wwwpadtinccom 1 1-800-293-PADT

March 30 2007 The Focus Issue 56

By Doug Oatis

If yoursquore like me and by that I mean super-control freak when it comes to your meshthe meshing in Workbench v10 sometimesfell short This was especially true when itcame to complex hexsweep meshingThere was no way to explicitly define thesource and target of your mesh This alongwith many other things has changed in v11

If you had played with the beta version orhave installed and are running the releasedv11 there are too many additions to coverin a single article One of the first thingsyoursquoll notice in Simulation is the meshingcontrols that are now available

Using the lsquoMethodrsquo control you can spec-ify whether you want to use the automaticmesher Tetra (from ICEM) sweep orCFX-Mesh (which used to be its own inde-pendent module) Using the method youcan also control mid-side node usage on apart-by-part basis

If you set the lsquoMethodrsquo to be sweep thereare several options that are available Theone that I like the most is being able tomanually define either just the source orboth the source and target This controlallows you to specify

April 20 2007 A Publication for ANSYS Users Issue 56

This One Goes to 11

By Eric Miller

PART 1 ANSYS EnhancementsWe have been talking up Version 110 ofthe ANSYS Inc products for some timeand now that it is released we can finallystart to share with ldquoThe Focusrdquo readers whatour engineers find in this major releaseConsider this article an executive overviewof new features and capabilities focused onpointing out what we feel is significant andletting you know how it can help you Weare skipping about 75 of the changes andonly highlighting two or three in each partof the software For the remainder of theyear we will have detailed technical articlesdelving deeply into key areas but if you

want to know more come to a PADT up-date seminar attend an ANSYS Inc We-bEx seminar or read the release notes thatcome with the 110 documentation

Before we start marching through featuresand functions and to understand a littlebetter why things are the way they are weshould step back and look at some keynon-technical factors that had an impact on110 1)The entire development teamwanted to take the time to make 110 morerobust and to track down as many issues aspossible before release 2) ANSYS Inc pur-chased FLUENT Inc a company that actu-ally had more people than ANSYS Inc did3) a strategic decision to provide users witha ldquosystem levelrdquo solution for simulationand 4) a business decision to increasegrowth in large enterprise size customers

All of these combined to force a change partway through the development process in-cluding being more cautious about rushingnew features out the door supportFLUENTrsquos organizational needs add more

time-transient and rigid-body capabilitiesand address legacy model and interoperabil-ity with competitive codes Not only wasdevelopment direction changed but the de-velopment schedule was stretched out and amuch longer testing cycle was added to theend In the end PADT feels pretty stronglythat these changes made sense from a busi-ness perspective and resulted in a muchbetter technical solution

If we covered ANSYS Workbench TASICEM CFD AUTODYN etchellip we wouldfill up 20 pages So in this issue we willonly be looking at significant enhancementsin ANSYS Workbench will be in the nextissue then we will cover all the other prod-ucts A key thing to notice while reviewingthese changes is that fact that developmenton the core ANSYS product is still veryrobust and technically aggressive Almostall of the features listed here were added toadd capability that users have asked fordirectly

ContentsANSYS Release 1101Simulation Meshing 1Modeling Bond Failure3Advertising 7

(Cont on Pg 6)

110 In Depth Simulation Meshing(Cont on Pg 2)

wwwpadtinccom 2 1-800-293-PADT

March 30 2007 The Focus Issue 56

DebondingThis capability started out as a separateelement but users soon asked if it could bemoved to the generalized contact elementsand development listened So now youdonrsquot have to have a continuous meshacross your bonded joint and you simplyadd a new material property (TB CZM) toyour contact elements You can specifydifferent criteria for mode 1 (separation)and mode 2 (shear) failure and can useenergy or gap distancesliding modelsNon-Linear StabilizationHave you ever watched a model with localbuckling or wrinkling diverge and daydream about hooking up little dampers toevery node in the model In version 110you can now tell ANSYS to do that for youand it will add small damping to every freeDOF using the new STABILIZE com-mand What is cool is that it does not use adamping matrix (that would really slowdown the solution) but instead calculates theforce from damping and applies it to theDOFrsquos There is no such thing as a freelunch so turning this on can slow down ananalysis It is recommended that you runwithout stabilization and if you model di-verges use a restart with it on and stepthrough the ldquotrouble areardquo doing a anotherrestart with it off once you have passedthroughElement and Solver Support for Multi-bodyDynamicsTo better support system level models a lotof changes were made in Workbench andANSYS On the ANSYS side the mostsignificant enhancements ldquohinged aroundrdquoupdating and improving the MPC184 fam-ily of joint elements ANSYS now supportsspherical revolute universal slot point-in-plane translational cylindrical planer ori-ent rigid and general joints Another nicechanges is that you use SECTION com-mands to define the joints making it easier

