Parallel Internal Fan Problem

[nathan.alday]

I have a recurring (but intermittent) problem running an internal-flow problem with two parallel (identical) circulating fans.  I’ve successfully run dozens of design variations, but occasionally I’ll tweak the design, and then one of the fans will show abnormal flow, showing a weird pressure differential.  After that, it seems like I need to start a completely new design study, in order to get it to work again.  Even if I specify a fixed flow for the fans (i.e. not a curve), it’ll still show abnormal flow (although the status file will lie and list the specified the flow volume).  The pressure aberration is there as well.

This problem is obvious from iteration 0, looking at a cut-plane through the fans.  It is not something the solver converges to, but is an error from the start.

I’ve searched the forum and found at least one similar occurrence.  Is this a known problem, and is there a work-around?

[Jon.Wilde]

Are the fans uniformly meshed with at least 5 elements from inlet to outlet?

It can also help to give them unique names.

Advection Scheme 5 may be very useful also.

[r-taylor]

Are the fans shrouded/ in housings?  I've seen similar behavior if the fans are bordered all around by fluid.  I've also noticed similar behavior if there are shrouds/housings around the fans, but if there is a very small offset between the surface of the fan and the housing.

[nathan.alday]

Yes, I’ve tried all of those set-up approaches (based on previously-posted, similar problems) – but as I alluded to, this doesn’t seem to be a set-up problem.  The model will run fine (even with the fans named the same, and meshed very coarsely), until some minor geometry tweak is made in some area of the model not even close to the fans.  It’s almost like something gets corrupted when the geometry is updated (i.e. a volume is added, or the number of surfaces changes).

Thanks anyway for your suggestions.  I’ve started a new model from scratch again, and it’s running fine for now. 

[heath.houghton]

Nathan,

When you make the tweak to your geometry, are you starting the solution from zero or restarting from last saved iteration (interpolating the results onto the new mesh)?

Heath

[nathan.alday]

Heath,

I'm starting from zero every time...

[apolo_vanderberg]

When you do this. Can you first run for zero iterations and check the fans to ensure that they are properly initializing.

What I mean by this is that when fans have an RPM, and we assign the material to each individual fan volume, we will use the centroid of the fan as the center of rotation to determine swirl.

If for some reason one of your fans has an incorrect centroid specified (such would be the case if we assigned the fan material to 2 volumes at the same time), you'll see that the angular as incorrect.

Each fan should appear as a dart board (or concentric rings of increasing velocity) at zero iterations. If one fan is incorrect there's no sense in running withouth correcting the problematic fan by assigning a generic material and then re assigning the fan material to that single volume.

Apolo

[nathan.alday]

Yes – I’ve run into the RPM problem too – but I don’t think that’s the case here.  I’ve set RPMs to zero, and specified a constant flow rate (to try and work around the problem), but there are still issues right from iteration 0.  Although the summary file tells me that the fans are operating at the set rate, a cut-plane shows completely different velocities through the fans (see enclosed images).