Rate Budget Fluid Imbalance

Hello,

I'm trying to run a transient coupled heat and mass transport model and have no problem with convergence, but notice some imbalances in my rate budget. Both heat and mass are fine, with imbalances around 0%, but my fluid imbalance is close to 4%. The majority of this discrepancy is in the Distributed Sink/Source, but I'm having trouble finding information on what this actually means. I did find the following thread in this forum, but don't think the help was ever updated.

http://forum.mikepoweredbydhi.com/index.php?topic=1078.msg2625#msg2625

Could someone please explain what this Distributed Source/Sink is, and how I might be able to go about reducing the fluid imbalance of my model?

Thank you!

Actually the help was updated. When you look at the 'Rate Budget Panel' page in the help system, you'll find exactly the three components listed that are mentioned in the other post. Regarding your model, it is hard to give you a recommendation without seeing it. Maybe you can narrow down the imbalance spatially by using the Subdomain Rate Budget / Subdomain Period Budget panels for parts of the model.

Hi Peter, thanks for the response.

I see the updated page now, I had been looking at the online help as opposed to the help system incorporated into FEFLOW. The page says that, in density-dependent models, the buoyancy term is considered in the Distributed Source/Sink term of the rate budget. After reading this, I decided to experiment with the Fluid Density options in the Transport Settings tab. I switched from the 3rd option to the first, invoking the Boussinesq approximation, and my imbalance was solved. However, I'm not sure that I'm justified in using this approximation, as my model deals with high density fluids (with solute concentrations on the order of 200,000mg/L, density around 1140kg/m[sup]3[/sup]). I'm having a difficult time finding a source that supplies a numerical range of density gradient over which the Boussinesq approximation is valid, everything I've found so far just says it's valid at low density gradient.

Do you think that I would be justified in using this approximation, or should I find another way to solve my imbalance?

The degree of approximation should be related with the degree of balance accuracy you are interested in. For example, if the free convective system is superposed by forced convective processes (e.g. triggered by pumping to keep a specific threshold of the primary variable(s)), then the required amount of water may change between these approximations. This may be attributed to the fact that density may change the volume of water.

Indeed, there is no "guideline" which states which degree of approximation should be used for which case. A "guideline" is ventured, because the different cases may vary drastically. In other words, each numerical model is unique. Accordingly, if your system is characterized by sharp gradients in density, then you may try different degrees of approximations.

Beside the spatial discretization, the temporal discretization may also play an important role. If you simulate density-driven flow phenomena, you may try to constraint the time-step size in addition to the automatic predictor-corrector time-stepping scheme by adopting a [b]growth factor between subsequent time-steps[/b] and/or a [b]maximum time-step size[/b]. In this context, a stricter [b]error tolerance[/b] and/or [b]error norm[/b] may also help to reduce the imbalance.