to create and store definitions On thesolver side a lot of work went in to speed-ing up implicit transient dynamics to getbetter performance on rigid-flexible systemmodels In addition a new solver was intro-duced to solve all-rigid models This evolv-ing technology combined with usersrsquocontinues to shape the direction of simula-tion toolsPCG LanczosEvery once in a while ANSYS Inc intro-duces a new solver that changes everythingThis time they replaced the sparse solverinside the Block Lanczos solver with thePCG solver and created a new solver forEigenvalue solution of large (greater than 1Million) DOF that is very fast The basicdifference is that the sparse solver does aton of IO whereas the PCG solver usesmemory So if you have a ton of RAMyou can crank through big model runs veryquickly To prove that point we solved a 20MDOF problem with 30 GB of RAM inabout 12 hours It simply filled up the diskand crashed when we used LANB

RotordynamicsSince about half of PADTrsquos staff comesfrom the turbomachinery world we arecrazy about Rotordynamics (RD) At 110a lot of critical details have been worked outbased on feedback from users around theworld who have really taken to the enhance-ments made at 100 To summarize a fewkey changes at this release Most of yourcommon structural elements now supportCORIOLIS effects a new bearing element

(COMBI214) was introduced to make iteasier to specify non-linear bearing behav-ior orbit printing plotting and animatingwas added and major changes were madeto the solution process and Campbell plot-ting to allow for the inclusion of prestresseffects during post processing Based uponour usage and discussion with a couple ofheavy users the consensus is that mostpeople should no longer feel the need to usein-house RD tools If you find somethingyou need is missing please let us knowwhat so we can get feedback to develop-ment For more info on this feature checkout the article in the last ldquoThe Focusrdquo Spin-ning up RotordynamicsCMS EnhancedAnother major area of focus to support sys-tem modeling is support for the ComponentMode Synthesis method (See issue 43 for adescription of the method and ANSYS ca-pabilities at V 100) At 110 CMS nowsupports static transient harmonic andspectrum analysis In a key enhancementfor modeling large mechanisms CMS nowsupport large deflection and substructureprestress And if you are a CMS junky youwill also be happy to hear that Residual-Flexible Free interface conditions are nowsupportedNew Element CONTA177Rounding out the generalized contact fami-ly 177 is a line-to-surface element that letsyou model contact between beams and sur-faces or the edge of shell surface to a gen-eral surface It works like the rest of thefamily and replaces the need to model suchsituations with node-to-surface contact Weexpect to see heavy use with MPC bondedcontact to connect the ends of shells to shellsurfaces at T-JointsNew Element SHELL281This is basically an 8-noded version of theShell 181 and brings the new element tech-nology to the 8-noded shell topology Itwas added based upon

(This One Goes to 11 Cont)

Want More information about Version 110 Besides the thor-ough documentation in the help manual under Release Notesvisit the ANSYS inc website and download the webex semi-nars you may have missed in March

Under the Customer Portal Left Menu select WebEx Presen-tations under the Product Information menu

httpwww1ansyscomcustomer

Resources

Review ANSYS Incrsquos Quick Start Guides which are multi-mediatraining sessions similar to a slide show but with user controlledpace downloadable files and index etc

So far they are all Release 110 related focusing primarily onthe Workbench Interface

