CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Weighted Averages Disagree (http://www.cfd-online.com/Forums/fluent/47256-weighted-averages-disagree.html)

Ryan February 6, 2008 16:15

Weighted Averages Disagree
 
I'm modeling the flow of water through a 6ft pipe that has a 1ft pipe diverging from it. Roughly 1% of the flow goes through the small branching pipe. About 16inches downstream from the junction there is a 12" butterfly valve that is fully open.

I'm running a specified velocity profile at the inlet of the large pipe, pressure outlet at the exit of the large pipe and a pressure outlet(with target mass flow rate enabled) at the exit of the branch. The simulation converges; however I'm having difficulties comprehending the results.

I have bounded planes at various locations with their normals parallel to the axial direction of the pipe section. At these planes I'm looking for an average total pressure value and the mass-weighted average is different than the area-weighted average. I thought they would be the same considering this case has incompressible flow.

Also, there is great concern over the fact that the average total pressure in one section of the pipe (where recirculating flow occurs) increases in the downstream direction which is not physically correct.

Any input is greatly appreciated.

Carlos February 8, 2008 04:16

Re: Weighted Averages Disagree
 
I can't help you with the differences between mass and area weighted, I can only guess that mass weighted is just area weighted mutliplied by the pipe length and then density to get mass.

As for the differences in total pressure, I understand that the higher the total pressure, the greater the losses in the flow. Surely viscous effects will be doing this as you move down the pipe? I'm not a pipe person though to might be talking rubbish.

One thing I do know is that when you create planes in the post processor then often the flow rates do not add up. I model flow in and around a livestock trailer which only has side vents for the flow to mix inside the trailer. By creating planes in Fluent in these vent apertures I found a difference between flow-in and flow-out by over 15%.

By continuity this is utter garbage. This is all do to interpolating and errors build up and cause the descepancy - it even states this in the Fluent manual! The only way around this is to decide at the preprocessor stage where you want to measure the flow in the pipe.

Next split your domain using these planes and set their boundary type to interior. When you mesh, export to fluent and then run the simulation you'll find that the mass flow rates are bang on. It should follow that all other quantities will show physical results.

Good luck!

Carlos.

Ryan February 8, 2008 14:04

Re: Weighted Averages Disagree
 
Carlos,

Thanks for the response! Your input made me realize that I cannot trust the results I obtained by taking the average across a plane. I did however obtain an average mass flow rate within 1% from plane to plane in a single pipe, but the pressures don't make sense and the velocity decreases downstream as it should.

I've come to the conclusion that because there is recirculating flow, that is why the total pressure is increasing (total pressure is suppose to decrease downstream in pipe flow due to viscous effects.)

Looking at a single plane, there is mass flowing to the right and to the left (recirculating/backflow). The average mass flow rate is the desired value, but the total pressure of the plane is greater than upstream because there is mass (thus energy...) going through the plane twice. This doesn't explain why there is a difference between the mass and area weighted averages.

Thanks again,

any input on the matter is still welcomed!

Carlos February 8, 2008 15:55

Re: Weighted Averages Disagree
 
Ryan,

I see your problem but I can't really comment as to what the descrepancies are. If you have time it is worth remeshing with interior boundaries to see if the results are vastly different, although from what you describe this is unlikely to make the difference.

If you are still studying it may be worth asking a lecturer in fluid mechanics to comment or have a look at a typical representative probelm in a fluid mechanics text book to see what a physical result would look like.

One other thing which springs to mind is what order is the descretization of your solution? If you are using the default settings then effectively it is a first order solution which isn't accurate enough for decent results. If this is the case try going for 2nd order for pressure, and QUICK (third order) for all other discretization quantities, namely momentum and turbulence. This is under solution->controls.

Also make sure you have proper convergence i.e. flat residuals after say 3000-5000 iterations (lower the residual monitors to 10-^9 to stop the solution 'converging' early from the default setting). You can only judge convergence when the residuals have more or less zero gradient.

Make sure the above two points are invetsigated as the solution is highly dependent on the order of the solution and convergence. All the best,

Carlos.


All times are GMT -4. The time now is 21:10.