various European environmental legisla-tive drivers — which include urban wastewater treatment,fresh-water quality standards for protection of fish andwater framework directives — U.K. water companies haveembarked on a new asset management plan. Part of thisplan requires the treatment of significantly greater amountsof wastewater, either by building new treatment plants or by increasing flows through existing plants or works. At thesame time, many sites face additional tighter constraints foreffluent discharge. The majority of wastewater is treated inmodern, large-capacity activated sludge process (ASP)plants. Water companies have been making increased useof analytical process modeling tools, such as computationalfluid dynamics (CFD), to find capital cost savings, achieveperformance improvements and improve energy savings forthese plants.

A modern wastewater ASP includes several operationalstages that may be modeled with CFD. However, using CFDto investigate these unit operations successfully requires

ENVIRONMENTAL DESIGN: WASTEWATER TREATMENT

some process knowledge. This article illustrates a few of theprocesses and explains how they are best addressed withANSYS CFX software and multiphase modeling techniques.

The basic sequence of operations at a wastewatertreatment site with an ASP plant includes the followingstages:

■ Inlet works with de-gritting and flow balancing ■ Primary settlement■ Activated sludge treatment in aeration lanes ■ Secondary settlement■ Tertiary treatment

Inlet WorksIn most U.K. works, the wastewater enters from an

upstream combined sewer system. This wastewater is amixture of rain water and sewage loaded with solid particlesof irregular size, shape and density. A large inlet worksremoves gross solids and delivers equal flows and loads tothe multiple lanes of an ASP; otherwise, the lanes may

Inlet WorksPrimary

SettlementSecondarySettlement

TertiarySettlement

Discharge

Sludge

Air

Sludge Treatment

DigestionDewatering

Land application

ChemicalAddition

AerationLanes



Storm settling tank influent

A primary settling tank was modeled with multiple drift fluxes. Thisplot shows the stratified distribution of solids through a typical 30-mdiameter tank.

become overloaded or underloaded, and, subsequently,they will not work as well. CFD modeling of the inlet workscan be used to determine the equality (or inequality) of theflow distribution among the lanes, as well as the trajectoryand final resting place of solids that move independently ofthe bulk fluid. For example, a discrete particle trackingmodel may be used to determine the solids retention efficiency of grit traps and balancing tanks, whereas a continuum multiphase model may better show how solidsmove independently of the water down the different lanes ofa distribution chamber.

Primary SettlementAfter removal of the larger solids in the inlet works, the

wastewater passes into a primary separation zone. The pri-mary tanks are often circular with a central influent, or riserpipe, at the center of the tank. Separation of solids occursby settlement. The ability to retain solids depends on thebalance between the radial up-flow velocity in the tank andthe solids’ settling velocity. In order to model settlement in a primary tank with CFD, a multiple drift flux model is used inwhich the influent solids particle size distribution is definedas a series of size groups (mass fractions). Each size grouphas a drift settling velocity pre-calculated from knowledge

The inlet works for a large ASP illustrates the typical scale.

The interstage chamber was modeled with ANSYS CFX software. Streamsfrom an inlet culvert demonstrate the typical flow patterns at the inlet distribution chamber. The streamlines are colored by time, with blue representing the initial time at 10 seconds.

of the wet solids density. The total solids concentration thusis determined from the sum of the size groups progressingthrough the tank. This multiple drift flux modeling techniquehas been used to determine the optimum number of primary tanks and their required side wall depth for newbuild sites in the U.K., thus minimizing the land use require-ment and reducing overall civil engineering costs.

Activated Sludge Treatment in Aeration LanesAfter primary separation, the wastewater stream passes

into a series of aeration lanes in which bio-chemical reac-tions occur that convert the solid particulate waste intoactivated sludge. The sludge then can agglomerate (or flocculate) into large clusters of particles that can be morereadily separated by sedimentation. The bio-chemical reaction rates depend on the levels of dissolved oxygenpresent within the wastewater. These levels can be modeledwith multiphase CFD. A surface aerator, which draws liquidand solids from the lower region of the tank up through adraft tube and then sprays them back across the surface ofthe tank, influences the solids distributions within the tankand also introduces oxygen into the aeration lane. A studythat varied the length of a draft tube diffuser was performedto investigate how the geometry affected the sludge bed



The surface aerator of this bio-reactor is used to resuspend the solids bedfrom within an aeration lane and to entrain air into the reactor.

Activated sludge is settled out in this clarifier. Flow enters the tank fromthe top and flows radially outward.

Deflection ring Stilling well

Inlet pipe

McKinney baffle

CFD was used to determine the influence of the draft tube depth on sludgebed entrainment for the surface aerator in the bio-reactor modeled here.Iso-surfaces of solids concentration are shown using blue at 3,000 mg/land yellow at 20,000 mg/l. Streamlines identify flow patterns that pass upthrough the draft tube and are projected out, by way of the aerator, across the tank surface.

entrainment. This research found that a longer draft tubeshould be used with the surface aerator under investigation.This change was shown to maximize the aeration and mixing capacity.

Secondary SettlementAfter traveling through the aeration lane, the wastewater

undergoes secondary treatment in a clarifier. The activatedsludge settles out and the effluent passes over a v-notchedside weir. The secondary clarifier may be modeled with anextended drift flux model incorporating both sludge settlement and rheology models defined as functions of localconcentration. The results of simulation provide both thegradient of solids within the tank ranging from less than5mg/l in the surface water to greater than 20,000 mg/l in thecompressive zone near the bottom of the tank and a measureof the likely effluent solids concentration (solids going overan exit weir, typically in the range of 10 to 30 mg/l). Thismethod has been used extensively to prove clarifier performance — as compared with idealized mass flux theory — and to optimize the position of retrofit baffles toallow a higher flow throughput for the same effluent solidsconcentration on existing units. MMI Engineering has usedthese techniques to design optimum clarifier influent

arrangements that increase throughput and maintain theeffluent solids at more than 40 sites.

This article illustrates four examples of applying CFD towastewater systems. Many other unit operations may beexamined with similar models to those described here. Theextension of aeration lane modeling to include microbialpopulation balances and bio-kinetic reactions (the ASM1model) currently is being investigated at MMI Engineering. ■

For further guidance on using CFD for wastewater modeling, consult theAqua Enviro training course “Introduction to CFD Modeling for Water andWastewater Treatment Plants” at www.aqua-enviro.net/calendar.asp.

References

[1] Burt, D.; Ganeshalingam, J., “Design and Optimisation of Final Clarifier Performance with CFD Modelling,” CIWEM/Aqua Enviro joint conference, Design and Operation of Activated Sludge Plants,April 19, 2005.

[2] Robinson C.; Wilson R.; and Hinsley S., “Calculating Primary SettlingTank Performance with Computational Fluid Dynamics,” 4th AnnualCIWEM Conference, Newcastle, U.K., September 12–14, 2006.

These simulation results depict solids concentration for a radialcross section of a wastewater clarifier. Red indicates areas of higher concentration.


ENVIRONMENTAL DESIGN: POWER GENERATION

With pure water as the only byproduct, fuel cells are oneof the most environmentally safe alternatives for providingpower for vehicles and stationary applications: Stacks ofthe devices generate electricity directly from hydrogen andoxygen. One major concern in designing fuel cell stacks isdissipating heat created during the electrochemical conver-sion process. Thermal hot spots within the fuel cell stackmay degrade performance, induce thermomechanicalstresses and shorten the useful life expectancy of the stack.

Temperature distributions within the stack depend onmany variables, including non-uniform heat generation, fluidproperties and flow quality, fuel cell geometry, and the configuration of cooling plates between the cells. To arriveat a suitable design, engineers may resort to numerous prototype build-and-test cycles that are lengthy and costly —not to mention how they stifle innovation — because of theprohibitive time and expense of evaluating new ideas andwhat-if scenarios. These limitations can be alleviated somewhat with “deterministic” computer-aided engineering(CAE) methods that perform a series of individual analyses. Even in this scenario, engineers must run hundreds or even thousands of individual simulations to arrive at a satisfactory design.

Thermal model of a four-cell stack was created withcoolant flow and convective heat transfer modeledwith pipe elements. Pipes also were used to modelthermal contribution of air and hydrogen flow.

Structural analysis and shape optimization of the fuel cell end-plates were performed to optimize the stiffness within space limitations.

Cooling Down Powered-Up Fuel CellsBy Andreas Vlahinos, Advanced Engineering Solutions, Colorado, U.S.A.

Researchers use probabilistic methods and design optimizationto improve heat-transfer characteristics of fuel cell stacks.



A more efficient way to optimize a design with manyvariables and uncertainty is to account for variation usingadvanced computational and probabilistic tools early in thedesign process. This approach is being used extensivelyon research for market-viable alternative energy solutions.In some of this leading-edge work, the ANSYS Workbenchplatform and ANSYS DesignXplorer software have beenimplemented for performing design of experiments inaccounting for uncertainty and variation in materials, manufacturing and load conditions. Simulation tools also are used to streamline laboratory experiments bynumerically evaluating the design space to assess anddetermine which variables have the largest impact onresults. Laboratory tests validate the results and are fedback into the model to improve its predictive capabilities.

In one project studying fuel cell design, the engineeringconsulting firm Advanced Engineering Solutions, based inthe United States, used an approach that was aimed atestablishing optimal design methodologies for fuel cells.The company also was charged with improving productdevelopment time and costs by reducing the number ofphysical prototypes and laboratory tests required. In onecase in particular, the research team used tools fromANSYS to develop a fuel cell stack thermal modelingprocess to assess design sensitivity on fuel cell thermal performance. The models were used to evaluate new cooling plate flow paths and to assist in the development of improved heat transfer characteristics.

