wrong result due to cantact set
help!help!help! I need your help.
when I used Explicit solver in Radioss to solve large deformation of Fabric .
I set MAT=LAW19 、N=1 in Pshell=TYPE9 (SH_ORTH) and set contact=type7 like this.
result :
the Slave nodes group and Master surface are both the component itself.
how could I set the interface to fix this wrong?
Answers
-
hi zifan,
i'm sorry i can't help you so much here, but anyway, can you please detail better what exactly is going wrong in here?
It would be nice if you describe a litte further what you're trying to do, what do you consider it's not ok, and so on.
Hope others can help you better.
0 -
Adriano A. Koga_21884 said:
hi zifan,
i'm sorry i can't help you so much here, but anyway, can you please detail better what exactly is going wrong in here?
It would be nice if you describe a litte further what you're trying to do, what do you consider it's not ok, and so on.
Hope others can help you better.
Thank you so much for getting back to me.
Because I don't know much about Radioss and FEA.
What I am doing now is the analysis of the unfolding process of the umbrella cloth.
I set the contact type as type 24(surface-to-surface contact ),but the calculation is too slow. It spends too much time on CONTACT SORTING.
0 -
zifan kuang said:
Thank you so much for getting back to me.
Because I don't know much about Radioss and FEA.
What I am doing now is the analysis of the unfolding process of the umbrella cloth.
I set the contact type as type 24(surface-to-surface contact ),but the calculation is too slow. It spends too much time on CONTACT SORTING.
what are your units?
40 seconds of runtime? Looks a lot for explicit models. Or is it 40ms?
I would suggest you to go through the Radioss ebook in Altair University as wel.
https://altairuniversity.com/free-ebooks/free-ebook-crash-analysis-with-radioss-a-study-guide/
0 -
Adriano A. Koga_21884 said:
what are your units?
40 seconds of runtime? Looks a lot for explicit models. Or is it 40ms?
I would suggest you to go through the Radioss ebook in Altair University as wel.
https://altairuniversity.com/free-ebooks/free-ebook-crash-analysis-with-radioss-a-study-guide/
My units is (g mm ms).The total runtime is 130ms.
I just took a screenshot between 30ms and 40ms.
The contact sorting takes too much time.
this screenshot is between 50ms and 60ms.
Maybe I set the range of contact surface too wide
0 -
zifan kuang said:
My units is (g mm ms).The total runtime is 130ms.
I just took a screenshot between 30ms and 40ms.
The contact sorting takes too much time.
this screenshot is between 50ms and 60ms.
Maybe I set the range of contact surface too wide
how many core's do you have available, or how many are you using?
Explicit analysis can paralelize quite well in general.
In order to speed up computations, you could try:
- using more core's for running
- forcing mass scaling to increase the timestep, while watching/controlling for additional mass
- increase mesh size to get a larger timestep
0 -
Adriano A. Koga_21884 said:
how many core's do you have available, or how many are you using?
Explicit analysis can paralelize quite well in general.
In order to speed up computations, you could try:
- using more core's for running
- forcing mass scaling to increase the timestep, while watching/controlling for additional mass
- increase mesh size to get a larger timestep
0 -
zifan kuang said:
as you mentioned before, the contact sorting is taking most of the time, maybe playing with the GapMin and Igap settings could help speed-up this part of computations.
i'm not quite sure of the option -nt.
Are you running in a cluster? 108 core's is a lot.
0 -
Adriano A. Koga_21884 said:
as you mentioned before, the contact sorting is taking most of the time, maybe playing with the GapMin and Igap settings could help speed-up this part of computations.
i'm not quite sure of the option -nt.
Are you running in a cluster? 108 core's is a lot.
CPU: 56 X 2
My options:
To speed-up this part of computations, may I increase or decrease the GapMin ?
0 -
zifan kuang said:
CPU: 56 X 2
My options:
To speed-up this part of computations, may I increase or decrease the GapMin ?
parallelization has some limits.
I believe the rule of thumb was around 5k elements per core.
If you increase too much the core count, the solver will spend more time exchanging information, than running the model.
Also, check if you have 108 physical core's or if it's HT (56 physical with 2 core's each).
You should rely on your physical count.
I'm not an expert in Radioss contact models, but I believe by reducing the GapMin you have less contact candidates, BUT you might loose some contacts, if it is too small. The gapMin is the distance where contacts will be monitored, if I recall it right.
I believe it is described in the help.
0 -
Adriano A. Koga_21884 said:
parallelization has some limits.
I believe the rule of thumb was around 5k elements per core.
If you increase too much the core count, the solver will spend more time exchanging information, than running the model.
Also, check if you have 108 physical core's or if it's HT (56 physical with 2 core's each).
You should rely on your physical count.
I'm not an expert in Radioss contact models, but I believe by reducing the GapMin you have less contact candidates, BUT you might loose some contacts, if it is too small. The gapMin is the distance where contacts will be monitored, if I recall it right.
I believe it is described in the help.
Maybe I should send you my model to make you understand the problem I'm talking about better .
Can I get your email address?Maybe you can send me a hello via my E-mail（1961035822@qq.com）.Because my model file is 600Mb.
0 -
zifan kuang said:
Maybe I should send you my model to make you understand the problem I'm talking about better .
Can I get your email address?Maybe you can send me a hello via my E-mail（1961035822@qq.com）.Because my model file is 600Mb.
Hi,
Honestly, i believe it would be better to try get in touch with some specialist from your closest Altair representative and ask them for support.
I don't have the expertise to help you, neither the resources to run this model in here.
0
Altair Employee