Learning Forum

[Kevin Bergmann]

Hello,

I am currently facing problens in trying to connect the nodes of a superelement to my model. Envisioned is to define a pair of contact-target-elements at the interface between the superelement and the rest of my model and to connect them vie a shared real constant set in order to create a bonded contact. Working with a superelement forces me to define the contact using APDL commands only. I am using ANSYS 2023 R1.

So far I am doing the following:

-----------------------------------------------------------

/Prep7

[import superelement]

R,1300                                                     !define real constant set for contact-target-pair (highest set-number +1)
REAL,1300                                              !activate newly defined set
RMODIF,1300,6,-0.0075                         !manually overwrite the current pinball radius (act of desperation)

ET,10172,CONTA174,,2,,1                       !Define contact-element type as MPC, force-distributed
KEYOPT,10172,12,5                                !Behavior of contact surface: always bonded
TYPE,10172                                             !Activate current element type
NSEL,,,,ContactNodes                             !Selection nodes for contact elements via named selection containing the nodes of the contact surface
                                                                 !Selection generated by selecting contact surface in GUI and convert it into nodal selection using worksheet tool
                                                                 !All nodes are present as anticipated (8 nodes per element)
ESURF,,BOTTOM                                    !Generate contact elements - I can see them being generated in the APDL GUI
                                                                 !Also, I find them in ELIST with REL = 1300, TYP = 10172
ET,10173,TARGE170                              !Define target-element type
TYPE,10173                                            !activate target element type
TSHAP,QUAD                                         !targets shall use only four nodes, choosing four-node quadrilateral (3-D) definition
/INPUT,target_elements,inp                    !import target elements
                                                                !target elements are simply defined as "E, Node#1, Node#2, Node#3, Node#4" from superelement nodes
                                                                !I also verified that the import works by using ELIST - elements are shown, belonging to REL = 1300, TYP = 10173
ALLSEL

[apply loads on superelement]

FINISH
SAVE

-----------------------------------------------------------

As to my knowledge, after defining the pair, Ansys should check whether there are contact-target pairs present sharing the same real constants sets and if so it should couple their degrees of freedom which should again be visible by using CPLIST command. However when I do so, the list is empty ("No coupled sets to list").

The result of this is that my model behaves unphysical: apparently torsion and bending torqures are transferred between superelement and the rest of the model, but tensive forces only result in the contact surface being distorted, whereas the rest of the model shows stresses close to zero. As soon as I remove the superelement and apply a tensive force on the contact surface directly, the model works without issue - therefore I presume the error lies in the contact definition described above.

Does any of the here present ladies and gentleman maybe have a solution for my troubles - or at the very least an idea where else I could look for one?

Best regards.

[wrbulat]

Hi Kevin,

I think I would need a few more details from you in order to setup a relevant test case that would allow me to devise a setup procedure that I could then share with you. 

Are you using CMS (component mode synthesis) superelements, or are you instead using the older legacy superelements created via Guyan reduction? Note that the latter is often a poor choice for structural dynamics due to the approximate representation of the distribution of mass. Guyan reduction is fine however in static analyses. Is your intended use pass analysis type static or dynamic (e.g., harmonic response or transient)? 

Are you intially modeling in Mechanical (defining "condensed geometry" and partially defining contact in that environment), and then reading the ds.dat input file Mechanical generates into MAPDL where you are attempting to complete setup? Or are you instead working completely in MAPDL from start to finish?

Thanks!

Bill

[Kevin Bergmann]

Dear Bill, 

Thank you very much for your answer. I am indeed using a CMS superlement. It is intended to contain the load attachment point for our simulations, and transfer bending torques, torsion torques and longitudinal forces into structure it is attached to. The background is a static analysis (for dynamics we do not regard the superelement - the model also works as intended in that case).

Initally I am building the 3D-FEM in Mechanical (first importing a CATIA-CAD file, defeaturing it in Spaceclaim and then editing the contact and solution settings for the 3D-FEM in Mechanical). Prior to entering the Solution however, I inserted a "Commands" element where I insert the APDL-script from my original post. The SE is imported using the following commands:

--------------------------------------

LOCAL,120000,0,7.000,0,0                       !generate new local coordinate system for SE integration. X coordinate adjusted to ensure close proximity of SE and 3DM nodes/surfaces 
CSYS,0                                                      !(Re)activate global Coo-Sys (this command might be superfluous) 
SETRAN,'CMS',120000,,'CMS_1','sub'     !Generate a copy of the original SE, where the SE is moved to the origin of coordinate system 120000
ET,1500,MATRIX50                                   !SE-Elements will be of type 1500 
TYPE,1500                                                !Set current type to 1500
SE,'CMS_1'                                               !Import updated SE
CSYS,0                                                     !(Re)activate global Coo-Sys (this command might be superfluous)
ALLSEL

--------------------------------------

This segment is followed by the contact definition stated in my original post and followed by the load application. Not sure what files Mechanical imports, but when cross-checking my scripts in MAPDL, I always import a file.db generated from executing my 3D-FEM with the commands block deactivated (if I find my scripts working I then enter the correct command sequence in the COmmands element and activate it in Mechanical.