The thermal modeling process incorporated an ANSYSMechanical 3-D multi-cell stack thermal model that

reflected real-world stack geometry and non-uniform heatgeneration in the membrane. ANSYS DesignXplorer technology was used for design space exploration andprobabilistic design methods. Classical design of experi-ments techniques integrated with the model were used todefine response surfaces and perform sensitivity and trade-off studies on heat generation rates, heat-sink fin geometry,fluid flow, bipolar plate channel geometry, fluid propertiesand plate thermal material properties. A Taguchi screeningstudy was used to identify the most sensitive input para-meters; robust design was used to understand the impactof variation on thermal performance.

Researchers at Advanced Engineering Solutions thenused the ANSYS thermal model to develop an alternativecoolant flow path design that yielded improved thermal performance. The team found that this approach shavedmonths off the development process and led to innovativedesigns through improved understanding of fuel cell behavior, especially the impact of a wide range of design variables. ■

References

[1] Vlahinos, A.; Kelly, K.; Mease, K.; Stathopoulos J., “Shape Optimizationof Fuel Cell Molded-On Gaskets for Robust Sealing,” ASME paperFuelcell2006-97106, 2006 International Conference on Fuel CellScience, Engineering and Technology, Irvine, CA June 19–21, 2006.

[2] Kelly, K.; Pacifico, G.; Penev, M.; Vlahinos A., “Robust DesignTechniques for Evaluating Fuel Cell Thermal Performance,” ASME paperFuelcell2006-97011, 2006 International Conference on Fuel CellScience, Engineering and Technology, Irvine, CA June 19–21, 2006.

Response surfaces show the relationship between multiple variables,in this case visualizing the impact of fin thickness and base thickness on the maximum temperature of a cell stack.

After generating 10,000 virtual experiments, engineers create a scatter plot of performance requirements showing maximum temperature versus pressure drop.Dark blue squares represent data points that meet all design requirements and have minimal temperatures.

DS_DPDS_TBASE

1.24E+1 2.50E+2 5.00E+2 7.50E+2 1.00E+3 1.25E+3 1.59E+3

8.55E+1

8.25E+1

8.00E+1

7.75E+1

7.50E+1

7.25E+1

7.07E+1

MAX

T_FI

N

DS_TF

IN

MAX

T_TF

IN

7.31E+1

7.30E+1

7.28E+1

7.27E+1

7.26E+1

7.24E+15.00E-34.30E-33.60E-32.90E-32.20E-31.50E-3

2.00

E-3

2.80

E-3

3.60

E-3

4.40

E-3

5.20

E-3

6.00

E-3

www.ansys.comANSYS Advantage • Volume I, Issue 3, 2007s8 www.ansys.comANSYS Advantage • Volume I, Issue 3, 2007s8

PEM fuel cell

Contours of the mass fraction of water within the anode (left) and cathode (right) channels for a commercial parallel geometry PEM fuel cell


Making Electricitythrough Chemistry Analysis helps power fuel cell design.

By Laura Ambit and Esther Chacón, Instituto Nacional de Tecnica Aeroespacial, Madrid, Spain

Monica Pardo and Eva Novillo, Compañía Española de Sistemas Aeronáuticos, Madrid, Spain

Fuel cells are electrochemical devices that produceelectricity from an external supply of fuel and oxidant. Manycombinations of fuel and oxidants are possible; however,the fuels most often used are hydrogen, hydrocarbons andalcohols, while oxygen typically is the oxidant. The conver-sion of the fuel to energy takes place via an electrochemicalreaction in which the only byproducts are water (whenhydrogen is the fuel) and heat. The process is clean, quietand highly efficient. For these reasons, fuel cells are highlyregarded in the search for sustainable energy sources.

There are several types of fuel cells, and their differ-ences are dependent on the nature of the electrolyte.Polymer electrolyte membrane (PEM) fuel cells operate atlower temperatures than other types, can supply up to 10 Wof power per cell and can be stacked to handle higherpower loads. The principal applications of PEM fuel cells arein transportation — some experts believe fuel cells will revo-lutionize the automotive industry — and decentralizedstationary electrical applications, which range from poweringhome co-generation systems to vacuum cleaners and notebook computers.

In work done by the Instituto Nacional de TecnicaAeroespacial (INTA) and Compañia Española de SistemasAeronáuticos (CESA) in Madrid, Spain, researchers chose tomodel a single 7 W PEM fuel cell with the intent of under-standing how different geometric configurations and

operating conditions affect the cell’s performance. The per-formance depends on a variety of structural and functionalparameters, such as the geometry of the flow paths in thebipolar plates, along with the humidity, temperature andoperating pressure. To improve the performance of a PEMfuel cell, it is necessary to understand the behavior of vari-ables such as velocity, flow distribution, condensation ofwater and current distribution. Numerical simulation thusbecomes an important tool for understanding the physicalphenomena that take place.

The work at INTA and CESA set out to utilize the fuel cell module of FLUENT computational fluid dynamics (CFD) software to capture the fundamental processes of thefuel cell and to optimize the flow path design of the bipolarplates to improve efficiency. To achieve these objectives, a number of simulations were carried out, ranging from the simplest models of fluid flow analysis to more complexones that included modeling electrochemistry and multiphase flow.

The simulations were conducted using both a commer-cial geometry with parallel channels and a prototypegeometry with two serpentine path flow channels. Theeffect of operating conditions such as inlet flow humidity,mass flow rate and the influence of geometric parameterssuch as channel width also were studied in a simplifiedmodel of a single serpentine channel.


Within the parallel geometry, the researchers simulateda laminar, incompressible single-phase (gas only) flow usingFLUENT 6.2 technology. In these simulations, the electro-chemical phenomena were ignored, and only an analysis ofthe fluid flow was carried out. The computational domainwas restricted to the flow channels of the anode and cath-ode. Different mass flow rates were simulated, from a highexcess of both reactants to the minimum flow that guaran-tees the electrochemical reaction will occur. The resultsshowed a non-uniform distribution of the flow in all simula-tions, meaning that a large part of the membrane surfacewas being wasted. Similar fluid-flow analysis of the serpen-tine geometry concluded that it allowed a more uniformdistribution of flow, and, thus, better electrical conductivity,than the parallel geometry.

In addition, a simulation of the simplified single serpen-tine channel model was used to better understand allvariables involved in the electrochemical reactions and alltransport phenomena inside the gas diffusion layer and themembrane electrode assembly (MEA). The FLUENT fuel cellmodule accounts for reacting flows in contact with the MEA,


How PEM Fuel Cells WorkThe PEM fuel cell consists of an anode and a cathode, separated by

a polymer electrolyte membrane (PEM). Simply put, the fuel cell works as follows: Hydrogen and oxygen molecules enter into the device, the hydrogen is broken down in order to produce electricity — and water is created as the byproduct. A catalyst layer is placed between the anode (or cathode) and the PEM.

Oxygen enters the fuel cell on the cathode side of the device. Hydrogen entersthe anode side of the device, and, as it comes into contact with the catalyst layer, it splits into two hydrogen ions and two electrons. The hydrogen ions areconducted through the PEM. When the hydrogen ions come into contact with thecatalyst layer on the cathode, they join together with the oxygen atoms andrecombine with the electrons that have driven the energy-producing current,forming water as the only byproduct for the entire process.

To ensure efficiency in the process, the channels through which the oxygenand hydrogen pass in the anode and cathode should be designed to create asmuch contact area as possible between the gas molecules and the catalyst layers.

Contours of current density are plotted in order to comparethe effect of varying channel widths on fuel cell design.

Basic design for a PEM fuel cell. Hydrogen entersthe fuel cell at the anode while oxygen enters thefuel cell at the cathode.

Anode

Polymerelectrolytemembrane (PEM)

Cathode

Polarization curves based on the FLUENT results showing the influence of relativehumidity at the inlets both with and without multiphase flow simulation

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

Current (A/cm2)

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

Without-multiphase HR 75%With-multiphase HR 75%Without-multiphase HR 100%With-multiphase HR 100%

Volta

ge (V

)

heat transmission between reactants and bipolar plates,diffusion of reactants through porous media, and liquidwater formation via multiphase flow. Using this module,complete polarization curves of the single serpentine channel were calculated. The simulations allowed INTA andCESA to observe the influence of such parameters as thehumidity of the mass flow inlet, the mass flow rate and thechannel width over these polarization curves, and, subse-quently, the electrical current density of the cell. The CFDfindings demonstrated that the change of flow directionresults in an increase of current density in the local region.

From both the full electrochemical simulations of thesimplified serpentine geometry and the fluid flow simula-tions of the complete serpentine geometry, it is apparentthat a serpentine design can improve fuel cell performancewhen compared to conventional parallel geometry. Buildingon this research, the current intention is to achieve a fullelectrochemical simulation of the complete serpentinedesign using FLUENT software. ■

The authors would like to acknowledge Carla Vico for her assistance with this article.


ENVIRONMENTAL DESIGN: AUTOMOTIVE

The Future of FuelA European research project is developing internal combustion engines powered by hydrogen.

By Jorge Ferreira, ANSYS, Inc.

is that it can be produced from waterand a renewable energy source, suchas solar power. The main emission thatresults from this type of process iswater vapor, making hydrogen a positive alternative fuel that has thepotential for reducing carbon dioxideemissions.

