Learning Forum

[Amrith Mariappan]

Hi,

I'm modeling hypervelocity impact (2.22 km/s tungsten carbide sphere on HSLA-100 steel) with coupled thermal-structural analysis using S-ALE (Structured ALE) mesh.

Initial Attempt & Issue:

I first tried the implicit thermal solver using:

- *CONTROL_SOLUTION (Thermal-Structural Analysis)

- *CONTROL_THERMAL_NONLINEAR

- *CONTROL_THERMAL_SOLVER

- *CONTROL_THERMAL_TIMESTEP

However, I observed physically unrealistic temperature rates (dT/dt on the order of billions of °K), which is physically impossible. I switched to the Explicit Thermal Solver (*CONTROL_EXPLICIT_THERMAL_* cards) as an alternative, but encountered a fatal initialization error.

Current Error:

forrtl: severe (164): Program Exception - integer divide by zero

Image PC Routine Line Source

lsdyna_mpp_dp_imp 00007FF731E6B4CC XPLCTH_INIT 1148 dyn20x.F

...

The crash occurs during initialization (XPLCTH_INIT), suggesting a division by zero in the thermal data setup.

Specific Questions:

1. Explicit Thermal Setup with S-ALE:

   For cards like *DATABASE_ALE, *CONTROL_EXPLICIT_THERMAL_INITIAL, and *CONTROL_EXPLICIT_THERMAL_OUTPUT, the manual states SETID can be left blank for "all parts," but LS-PrePost/LS-DYNA requires an integer. Since S-ALE mesh elements are generated internally at runtime (not predefined in *ELEMENT_SHELL), I cannot reference them via traditional *SET_SHELL_LIST.

   - How do I properly define SETID for S-ALE domains in these cards? Should I use *SET_MULTI_MATERIAL_GROUP instead?

2. Phase Change Modeling:

   My material definition includes *MAT_THERMAL_ISOTROPIC_PHASE_CHANGE for the steel target (solid-to-liquid). Does the Explicit Thermal Solver support phase change energy (latent heat), or is this capability restricted to the implicit thermal solver? If supported, are there specific considerations for S-ALE multi-material groups?

3. Divide-by-Zero Cause:

   The error trace points to XPLCTH_INIT (dyn20x.F:1148). This typically indicates zero thermal mass (density × specific heat) in a thermal part definition. Given that my vacuum/void material (*MAT_ALE_VACUUM) has near-zero thermal properties, could this be triggering the crash? Should vacuum regions be excluded from the thermal solver via *SET_PART_LIST, or is there a minimum non-zero thermal density requirement?

Model Context:

- 2D Axisymmetric S-ALE mesh (fine resolution)

- 3 ALE Multi-Material Groups: HSLA-100 Steel, Tungsten Carbide, Vacuum

- Thermal properties defined via *MAT_THERMAL_ISOTROPIC and *MAT_THERMAL_ISOTROPIC_PHASE_CHANGE

- Using MPP R14.1.1 on Windows

Keyword file attached for reference:

https://buffalo.box.com/s/y7rz3bqg4xsvuir6z94b50mcg2g3jp1s

Any guidance on proper explicit thermal initialization with S-ALE and phase change setup would be greatly appreciated.

Thanks,

Amrith

 

[ErKo]

 

 

 

Hello

 

Ansys empl. can not download any files , but please wait for others (peteroznewman or DChen) perhaps to have a look and provide feedback.

Also it would be useful not to post or share any ITAR export controlled models/appl. – is this project of such nature?

 

In general *CONTROL_EXPLICIT_THERMAL_SOLVER: the elements supported by this thermal solver are beams, shells, solids, 
and multi-material 3D ALE elements. So not 2D ALE.

All the best

Erko

 

 

 

[Ian Do]

 

Hello Amrith,

You have provided good attention to details. In HVI (hyper velocity impact) problems, typically the event is over in the microsecond range. whereas if you look at the thermal diffusion characteristic time for metals, it may be in the 50-250 ms range (?). Most HVI problems ignore conduction for this reason, and I think that is a reasonable assumption, unless your impact event goes on for a few hundreds of ms. We can consider thermal softening in the metals via their mat+EOS models. Johnson-Cook, Steinberg-Guinan, … I think that has worked well in my experience. No phase change feature is generally available in S-ALE specfically. If exists, it likely resides in the mat and EOS models. 

Regarding the general thermal conduction consideration for ALE, I must admit that it is still very rudimentary. It is quite a logistic challenge to handle multi-material heat transfer. Here’s why. LAG parts has 1 elm containing the same mat. Each node of that elm can store 1 T. So we can store the mat T at the node easily. Now if 1 elm (say, 4 nodes, 2D) contains 3 materials, we still have 4 nodes, we can’t store the T of each mat readily and still tie the T’s to the nodes.   

Regards,

Ian Do

 

[Amrith Mariappan]

Dear Ian,

Thank you for the detailed explanation and for taking the time to clarify this. I really appreciate your insights.

Your comments about the timescale differences between hypervelocity impact events and thermal diffusion were very helpful and helped me better understand why thermal conduction is usually neglected in these simulations. From what I understand, because HVI events occur over extremely short timescales, they are often treated as nearly adiabatic processes, with temperature rise handled through the material model rather than conduction.

Your explanation about the limitations of thermal conduction in ALE/S-ALE due to multi-material cells also clarified a lot for me.

Thanks again for your guidance.

Best regards,
Amrith

[Ian Do]

Hello Amrith,

You are welcome! 

Ian