Learning Forum

[dvirnyk]

Hello,

 

I am simulating twist extrusion (helical/ECAP-type die) of Ti powder using SPG

(ELFORM 47) with LS-DYNA R16.1.1 MPP double precision, 12 cores.

 

MODEL

- Billet: SPG solid, MAT_CSCM (cap model), relative density 0.70,

  rate effects disabled, softening disabled.

- Die: rigid shells; contact via *DEFINE_SPG_TO_SURFACE_COUPLING (SPGP=powder,

  SURF=die segment set).

- Helical section: cross-section rotates ~64 deg over 18 mm of channel length.

- Loading (both tried): (a) pressure on null-shell skins on the billet end faces,

  (b) prescribed velocity through a soft elastic spacer merged to the SPG nodes.

- Back pressure 200 MPa on the exit face in both cases.

 

PROBLEM

The run is perfectly healthy for a long time: time step stays constant

(~1.2e-7 s), KE/IE ~ 0, internal energy grows smoothly, no erosion at all.

Then, within less than 1 ms, an entire transverse layer of background elements

inverts (negative volume), and the job dies with either

 

    "gt ispgnumextn"

 

or a hard crash:

 

    forrtl: severe (157): Program Exception - access violation

    SPGBSORT (spgbsort.f) / SPGMGBUCKET_SMP / SPGFINOD3 / SPGFINOD / SPGSMOOTH

 

Checking the deleted nodes, the first failures are all from one original element

row, located at the EXIT of the helical section (where the cross-section stops

rotating and the material "unwinds"), and they are spread across the whole

cross-section (r = 2...11 mm, channel radius 14.7 mm) - i.e. in the core, NOT at

the die wall.

 

SETTINGS ALREADY TRIED (none of them removes the failure, they only delay it)

- DX = DY = DZ: 1.5 / 1.6 / 1.8 / 2.0

- KERNEL = 1 (Eulerian); ISPLINE = 0

- SMSTEP: 15 / 5 / 3

- MSC = 1 together with ITB = 3

- IDAM = 1, STRETCH = 1e10 and 6

- PDAMP = -0.005 / -0.01, BOXID active

- Coupling SFP = 0.5 / 0.7, THK = 0.25 / 0.5, friction = 0 / 0.05 / 0.1

- CONTROL_TIMESTEP: TSSFAC 0.5, DT2MS = 0, ERODE = 1; DTMIN = 0.1

- Mesh size of the billet: 2.8 mm / 1.24 mm / 1.0 mm

- Drive speed 100 mm/s and 50 mm/s

 

QUESTIONS

1. Are there recommended SPG settings for large TORSIONAL/shear deformation of

   this kind (background mesh must follow a ~64 deg rotation)? Any guidance on

   DX and SMSTEP for this class of problem?

2. What controls the internal array dimension behind "gt ispgnumextn", and can

   it be increased (keyword/parameter/memory)?

3. For a helical rigid wall, is *DEFINE_SPG_TO_SURFACE_COUPLING the recommended

   approach, or is *CONTACT_AUTOMATIC_NODES_TO_SURFACE preferred?

4. Is *CONTROL_SPG required or recommended here? I have not used it so far.

5. For powder consolidation I do NOT want material separation. What are the

   recommended IDAM / STRETCH / FS settings for a "no bond failure" setup?

 

Any advice would be greatly appreciated.

[ErKo]

 

 

 

 

Hi

Not sure why, but also not sure you can use MAT_CSCM/159.


Checked now and actually the mat159 is ok with SPG so all good there.

Perhaps turn off erosion (control time step) – also wait for other members to provide more feedback.

All the best

Erko

 

 

 

 

[Dennis Chen]

can you check your control_termination and control_timestep?   dtmin and erode there may be what termianted the job - first to check. 

secondly, I am not sure ( need to check) concrete should be modelled wtih particle methods since you can't expect it to see really large deformations.

I am also not sure CSCM allow you to add erosion crtieria since tension is drastically lower than compression.   These are unknowns to me but maybe you can try starting simple, with just continuum elements, and turn off any termination condition for element becoming very skewed and see why that deformation is happening and if that makes sense. 

After that, you can turn on more complexity to somehow model the intended failures.