Using trial runs and miniature models, many truck and automobilemanufacturers, including BMW, MANand Ford, have examined what can be done with alternative fuels. Theresearch to date has been conductedmostly on bivalent systems — enginesthat can use two types of fuels — withmost experiments using fossil fuelsand hydrogen. The BMW Hydrogen7™ luxury performance automobile is

capable of running on either hydrogenor gasoline. Such mixed use of fossilfuels and hydrogen has the advantageof extending the range of today’s carsas compared to pure hydrogen-fueledcars. However, a number of disadvan-tages arise when using one enginedesign for multiple and very differentfuels. The engine is not optimized for hydrogen nor for gasoline or dieselconsumption, meaning that efficiencycannot be optimized. When comparedto gasoline and diesel, hydrogen has a good deal of variation in physicalattributes, such as density, evapor-ative characteristics and combustion behavior. As these types of factorshave a direct effect on engine perform-ance, it is clear that if hydrogen were

BMW Hydrogen 7 with bivalentengine for hydrogen and gasolineImage © BMW Group

Commercially available reserves offossil fuels are fast running out, and theinfluence of harmful automobile emis-sions on the global climate is anongoing debate. For these and otherreasons, researchers and developershave been involved in investigatingalternative fuels for the automotiveindustry. Fuel cells and electric cars arepossible alternatives to today’s carsand trucks, which are powered by fossil fuels; however, these tech-nologies face some disadvantages,such as limited power dynamics andunsatisfactory power–weight ratios. As another alternative, the internalcombustion (IC) engine itself offersmany promising solutions if it is fueledby hydrogen. One benefit of hydrogen

GH2 Feed Line

Boil-Off Pipe

Safety Blow-Valve Feed Line

Exhaust Pipe BMS

Air Inlet BMS

Water Cooling Cycle

Gasoline Pipe

1 LH2 Fuel Tank

2 LH2 Tank Cover

3 LH2 Tank Coupling

4 Safety Line to Blow Valve

5 Auxiliary Units Capsule containing Heat Exchanger for H2 and Control Unit of the Tank

6 Bivalent Internal Combustion Engine (H2 /Gasoline)

7 Intake Manifold with H2-Rail

8 Boil-Off-Management-System (BMS)

9 Gasoline Tank

10 Pressure Control Valve

used to replace gasoline or diesel fuelin an engine designed for those fuels,there would be a loss of fuel efficiencyand engine effectiveness.

In order to properly take advantageof the characteristics of hydrogen as afuel, a hydrogen-powered engine mustbe built from the ground up. This wasthe goal of the European Commission-funded Hydrogen Internal CombustionEngine (HyICE) research project, athree-year effort aimed at designing a clean automobile engine. This initia-tive led to the development of ahydrogen-powered IC engine thatoffers significant advantages in termsof cost and power as compared withother systems. The project team, co-ordinated by the BMW Research andTechnology Group, comprised auto-mobile manufacturers, automotivesuppliers and two universities. Already,the group has shared its results andexperiences with partners in the UnitedStates; in 2003, the United States andthe European Union agreed to collabo-rate on speeding up the developmentof the hydrogen economy.

An important consideration for theproject was the customization andimprovement of appropriate simulation

tools for the hydrogen-based combus-tion process, in order to support thefuture mass production of engines.ANSYS CFX computational fluiddynamics (CFD) software was selectedas the main commercial CFD platform,because it already was employed bymany of the project participants andbecause the software could be customized for the specific needs ofthe effort.

When designing the simulationmodel, special attention was paid tothe specifics of hydrogen combustion.HyICE studied two different methodsfor fuel injection and, therefore, devel-oped two different simulation models.For the cryogenic method, already inuse as a bivalent solution, hydrogenwas mixed with oxygen in the inletport before it entered the cylinder, atwhich point it was compressed andignited. In the direct injection method,hydrogen was injected directly and athigh pressure into the cylinder andsubsequently ignited. Hydrogen com-bustion is much faster than that offossil fuels and occurs under higherpressures. Experts from ANSYS, Inc.implemented and tested different setsof models for their numerical stabilityand accuracy. These were comparedwith experimental data.

Because the ignition process takes only a few milliseconds to occur, the team developed a quasi-one-dimensional combustion/ignitionmodel to simulate this behavior. A full 3-D simulation of the combustionprocess then was developed based onthe solution provided by the ignitionmodel.

The goal behind using CFD for theHyICE project was to accelerate thedevelopment process, though anothergoal focused on the design of a reliableand validated simulation solution forfuture development projects. Thesegoals were achieved. The agreementof the simulation results with experi-mental data, especially in the areas oftemperature and pressure distribution,was excellent. The extensions andmodifications to the software made inthe course of the HyICE project can be applied to conventional engines as well. ■

Two injection methods examined under the HyICE project:(top) direct injection of high-pressure hydrogen and (bottom) port injection using cryogenic hydrogenImage © BMW Group

Cryogenic port injection plots show hydrogen massfraction and motion of hydrogen as it flows into thecylinder during the intake stroke of the engine.As the intake valve opens and the piston drops downaway from the valves, the piston motion draws thefuel air mixture in through the intake port.


ENVIRONMENTAL DESIGN: AUTOMOTIVE

www.ansys.com s11

390 degrees ATDC

420 degrees ATDC

430 degrees ATDC

460 degrees ATDC


Special DeliveryResearchers use simulation and medicalimaging to explore new options formanaging pain. By Malisa Sarntinoranont, Xiaoming Chen, Jianbing Zhao

and Thomas Mareci, University of Florida, U.S.A.

The latest therapeutic agents forchronic pain, spinal injury and otherneurodegenerative diseases are char-acterized as macromolecular (large)proteins. Delivering these drugs at thesite of action is gaining popularity.However, the transport environment inthe spinal cord and other nervous tissuemust be taken into account whendesigning direct infusion therapies.

Since macromolecular drugs dif-fuse relatively slowly, transport factorsaffect the effectiveness of deliverygreatly. The delivery of drugs throughthe spinal cord is dependent on a variety of factors, including variationsin material properties and flow regionswithin the cord itself. Specializedanalysis methods that correctly predictthe related transport behaviors arerequired before one can develop general, and possibly patient-specific,delivery protocols.

By coupling medical imaging withcomputational fluid dynamics (CFD)analysis, a research group at the Uni-versity of Florida in the United Statesrecently developed methods for pre-dicting the distribution of a drug tracer

injected directly into the rat spinal cord[1, 2]. Traditional magnetic resonanceimaging (MRI) was used to determinethe geometry and structure of thespinal cord. Diffusion-tensor MRI (DT-MRI), which provides information onhow water molecules spread throughtissue, was used to determine the preferred and most likely transportdirections in the cord.

Analyses of interstitial pressure,velocity and tracer distribution withinthe porous media in the spinal cord were performed using FLUENTsoftware. An anisotropic hydraulicconductivity (K) was applied in thewhite matter, a transport region located at the periphery of the spinalcord, to model the flow through it. The magnitude of K was assigned based on experimental data [3]. DT-MRI technology was used to identify thedirection of maximum water diffusivity,which, since it was assumed to be parallel to the local fiber orientation,was used to determine fiber tractdirections. This alignment data wasused to assign behavior properties tothe model.

Infusion site

gm

wm

Maximum eigenvectors (identified by arrows in theimage) identify the locations of maximum water diffusivityand preferred tissue transport for a fixed rat spinal cord.The red arrows represent aligned eigenvectors.

MRI-derived three-dimensional geometry of the rat spinalcord. White matter (wm) is in grey and grey matter (gm)is in green. Drug is delivered into the white matter, nearthe boundary with the grey matter.

Predicted albumin tracer distribution in the spinal cord20 minutes after a 2.0 µl infusion. Red areas represent sites of higher concentrations.

0.004

0.003

0.002

0.001

0.001 0.002 0.003 0.004x(m)

y(m

)

BIOMEDICAL

Using FLUENT technology, the distribution of a small volume infusion of the tracer then was predicted. Convection-dominated transport alongwhite matter tracts was found, and thepreferred distribution was identifiedalong the cord axis with little penetra-tion into adjacent gray matter zones.These results correspond well withsmall volume distribution trends foundexperimentally [3].

A model of this type could be usedfurther to analyze the effectiveness ofdifferent injection protocols, such ascontinuous versus discontinuous injec-tions or the effect of injection site ondrug distribution. Eventually, this type ofimage-based modeling effort may allowcustomized medical care that inherentlyfactors in patient-specific physiologicaldifferences. ■

References[1] Sarntinoranont, M.; Banerjee, R.K.; Lonser,

R.R.; Morrison, P.F., A Computational Model of Direct Interstitial Infusion ofMacromolecules into the Spinal Cord,Annals of Biomedical Engineering, 2003,31, pp. 448–461.

[2] Sarntinoranont, M.; Chen, X.; Zhao, J.;Mareci, T.M., Computational Model ofInterstitial Transport in the Spinal Cord Using Diffusion Tensor Imaging, Annals of Biomedical Engineering, 2006, 34,pp.1304–1321.

[3] Wood, J.D.; Lonser, R.R.; Gogate, N.;Morrison, P.F.; Oldfield, E.H., ConvectiveDelivery of Macromolecules into the Naiveand Traumatized Spinal Cords of Rats,Journal of Neurosurgery (Spine 1), 1999,90, pp. 115–120.

More Certaintyby Using UncertaintiesEngineers apply probabilistic methods to historically deterministic problems.

Rail cars carry enormous loads,often triple that of the largest 18-wheeler trucks. These loads passentirely through the wheels, which notonly bear the weight but are subject to a number of other structural, ther-mal and fatigue loads. Griffin Wheel Company, a division of AMSTEDIndustries in the United States, produces 90 percent of rail wheels forthe North American railroad industry.The company recently applied proba-bilistic tools from ANSYS, Inc. to theirwheel design process.