See menu on the left side of the Customer Portalhttpwww1ansyscomcustomer

Resources(Cont on Pg 3)

wwwpadtinccom 3 1-800-293-PADT

March 30 2007 The Focus Issue 56

in-house testing and user input to providemore accurate and stable modeling of buck-ling and wrinkling problems when usingshell elementsNew Element USER300A long time request from users has been tomake the creation and control of user ele-ments easier After a lot of research anddiscussion development has added theUSER300 element You still have to createsome FORTRAN code but it is much morestraightforward and you control the ele-ments behavior with APDL and the newUSERELEM and USRDOF commandsBasically they have created an element APIthat greatly simplifies the definition andcontrol of the element so you donrsquot have toget down-and-dirty anymore Check out theProgrammerrsquos manual Section 611 for de-tailsDrucker-Prager PlacticityEvery once in a while a new material comesalong that everyone gets excited about Westarted getting requests for Drucker-Pragerabout 2 years ago and it became quite thetrend So development did their researchfound some verification problems andplugged away to deliver this rather interest-ing model for materials that have pressure

dependent material behaviorJ-Integral CalculationsIn the past if users wanted to carry outfracture mechanics assessment on theirmodels they had to use 2D geometry or a 3rd

party tool for 3D At 110 this area has beenredone to support 2D and 3D with the 18xfamily of elements Results can be viewedwith the new PRPLCINT commands anddetailed info can be extracted with GET

Coupled-Field for Electrostatic-StructuralIn the past electrostatic MEMS devicesrequired an iterative solve to capture dis-placement and electrostatic behavior At110 the PLANE223 SOLID226 andSOLID227 elements have been upgradedto support matrix coupling of UX UY UZand VOLT DOFrsquos This should speed upand simplify a large number of MEMS sim-ulationsNew Solver PCG LanczosModal extractions can now be done usingthe Block Lanczos method with PCG as thematrix solver This is much faster and morerobust than the standard Block Lanczoswhich uses the Sparse solver because PCGdoes much less disk IO It is preferred formodels with more than 1 MDOF but bewarned if you donrsquot have a lot of RAM

available it will swap to disk and you wonrsquotsee much of an advantage If you do haveRAM it is very fast and robustReset Picking ButtonIf you use the ANSYS GUI you may noticethat sometimes picking ldquogoes awayrdquo A newbutton in the GUI should allow you to fixthis problem Not a major technical en-hancement but something that will help alot of peopleAutomatic DIM and no More Need for Com-pressYou no longer have to create arrays withDIM if you are filling them with vectorand matrix commands it makes arrays onthe fly and dimensions them as neededAnother APDL change was a new VGETParRENTNLIST which creates acompressed array of selected entity num-bers So no more GET VMASKVFUNCOMP Both of these changesshow that APDL is alive and well and de-velopment is listening to users and makingthings better for us

We hope you have found this informationuseful As was stated earlier this is only asampling of the changes and we have noteven touched Workbench yet Look formore in future issues of ldquoThe Focusrdquo

(This one goes p to 11 Cont)

By Rod Scholl

Prior to ANSYS version 100 if we wantedto model delamination of two solids ourbest option was to use standard contact anda script which would check the tensionbetween elements and then kill those thatexceeded a given criteria

This type of script was highly sensitive tomesh size and was a little inefficient be-cause one had to build a do loop to go intopost1 to evaluate the contact elementskill the ldquofailedrdquo ones and then do a restartin between each substep Also this methodis highly path dependent meaning onemight need to model the whole ldquounzippingrdquoprocess to get to a final stateInterface Technology of version 100At 100 the interface elements (INT20X)were introduced which not only automatedthis interface failure based on criteria italso implements a function to describe theseparation before failure of the interfacesubstance which we will call ldquoepoxyrdquo

This function allowed for capturing the loadtransmission deep into an unzipping seamusing a very accurate exponential function

As we explored in the past article thismethodology was easy to implement al-though convergence required some criticalmesh density as shown to the right

Thus for the epoxies I typically encounterthe required mesh density made this ap-proach too expensive for 3-D analysis

110 In Depth Modeling Bond FailureCohesionElementStress

Shape determined byMaterial Constants (andloading) usingexponential model forsurface potential

ldquoFailedrdquo elements

Achieving Convergence

0010203040506070809

1

00 05 10 15 20 25

Mesh Density Factor

Tim

e

e

Converged atTime=10

Unstable Stable

(Cont on Pg 4)

wwwpadtinccom 4 1-800-293-PADT

March 30 2007 The Focus Issue 56

Version 110 EnchancementNow in 110 we have another option formodeling separation This new approachconveniently uses the existing contact ele-ment technology One big advantage isunlike the INT20X method one doesnrsquotneed mesh continuity at the interface Infact the implementation is a snap requiringonly a couple commands beyond what nor-mal contact requires As usual the helpdocumentation is fantastic and can step youthrough the implementationhellip you willlikely need to go to the theory manual 411to understand the material property imple-mentation ndash but herersquos my basic instructionoverview

