Topological Optimization

mf

jacob
Maybe later, I have OpenFOAM running all day also. My sims are mostly nonlinear, maybe it's that. Also benefits are connected to certain minimum numbers of DOFs per CPU.

And then what? Should I record times? Does it stop automatically at a specific time?

EDIT: there are 2 missing groups 'Group_Of_All_Faces' and 'Group_Of_All_Volumes', I deleted the call for ...Faces and added ...Volumes. Then I get

║ ║
║ les 20592 mailles imprimées ci-dessus n'appartiennent pas au modèle ║
║ et pourtant elles ont été affectées dans le mot-clé facteur : PRES_REP

I can't use this.

jacob

when saving check

Yeah, I am interested in the time difference. It's just simple statics with 1 step.

It doesn't matter if it is linear or nonlinear, MPI is about solving KU=F, in nonlinear you just do it many times, so the speed up is in one iteration.

mf

jacob
That's not the reason for the missing groups.

It's probably WAY too small, storing and reading data will take more time than STAT_NON_LINE.

Still getting the PRES_REP error. Can't correct the geometry, have different shaper version (error upon changing to shaper).

jacob

yes it is, checking will create 'Group_Of_All_Faces' and 'Group_Of_All_Volumes',

well 2e6 dof takes for example 2 min, but if I have a nonlinear study I have 2 min * 1000, so I would need to speed up even these "small" studies.

mf

jacob
Ah this is a kind of standard group, I don't work with these. Good to know, I created them by hand in the last examples.

The benefit here seems to be small in the lower counts, indeed:
1/4: 6m35s
1/8: 5m13s
4/2: 4m26s
8/2: 2m54s --> sweet spot
16/2: 2m43s

jacob

I have tried what you suggested, but as you can see the gain is only about 10 - 20%. Maybe it's useful only for really large studies. But even this size would take tens of hours with contact. So to have even larger studies is not practical.

mf

jacob
That's what the CA docs also say. It's above a certain size, when the benefits come. It's also possible, that other things happen like throttling or other bottlenecks start to appear. MUMPS can be very hard on the CPU.
Take a look at our numbers: yours is quicker on the 1/4, almost the same at 4/2 but mine is quite fast on 8/2. Your 6/2 doesn't look good although you have cores left? Maybe some BIOS setting is in the way.

I've ran into a new problem: I did 21 iterations, but all_density.csv is only written until INST=12.
I repeat Postprocess_run.py and it contains all of them, weird.

jacob

I don't know, you interfered a lot in it.

If your mesh is in ORG, then probably would be better cp ./ORG_3D/.comm ./ORG_3D/.export .
Also, the element size should be uniform and radius > element size.

jacob

I had no problem

mf

jacob
I don't know what you mean, exactly.

The missing data in all_density.csv? Oh, I've seen it quite a few times now. If you do a

watch -n 5 tail all_density.csv

and compare that with the timestep you're in, you will see it doesn't always write or somehow misses the command. The Postprocess_run.py works perfectly but somehow it's not always executed. It could be because my system was very busy yesterday. No biggie, I just execute it one more time before looking at the data. It could be a timing issue, in OpenFOAM scripts I therefore often insert a short

sleep 0.1

Mario.

jacob

extended with thermal analyses

mf

jacob

Hello,

very nice work, looks very promising. Just for my understanding: it's about maximum heat conduction from the source to the temperature whilst reducing the amount of cells, right? So, shouldn't the 'branches' fully extend to the source and not stop before it?

Thanks,

Mario.

jacob

If the mesh would be much more refined I think I would get something like this. However, the smaller branches are filtered out.

mf

jacob

Oh, I think I misunderstood. The whole domain takes the heat away, not just the border on top of it (in your calculation)?

Mario.

jacob

yes, the whole domain is a heat source.
And there is no BC on the boundaries so it is equivalent to isolation.

Amin LAK

Hi @jacob
I am a structural engineer and new in Code-Aster (while have some years of experience with other FE programs like Sofistik and Ansys).
Seeing your optimization study using Code-Aster, I am thinking to use Code-Aster for the topology optimization of a problem that I am busy with. (First, I was thinking to use Topopt programs of DTU for that purpose. https://www.topopt.mek.dtu.dk/apps-and-software)

The problem is a structure resting on an elastic bedding, which means that the boundary conditions also need to be updated during the optimization process according to the geometry of the structure and thus the geometry of the contact interface between the structure and the bedding.

Since I am new in Code-Aster and I also don't know Python (I know C, Fortran and Matlab), I think it will be a difficult task for me. I was wondering if you or someone in the Forum will be willing support me on this task.

Thank you,
Amin

jacob

Hello,
try to describe your problem on some simple geometry.
Maybe for some simple mesh I might be willing to try to create some test case.

My code has no restriction for code_aster, so it can be used even with contact or other nonlinearity, however I have not tried it.

Amin LAK

jacob
Hello Jacob,

Thank you for your reply.
Imagine the structure may occupy an space as shown in the picture below. It is composed of
A) A steel cylinder with a diameter of 38 and a height of 8, which is resting on an elastic bedding with a stiffness of 1e9. The lower face of this cylinder is also subject to a downward uniformly distributed load of 1,5e5.
B) On top of cylinder A there is a steel cylinder with a diameter of 10 and height of 5.
C) On top of cylinder B there is a hollow cylinder with an outer diameter of 10 and an inner diameter of 9,8 and a height of 20.
Cylinder B is actually acting as a transition piece between Tube C and Cylinder A.
The top of tube C is subject to a moment of 1,3e9.

The goal is to optimize the cylinders A and B.

Thank you,
Amin

jacob

I don't understand how the downward load should act or how the structure is fixed in space. Stiffness 1e9 is elastic modulus in MPa?

Regarding optimization. What is the goal? To reduce weight to what percent? However, optimized topology of a bar loaded just with a torque is just a tube.

Amin LAK

jacob
The downward load is pulling down the lower skin of the pink cylinder (dia=38).
The stiffness of 1e9 has a unit of force/length, for example N/m
The first goal of the optimization is to find an overall concept of the structure.
The tube is not loaded by a torque but by a bending moment, thus one side of the large magenta cylinder is pushing against the elastic bedding and the other side is trying to lift up from the bedding (the bedding is only-compression) where the distributed load is resisting against the uplift.

« Previous Page Next Page »