A basic freight car wheel is relative-ly simple, yet the wheels are subjectedto extreme forces and, therefore, mustwithstand tremendous amounts ofabuse. The wheel not only bears theload of the car, but also its tread surface is used as a brake drum,absorbing varying loads under con-stantly changing thermal conditions. In addition, the flange guides the trainon the track, conveying lateral loadsthroughout the wheel. Althoughdeceptively simple in construction, themulti-faceted character of the freightcar’s load environment makes for anextremely complex analysis.

Freight car wheels are solid steelcastings. Heat-treating strengthensthem, improves wear resistance andinduces circumferential residual com-pressive stresses in the upper rim toprevent fatigue crack formation. Heat-treating, however, generates axialtensile stresses in the lower part ofthe rim, causing vertical split rim, a

rare but catastrophic failure mode.Understanding the factors that canaffect these types of stresses isessential in effectively optimizingwheel design.

For an engineering analysis, manyfeatures are inherently variable anduncertain: operational loads, geometry,manufacturing processes, materialproperties and operational environ-ments, as well as testing. Theseuncertainties lead to uncertainty inproduct development and manufactur-ing. The traditional deterministic designapproach accounts for variations byusing safety factors. But this approachdoes not account for the random

Computer-aided design (CAD) model (right) of railroad freight car wheel (left)

By Kexiu Wang, Griffin Wheel Company, Illinois, U.S.A.

nature of design parameters. Treatingthe various parameters as singly deter-mined values decreases predictablereliability. Without measuring this relia-bility, performance levels becomeinconsistent. Moreover, since commonpractice assumes the worst-case scenario for each singly determinedvalue, the resulting design is often lessthan optimal, and subsequent changesproduce undetermined effects in otherareas.

The probabilistic method makesuse of statistical tools as a more reliablemeans to account for these multi-faceted uncertainties. During ananalysis, parametric uncertainties are


TRANSPORTATION

As a complement to the ANSYS Workbench environment, ANSYS DesignXplorer softwareprovides a number of probabilistic analysis tools. Engineers use these tools to describe a para-metric model in terms of statistical distribution functions for variations in the input parameters.

The technology uses two methods to estimate analysis variations. The Direct Samplingmethod, based on Monte Carlo sampling, requires a large number of simulations and can benefitfrom the parallelization techniques offered by products from ANSYS, Inc. The other method, calledDesign of Experiments (DOE), is based on response surfaces. DOE requires fewer simulations thanDirect Sampling and builds an approximation of the system response from which probabilisticresults are drawn.

Both methods analyze results variability and allow standard statistical analysis techniques(mean, standard deviation, kurtosis, etc.) as well as statistical sensitivity measures, with the latteractually identifying critical parameters driving the design.

ANSYS DesignXplorer software also provides information about the probability to reach agiven performance. This data helps assess the risk of failure for a given design at a given targetvalue, such as maximum stress, maximum displacement and minimum eigenfrequency. An equivalent Sigma level also is given, based on the Six Sigma quality criteria.

Vertical split rim and contours showingthe residual tensile stress

The scatter plot compares maximum axial stress to the stress exponent.Correlation coefficients show strong sensitivity of the stresses to creep.

characterized statistically in terms ofprobability density functions (PDFs).These PDFs quantify the inherent risksin a system and allow evaluation ofinput parameter variations in relation to changes in output performances.Probabilistic analysis yields a morecomprehensive understanding of theentire system, allowing engineers todevelop a better understanding ofproduct behavior in, and responses to, real-life conditions.

When used in simulation, once the random variations of boundary conditions, geometry and materialproperties are specified for a specificanalysis case, the input variables arestudied simultaneously by using statistical sampling methods. The parametric finite element analysis (FEA)model then is invoked repeatedly, performing deterministic analyses overthe resulting input parameters.

Deterministic approaches haveshown that the residual stress (fromheat-treating) varies in relation tomany parameters. To investigate theeffects of different parameters in theheat-treatment process and to identifyparameters that have the greatestimpact on residual stress, Griffin engineers analyzed a CJ36 freight car wheel using the probabilistic tools from ANSYS. After performing a deterministic de-coupled thermo-

mechanical analysis on a baselinemodel, engineers performed a LatinHypercube sampling probabilisticanalysis. This determined the vari-ations in the residual stresses whengiven the uncertainty of the manufac-turing process parameters, boundaryconditions and material properties.The scatter plot showed that the residual stress was especially sensitive to creep.

The probabilistic analysis is beingused to identify future steps neededfor further optimization and eventuallywill lead to optimizing the Griffinwheel’s residual stress field, therebyimproving wheel reliability. Theprocess illustrates how simulationtechnology from ANSYS can be used

By Pierre Thieffry, ANSYS, Inc.

22000

20000

18000

16000

14000

12000

10000

8000

6000

4000

200011.6 11.8 12 12.2 12.4 12.6 12.8 13 13.2 13.4 13.6 13.8 14 14.2

MAX

σy

to understand production processuncertainties and related parametervariations in a manufacturing process,leading to increased product reliabilityand quality. ■

Probabilistic Analysis with ANSYS Workbench

Variation of the simulation results with respect to design parameters

1.20E+0

1.17E+0

1.13E+0

1.09E+0

1.05E+0

1.01E+0

2.02

E+1

2.01

E+1

2.00

E+1

1.98

E+1

1.97

E+1

1.96

E+1

-5.69E+3

-5.81E+3

-5.94E+3

-6.06E+3

-6.19E+3

-6.31E+3

Safe

ty fa

ctor

min

imum

Force Y component

DS_Depth


TRANSPORTATION

Overview of the mechanical analysis of a gypsy wheel (left to right): The geometric model of the gypsy wheel was meshed, loading of the wheel was defined and a mechanicalsimulation was executed in order to validate the structural integrity of the assembly.


MARINE

Designing Outthe Weakest LinkEngineering simulation validates the design of a mooring system component, a critical wheel/chain assembly that holdsships in place during oil and gas operations in the North Sea.

By Joel Thakker, Integrated Design & Analysis Consultants Ltd., Croydon, U.K.

Floating production, storage andoffloading (FPSO) vessels take oil orgas from deepwater offshore petroleumplatforms, process it and store thematerial until it can be offloaded ontowaiting tankers or sent through apipeline. To maintain a stable, fixedlocation — even in rough waters —these huge ships have on-board winchsystems that handle mooring chains,which can be hundreds of feet inlength and weigh thousands of tons.Critical to the winch system is a centralassembly called a gypsy wheel that,together with hydraulic chocks, grippers and interlocks, controls therelease, retraction and tensioning ofthe mooring chain.

To validate the structural integrityof a newly designed gypsy wheel,

Whittaker Engineering in Scotland, acompany that provides engineeringdesign to the offshore oil and gasindustry, approached the engineeringconsulting firm Integrated Design &Analysis Consultants (IDAC) Ltd. Theprimary reason for developing a newdesign stemmed from an earlier designfailure that resulted in a mooring chaindislodging from the assembly and consequently sinking to the oceanbed. IDAC was tasked to analyzestresses and deformation associatedwith the gypsy wheel under both pre-load and operating load conditions.The cause for the failure fell outside thescope of this analysis.

The challenge in this project was inevaluating whether the wheel andchain locker components of the gypsy

wheel, which hold the retracted chain,could withstand various loading condi-tions as well as whether the size andweight of the assembly could be mini-mized without compromising structuralintegrity. The design would have beenimpossible to load-test safely and difficult to analyze precisely by con-ventional hand calculations. For thesereasons, engineering simulation pro-vided a useful evaluation pathway.Upon validation of the design, the newgypsy wheel assembly was to beinstalled on board the FPSO Captain, avessel operated by Chevron Texaco inthe North Sea.

The simulation process beganwith the finite element analysis (FEA),in which the geometric model of thenew gypsy wheel was meshed using


MARINE

www.ansys.com 19

ANSYS Mechanical software. Nonlinearcontact elements were generatedbetween the shaft and the wheel, andbonded contact elements were gener-ated between the bolts and the wheelto simulate the bolt preload. As part of the meshing capabilities, contactelements automatically detect contactpoints and allow for dissimilar meshesbetween contacting parts. In addition,the mesh is configured to account forjoined parts, thus avoiding the task ofmanually adjusting mesh densitiesand selecting element types — aprocess that can be time-consuming.

In the new gypsy wheel design, themooring chain was guided into thechain locker, thereby relieving theassembly of excessive loads duringthe mooring process and intrinsicallyreducing the chances of failure. Perdesign mandates, the frame of thegypsy wheel was designed to with-stand transient and impulse loads for asmall period of time. The transientloading cases were conducted tomimic operating conditions, while theimpulse loading was done to designthe structure for accident scenarios.(An accident had caused the system to fail in the first place, thus demandingFE structural evaluation.) In addition,due to the limitations associated withaccess and mechanical handling,effort was put into keeping the weightand size of the assembly at a minimumwithout compromising the structuralintegrity of the assembly.

Three separate load cases werestudied as part of the investigation.The simplest load case was used to evaluate whether the assembly suffered damage under a pre-loadingscenario. The other two cases ana-lyzed a normal and angular downwardforce independently as well as in com-bination with an out-of-plane force. Ineach case, the simulations confirmedthat the new gypsy wheel design couldundergo the pretension and opera-tional loads, with resulting deformationand stresses falling well within thedesign parameters.

In the end, structural analysisusing ANSYS Mechanical softwareeffectively evaluated the new designof the gypsy wheel under the variousloading conditions and failure modes.Simulation overcame the inability toperform load tests safely as well asthe difficulty of time-consuming andless accurate manual calculations.

IDAC then worked with Whittaker Engineering to produce an engi-neering package that was acceptableto the ship’s certifying authorities. Subsequently, Whittaker Engineeringmanufactured and installed the eightnew chain gypsies, which currently arein service in the North Sea. ■

Total deformation (top) and von Mises stress (bottom) for a gypsy wheel that is supported by a central shaft.From the von Mises plot, it can be seen that the area that supports the chain on the top of the wheel experiences high stresses, as does the shaft that supports the wheel.


