December 21,2005 The Focus Issue 43

December 21, 2005 A Publication for ANSYS Users Issue 43

Modal Analysis in Bite Sized Chunks: CMS

By: Eric Miller additional step of conducting a modal
The scene is typical for many large compa- analysis and using the results to modify the
nies that have standardized on ANSYS. Ev- transformation matrix that converts stiffness
ery year management looks at the software into the reduced matrix. See the theory
renewal budget and sees that lone seat of manual, 17.6.5, for a good description of the
NASTRAN. They make a call to see why it process and the actual equations used. This
needs to be renewed and a crusty “guru” also includes a bibliography of the basic
emerges from his cubicle, blinking at the work done to develop this method.
brightness of the lights and shying away Once the sub-structures are created, you
from contact with humans. A slight funk assemble them together at the system level in
follows him as he enters the boss's office and what is called the “use” pass. At this point
states: “we have to keep that seat. ANSYS you conduct a modal analysis and obtain the
can't do CMS” He turns and stumbles, his “modal” response of the entire system by
eyes never leaving the ground, back to the “synthesizing” it from the “component”
safety of his cozy home, surrounded by responses. simple 6 limbed bug with a head and a tail,
yellowing stacks of printout and his made out of beams so it runs fast. As you can
most prized possession, the aged Another difference between CMS see by following the APDL scripts, much of
MSC manuals from the early and basic sub-structuring is that the process is repeatable and can be easily
80's. you have to determine what type automated with generic macros. As you can
of boundary condition you want see in the table, the results line up very well.
But extinction for this rare to force on the MDOF's. When
breed is forthcoming. With the generation pass does the Even though it is still developing, the AN-
little fanfare, ANSYS modal analysis, you can leave SYS implementation of CMS exceeds that
began to add Component the MDOF's free (called free- of the NASTRAN world in a couple of
Mode Synthesis (CMS) to free) or you can force zero ways. CMS analysis is pretty simple if your
its growing list of features displacement on them (called structure is simple, but can get confusing
that were once the domain fixed-fixed). The free-free is fast for complex structures. ANSYS pro-
of the NASTRAN user. more accurate for predicting vides tools to help with bookkeeping as
CMS is a very clever way middle and high frequency well as copying and translating components
to conduct an efficient and modes and the fixed-fixed for in the assembly coordinate system. This
accurate simulation of the lower modes. There are other seems trivial but can make analysis much
modal behavior of large methods used in the easier. In addition, as with anything in
and complex structures by NASTRAN world that ANSYS ANSYS, you can script the whole thing
breaking the structure into pieces, solving will be adding per user requests. with APDL.
them, then combining the results together at
the system level. It is a very common After conducting the “use” pass you can So, next time you have a large system you
technique for any large frame structure such view the system response by plotting system need to do a modal analysis on, consider
as automobiles, airplanes, rockets and mode shapes and looking at modes. If using CMS. The documentation is very
satellites. CMS is basically the extension of needed, you can also move on to an clear and well organized. As usual, start
sub-structuring from stress to modal analysis. “expansion” pass and get the response of with a simple model and use APDL heavily.
each DOF in a component at any mode you (Continued on Page 2)
If you remember sub-structuring (if you don't want. Once you have expanded modes of
you should, it can save hours if not days of interest, you can combine the results in the
simulation time) what you do is define nodes post processor and view the entire system Contents
at an interface, and use Guyan reduction to response, or any subset of components, as a Modal Analysis in
create a reduced stiffness matrix in terms of system in POST1. Bite Sized Chunks..........................1
those interface nodes. These nodes are given
the politically incorrect name of Master This process is shown in the included
example “The CMS Bug” We don’t have Dr. Jargon Discuses: Locking.........2
Degrees of Freedom, or MDOF. In CMS the
same reduction is done but it includes the any models of rockets, airplanes or cars that
we can share with the world, so we made a Improvements to PCG Solver.........4

