Discussion Forum

Hi

I suppose you are in V4 (use the tags on the forum its easier for us all).
In V4 COMSOl is proposing the local normal and tangeant reference frame relaitf to any surface/edge.
t1 and t2 are the two tangeant vectors check the doc how they are defined (see model>definitions>corrdiante system

--
Good luck
Ivar

Hi,

sorry for digging out this thread but I think my problem fits here.
What I try to do:
- 2D Model with time dependent growth of a geometry (preferably non parameterized, should work with 4.x)
- 2 Physics, ht & chds

To the problem:
If i bring any variable (z or t) in the size parameters of the geometry (for example height increases with time), the structure can't be built. I guess z/t is in this case uninitialized. A way to set this initial parameter is not known to me, only through Definition -> Parameter (Variable does not work) and sweeping is not the way to go.

What works:
Build it up in 3.5a. I can easily access the coordinates through functions there. But I had to use a trick, because varying the geometry through time is not possible afaik. So I took the final geometry and varied the boundary conditions through some step functions depending on time and position.
But this is only an intermediate result. The next step has to be done in 4.0a and if possible without the trick. At the moment I use a base vector system with orthonormal axes (x, z) and have tried several schemes to access the coordinates like mod1.sys1.z or sys1.e_z like above. But nothing works (also didn't find any information regarding that).
Just for information, Matlab is not available here.
Any help is welcome.

Kind regards,
Bernd

Hi

I'm not sure you can change a geometry based on a time variation, certainly not with a transient analysis alone because the time stepping will not "loop" around the geometry node and update it, neither not the mesh node.

For that you need a full (external) "Parametric Sweep" node (and not a solver continuation sweep), but then how to call the transient one ? perhaps with a transient for one time t defined by the parametrical sweep


Note when you change the geometry, do not change the topology (do not collapse or create new boundaries or domains, as then COMSOL will renumber the internal "id" and you boundaries conditions will change place
--
Good luck
Ivar

Hi

I'm not sure you can change a geometry based on a time variation, certainly not with a transient analysis alone because the time stepping will not "loop" around the geometry node and update it, neither not the mesh node.

I thought on something about scaling. Should be the easiest way without remeshing and new borders. I'm not sure, if sth. like this is implemented.

For that you need a full (external) "Parametric Sweep" node (and not a solver continuation sweep), but then how to call the transient one ? perhaps with a transient for one time t defined by the parametrical sweep

Does the parametric sweep use the results from the last calculation? I never read about that. This is necessary because the diffusion depends on its history.

Hi

I believe scaling is as any new geoemtry

no a full Parametric Sweep restarts normally at initial conditions (if not specifically told to do something else)

--
Good luck
Ivar

I've tried a parametric sweep with two parameters. Now I have a inner (transient) analysis of short time span and two outer sweeps. One for position and another for overall time. So it should work..

But Comsol says: Parameter list size must be divisible by number of parameters.
I have no idea, what this means. Without sweeps activated, the computation runs fine.

As there is a "load parameter value" field available in the sweep options, I guess there is no "save load parameters" for each cycle. This can be seen as an recommendation for a next release.
Such feature would make Matlab needless.

You mentioned already in other threads, that you aren't an expert on the variable naming scheme issue. Maybe Magnus can tell something about that, if he reads over.

Hi

check if you have enough values: if you have 2 parameters you need an even number of values, you probably have an odd number and COMSOL is lost (if 3 parameters you need full triplets etc)

--
Good luck
Ivar

You mean the range values for the sweep step? The range goes from 0 to 648 with stepsize 1. That makes 649 steps. I'll try later. At the moment, I have other things to do.

Hi

if you have two parameters "param1" and "param2" then your set values are to be grouped in this way (if I remember right also for V4):

{first_param1, first_param2, second_param1, second_param2, third_param1, thrid_param2, ...}

--
Good luck
Ivar

Hi, I have a similar problem....

I have a 2-d model of bedrock below a mountain topography. I use a parametric sweep to flatten the mountain surface geometry, and use a Mohr-Coulomb material model to calculate resulting stresses and deformation in the rock below. The parametric sweep also changes the material strength properties from stage to stage. I need the stresses from one stage to continue through to the initialization of the next, but this doesn't seem to happen with the current model setup.

Do I have to update the Material's Initial Stress and Strain state each step? Can you suggest a simple way to do this?

Thanks,
Kerry