As however I am unable to use the SE inside of the Mechanical GUI (at least as to my knowledge) I am limited to define the contact between SE and 3D-FEM in MAPDL completely. 

 

Generally asking: Is the contact definition from my original post correct per se? Or can you already spot an issue with the provided code?

I am asking, because I received the SE from a third party who are using NASTRAN themselves. I am not sure whether this issues may be the reason for some issues. Before trying out the appraoch above, my first idea was to directly link the nodes from SE and 3D-FEM using the CPINTF command and linking their UX, UY, and UZ degrees of freedom. This approach worked fine for longitudinal forces, but caused an underestimation of bending and torsion torques (their maximum stresses turned out to be lower than the values that you would expect using analytical formulae). Therefore I am pusruing the idea of creating a surface contact and let Ansys handle the force and torque transfer directly.

Best regards.

[wrbulat]

Correction: there was a typo in my previous version of the LINK180 test case (I misspelled SECCONTROL). Sorry for the confusion. In the version pasted below, the compression in the forward cable appears to have disappeared (it went slack, as intended). So while I don't know why it appears to not work for you, I can at least provide you with an example in which SECCONTROL,,1 does work. By the way, I still do see a problem with SECCONTROL,,,,,CV3 for CABLE280 elements - I have so far not seen it affect compresion/tension stifness ratio.

fini

/cle

 

/vup,1,z

/vie,1,3,2,1

/esha,1

 

 

C********************************************

C*** PARAMETERS

C********************************************

pi=acos(-1)

 

h_tower=25 ! TOWER HEIGHT

a_tower=0.1 ! TOWER SQUARE CROSS SECTION EDGE LENGTH

E_tower=2e11 ! TOWER ELASTIC MODULUS

nu_tower=0.3 ! TOWER POISSON'S

 

n_cables=3 ! # OF CABLES

h_cable=15 ! HEIGHT OF CABEL ATTACHMENT TO TOWER

r_cable=10 ! RADIUS OF CABLE ANCHORS TO GROUND

rcross_cable=0.01 ! RADIUS OF CABLE CROSS SECTION

E_cable=2e11 ! CABLE ELASTIC MODULUS

nu_cable=0.3 ! CABLE POISSON'S

 

esz=1 ! MESH SIZE

 

FX_top=1000 ! X COMPONENT FORCE AT TOP OF TOWER

 

 

C********************************************

C*** MODEL

C********************************************

/prep7

 

et,1,188 ! TOWER ATTRIBUTES

mp,ex,1,E_tower

mp,nuxy,1,nu_tower

sect,1,beam,rect

secd,a_tower,a_tower

 

et,2,180 ! CABLE ATTRIBUTES

mp,ex,2,E_cable

mp,nuxy,2,nu_cable

sect,2,link

secd,pi*rcross_cable**2

seccontrol,,1 ! TENSION-ONLY

 

lsel,none ! TOWER GEOMETRY

k,1

k,2,,,h_cable

k,3,,,h_tower

l,1,2

l,2,3

dk,1,ux

dk,1,uy

dk,1,uz

dk,1,rotz

fk,3,fx,FX_top

latt,1,1,1,,,,1

 

lsel,none ! CABLE GEOMETRY

csys,1

*do,i,1,n_cables

 k,3+i,r_cable,(i-1)*(360/n_cables)

 l,3+i,2

 dk,3+i,ux

 dk,3+i,uy

 dk,3+i,uz

*enddo

 

latt,2,2,2,,,,2

 

alls ! MESH

esiz,esz

lmes,all

 

fini

 

 

C********************************************

C*** SOLVE

C********************************************

/solu

acel,,,9.81

nsub,5,25,5

outr,all,all

nlge,on

pivc,off

alls

save

solve

 

 

C********************************************

C*** POST PROCESS

C********************************************

/post1

/pbc,rfor,,1

/pbc,f,,1

/esha,10

/vsc,1,2.5

plns,u,x

 

/eof

 

 

[Kevin Bergmann]

Dear Bill,

The answer provided above seems to belong to a different post, if I'm not mistaken?

Best regards.