www.padtinc.com 1 1-800-293-PADT
 December 21,2005 The Focus Issue 43
(CMS, cont.) Comparison of Modal Results:
Also, when that NASTRAN guy emerges Full Model vs. CMS Model
from his cubicle, let him know he should be Full Bug CMS Bug Delta Delta
looking for a nice retirement community. In (Hz) (Hz) (Hz) (%)
a couple more releases ANSYS will have 40.52 40.41 -0.11 -0.27%
even stronger and easier to use capabilities in 65.44 65.11 -0.33 -0.50%
this area. 75.94 75.77 -0.17 -0.23%
85.71 85.42 -0.28 -0.33%
86.51 86.35 -0.16 -0.18%
101.5 101.24 -0.25 -0.25%
102.99 102.72 -0.28 -0.27%
111.26 111.04 -0.23 -0.21%
111.4 111.18 -0.22 -0.20%
113.75 113.62 -0.13 -0.11%
120.03 119.85 -0.17 -0.15%
122.57 122.37 -0.21 -0.17%
125.58 125.31 -0.27 -0.21%
- ANSYS Workbench Integrated With CoCreate CAD 129.77 129.52 -0.25 -0.19%
131.71 131.21 -0.5 -0.38%
- ANSYS and VISTAGY Form Strategy for Composites 143.16 142.82 -0.35 -0.24%
150.69 150.45 -0.24 -0.16%
- ANSYS Releases new AWQA and ASAS Versions 171.04 170.51 -0.53 -0.31%
News 175.51 175.25 -0.27 -0.15%

FE Term of the Month: Locking

By: Professor F. E. Jargon must ensure that you have eliminated both forms Nonlinearities training class.
Hello friends. I apologize for my extended ab- of locking. Volumetric locking can affect a plastic or hyper-
sence. Things get pretty busy here at the jargon Shear locking is the first form of locking I’ll dis- elastic analysis. ANSYS calculates hydrostatic
school on occasion. Just because I’m out for a cuss here. Imagine a four-node rectangular ele- and deviatoric stresses and combines them to pro-
couple of months doesn’t ment (e.g. SOLID45). Bend the element inward duce the total stress state of the model. The prob-
mean I’m gone — oh no. so that it forms a trapezoidal shape as shown in lem occurs when calculating hydrostatic stresses.
I do appreciate your let- Figure 1. A “real life” object undergoing this sort When you have plasticity or hyperelasticity,
ters of concern, although of bending is going to experience curvature along you’re dealing with a material that is mostly in-
I’m wondering how its top and bottom surfaces. compressible, i.e. your Poisson’s ra-
some of you got my However, with the four-noded tio is approaching 0.5. When this
phone number. As to the quad, the top and bottom sur- happens, the equation used to calcu-
person who showed up faces remain straight, inducing late von Mises strains approaches a
on my front patio at three artificial shear stresses along zero denominator, resulting in mathe-
in the morning, obvi- the left and right boundaries. matical errors in the stress calculation
ously inebriated—not For small deflections, this isn’t and, again, over-stiffening of the
cool. much of an issue. However, model. To check for the existence of
A logical follow up to my last article is the topic of with large deflections, shear volumetric locking, plot the hydro-
locking, by which I mean shear locking and volu- locking can result in over-stiff- static stresses and look for a checker-
metric locking. Both of these items, while trivial ening of the model and there- Figure 2: Hydrostatic Stresses board pattern as shown in Figure 2.
as far as a linear analysis is concerned, become fore under-conservative To remedy this, consider using
quite critical when one is performing a nonlinear stresses. To alleviate this problem it is wise to Mixed u-P formulation (where ‘u’ represents dis-
analysis. To neglect the possibility of either form choose elements with midside nodes or changing placements and ‘P’ represents hydrostatic stresses
of locking can give you erroneous results. In fact, the element formulation. Using the 18x series is or pressures), so that hydrostatic pressures are
if locking is present in your model, your results highly recommended. These elements have five calculated as a degree of freedom, rather than
will be wrong. If you are performing analyses different forumulations available, B-Bar(default being a derived quantity. You may also consider
with material or geometric nonlinearities, you for low order elements), URI (default for high- the other element formulations, with them each
order elements), Enhanced Strain, Simplified En- having pros and cons.
hanced Strain, and
I hope that you’ve not only received a vocabulary
Mixed u-P. Learn
lesson but also some tips on performing a better
more about these op-
nonlinear analysis. For even more nonlinear analysis
tions in PADT’s Ad-
help, check out the Focus archive for numerous arti-
vanced Structural
cles related to nonlinear modeling or consider taking
Figure 1: Shapes that Cause Locking
a training course.