SPORTS

Going for the GoldSimulation helps design low-drag canoesfor Olympic-medal performance.

By Nicolas WarzechaInstitute for Research and Development of Sports Equipment, Berlin, Germany

Andreas Spille-KohoffCFX Berlin Software GmbHBerlin, Germany

Competition among world-class athletes at theOlympics has become so intense that tiny variations in performance mark the difference between the gold medalwinner and the also-rans. Relying heavily on computer simulation to reduce the air resistance of their bobsleds, theGerman national team leveraged their win in that sport toemerge triumphant in the 2006 Winter Olympics at Turin,Italy. Germany edged out the United States in overall goldmedals, 11 to 9, and in total medals, 29 to 25.

The work in optimizing the performance of bobsledswas carried out by engineers at the Institute for Researchand Development of Sports Equipment (known by its German acronym of FES) in Berlin, one of the world’s leading centers for the development of sports equipment.Today, FES engineers are hard at work designing skiffs,canoes and sailboats that they hope will help produce a similar triumph at the 2008 Summer Olympics in Beijing, China.

To gain an edge in the canoe competition, FES engineers are using ANSYS CFX fluid simulation softwareto simulate the performance of various canoe designs. Theyselected this technology largely because it provides thepowerful CFX Expression Language (CEL), which allowsusers to create their own physical models quickly from within the user interface, to add new variables, and to

Image ©Camera 4/Thonfeld.


SPORTS

define property relationships and boundary condition profiles. CEL goes beyond similar languages by allowing FORTRAN™ routines to be called, allowing other FORTRANapplications to be coupled to ANSYS CFX software.

The engineering team at FES began by simulatingexperiments involving towing a canoe, which eliminated thechallenge of simulating the effect of the paddlers’ strokeson the boat’s motion. They used ANSYS ICEM CFD Hexasoftware to create a block-structured grid model with 3 million elements and used the free-surface multiphasemodel of the ANSYS CFX product to analyze both themotion of the water and the air trapped by the movement ofthe boat and the water. This simulation showed a very goodcorrelation with the drag measured in the towing experi-ments. All cases were simulated in parallel on a 64-bitLinux® cluster, whose installation was supported by CFXBerlin. The best results were acquired by running the CFXsolver on 10 to 20 mesh-partitions, depending on the size ofthe grid. Using this approach, a transient simulation representing a 10-second real-time interval for a movingboat could be performed in one to two days.

The wetted surface of the canoe in this model did notmatch the experiments, a conclusion that was expected

since this simplified model did not account for the forcesand moments acting on the boat. So FES engineers simu-lated the boat at its position with the ANSYS CFX solver tocalculate these forces and moments and estimated the newposition of the boat — that is, sink and trim values. Theycontinued with a series of manual steps that slowly con-verged to a final position showing good agreement withexperiments. Once they had determined that this approachprovided realistic results, FES automated the analysis, withthe assistance of CFX Berlin, by writing CEL expressionsand some pieces of FORTRAN code that performed all ofthese steps in the same way but much faster and automati-cally within the ANSYS CFX solver. With this approach, itwas possible to evaluate and compare the performance ofseveral alternative designs.

The analysis performed by FES provides the drag aswell as complete information on the movement of the water around the boat, the position of the boat and theforces acting on the boat. In particular, computational fluiddynamics (CFD) makes it possible to measure the bowwave, aft wave and wake of the canoe to a high level of precision. Simulation provides engineers with good indica-tions of what is causing drag in a particular design and whataspects of the design they should change to improve it.

In the past, designing these boats was based largely ontrial-and-error prototyping. Using simulation, FES engineerswere able to quickly design a new skiff that ANSYS CFXsoftware predicts will provide a 3 percent improvement indrag. A prototype of this boat is currently under construc-tion. After completion of the prototype, physical testing willbe used to verify the simulation results. In the meantime,FES engineers are working to expand the scope of CFD analysis to analyze other effects that are difficult orimpossible to measure, such as the effect of initiating paddling forces at different times and variations in thevelocity of the boat. Including these effects in the analysismay make it possible to advance the performance ofcanoes and athletes to even higher levels. ■

Experimental testing area used to study water flow and wave formation around a boatas well as total drag


CFD results depict the wave pattern as it develops along a canoe body. Wave height is indicated by color, with blue denoting lowest areas and red indicating the highest areas.

A series of Poincaré planes in the downstream section of a 90-degree tube bend; rows are colored by tracerspecies, residence time and frequency (1/residence time), respectively, and x represents location in the tube afterthe bend, which ends at x = 8.0.

Geometry and mesh for a lean premixed natural gaspower turbine CFD case, based on the General ElectricAircraft Engines LM6000 engine


PARTNERS

Something in the MixResearchers use the Poincaré plane method to obtain quantitative time scale information from CFD simulations.

Mixing processes are involved at some level in nearly all chemicalmanufacturing processes. They arefundamental to the successful opera-tion of combustion-driven systems.Today, many computational fluiddynamics (CFD) practitioners in thechemical process industry are able touse simulation to obtain detailedinsight into overall performance of theirprocess equipment. However, it is stilldifficult to relate CFD data to the effective management and control of aparticular process. In addition, the costof production delays due to sudden,unexpected changes in product quality provides strong motivation to understand the impact and relevanceof CFD studies that are focused onthese areas.

While CFD continues to be moreaccessible to analysts, managers andoperators, problem complexity andsophistication also has increased.Relating flow data, such as mixing timescales to device performance, now is a major challenge. Flow visualizationmethods, which use iso-surfaces andcutting planes, can be used to helpvisualize flow topologies in an ad-hocway. Streamlines and time-dependentstreaklines also are effective at eluci-dating flow patterns. However, theseapproaches are limited in that they provide very little quantitative informa-tion on how flow patterns affect overallperformance.

At Intelligent Light in the UnitedStates, engineers turn to the Poincaréplane method to obtain quantitativetime scale information from CFD simu-lations. Poincaré planes, placed atvarious locations within a flow domain,display the time and locations at which

By M.N. Godo, Intelligent Light, New Jersey, U.S.A.

Geometry and mesh for a laminar flow case through a90-degree bend

streaklines cross these planes. Timescales obtained from these plots relatedirectly to how effective a mixing tankis or how efficiently a furnace or incin-erator can be run. Being able to seetime scales within mixers and combus-tion chambers offers much easierinterpretation of the CFD data foreveryone involved in the productionprocess. For instance, Poincaré planesshowing holes or concentric rings indicate flow regions that are stronglysegregated, that is, poorly mixed. Ingeneral, this behavior is undesirable;knowing exactly where this occurs in a process vessel is a key step inresolving performance problems.

Since its introduction, the Poincaréplane method has been applied to mixing studies [1] and fundamental flow problems [2]. To create Poincaréplanes, FIELDVIEW [3], a CFD post-processing tool from Intelligent Light, is used to interpret CFD simulationresults that are generated by FLUENT

x=10.0 x=13.0 x=16.5 x=20.0 x=23.5

z = 0.0

Trac

er s

peci

es [

-]Re

side

nce

time

[s]

Freq

uenc

y [H

z]

z = 4.0

x = 8.0 x = 24.0

d = 4.0

Cross-sectionalmesh detail(enlarged)

Cooling air slot

Counter-rotatingswirl inlet

x/D = 0.18

A series of Poincaré planes used in the analysis of a gas turbine case (specifically the GEAE LM6000) modeledusing FLUENT software: (top) RANS turbulence model and (bottom) LES turbulence model

software. FIELDVIEW is able to readdata exported directly from FLUENTtools as well as from the ANSYS CFXproduct. Using the velocity field information from the CFD simulations,FIELDVIEW calculates the large number of streaklines necessary toobtain accurate results for Poincaréplanes. Because of the repetitive,quantitative tasks needed, the FIELD-VIEW programming language, FVX™,was used to automate streakline trajectory calculations, identify streak-line intersections with the Poincaréplanes and visualize the final results.

Of two cases studied, the first simulated simple laminar flow througha 90-degree bend. The second casewas a fully validated flow calculationfor a lean premixed natural gas power turbine, based on the General ElectricAircraft Engines (GEAE) LM6000engine. A counter-rotating swirl inletboundary condition was provideddirectly by GEAE. Both Reynolds-averaged Navier–Stokes (RANS) andlarge eddy simulation (LES) turbulencemodels were calculated using FLUENTtools, and GAMBIT software was usedto create the meshes for both cases.

For the 90-degree bend case, it was observed that flow details basedon either the residence time or frequency are highly structured, andthey exhibit significant local differ-ences as the fluid is rolled up by theaction of the vortices. Notably, the fluidin the center of the tube, which has a residence time roughly five times thatof the flow near the upper section, hasa significant impact on mixing effec-tiveness, as the flow has clearlybecome quite structured.

Within combustion chambers, akey goal in design assessment is toquantify mixing rates, particularly attime scales that are on the same orderof magnitude as the chemical reactionand energy and mass transfer rates.For the RANS turbine case, areas of strong flow isolation are clearly seen near the inlet. In addition, theRANS solution exhibits significantlymore structure than the LES solution.Time scales, observed in the Poincaré


PARTNERS

www.ansys.com 23

planes for the RANS case, cover a wide range. This strongly affects theextent of combustion predicted by this simulation. In contrast, Poincaréplanes for the LES case show a veryhigh level of chaotic mixing on a finespatial scale. Apart from the regionimmediately downstream from the swirl inlet, there were no significant differences in either the residence timeor frequency, and the central flameenvelope is nearly gone at the farthestdownstream plane for the LES case.Mixing time scales for the LES case are expected to provide more realisticpredictions of the combustion physicsin this particular case. ■

The author would like to express gratitude to Greg Stuckert and Graham Goldin of ANSYS, Inc.for providing the GEAE LM6000 combustion caseand for sharing their considerable knowledge ofbest practices concerning the setup of the partiallypremixed combustion routines as well as theparameters for the large eddy scale calculations.

