Plated structures and shell elements

Andrea

I'm working on a big plated structure modelled with shell elements.
I found that is very complicated to manage it in Mecway for the following reason:

To assign plate thickness is necessary to select faces and create a "new component". Generating a new mesh one has to delete the components creates before and repeat the procedure (select faces -> new component -> assign material)
I'm right or there is a faster procedure?

Another question: with the internal solver there is a simply law to estimate the required RAM linked to the number of nodes (or DOF)?

Because I need to buy more RAM

Andrea

I tried to get a solution changing the quad8 to quad 4. Mecway solver doesn't support tri-3 so the mesh is holed but with 100 knodes I get a solution in a few minutes. With the original mesh (quand8 + tri6) and about 300k nodes Mecway utilize all the RAM and after 1 hour is still on the "applying constraints" step. I read that with 16Gb is possible to solve until 1M nodes. I don't use generally the internal solver so I don't have a time estimation.
Tomorrow I will try with 32 Gb of RAM.

Victor

Here's some memory use data for solids:

Internal solver: https://mecway.com/forum/discussion/comment/1139/#Comment_689

CCX: https://mecway.com/forum/discussion/comment/3166/#Comment_3141

As for recreating components, I don't think there's a shortcut if it comes from a single STEP file. Only if it was multiple files would they automatically be assigned to different components. If you're not already doing it, you can assign the faces to named selections, then click on them to select them instead of doing it graphically.

Andrea

Thanks!
I saw that the problem is related to bonded contact. I have created a single step file and the solution time is acceptable with less memory usage:

About 300k nodes -> Linear static analysis with internal solver -> about 13 minutes with Ryzen7 2700x and 16 Gb RAM

I need another hint about graphical performance. In fact Mecway is slow on rotating the model, zoom etc. Does it depend from the GPU?

Victor

Graphics is mostly limited by the CPU or memory and I don't think there's much that better hardware can do about that. You can turn off some of the display options and avoid being in Select nodes mode since that spends time searching for nodes to highlight whenever you move the mouse.

Version 13 will have faster graphics for navigation, especially rotation and panning.

Andrea

OK...I think that all Mecway users are waiting for next release.

I have a issue with several load combinations because Mecway, when you are adding a new load, automatically makes this load active for all load configurations.
I think that is better the opposite way: a new load added on a specific load configuration is automatically suppressed for the others.

Obviously is my opinion and I found the procedure improper when you have to manage several LC and loads.

PS I noticed that installing Mecway in a new Windows installation (same machine but more RAM free) zooming, pamming etc. are faster

Regards

Victor

You might be right about configurations. I've made a note about that. It needs a (un)suppress in all configurations option too.

For the new install, are you sure all the display settings were the same?: Node/face/element selection mode and all options in the bottom section of View menu?

Andrea

@Victor wrote: It needs a (un)suppress in all configurations option too.

Better!

Yes, same display setting.

I tried again a solution with bonded contact and new RAM modules until 32Gb. Applying constraint step occupies until 24 Gb RAM (only for Mecway)

Andrea

I try to run my model on i9 9450x - Nvidia Quadro P4000 and 64 Gb RAM workstation.
Visualization is fluid also refinig mesh to about 1M nodes