www.padtinc.com 2 1-800-293-PADT
 December 21,2005 The Focus Issue 43
APDL Macros for BUG CMS Example Problem
!======================= k,2,x2,y3,z2 lmesh,lg1
!RUNBUG.MAC k,3,x2,y2,z1 secnum,304
!======================= k,4,x2,y1,z2 lmesh,lg2
! Creates BUG full and CMS models kgen,2,1,4,,x3-x2 lgen,3,all,,,x3-x2
! No Args kgen,2,5,8,,x4-x3 cm,eleg,elem !Save as component ELEG
! *do,i,1,12,4 esel,all
finish l,i,i+1 ksel,s,,,1
!------------ FULL BUG MODEL --------------- l,i+1,i+2 nslk,s,1
/clear !start with an empty model l,i+2,i+3 cmsel,a,nmst
/file,fullbug l,i+3,i cm,nmst,node
mkbug !Run macro that builds model *enddo allsel
save *do,i,1,4 /vup,1,z
/solu !Run a standard modal on the l,i,4+i /view,1,1,1,1
full model l,4+i,8+i /eshape,1
antype,modal *enddo Eplot
cpintf,all sectype,1,beam,csolid
modopt,lanb,30,0,20000,,off secdata,dd1/15,5,5 !==========================
mxpand,30,,,1 et,1,188 ! DO_SUB.MAC
solve lmesh,all !==========================
save cm,ebody,elem ! Save as component ebody ! Generic maro to make a substrcture based on
finish esel,u,,,all the name in ARG1
/post1 !----------------- Make Head ! Master DOF's must be defined in component NMST
wrt_frq,'fullbug' !Write frequencies to a file ksel,s,,,1,9,8 ! Elements must be defined in component E%ARG1%
so we can compare ksel,a,,,2,10,4 /filnam,%arg1%
/prep7 !Grab CP'd nodes, they are ksel,a,,,4,12,4 /solu
the interface nodes. nslk,s,1 antype,substr ! Going to do a substructure
nsel,s,cp,,1,9999 cm,nmst,node seopt,%arg1%,2 ! Sepcify the name (arg1) and
cpdel,all,all k,101,x4,y1,z2 that mass matrix should be there
cm,nmst,node ! make a comp called NMST k,102,x4,y2,z1 cmsopt,fix,30 ! Use fixed BC's on
finish k,103,x4,y3,z2 interface, 30 modes
do_sub,'body' ! Build substructure of each k,104,x5,y2,z4 cmsel,s,e%arg1% ! Select the element
part of the model k,105,x5,(y2+y1)/2,z5 componenet
do_sub,'head' k,106,x5,(y3+y2)/2,z5 nsle
do_sub,'tail' k,107,(x4+x5)/2,y2,z2 cmsel,r,nmst
do_sub,'leg' l,101,104 m,all,all
/clear ! Now use the substructures bsplin,102,107,104 nsle
to solve l,103,104 solve
/filnam,bug_use l,104,105 finish
/prep7 l,104,106 Save
et,1,matrix50 ! define substructure ET sectype,101,beam,csolid
se,body ! Include each substructure secdata,dd1/15,5,5 !===================
se,head secnum,101 ! DO_EXP.MAC
se,tail lmesh,all !===================
se,leg cm,ehead,elem !Save as component EHEAD ! Expands the mode shapes (7-30) for the
cpintf,all ! Couple the interface nodes esel,u,,,all substructured parts
finish ! Do a normal modal analysis !------------------Make Tail ! ARG1 is the substructure root name
/solu ksel,s,,,101,103 /filnam,%arg1%
