code_saturne User's Forum

Hi there,

I am simulating the flow behaviour (supposed incompressible) between two moving plates (in the normal direction) using the version 5.0.8 of Code_Saturne and its ALE module. In the case of a pressure Dirichlet boundary condition (itypfb(ifac) = isolib) applied to the inlet and the outlet fo the flow, oscillations (apparently corresponding to a checkerboard mode) have been observed when the flow is entering the duct (when the distance between the two plates is expanding) but these oscillations disappear when the flow is exiting the channel (when the distance between the two plates is shrinking). These oscillations are observable in the two screenshots attached to this post.

I have tried to impose other types of boundary conditions, such as non standard, for example by imposing a zero-flux condition for the velocity field (so by means of Neumann conditions activated by the flag icodcl = 3). The oscillations disappear but the velocity pattern is different and I supposed not to be the expected one by comparison with the results from Finite Element codes which have been used for the same configuration.

Thank you in advance for any feedback.

Kind regards,

So_nuc

Hello,
Is it a valve you want to model? If yes, this boundary should be considered as moving walls.

If it is inlet outlet boundary conditions, as mention in the user and the theory guide, isolib BC type retain the flow (imposing a 0 mass flux) when entering (it is a free outlet, but not inlet is allowed) which explain why you get cherkerboard effects.

In your try with a Neumann on the velocity, what do you impose on the pressure? I would suggest a constant Dirichlet.
You can also test the free inlet outlet BC (itypfb(ifac)=ifrent, or using the GUI), see the theory guide if you want more details on it.

I hope it will help.
Best Regards
Martin

Hello Martin,

Thank you for this feedback. Actually I just want to simulate a flow between two moving walls with pressure imposed at the left/right sides of the duct.

Within the test considering Neumann boundary conditions for the velocity, the pressure is indeed imposed by means of a constant Dirichlet boundary condition and this actually gives the same results compared to the case itypfb(ifac)=ifrent, which is expected since the Bernoulli theorem supposes the Euler equation for the fluid (no viscosity effect and as a consequence the normal stress at the boundary should be 0).

I have noticed your explanation in the second paragraph and have wondered why isolib BC type applies different BC for the velocity when the flow is entering or exiting a channel ? What would be the solution to avoid such checkerboard effects when the flow is expected to enter the duct (mass flux < 0) ?

Thank you in advance.

Best regards,

So_nuc

Hello,

Thank you, it indeed worked as you mentionned .

Best regards,

So_nuc