Velocity divergence error

[sanjeev784]

Hello,

I am solving an inflow-outflow type problem with variable density. The density is varying from inflow to outflow. But the velocity diverges after a few irritation. Can anybody let me know the solution how to rectify that? Many thanks in advance.

Regards,
SR

[mathrack]

Hello,

Considering the very small number of details provided, I think nobody can help.

Regards,
Mathrack

[pbartholomew08]

Which BC file are you using as a basis? If you look at the BC-Mixinglayer file for example the density field is given a smooth profile based on TANH(y), rather than using a discontinuous function such as

if (y > 0) then
  rho = dens1
else
  rho = dens2
end if

This may help resolve the issue. However, as said without more information it will be hard to help.

[sanjeev784]

The boundary condition and input file are attached for reference. I am already using tanh function.

input.zip

Regards

[pbartholomew08]

Having had a look at this, when you are computing the TANH profiles there is a division by the location x which results in a division by zero.

It also seems that you have a reaction module which implies there could be a change in the thermodynamic pressure, this type of flow isn't currently supported.

[sanjeev784]

Hi, I am using the setup of flow over a cylinder and the cylinder is removed from the geometry, therefore I am solving a convection-diffusion with variable density. Regards, Sanjeev ------------------------------------------------------------------- ----------------------------------------------------------------------- Dr. Sanjeev Kumar Ghai Research Associate School of Engineering Room 2.18, Stephenson Building Newcastle University Newcastle-Upon-Tyne NE1 7RU United Kingdom

…

On Thu, 29 Jun 2023 at 16:49, Paul Bartholomew ***@***.***> wrote: It looks like you're trying to do flow past a cylinder. This hasn't yet been implemented for the variable density solver I'm afraid. — Reply to this email directly, view it on GitHub <#199 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A6ANULYE7AFQBXZKWVZXKV3XNWPXXANCNFSM6AAAAAAZYRXP6Y> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[pbartholomew08]

Hi, I've tried setting up a somewhat similar flow here https://github.com/pbartholomew08/Incompact3d/tree/lmn-inflowoutflow

It uses a cos profile for density so that it goes from dens1 at the inlet, rising to dens2 around the middle of the domain and back to dens1. It's quickly put together so I don't claim it is correct, however it appears to behave sensibly.

[sanjeev784]

Hi Paul, Many thanks for sending this test case. I have run the setup that you have sent without any change. I have observed that velocity is not changing according to the change in the density but it linearly decreases from inlet to outlet. Can you comment on that? Regards, Sanjeev ------------------------------------------------------------------- -----------------------------------------------------------------------

…

On Thu, 29 Jun 2023 at 18:01, Paul Bartholomew ***@***.***> wrote: Hi, I've tried setting up a somewhat similar flow here https://github.com/pbartholomew08/Incompact3d/tree/lmn-inflowoutflow It uses a cos profile for density so that it goes from dens1 at the inlet, rising to dens2 around the middle of the domain and back to dens1. It's quickly put together so I don't claim it is correct, however it appears to behave sensibly. — Reply to this email directly, view it on GitHub <#199 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A6ANUL3TTYEWCMP6RNP2JZDXNWYIJANCNFSM6AAAAAAZYRXP6Y> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[pbartholomew08]

Hi Sanjeev, you're right that doesn't make sense and sounds like a bug I need to look into.
However I don't get the velocity decreasing towards the outlet using the input_DNS300_LR.i3d file

[pbartholomew08]

Actually for this case the rho*u should not be a constant, we have

drho/dt + u.grad(rho) = -rho div(u) \approx 0 (the divu constraint is ~10^-2 vs grad(rho)~1)

so the velocity should be approximately constant, atleast in early times.