antype,modal nslk,s,1 resume
modopt,lanb,30,0,20000,,off cmsel,a,nmst /solu
mxpand,30,,,1 cm,nmst,node expass,on
solve k,201,x2,y2,z3 seexp,%arg1%,bug_use
finish k,202,x1,y2,.8*z4 *do,i,7,30
/post1 k,203,x1,y2,.75*z5 expsol,1,i
wrt_frq,'cmsbug' ! Write the freq's out to a k,204,x2,y2,1.5*z5 solve
file bspl,201,202,203,204,,,1,0,0,1,0,0 *enddo
save finish
finish !---- Now go to each part and sectype,201,beam,csolid !=======================
expand secdata,dd1/10,5,5 !WRT_FRQ.MAC
/clear ! the mode shapes sectype,202,beam,csolid ! par1 = root name for file
do_exp,'body' secdata,dd1/30,5,5 !=======================
finish $/clear sectype,203,taper ! Simple utility macro to output frequencies
do_exp,'head' secdata,201,kx(201),ky(201),kz(201) *get,mxstp,active,,set,nset
finish $/clear secdata,202,kx(204),ky(204),kz(204) *cfopen,%arg1%.txt
do_exp,'tail' secnum,203 *vwrite,mxstp
finish $/clear lmesh,all %g
do_exp,'leg' cm,etail,elem !Save as component ETAIL *do,i,1,mxstp
!------- Time to plot the whole model esel,u,,,all set,next
/post1 !----------- Make legs *get,frq,active,,set,freq
cmsfile,add,body,rst ! Specify the expanded ksel,s,,,1 *vwrite,frq
mode shapes nslk,s,1 %g
cmsfile,add,head,rst ! for each part in the cmsel,a,nmst *enddo
model cm,nmst,node *cfclose
cmsfile,add,tail,rst k,301,x2,y1,z2
cmsfile,add,leg,rst k,302,x2,y1*2,z1 !===================
plt_mshp ! Make plots k,303,x2,y1*2,z1*4 ! PLT_MSHP.MAC
k,304,x2,y3,z2 !===================
!======================= k,305,x2,y3*2,z1 ! Simple macro to plot out modeshapes
!MKBUG.MAC k,306,x2,y3*2,z1*4 allsel
! Builds bug model. Change values below to lsel,u,,,all /plopt,info,2 $/plopt,leg3,off $/plopt,minm,off
change size bspl,301,302,303 /eshape,1,
! No Args cm,lg1,line /win,1,ltop $/win,2,rtop $/win,3,lbot
!======================= lsel,u,,,all $/win,4,rbot
/prep7 bspl,304,305,306 /view,1,0,0,1 $/vup,1,y
! Define locations cm,lg2,line /view,2,1,0,0 $/vup,2,z
y1 = -2 $y2 = 0 $y3 = 2 sectype,301,beam,csolid /view,3,0,-y,0 $/vup,3,z
x1 = -10 $x2 = -8 $x3 = -4 $x4 = 0 $x5 = 4 secdata,dd1/10,5,5 /view,4,1,1,1 $vup,4,z
z1 = -1 $z2 = 0 $z3 = 1 $z4 = 2 $z5 = 4 sectype,302,beam,csolid *get,mxstp,active,,set,nset
! Mat props secdata,dd1/25,5,5 *do,i,1,mxstp
ex,1,10e6 sectype,303,taper set,i
nuxy,1,.23 secdata,301,kx(301),ky(301),kz(301) plnsol,u,sum,2
dens,1,.001 secdata,302,kx(303),ky(303),kz(303) *enddo
dd1 = (z3-z1)/2 !Set a size variable, DD1 to 1/2 sectype,304,taper
the body height secdata,301,kx(304),ky(304),kz(304)
esize,dd1/2 !Set element size to be 1/2 DD1 secdata,302,kx(306),ky(306),kz(306)
!-------------- Make Body secnum,303
k,1,x2,y2,z3 cmsel,a,lg1

