Problem with flow_tools/evaluator?

[Daniel Lecoanet]

Hey,

I've come across a really odd problem.  I've been using flow_tools (in my repo) to calculate things on the fly like rms Re, peak Re, for low Ra number convective steady states.  I run a convection simulation which outputs the rms Re to the screen every 25 iterations.  I also dump the u and w fields every 1000 iterations to hdf5 files.  I've also written a script that reads in the latest u,w dump from a bunch of different simulations, and then calculates the rms Re, etc., as a way of seeing how these quantities change as a function of Ra, height of the box, and other control parameters.  Now the weird thing is that the rms Re I calculate in post-processing is different from what is output to the screen from flow_tools.  Furthermore, I've discovered that depending on what other quantities I'm calculating on the fly (like volume averaged KE, rms velocity), the rms Re and peak Re calculated by flow_tools changes, even though the flow is the same (same initial conditions and evolution equations)!  It seems that the quantities I calculate within my problem script are all being calculated correctly; it's only things that are being calculated using flow_tools that are having this problem.

Daniel

[Keaton Burns]

Hi Daniel,

Now the weird thing is that the rms Re I calculate in post-processing is different from what is output to the screen from flow_tools.

How are you calculating the statistics in your script, and in post-processing?  The flow_tools implementation takes the max, min, and mean over the grid points, and has an issue where the mean is skewed away from the correct value when the load balancing isn’t ideal.  I’ve fixed this in my fork, and made a wider set of changes that I’ll be issuing a pull for soon.  

In any event, this grid-based calculation will yield a different result from a computation of the mean using a volume integral, or something like that.

Furthermore, I've discovered that depending on what other quantities I'm calculating on the fly (like volume averaged KE, rms velocity), the rms Re and peak Re calculated by flow_tools changes, even though the flow is the same (same initial conditions and evolution equations)!

Strange!  Are your other calculations here are being done through a separate analysis handler?  I haven’t really used the flow_tools Re calculators, so I’ll take a closer look at the code and see if anything jumps out.

If it’s not too difficult, could you send a quick example script that shows this behavior?

Thanks,

Keaton

--
You received this message because you are subscribed to the Google Groups "Dedalus Development" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dedalus-dev...@googlegroups.com.
To post to this group, send email to dedal...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dedalus-dev/CAJoYf%3Dirm%2BVVfpw0sqJ0wvPJEC8wJaU1spzYpXsTerBoJYe2Hw%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

[Daniel Lecoanet]

Hey Keaton,

Just to update on this... Pleiades is down, so I tried testing this on Henyey... doesn't seem to be any problems/inconsistency.  I'll try testing again on Pleiades once it's back up.

Daniel