References[1] Zalc, J.; Szalai, E.; Alvarez, M.; Muzzio, J.,

“Using CFD To Understand Chaotic Mixing inLaminar Stirred Tanks,” American Instituteof Chemical Engineers Journal, 48(10),2002, pp. 2124–2134.

[2] Shariff, K.; Leonard, A.; Ferziger, J.H.,“Dynamical Systems Analysis of FluidTransport In Time-Periodic Vortex RingFlows,” Physics of Fluids, 18(4) , 2006,pp. 047104-1 – 047104-11.

[3] FIELDVIEW, CFD Postprocessor, Version 11,Intelligent Light, Rutherford, NJ, 2006.

x=0.010m x=0.015m x=0.020m x=0.025m x=0.030m

Flam

e su

rface

[-]

Resi

denc

e tim

e [s

]Fr

eque

ncy

[Hz]

x=0.010m x=0.015m x=0.020m x=0.025m x=0.030m

Flam

e su

rface

[-]

Resi

denc

e tim

e [s

]Fr

eque

ncy

[Hz]


PARTNERS

The disciplines of computer-aided engineering (CAE)and high-performance computing (HPC) have been closelyaligned and interdependent since the 1970s, when ANSYS,Inc. was founded. As software and hardware technologieshave evolved, engineers who conduct simulation analysishave been among the beneficiaries. Recent advances inHPC have been particularly valuable, bringing down the cost of entry for small workgroups in need of large-scalecomputing capacity. In particular, cluster-based computers— based on x86/64-bit processors from Intel® and AMD —now represent over 50 percent of HPC solutions and provideenormous computing capacity for a fraction of the cost ofprevious-generation solutions. Working with Microsoft® andother partners, ANSYS, Inc. now is making clusters a moreviable solution for Windows®-based customers through support of the Microsoft Windows Compute Cluster Server2003 (Windows CCS) operating system.

The Argument for ClustersEngineers who perform simulations in support of

product development are well versed in the business driversthat make clusters attractive. Simply put, more computingcapacity increases productivity along with the value that simulation brings to the product development process. By

reducing turnaround time, increased parallel computingcapacity helps ensure that simulation results are available ina time frame that can impact engineering decisions. Byenabling larger and more detailed simulations, computing systems with more memory (RAM) yield more accurate andmore reliable results. Finally, by increasing throughput, alarger computing capacity enables the engineering team tosimulate multiple design options while meeting schedulerequirements. Clusters provide all three benefits — parallelspeedup, large memory availability and capacity for highthroughput — in a form that can be expanded over time assimulation needs expand.

Given these benefits, it is not surprising that clusters arenow the dominant platform for computational fluid dynamics(CFD) simulations using ANSYS CFX or FLUENT software,since these packages have, for many years, been designedfor parallel speedup on clusters. More recently, with therelease of version 11.0 technology from ANSYS, clustershave become a much stronger solution for finite elementanalysis (FEA) simulations as well. The new distributed memory solver in version 11.0 provides improved parallelscale-up. In addition, clusters are being used to increasethroughput for parametric FEA analysis.

Cluster Computingwith Windows CCSNew clustering technology from Microsoft speeds upengineering simulation.

By Barbara Hutchings, ANSYS, Inc.

Early Adopters Alden Research Laboratory, Inc., based in Massachusetts, U.S.A., is an acclaimed fluids flow engi-

neering and environmental laboratory providing analytical, computational and physical flow modelingservices. When the CFD team at Alden wanted to expand their analysis capacity, they turned to a clusterrunning FLUENT 6.3 software on Windows CCS. “We needed to increase our computing power in order to increase the number of FLUENT simulations we perform as well as to consider larger, more detailedmodels,” said Dan Gessler, Alden’s director of numeric modeling.

Using FLUENT software, Alden engineers simulate flow in advanced hydroturbine designs. For example, the Alden/Concepts NREC turbine team used flow modeling to maximize generating efficiencyof the fish-friendly turbine.The unique turbine design has the lowest fish mortality for turbines in its class.

According to Charles Ulrich, Alden’s IT manager, “The ability to deploy a cluster using Windows CCSwas very attractive for us, as it leverages our expertise and fits into our current computing environment.The deployment was quite smooth: We had our cluster up and running FLUENT software within twoweeks. The integration of FLUENT with the Microsoft Job Scheduler is especially valuable, giving us theability to manage and monitor multiple simulations on the cluster.”

Flow around a vertical axis runner, simulatedusing FLUENT software and Windows CCSImage courtesy Alden Research Laboratory, Inc.


Windows Compute Cluster Server 2003Windows CCS was released by Microsoft in 2006 to

enable cluster computing within a Windows environment.Windows CCS is based on the Windows Server 2003 64-bitStandard Edition and leverages familiar Windows tech-nologies, such as Active Directory, to provide authorizationand authentication services on the cluster. In addition, Windows CCS provides cluster management utilities fordeploying and administering the cluster as well as a built-injob scheduler to control and manage multiple tasks on the cluster. The combination of support for clustering and 64-bit memory addressing has made Windows CCS a veryviable option for engineers using products from ANSYSwho want to leverage their existing Windows infrastructureand expertise.

A typical cluster configuration involves one or more clientsystems (for example, desktop workstations) running theANSYS Workbench platform. These clients submit computetasks (solver jobs) to the cluster via the ANSYS RemoteSolve Manager (RSM) and the Windows CCS Job Schedulerrunning on the cluster head node. For software not yet integrated within the ANSYS Workbench environment —such as the FLUENT 6.3 application — the process is very similar, with the FLUENT GUI running on the client systemsand solve requests submitted to the Microsoft Job Sched-uler via a new FLUENT launcher panel. Both version 11.0from ANSYS and FLUENT 6.3 packages are fully integratedwith the Windows CCS Job Scheduler, providing off-the-shelf management of jobs on the cluster.

PerformanceThe performance of ANSYS 11.0 and FLUENT 6.3

products on Windows CCS has been documented by Hewlett-Packard, and the results are very good. For FLUENT, parallel scaling is nearly linear with the number ofprocessors on a correctly sized cluster and similar to

performance on equivalent clusters running Linux®. Figure 1shows performance of a typical FLUENT simulation involving3.6M finite volume cells as the cluster size increases up to16 processors (32 cores). As CPUs are added to the cluster,the simulation speed scales well — with a speedup ofroughly 75 percent of ideal scaling on 32 cores. ANSYS 11.0scaling on Windows CCS is shown in Figure 2, in which thebenchmark suite yields a range of speedups depending onthe solver and physics involved.

ANSYS and MicrosoftMicrosoft and ANSYS have worked closely to ensure

that Windows-based clusters are a strong solution for engi-neers using solutions from ANSYS, Inc. This engagementbegan and continues in the technical arena, with supportfrom Microsoft for porting and tuning applications fromANSYS on Windows CCS. Feedback from ANSYS helps todefine requirements and improve the combined offering.The resulting performance makes Windows CCS an excellent choice for expanding simulation capacity.

The two companies also are working together at customer sites, building a combined understanding of thedetails required to successfully deploy software from ANSYSon Windows CCS. Working with Microsoft, ANSYS hasdeveloped detailed guidance for customers on sizing thecluster, setting up the cluster and deploying engineering simulation software on the cluster with connection to desk-top clients. Despite its very attractive price/performanceratio, cluster technology has a reputation for being a challenge to implement. By teaming up, ANSYS andMicrosoft have improved their ability to respond promptly toquestions and problems that customers encounter. Bothorganizations also are working with original equipment man-ufacturers (OEMs) — including Hewlett-Packard, IBM®, Dell®,SGI® and Sun®, as well as system integrators and resellers —to help streamline the delivery of complete Windows CCSsolutions to engineering simulation customers. ■

For more information, email [email protected] 6.3.33FL5L2 Benchmark (3.6M)

4000

3500

3000

2500

2000

1500

1000

500

0

HP CP3000(Xeon Dual Core 3.0GHzVoltaire Infiniband)

Number of Cores Benchmark Test

Per

form

ance

Rat

ing

(Big

ger

is B

ette

r)

1 2 4 8 16 32

8.0

7.0

6.0

5.0

4.0

3.0

2.0

1.0

0.0

Rel

ativ

e S

peed

up (H

ighe

r is

Bet

ter)

bmd-1 bmd-2 bmd-3 bmd-4 bmd-5 bmd-6 bmd-7

Figure 1. Performance of FLUENT 6.3 software using the HP-MPI (messagepassing interface) on a standard benchmark simulation of flow around a simple automotive sedan shape. The performance rating is comparable forWindows CCS and Linux. Results courtesy Hewlett-Packard.

Figure 2. Performance of ANSYS 11.0 benchmarks as the processors countincreases up to eight cores. The new Distributed ANSYS solver scales well onclusters. Results courtesy Hewlett-Packard.

PARTNERS

1 Core

2 Cores

3 Cores

4 Cores

CCS

Linux

Component Mode Synthesis inANSYS Workbench SimulationCMS superelements provide flexibility of simulation models while reducing the number of degrees of freedom for highly efficient solutions.By Sheldon Imaoka, ANSYS, Inc.

At ANSYS Workbench 11.0, theANSYS Rigid Dynamics add-on moduleenables users to define joint connec-tions in complex kinematic assemblies.