www.padtinc.com 3 1-800-293-PADT
 December 21,2005 The Focus Issue 43

PCG Solver Improvement in V10.0

By: Rod Scholl other than testing) use the PCGOPT,1 command. The default is
With all the speed en- PCGOPT,0 which lets ANSYS decide the stability vs. solution time
hancements over the dilemma, and in my testing seemed the best way to go. But if you
last couple years to the get into a particularly poorly conditioned problem try PCGOPT,4 –
solution phase, solve I suspect it would still give the sparse solver a run for its money.
time has become Thus my estimate is the above error message will quickly be a thing
much less of a consideration when doing FEA – and that is really of the past.
sayin’ something.
Previously I used PCG solver as a rule, and only used the Sparse for
the few types of analyses I did that required it – and the occasional Want to learn more about the program-
analysis where the number of PCG iterations exceeded 1000; on ming language used to make the ANSYS
about 1 in 5 analyses I also would switch to sparse after getting this GUI? Want to customize the GUI?
message that perhaps you are familiar with. Occasionally changing www.padtinc.com/support/ansys/tcltk
the MULT option on EQSLV would help as it suggests… but usually
not. Links Join the growing Workbench user
It could be the type of analyses I did in 2004-2005 were mostly community on the ANSYS, Inc.
ill-conditioned – or maybe the continued speed increases of element customer portal.
formulation and other steps just made me pleased to be solving faster www1.ansys.com/customer
than last year and I abandoned the 1.5X-3X benefit I usually saw Log in and navigate to the WB
using PCG over the Sparse. I do know that there was just something Community
disturbing about watching the iterations climb to 1500+, and then
debate whether to bail out now and switch to sparse or continue Upcoming Training Classes
waiting for the PCG to handle the poorly conditioned problem. I Month Start End # Title Location
don’t like such weighty decisions… I suppose I began to avoid the Jan '06 17-Jan 19-Jan 104 ANSYS Workbench Simu- Irvine, CA
lation - Intro
dilemma by leaving off the EQSLV,PCG command… 20-Jan 20-Jan 105 ANSYS Workbench Sim Irvine, CA
Struct Nonlin
However, in 10.0 the PCG solver has been made even more robust.
26-Jan 27-Jan 801 ANSYS Customization Tempe, AZ
I’ve done some testing with it, and I am again switching back to it as with APDL
the preferred solver. (It’s really a matter of choosing between two Feb '06 6-Feb 8-Feb 101 Introduction to ANSYS, Irvine, CA
great choices, which is Part I
a nice luxury.) 9-Feb 10-Feb 202 Advanced Structural Non- Tempe, AZ
linearities
It took me a little 13-Feb 15-Feb 104 ANSYS Workbench Simu- Tempe, AZ
lation - Intro
foolin’ around to make
16-Feb 16-Feb 105 ANSYS Workbench Sim Tempe, AZ
a macro of an ill-condi- Struct Nonlin
tioned problem. This 22-Feb 23-Feb 301 Heat Transfer Tempe, AZ
macro fails during Mar '06 6-Mar 8-Mar 101 Introduction to ANSYS, Albq, NM
PCG iterations in ver- Part I

sion 9.0 – but works 9-Mar 10-Mar 102 Introduction to ANSYS, Albq., NM
Part II
like a charm in 10.0. If 16-Mar 17-Mar 203 Dynamics Tempe, AZ
you want to revert to 23-Mar 24-Mar 501 ANSYS/LS-DYNA Irvine, CA
9.0 behavior (I’m not 27-Mar 29-Mar 902 Multiphysics Simulation Tempe, AZ
sure for what reason for MEMS

PADT’s “ANSYS Customization with APDL” is a great way to learn all about macro writing for
ANSYS

Need Material Properties? MatWeb is a huge online database of material properties that now
Resources supports ANSYS file formats.

The Focus is a periodic publication of Phoenix Analysis & 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: http://www.padtinc.com/epubs/focus/about

www.padtinc.com 4 1-800-293-PADT