Thatrsquos it Just by the presence of the CZMproperty those elements will now expandand fail according to the epoxy informationentered on the TBDATA command Alsothe data input is fairly intuitive such asmaximum tensile stresshellip and max separa-tion at failure

Converting Epoxy Material Data toTBCZM and FKN

The arguments for the CZM model are showbelow for the CBDD case (The CBDE argu-

ments are similar but use an energy crite-rion rather than gap at debonding)

Also for this article I think we can get agood understanding by considering onlyC1 C2 and C5 as though our epoxy onlyfails in tensile stress

So letrsquos look at the implementation for a testcase

With this data and a failure of the epoxy at50 elongation we have a C1 of 1000 psiand a C2 as follows

And thus we have

ANSYS implements this as follows

Following the red line is the implementedload path So we can see we have someflexibility to change Ubarn to have this bi-linear curve match our material data Thusthe more quickly the epoxy unloads afterreaching oacutemax ndash the closer Ubarn should beto C2(UCn)hellip

To control Ubarn and hence the curveshape of our epoxy we need to control theslope Kn which is simply the FKN realconstant for the contact pair

So for the case where Ubarn is 75 theway to (UCn) we calculate FKN as follows

Note that the negative value tells ansys tointerpret FKN as an absolute value and notbased on the underlying elementstiffnessdimensions

Presto You now have a bilinear implemen-tation of your epoxyrsquos loaddisplacementusing existing element technology and afew arguments on a TBCZM command

Test your PropertiesWhile investigating this new feature I builta script which pulls apart two blocks andmeasure the force required to do so Usethis script to enter your material propertiesand then verify you are getting theforcedisplacement behavior your expect Itseems to be stable over typical valuesWatch your Pinball SizeKeep an eye on your contact pinball radi-ushellip if the epoxy deflection allows the ele-ments to move outside the pinball radiushellipthe contact will let go no matter HOWstrong of an epoxy you are modelingDamping CoefficientSo stability is going to play a big role in theactual implementation of this methodologyAlthough it was not necessary for the sim-ple blocks in pure tension of the scriptabove you will likely needed it for peelloading scenarios The good news is thatno matter how stiff your epoxy given a

large enough stability numberconvergence will be a snap Thebad news is that accuracy of fail-ure stress of the elements de-grades rapidly with changing thisstability factor This C6 damp-ing coefficient implements aforce balance based on time-stepsize and change in deflectionThus changing C6 or the timebetween substeps will have animpact on results (See the newSTABILIZE command for whatI suspect is a very similar imple-mentation) Although the manual

recommends stability numbers like 01 and001 ndash I found that I was much happier withthe accuracy afforded by 0001 or 00001and below (I also note that the VM testcase uses 00001)

(Modeling Bond Failure Cont)

Failure Stress 1000 psiEpoxy Modulus 2E6 psiEpoxy Layer Thickness 0003rdquoMax Elongation 50

1) Build contact normally2) For the elements suspect to separationcreate a new material property usingTBCZM

a) TBCZMMatlCBDD or CBDEb) TBDATAMatlepoxy

infodamping coefficient

Constant Sym-bol

Meaning

C1 oacutemax maximum normal con-tact stress

C2 UCn contact gap at the com-pletion of debonding

C3 max maximum equivalenttangential contactstress

C4 UCt tangential slip at thecompletion of debond-ing

C5 artificial damping coef-ficient

C6 flag for tangential slipunder compressive nor-mal contact stress

C2 = MaxElongationEpoxyLayerthickness

C2= 00015rdquo

FKN = -C1(075C2)

(Cont on Pg 5)

wwwpadtinccom 5 1-800-293-PADT

March 30 2007 The Focus Issue 56

Do not use a default or 0 stability factor (C6)

