Learning Forum

[Daniel Dabic]

Hello, 
i got to build a digital twin for a milling process. My first idea was to build it with explicit dynamic. But i'm unsure if this ist the right choice, because i have to simulate multiple lines and the whole time period is lasts over a few seconds. I saw somewhere that the explicit method is used for periods maximum to 1s. 
Does someone has a solution for my problem or a good hint? 
Thank you!

kind regards,
Dani

[Reno Genest]

Hello Daniel,

You can simulate milling with LS-DYNA. Here is an example:

https://www.lstc-cmmg.org/grinding

You can click in numbers and letters between brackets for more information and to download the model:

 

This simulation too 15 hours on 48 CPUs.

 

Let me know how it goes.

 

Reno.

[Daniel Dabic]

Hey Reno,

thanks for response. Are you sure that it's possible to simulate time periods >1s with ls-dyna? I saw that it behaves similar to the explicit periods.

kind regards,
Daniel 

[Reno Genest]

Hello Daniel,

LS-DYNA has an explicit and an implicit solver. In explicit, it is possible to simulate events longer than 1 second; we have customers simulating 30s or longer events, but they run on hundreds of cores. Also, it depends on element size (which determines the timestep) and your patience :).

With SPG, we have done a simulation of a screw in wood, metal, and concrete that lasted 2.4 seconds or more. With these problems, we typically increase the density of the SPG material quite a bit and it does not seem to affect the results too much. You can do a parametric study where you increase the density of the SPG material, record the computation time, and your results of interest. Does increasing the SPG material density affect your results of interest for your model? Then, you can judge by how much you can increase the SPG density and still get good results.

Also, with SPG, we typically distribute the SPG part to many CPUs using the following keyword:

*CONTROL_MPP_DECOMPOSITION_ARRANGE_PARTS

In LS-PrePost, it looks like this (this is an example; it might look different for your model):

Have a look at the LS-DYNA user manual Vol I for more information on this keyword.

The default decomposition for MPP LS-DYNA might put the entire SPG part on only 1 CPU which is not effective. This is why we use the above keyword to make sure the SPG part is being distributed on many CPUs for faster solve time.

 

Let me know how it goes.

 

Reno.

 

[Daniel Dabic]

Hey Reno,

thank you! Maybe would some background informations be helpfull... I'm a student at a university and i got the project to implement our ansys into our HPC. I'm going to find out if it works next week. The big goal is, to build up a simulation for milling processes. I tried this with LS-Dyna and with explicit Dynamics. I think the difference between this models is not that big right? The Problem is that its not working... do you have some ideas or things i have to pay attention? 
I have to compare the ls dyna results with the explicit dynamic results. Is it possible to calculate longer than 1 s with explicit dynamics? 

Kind regards,
Daniel 

[Reno Genest]

Hello Daniel,

Yes, you can simulate longer events than 1 second with explicit dynamics. The limit of 1second is a general guideline, but if you have the computer power, you can certainly simulate longer than 1 seconds with explicit dynamics (LS-DYNA). In our training, we have the following slide:

We have customers simulating events up to 30-40 seconds with LS-DYNA explicit dynamics.

 

I don't understand when you say "I have to compare the ls dyna results with the explicit dynamic results.". LS-DYNA is an explicit dynamics code. You have results from another explicit dynamics code (or solver) and you want to compare these results with LS-DYNA?

You say the simulation is not working. What do you mean by it is not working? Do you get an error message? The results are not what you are expecting? Do you have experimental results to compare? 

 

Reno.

[Daniel Dabic]

Hello Reno,

okay perfect! 

The task is, to compare the analysing systems, LS-DYNA and Explicit Dynamics, from Workbench. But i think this is unneccecary because both tools are working with the same systems, right? 

Yes some times i get the error message "Energy error too large". But i don't understand why, because the only thing i changed regarding a previous model is the rotation and the  end time. (rotation from 90° to 360°).
Another error is "time step too small". Then i changed the time step to a larger value but it's still the same error. 
Here are some pictures of my model:
 
So in general we could say that, if i set up a rotation >90° and a displacement >0mm i get a error message. 
What am i doing wrong?

Kind regards,
Daniel

[Reno Genest]

Hello Daniel,

I recommend to use LS-DYNA in Workbench. The 2 systems (LS-DYNA and Explicit Dynamics) don't use the same solver. The LS-DYNA system uses the LS-DYNA solver while Explicit Dynamics uses the Autodyn solver.

 

Let me know how it goes.

 

Reno.

[Daniel Dabic]

Hello Reno,

i tried it with LS-Dyna but in the results there is something missing. The workpiece is behaving like rubber, so there is no material seperation. Is there a option where i can turn on the failure on material strain as in explicit dynamics? Sadly i couldn't find anything like this... 

 

Dani

 

[Daniel Dabic]

Hey Reno,

do you have some tips for solving this problem?

Kind regards,
Dani

[Reno Genest]

 

Hello Daniel,

Material failure is activated at the material level in LS-DYNA. Go to Engineering Data and add a “Plastic Strain Failure” to your material:

 

Note that material calibration is very important to model failure properly. Have you done material testing? Do you have stress-strain curve? Do you know the plastic strain at failure for your material? Plastic strain is the most common failure criterion for metals. 

Let me know how it goes.

 

Reno.

 

[Daniel Dabic]

Hey, 

okay i am going to try this, but i already have johnson cook failure model in the properties. Shouldn't it work with this model too?

Kind regards,
Dani

[Reno Genest]

Hello Daniel,

 

Yes, the Johnson-Cook failure model should work. Can you show a screenshot of your Johnson-cook material model defined in Engineering Data with all the details?

Also, if you open the input.k file (generated by Mechanical) in a text editor, do you see your Johnson-Cook material model properly defined? 

 

 

Reno.

[Daniel Dabic]

Hello Reno,


so far so good i would say. Then i took a look in the input file and i'm not sure if the material model ist properly difined. Is this the right spot to look?


Dani