Topology optimization not converging and high violations

Altair Forum User

Hey all!

 

I hope you can help me! I´m doing an optimization for a university project of a bracket but get no convergence and high constraint violations.

 

 

The model:

See the design volume (blue) in Screenshot 1. The yellow areas are not being optimized. There the loads are applied and the model is constrained at the two yellow areas on the top.

I had to use TET10 elements because auf geometrical issues. 6 forces apply to the model. I use quite big elements but this is for doing quick calculations...I don´t want to wait 2 hours to see if it is working... but I think this should not be a problem here?!

<?xml version="1.0" encoding="UTF-8"?>

<?xml version="1.0" encoding="UTF-8"?>

 

 

Optimization:

I want to reduce the weight/volume. So objective is to minimze volume (I have tried Mass, volfrac). 

At the nodes, where I´ve applied the forces (basically in the middle of the yellow areas), I have set displacement constraints (betweent 0.5mm and 2mm).

I use min member size controll of 0.4 and a stress constraint of 450(N/mm²)

 

 

My Problem:

The solver always reaches the iteration limit of 80 without converging below the tolerance and ist violating the constraints massively. (screenshot).

I have tried to let the solver run 120 iterations but without any usable result.

<?xml version="1.0" encoding="UTF-8"?>

<?xml version="1.0" encoding="UTF-8"?>

<?xml version="1.0" encoding="UTF-8"?>

<?xml version="1.0" encoding="UTF-8"?>

<?xml version="1.0" encoding="UTF-8"?>

 

 

Here are the ones with 120 iterations:

 

<?xml version="1.0" encoding="UTF-8"?>

<?xml version="1.0" encoding="UTF-8"?>

<?xml version="1.0" encoding="UTF-8"?>

 

 

 

When I view the results given by the solver at iteration 80 it looks like it is about to find a solution that looks acceptable in my eyes. but the strange thing ist that the density values of the structure often are below 0.1 and not almost 1 as they shpould be. I have tried to check if it ist because of quite low loads (max about 165N) but with a scale factor of 10 the same element densities appear.

 

Results with density values > 0.05:

<?xml version="1.0" encoding="UTF-8"?>

 

Results with density values > 0.2:

<?xml version="1.0" encoding="UTF-8"?>

 

Results with density values > 0.5: 

<?xml version="1.0" encoding="UTF-8"?>

 

Has anybody an idea what I am doing wrong? Should I increase the number of iterations further?

 

 

Altair Forum User

Hi,

 

I am not sure what the optimization constraints are, but I can see that the displacement constraint is contributing to the failure of optimization.

 

Can you share the model using the dropbox link available in my signature?

 

Altair Forum User

Okay I have uploaded it. Maybe you can help me.

Now I tried an optimization with double the amount of elements and 150 iterations. It´s still not converging but seems that the density distribution looks quite good from iteration 120 upwards. And there are no big constraint violations past iteration 70.

 

Thanks so far!

Altair Forum User

Hi,

 

Instead of volume frac can you use volume as objective function?

 

Also, it is not recommended to use stress constraint for topology optimization.

Altair Forum User

Hi Prakash,

 

I have tried volume as objective - no change!

Altair Forum User

Ok... Now i tried with HEXA Elements, some simplifications in the geometry and much smaller allowable movements.

I´ve set the displacement constraints to 0.1 at every node of load application.

And it CONVERGED!  With both objectives, volume and volume fraction.

 

My assumtion is that with wider constraints the solver needs much more iterations to find a solution. Is that right?

Altair Forum User

Altair Forum User said:

My assumtion is that with wider constraints the solver needs much more iterations to find a solution. Is that right?

 

Yes. But I dont think it has anything to do with mesh as I tried with 1st order tetra, the solver took 150+ iterations and still failed.

 

What I was trying to find out is in the out file, as the iteration progresses the % violation on constraints kept on increasing /emoticons/default_biggrin.png' srcset='/emoticons/biggrin@2x.png 2x' title=':D' width='20'> 

Altair Forum User

Yeah... but at around iteration 40-50 it optimization always gets a huge constraint violation and aftarwards it decreases again.

And I get convergent optimizations with hexa-elements only 

Altair Forum User

True, at the same time it looks like more material is removed after 40-50 iterations. 

 

 

Altair Forum User

Do you have any recomendations on what I could try?

 

Another question: In the results above I had to set the density threshold to 0.1 to get at least a satisfying design - but shouldn´t it be set to 0.5 or at least 0.3? Caus now I have al lot of elements calculated to have a density around 0... Can you explain me this?

Altair Forum User

The density threshold actually depends. As you increase the slider to say 0.3 more material will be removed and the structure may not be stiffer.

 

0.3 will be the starting point. If the re-anlysis results are acceptable the better to go with the density threshold 0.3 optimized design or bring it down to 0.2 and try again /emoticons/default_smile.png' srcset='/emoticons/smile@2x.png 2x' title=':)' width='20'>