[verboomj]

Hello,

First of all I'm very new to programming/adjusting existing OF solvers.

I'm currently working with the solver adjointShapeOptimizationFoam but I'm looking to implement a volume constraint. As it will only appear in the sensitivity and thus only in the steepest descent line of code It shouldn't be to much of coding.

The constraint to be implemented can be seen in the attachment. The SD3 integral is the one to be added in the steepest descent line.

The question:
However I can't seem to find how I should start with volume integrals, in particular the upper integral. I can imagine I need to replace them with a summation of all cell volumes but can use some help in what direction I should look.

volScalarField.V() appears to be a method to get the cell volumes?

Any help is much appreciated.

Kind regards,
Jeroen

[verboomj]

So, after three days of digging up a ton of posts I've seem to manage to implement a volume constraint for the solver, still work to be done.

With this constraint it is possible to also find Stokes flow solutions. Initially it would give a full fluid field as a solution, alpha zero everywhere. Furthermore it should give some more freedom to what space your flow can use.

Some results in the attachment. If anyone wants some details feel free to contact me.

[C. Okubo]

Hi, verboomj!

Nice work! By your pictures, it really seems to be working! Could you share some details about this implementation? I am working on the same problem right now. I was trying to implement a Constrained Steepest Descent as per "Introduction to Optimum Design by Arora", but my alphas get stuck. Obviously, I am doing something wrong.

Also, are you using the volume constraint as equality or inequality?

Regards,

Okubo

[C. Okubo]

I think I done it!

Was able to do it with verboomj thesis + book and paper below:

https://repository.tudelft.nl/island...0-d4558a5fd1dd
https://link.springer.com/article/10...831-014-9141-9
https://www.springer.com/br/book/9780387303031

Thanks, verboomj!

[Cambridge]

Originally Posted by C. Okubo

I think I done it!

Was able to do it with verboomj thesis + book and paper below:

https://repository.tudelft.nl/island...0-d4558a5fd1dd
https://link.springer.com/article/10...831-014-9141-9
https://www.springer.com/br/book/9780387303031

Thanks, verboomj!

Hi C. Okubo

I am new in topology optimization, do you have any tutorials on that using OpenFOAM? How can you replicate the results of verboomj? Could you please help me?

Many many thanks in advance.

[C. Okubo]

Hi, Cambridge!

I don't have any tutorials on this case, sorry. I have some cases I ran in the past, but my code is a mess since I've been changing and trying some new stuff. I suggest you to study the references I passed and also this report:

http://www.tfd.chalmers.se/~hani/kur...ortAdjoint.pdf

I do not completely agree with everything on the report, but it gives you a good idea of how to make changes inside the solver.

Also, I think a good beginning is to start all your tests with laminar flow. Use nu = 1 and turn off turbulence models. The adjoint solver can be very unstable when turbulence is on (because of the frozen turbulence hypothesis maybe?) and you get no useful results.

Okubo

[Cambridge]

Originally Posted by C. Okubo

Hi, Cambridge!

I don't have any tutorials on this case, sorry. I have some cases I ran in the past, but my code is a mess since I've been changing and trying some new stuff. I suggest you to study the references I passed and also this report:

http://www.tfd.chalmers.se/~hani/kur...ortAdjoint.pdf

I do not completely agree with everything on the report, but it gives you a good idea of how to make changes inside the solver.

Also, I think a good beginning is to start all your tests with laminar flow. Use nu = 1 and turn off turbulence models. The adjoint solver can be very unstable when turbulence is on (because of the frozen turbulence hypothesis maybe?) and you get no useful results.

Okubo

Hi Okubo,

Oh, thank you so much for your help. I will follow and try what you advise. Have you ever done topology optimization by OpenFOAM? Do you any suggestions? Thank you.

regards

Cambridge