OutletBoundary Condition Causes Mass Imbalance
 

Hello there,

I have been trying to validate a CFD solver using the flow through an ordinary pipe problem (at very low Reynolds numbers).

The simulation is as expected until the flow reaches the outlet where a large velocity accumulates causing mass imbalance.

I am using the outlet boundary condition, specifying pressure equal to zero at the outlet.

I have found the following information whilst surfing the internet:

INTERNET INFORMATION

If significant gradients are calculated near a boundary, a mass imbalance can occur. This is due to the implied condition of fully developed flow for a constant pressure boundary. If the flow has not fully developed, FLOTRAN is forced to adjust it across the last row of elements to satisfy the boundary condition. Occasionally, this adjustment may cause a mass imbalance.

To prevent this from occurring, you can add a development length to the exit. This addition does not actually need to match the physical flow geometry. You can subtract the pressure drop in this "chimney" to obtain the desired pressure drop. The pressure drop in the "chimney" will vary over its full cross section. You will have to choose a representative value of the pressure drop to match the desired outlet conditions. You will not know this pressure drop, of course, until after you examine the results from the chimney. You can opt to apply an exit pressure profile, based on these results, to the original geometry.

You can calculate the required length of the chimney (L) using the characteristic diameter (D) from the development length formula:

Laminar flow: L/D ~ 0.06 Re Turbulent flow: L/D ~ 4.4 Re1/6

END OF INTERNET INFORMATION

I have tried using a 'chimney' as the above information suggests however, this does not remove the mass imbalance instead, a large velocity occurs at the new outlet position.

Has anyone come across this problem before?

Does anyone have any ideas on how to overcome this problem?

How can I reduce gradients close to the outlet boundary?

Could it be something to do with my mesh, I am using under 1000 nodes and no clustering?

Kind Regards

Andy