Discussion Forum

Hi

in any case do not try two solid nodes, but rather several chained solver sequences, a stationary to stretch ,then a time /transient for the impulse with a smoothen gaussian or rect() function for the impulse. Then do not forget to set the time stepping to strict or intermediate with enough fine steps to correctly resolve the impulse shock.

If you check carefully the equations of the Damping node, you will see that both Rayleigh and Isotropic damping factor are added to the equations for eigenfrequency and for frequency domain, but NOT for time stepping transient mode, for which ONLY the Rayleigh damping (alpha beta) will contribute !! check the doc too

--
Good luck
Ivar

Hi Ivar,

As always you reply to the problems very quickly!

Thats a good idea to use multiple studies.

After I wrote the message I realised that there is a solution to my problem, but it only works for Rayleigh Damping. If I define the Rayleigh Damping factors as a function, both alpha and beta can be zero until time t1. Where t1 is the time that the initial stretch is over.

I have also tried your idea.

I have however simplified your suggestion into a very simple model. In the model I have a disc again and I wish to initially stretch it, and then I wish to apply a displacement normal to the plane of the disc, this is just an arbitary loading scenario so I know how to use multiple steps.

I now have 2 stationary steps.

Step 1: Apply radial stretch.
Step 2: Apply surface displacement.

I then add a Prescribed Displacement 1, which is where I define the radial stretch and in the Equation option I select "Show equation assuming "Study 1, Stationary"

I then add a second Prescribed Displacement 2, which is where I define the surface displacement and in the Equation option I select "Show equation assuming "Study 2, Stationary"

However I now notice that Prescribed Displacement 1 has been overwritten. So I am a little confused.

EDIT: I should be going to sleep but I had to try one more thing, instead of creating a second independent stationary step, I also tried adding a second study step (Which I think is what you were talking about above). Again, in the Prescribed Displacement 1 I select in Equation, Study 1, Stationary 1. And for Prescribed Displacement 2 I select in Equation, Study 1, Stationary 2.

But again Prescribed Displacement 1 then shows as being overwritten.

If you wish I could attach the model?

Thanks for looking at this for me, I very much appreciate it. I am a very confused young man at this point!

All the best,

Rob

Hi

One or two study nodes or one or two solver sub-nodes under one study nodeis, for me a question of stacking, in manual GUI mode one study node is executed completely, two in two manual "compute"

The issue you will have, if you use prescribed displacement, is that then you block some boundaries to build-up your stress, if you release them the part will release back, an adding severla prescribed displacements can sometime be conflicting. I prefer eiter boundary or volume loads, or thermal stress build up, as these are cmmulative and often what happens truely when you assembly or manufacture your parts

--
Good luck
Ivar

Hi Ivar,

Thanks for the feedback.

I did think about using a thermal step and maybe its the best way to go, I will explore this today I think.

I do think that there is still a little confusion over using two sub-solver nodes (Stationary 1 & Stationary 2).

Could you perhaps suggest how I would implement a static and a transient study in one solver?

As in;

Solver 1
- Study 1 - Stationary (radial displacement)
- Study 2 - Transient (surface displacement or load)

Or would this have to be done as a two solver model, as in;

Solver 1
- Study 1 - Stationary (radial displacement)
Solver 2
- Study 1 - Transient (surface displacement or load)

Again, thanks,

Rob

Hi

for me it's the same, but one study node you launch it once and both solvers run sequentially, the two study node you need to manually launch each (2 clicks with some waiting, but you can calmly check the first results before you launch the second study ;)

--
Good luck
Ivar

Hi,

I can see the benefit of running two solver subnodes over running two solvers, but I am still not sure if this answers my questions.

Do I need to run subnode 1. Then, change the loads/BC in the model and then run subnode 2?

Rob

Ivar,

You said this some time ago...


"Chaining simulation are an interesting way to decompose a loading case, but it's not always trivial. By default if you link in 2-3 or more stationary solvers one after the other and ask for the default COMSOL solver set-up, you will see that COMSOL stores the result after each solver sequence and uses the next solver initial conditions from the stored previous solved step. Well at lqast that is what you should use to link your models. When you hit the dependent variable node in the solver sequence you can change the settings for the initial variables, from the default to the current solution to the stored solution ...

Remains one difficulty, how to change the BC (boundary conditions) to step through your loading case, as i.e if you load a bar with a twist (moment) then remove the moment and compress it, the bar will release it's twist strain and return to its initial condition (except if you keep the moment load). Then today you cannot turn on and off BC from the solver nodes, these are not linked, so such a model you must run manually (right click the study node compute to selected) each node, then change manually the BC run next node etc.

Again there is an exception, you use Matlab to script the sequence and perform the intermediate changes to your BCs

I hope I have given you a few inputs, to test out. Do it on siiiimple problems so you really manage to see the different implication of each solver variable, that is what I find the trickiest"

I will see if I can make the model work now, thank you.

Rob

Hi indeed

when you need to change BC in beteween two studies are reaquired. Eception: when you cn use a parameter such as stationary Continuation and use the Parameter as a boolean flag to change the BCs via some logical function. Somewhat cumersome, but it works.

When you have a stationary + time depednent soler sequence, then "t" is not defined during the stationary, I then add a Parameter t=0[s], in this way "t" is defined during the stationary solver sequence and I can use it as a boolean flag too

--
Good luck
Ivar

Thats a very interesting way of doing it, even if its feels like its a bit "fiddly".

But thank you for the suggestions, I will let you know how I am getting on.

I think this discussion could be of use to others looking to do pre-stressed time domain studies.

At the moment I am going about this by using a thermal step to induce the strain into the surface.

Thanks again,

All the best from sunny Bristol,

Rob

Hi

good to hear, thi year a sunny and +17 here in Yverdon, >30[K] thermal strees in hardly more than a week ;)

--
Good luck
Ivar

Looks like you live near the lake, it looks very nice...

I have a follow up question but have created a new question as I think its quite different.

Here is the link if you are interested.

www.comsol.com/community/forums/general/thread/26603/#p72389

Rob