This ability to model rigid and flexible parts in ANSYS WorkbenchSimulation via joints is beneficial sincethe rigid parts are modeled with massand rigid links (that is, rigid contact elements), thus providing load transfercapabilities and allowing users to eval-uate system-level performance withless computational cost. However, thesolution time is dictated by the size ofthe flexible mesh, and in some casesthe rigid assumption may not be accu-rate enough for specific applications.

A better approach in these cases isthe use of Component Mode Synthesis(CMS) superelements, in which theflexibility of the model is retained yetthe number of degrees of freedom(DOF) is reduced, thereby providingefficient solutions.

A superelement has a reducednumber of DOF compared with the

“full” model, yet it still can accuratelymodel the flexibility of structures.Superelements can be created by reg-ular substructuring or by componentmode synthesis, in which DOF at inter-face nodes are retained while all otherDOF are eliminated by condensation ofthe matrices. CMS appends a regularsubstructure with generalized coordi-nates, thereby improving the accuracyof the superelement response indynamic applications.

Since CMS superelements can beused in large-deflection nonlinearanalyses, they are especially attractivefor nonlinear transient problems. Thisis because of the low number of DOF(that is, interface nodes and general-ized coordinates) and the accuratedynamic representation. CMS super-elements, however, also can be used instatic, modal, harmonic and responsespectrum analyses.

If contact is used with super-elements (regular substructure orCMS), the number of interface nodes

Figure 1. Flexible Dynamic analysis setup with all nineoriginal assembly parts as rigid components

Figure 2. System-level analysis with joints defined Figure 3. Reference Coordinate System location and orientation

increases dramatically, depending onthe size of the contacting region.Hence, CMS superelements should beused in Flexible Dynamic analyses inANSYS Workbench Simulation withjoints. Joints are defined at a singlenode, so the number of interfacenodes used in CMS is reduced to abare minimum.

Example CaseA sample Autodesk® Inventor®

assembly was used for this example. AFlexible Dynamic analysis was set upwith all nine parts as rigid, as shown inFigure 1. Interaction between the partswas defined by joints using rigidbehavior for the associated surfaces,and two Joint Conditions were usedfor loading of the sample engine.

In a separate model, the connectingrod was set up in order to create a CMS superelement. Figure 2 showsthe original system-level analysis withthe joints. The individual model needsto reproduce the Joint Reference


TIPS & TRICKS

The rigid body is connected tojoints via rigid links (that is, rigid con-tact elements). The reason the authorprefers to delete all but one rigid link isso post-processing can be conductedas usual. In ANSYS Workbench Simu-lation, the original rigid mass is used totrack the position of the part, so if onlyone of the rigid links is left, the masselement moves with the connectedjoint, and the user still can visualize theresponse of the system.

Figures 4 and 5 show relativeangular velocity at the crankshaft–connecting rod joint for both the rigidassembly and the assembly that usedCMS superelements to represent thepiston, connecting rod and crankshaft,respectively. For relatively low loading,as illustrated in Figure 4, both resultsmatch well, as expected. For higherloading in which there is some relativedeformation, as shown in Figure 5, therigid-only and CMS cases start toshow differences due to the flexibilityof the parts.

It is important to note that the solution times for both system-levelruns were relatively quick. The rigid-only case ran for 303 iterations, whichrequired 19 seconds on a 3.2 GHz PC.The assembly with three CMS components ran for 1,340 iterationsthat required a total of 191 seconds.The CMS assembly case requiredmore iterations to account for the flexibility behavior that was included inthe analysis with the additions of thethree CMS superelement components.

Figure 4. Comparison of simulation results for the rigid and CMS cases match for lowloading. The green line indicates CMS results while the red line represents the rigidcase results.

Figure 5. For higher loading, the rigid case and the CMS one differ due to the flexibility of CMS superelement components. The green line indicates the CMSresults while the red line represents the rigid case results.

Figure 6. Results, with CMS parts, are post-processed asthey normally would be in ANSYS Workbench Simulation.

Coordinate System location and orienta-tion with a Remote Displacementsupport, as shown in Figure 3. With thatcompleted, the model can be meshed,and a superelement can be created.

To accomplish the generation of thesuperelement, a Commands object isinserted that selects constrained nodes(from the Remote Displacement support), deletes the constraints anddesignates those nodes as master DOF.A CMS generation pass is performedwith a user-specified number of request-ed modes, and the resulting file.sub isthe superelement file needed for the sys-tem-level analysis.

This procedure is repeated for eachrigid part that will be converted to a superelement. Note that since multiple superelements may exist, each.sub superelement file should berenamed to have a unique name. Lastly,the user should ensure that the sameunit system is selected as is used in thesystem-level assembly.

Once the superelements are generated, they can be incorporated intothe system-level model. The authorprefers doing this by adding two Commands objects, one under the partto be replaced and another under theFlexible Dynamic branch. The first has a single line, PART1_ = MATID,allowing the user to reference the element type ID a priori. The secondCommands object would delete all rigidlinks of the rigid part except one, add the superelement and then couple coincident nodes together.

Once overall results (deformation,joint results, spring results) arereviewed, the deformation, stressesand strains for each part can bereviewed by expanding the superele-ment solution, then using “Tools menu> Read ANSYS Result Files.” The usercan expand either results at particularpoints in time or all of the results, then post-process as usual using theANSYS Workbench platform, asshown in Figure 6. ■

Contact the author at [email protected] the full article from which this was excerptedas well as demonstration input files.

-17.086

-250.

-500.

-750-

-1000.

-1250.

2.e3 -3 2.5e-2 5.e-2 7.5e-2 0.1 0.125 0.15 0.175 0.2

Angu

lar V

eloc

ity [r

pm]

Time [s]

-1303.9

-17.086

-250.

-500.

-750.

-1000.

-1250.

2.e3 -3 2.5e-2 5.e-2 7.5e-2 0.1 0.125 0.15 0.175 0.2

Angu

lar V

eloc

ity [r

pm]

Time [s]

-1303.9


TIPS & TRICKS

www.ansys.com 27


ANALYSIS TOOLS

Accelerating toConvergenceANSYS VT Accelerator technology can help solvenonlinear transient and static analyses faster.

By Ray Browell, ANSYS, Inc. ANSYS, Inc. has offered leading-edge nonlinear solutions for a number ofyears. Today, by using a unique featureknown as ANSYS VT Accelerator, engineers are uniquely positioned tosolve their nonlinear analyses faster. Anynonlinear problem that is being solvednumerically typically needs some methodby which to iterate to a converged solution. One solution is the Newton–Raphson method, which is an iterativeprocess of solving nonlinear equations —the method used by most computer-aided engineering (CAE) software toolsincluding those from ANSYS, Inc. Usually,the solution used as the starting point isthe previously converged solution.

Technology from ANSYS includes apredictor method that extrapolates thesolution by using the previous history in order to get a better estimate of the solution. The ANSYS VT Accelerator feature greatly expands upon the predictor method by using a veryadvanced predictor–corrector algorithm.The algorithm is so powerful that sometimes no equilibrium iterations areneeded. In other words, the solution isconverged at the first equilibrium iteration, thereby greatly increasing thespeed of nonlinear solutions. And as afeature unique to ANSYS, ANSYS VTAccelerator technology enables efficientSimulation Driven Product Development.

Solver SpeedupVariational Technology (VT) has been

applied to two distinct types of mathe-matical problems: nonlinear solutions forstructural and thermal analyses as well asharmonic analysis. These capabilities arereferred to as ANSYS VT Accelerator. This

In this example from Florida Turbine Technologies, parameter modificationsinvolved variations on film coefficients and bulk temperatures.

feature provides a two times to fivetimes speedup for the initial solution,depending on the hardware, modeland type of analysis. ANSYS VTAccelerator technology makes re-solves three times to ten times fasterfor parameter changes, allowing foreffective simulation-driven parametricstudies of nonlinear and transientanalyses in a cost-effective manner.Users can make the following types ofchanges to the model before anANSYS VT Accelerator re-solve:

■ Modify, add or remove loads(constraints may not bechanged, although their valuemay be modified)

■ Change materials and materialproperties

■ Change section data and realconstants

■ Change geometry, although themesh connectivity must remainthe same (that is, the meshmust be morphed)

Nonlinear Solution SpeedupANSYS VT Accelerator technology

for nonlinear solutions speeds up thesolution of applicable nonlinear analysis types by reducing the totalnumber of iterations. The technologysupplies an advanced predictor–corrector algorithm based on Varia-tional Technology to reduce the overallnumber of iterations for nonlinear static and transient analyses. It isapplicable to analyses that includelarge deflection, hyperelasticity,viscoelasticity and creep nonlinearities.Rate-independent plasticity and nonlinear contact analyses may not show any initial improvement inconvergence rates; however, usersmay choose this option with these

nonlinearities if they wish to resolve theanalysis with changes to the inputparameters. In general, ANSYS VTAccelerator technology can be usedfor:

■ Nonlinear structural static ortransient analyses not involvingcontact or plasticity

■ Nonlinear thermal static or transient analyses

Harmonic AnalysisThe harmonic sweep feature of

ANSYS VT Accelerator provides a high-performance solution forforced-frequency simulations in high-frequency electromagnetic problemsand structural analysis. For a structural

harmonic analysis, the material mayhave frequency-dependent elasticity or damping.

