[verboomj]

So, I've read a few posts about the prefcell and prefvalue but apparently I don't understand it yet.

I'm using adjoinShapeOptimizationFoam, which uses simpleFoam to solve pressure and velocity.

However, from iteration 4 to iteration 5, suddenly my pressure at the inlet becomes negative. This behaviour is also visible in my objective function because it has a pressure term included.

At the inlet I assign a velocity and at the outlet I set the pressure to zero.

So does anyone know where this behaviour comes from?


EDIT: In the plot and file I used P = 1 at the outlet

[floquation]

prefcell and prefvalue fix the value for pressure, if it is not fixed by anything else.
The reason is quite simple: if the pressure differential equation is solved, you will get an integration constant. If your BCs only demand the gradient of pressure to be a certain value, your BC do not impose any condition on this integration constant. Numerically you do, however, need such a constraint, which is achieved by simply fixing the value for pressure in one cell.

In your case, your BC demands a value for P at the outlet, therefore prefcell and prefvalue do nothing.

Your BC seem appropriate.

It seems like your simulation is not properly converging, or diverging even.
Post your fvSolution and fvSchemes files and the solver output and someone may be able to diagnose the issue.

[verboomj]

Hi Kevin,

Thanks for your reply, good to know how the reference value works! However this was my first idea what could be wrong so I have to look further now.

In the attachment are my fvSolution and fvSchemes, which I left 'as downloaded' from OF and the log file from my solution. Only 25 iterations, but the behaviour already happens from iteration 4 to 5.

Also in the residual plot you can see p_resiudual is increasing significantly from iteration 2 to 5. Any idea?

Thanks in advance!

Jeroen


EDIT: In the plot only p, Ux and Uy are of interest, the adjoint solutions shouldn't have an effect on the primal solution.

[verboomj]

Solved it by rewriting the entire solver based on simpleFoam instead of adjointShapeOptimizationFoam. Must've been something wrong with the latter.