It seems there is a requirement for someentry on the stability factor Even some-thing inconsequentially small like C6=1e-15 will do Just not a zero or you may getan odd increase in load to debond asshown below

To give us a feel for what stability factorsmay be necessary and the domain of epox-ies that require small factors (and hencegive high accuracy predictions) I explored atest case for various factors using this script

Note that some scenarios require hundredsof equilibrium iterations for convergenceYou might alter the test script for yourparticular epoxy properties applicationthickness and then determine what damp-ing coefficient (C6) meets your requiredaccuracy Note how in cases A-D that theactual failed bond length is quite dependenton C6 but the last little bit of accuracyrequires many iterations to achieve

As Case F indicates you also might leantowards a bilinear curve of your epoxyrsquosloaddisplacement that leans towards asmall ratio (UbarnUnC)

Good luck and happy debonding

C6 = 0

(Modeling Bond Failure Cont)

wwwpadtinccom 6 1-800-293-PADT

March 30 2007 The Focus Issue 56

The Focus is a periodic publication of Phoenix Analysis amp Design Technologies (PADT)Its goal is to educate and entertain the worldwide ANSYS user community More informa-tion on this publication can be found at httpwwwpadtinccomepubsfocusabout

To keep up to date with global and Europeanevents and news read ANSYS Incs newslet-ter E-News

httpwww-harwellansyscomnewslettersindexhtmLinks

Upcoming Training ClassesMonth Start End Title LocationApr 07 25-Apr 27-Apr 152 ICEM CFDAIEnvironment Tempe AZMay 07 2-May 4-May 104 ANSYS WB Simulation - Intro May 07

7-May 8-May 100 Engineering with FEA Tempe AZ10-May 11-May 203 Dynamics Tempe AZ18-May 19-May 106 ANSYS WB Design Xplorer Tempe AZ21-May 22-May 105 ANSYS WB Sim Struct Nonlin Tempe AZ

June lsquo07 4-June 6-June 101 Intro To ANSYS Part I Tempe AZ11-Jun 12-Jun 201 Basic Structural Nonlinearities Tempe AZ13-Jun 14-Jun 204 Adv Contact amp Fasteners Tempe AZ21-Jun 22-Jun 301 Heat Transfer Tempe AZ25-Jun 26-Jun 107 ANSYS WB DesignModeler Tempe AZ

July lsquo07 9-July 10-July 104 ANSYS WB Simulation - Intro Albuquer-que NM

ANSYS partners with PCA EngineersLTD To integrate VISTA (TM) (VISualTurbomachinery Analysis) with AN-SYSreg solutinos

Online news releaseNews

(Meshing cont)

whether you want to mesh ldquothroughrdquo a partor ldquoalongrdquo a part

The lsquoThin Modelrsquo options are a very nicefeature because thatrsquos how you specify theusage of the SOLSH190 element You willalso see the ability to specify the number ofsweep divisions and define a sweep bias Inmy experience Irsquove found it easier to spec-ify the biasing on the lines in the sweepdirection

My preferred method is to insert a sizingcontrol on a line specify the number ofdivisions and then enter my biastypespacing The benefit of doing the bias-ing on the line level is that it actually showsyou the line divisions in the graphics win-dow

This is only one enhancement made to themeshing branch of Simulation You canalso use patch independent meshing totackle sloppy geometry (just like you couldin ICEM) There are also new automatedsettings depending on the type of analysisyoursquore doing If you select explicit Simula-tion will automatically drop all midsidenodes and use a slow transition value to gofrom small to large element sizes

Look for future articles going over othermodule enhancements in Workbench v11

The best resource for Workbench users out there is the ANSYSWorkbench Community Files discussion groups and related re-sources help WB users share with each other and the developersrsquoat ANSYS Inc It just passed 1400 members

httpwwwResources ansyscomcustomerResources

Therersquos a new ANSYSPublication Subscribeto the ldquoANSYS Advan-tagerdquo A CFD and FSIrelated publication

httpwwwansyscomadvantageindexhtm

Links

wwwpadtinccom 7 1-800-293-PADT

March 30 2007 The Focus Issue 56

The Shameless Advertising Page

Be Cool andKnow whatDay it Is

Download the 2006 PADTANSYS Calendar

Ready to print prototypes