For a high-frequency electromag-netic harmonic analysis, ANSYS VTAccelerator technology computes S-parameters over the entire frequencyrange. In practice, the harmonic sweepfeature completes one normal run in ANSYS Mechanical software at the mid-frequency of the specified frequency range. It then performsaccurate approximations of the resultsacross the frequency range (in user-specified steps). In addition tocontrolling the steps and the frequencyrange, users can specify the accuracyof the approximations. Two harmonicsweep solution methods are available:Variational Technology and VariationalTechnology Perfect Absorber. The Vari-ational Technology Perfect Absorbermethod provides about a 20 percentfaster solution, but it is somewhat less accurate. ■

Turbine model speedup with full PCG solver time as a baseline. Initial ANSYS VT Accelerator solution was morethan two times faster, and subsequent re-solves improved, with the third re-solve being 10 times greater.

CPU Ratio (vs Full PCG)

12.00

10.00

8.00

6.00

4.00

2.00

0.00

1.03

Std

Run

- Qu

asi I

CCG

1.00

Std

Run

- Fu

ll PC

G

2.07

Initi

al S

X -

(mos

t sec

ure)

3.29

Initi

al S

X (fa

stes

t)

8.59

Para

met

er c

hang

e 1

9.53

Para

met

er c

hang

e 2

10.67

Para

met

er c

hang

e 3


ANALYSIS TOOLS

www.ansys.com 29

In order to provide “innovations that work,” Florida Turbine Technologies, Inc. — whichexecutes all aspects of turbine engine design and development in the military and commer-cial aircraft industry — desires transient fidelity early in the design process. “Due to longrun times, we usually reserve transient analyses for detailed final design,” says Joseph T.Metrisin, lead structures engineer at Florida Turbine Technologies, Inc. “Faster solutionoptions will allow us to perform detailed transient analyses early on in the design process,resulting in more robust designs.”


Seeing is BelievingTired of looking at your CFD images inside out? Developments in version 11.0 software from ANSYS allow inclusion of solid parts during pre- and post-processing, makingfor more intuitive problem setup and results visualization.By Judd Kaiser, ANSYS, Inc.

When engineers are first introduced to computationalfluid simulation, one of the challenges they face is learningto see the world inside out. In the real world, we’re accus-tomed to focusing our eyes on the parts that fluid flows inand around. But unlike what we usually see in the real world,models created for computational fluid dynamics (CFD) simulations include only the regions in which the fluid flows— not the solid components that surround the fluid. Thisoften makes it difficult to visually communicate the relation-ship between the CFD results and the related solid geometryunder analysis, especially to the untrained eye. With somesurprise, the author discovered that a confluence of recent developments that came along with the ANSYSWorkbench 11.0 release gives an unexpected solution tothis visual dilemma.

First, the ANSYS Workbench 11.0 platform brings a newmeshing application that offers physics-based, part-by-partmeshing with multiple methods. For CFD users, the primarymethods of interest are swept meshing with inflation andtetrahedral meshing with inflation. The inflation layer allowsfor finer mesh resolution to better accommodate the near-wall physics details that are critical for accurate prediction of wall-bounded flows. This new application also offershighly automated, fault-tolerant meshing methods that weretuned for use in mechanical finite element analysis (FEA)

ANSYS DesignModeler software is used to prepare the model for analysis.

simulation. These methods are extremely fast and producemeshes that resolve the geometry with a minimum number ofelements. So why would they be of interest to the CFD user?

When preparing meshes for CFD simulation, engineersoften work with computer-aided design (CAD) models thatwere created for manufacturing purposes. These models areusually complex assemblies that describe the solid compo-nents that bound the fluid region. For CFD analysis, theseparts must be turned inside out to create a part that repre-sents the fluid domain. Typically, the CFD engineer discardsthe CAD model’s solid parts, since they are not of directinterest in the fluid simulation.

There is some additional “cost” to meshing the solidparts, carrying those meshes into the pre-processor andloading multiple files in the post-processor — so this maynot be something a user will do for everyday design work.However, if you are generating results that can be used formarketing or to share with individuals who are not experi-enced in looking at CFD simulations, you may find thatincluding the solid parts visually helps to convey your analysis and results more clearly.

Because these solid parts can be meshed “cheaply,” consider bringing them along for the ride. Foryour next CFD project using ANSYS Workbench 11.0 tools,give the following steps a try.

ANALYSIS TOOLS

1) After importing the CAD assembly, use ANSYSDesignModeler software to prepare the model foranalysis. For CFD simulation, this often involves theuse of fill and/or enclosure operations to prepare thefluid domain. (This step also could be performed withCAD software.)


Meshing occurs in the ANSYS Workbench 11.0 Meshing Application.

When defining physics for CFD pre-processing, it is easy to selectonly those parts that you want to include in your CFD domain.

The solid parts as well as the fluid simulation can be shown in the post-processor.

ANALYSIS TOOLS

2) Bring the fluid domain and all solid parts from the originalCAD assembly into the new ANSYS Workbench 11.0Meshing Application. You can use your meshing methodof choice for the CFD domain, but in general it should bea method that includes an inflation (prism) layer toresolve the near-wall physics. Next, instead of discardingthe solid parts, mesh them using the automatic meshingmethod in the Meshing Application, which will result in a mesh that resolves the solid parts with a minimumnumber of elements. Meshing these parts typically isvery rapid, so this comes at low computational cost.

3) Import the mesh assembly into CFX-Pre (the AdvancedCFD tab in the ANSYS Workbench platform). In CFX-Preyou may find that having these solid parts available helpsyou understand the problem visually. When definingphysics in CFX-Pre, it is now easy to select only thoseparts that you want to include in your CFD domain. (In this case, include the fluid part but not the solid parts.)The solid portions of the mesh will be saved in a .cfx filebut not in the .def file.

4) Solve the CFD simulation as usual.

5) In ANSYS CFX 11.0 software,CFX-Post allows users to loadmultiple files — so load both thesimulation result (the .res file)and the physics setup (the .cfxfile). You then will be able to create graphics to illustrate thenature of the flow field as usual.In addition, you will be able toshow the solid parts. Being ableto render the solid parts shouldmake it easier for the untrainedeye to comprehend the nature ofthe simulations. ■

The predicted spray resulting from atomization. Unstablewaves, called Kelvin–Helmholtz waves, are apparent on thesurface of the liquid. Downstream, the sheet disintegrates intoligaments and further into droplets. This image corresponds toa surface of constant liquid volume fraction (a = 0.05) and iscolored by the magnitude of the tangential (swirl) component of velocity.

Liquid atomization processes such asthose associated with pressure-swirlatomizers can be simulated using the volume of fluid (VOF) multiphase model inFLUENT computational fluid dynamicssoftware. This model is preferred when an engineer desires to predict the location of the interface between twoimmiscible phases or fluids. In the case ofthe atomization process occurring in a pressure-swirl atomizer, the VOF modelpredicts the gas–liquid interface locationduring the formation and disintegration of the liquid film, the formation and tearing of ligaments, and, ultimately, theformation and transport of droplets.

In this study, the research team devel-oped a FLUENT model to predict theliquid atomization and spray formation.Additionally, the spray was characterizedexperimentally for model validation. Ahigh-resolution camera combined with alaser flash was used to visually capture

CAD geometry of typical pressure-swirl atomizer

Predicting Liquid AtomizationSimulation can be used as a predictivespray characterization tool.By Lisa Graham and Kumar Dhanasekharan, Bend Research Inc., Oregon, U.S.A.

John Widmann and Birendra David, ANSYS, Inc.

Pressure-swirl atomizers, also knownas simplex atomizers, are used commonlyin many industries, including aerospace,automotive, pharmaceuticals and others.These nozzles work by forcing a liquidunder high pressure into a swirl chamberin which the fluid gains tangential momen-tum and exits through a small orifice ornozzle. The liquid exiting the nozzle formsa sheet that thins as it disperses radiallyoutward. The thin sheet becomes unstableand breaks up to form ligaments and thendiscrete droplets. The ability to tailor spraycharacteristics is important, for example,in controlling the evaporation rate of fuelsprays in gas turbine combustors or thetransport of drugs administered throughinhalation. A fundamental understandingof spray formation can provide usefulinsight into the design and operation of theatomizer — in order to produce sprayswith desired characteristics such as thedroplet size and spray pattern.

the spray. The model predicted a coneangle of 60.2 degrees, which comparesfavorably to the experimental values inthe range of 69 to 75 degrees. For the liquid flow rates investigated, the modelpredicted the atomization pressure within 10 percent of published data(spraying systems water capacity data).Additionally, the model predicted all thesalient features of the flow, including theair core that develops within the swirlchamber in response to the swirling liquidflow. As the liquid exits the nozzle orifice,the tangential momentum of the swirling liquid causes the sheet to move radiallyoutward, thin and, ultimately, disintegrate.

Under the conditions considered inthe model, the disintegration of the liquidsheet is preceded by the formation ofunstable waves, called Kelvin–Helmholtzwaves, on the liquid surface. The modelpredictions demonstrated that the devel-opment of Kelvin–Helmholtz waves led todisintegration of the sheet and formationof ligaments. The ligaments experiencedfurther breakup until surface tensionforces exceeded aerodynamic forces,resulting in spherical drop formation.

The research team used the validatedmodel to study liquid atomization undervarying operating conditions, includingsolution viscosity, surface tension andflow rate. This enabled engineers to map out a design space for successful operation of the nozzle. Such simulationsprovide a fundamental understanding ofhow operating parameters affect spray characteristics and help in tailoring nozzledesign and operation to obtain sprays ofdesired characteristics. ■

Spray pattern image from camera


ANALYSIS TOOLS

Inlet

Swirl chamber

ADVANTAGE - Ansys · ANSYS Advantage • Volume I, Issue 3, 2007 ... ANSYS, ANSYS Workbench, CFX,...

Documents

Transcript of ADVANTAGE - Ansys · ANSYS Advantage • Volume I, Issue 3, 2007 ... ANSYS, ANSYS Workbench, CFX,...