Poor Gradient Validation for Streamwise Periodic Flow solver with type MASSFLOW.

[NAnand-TUD]

Hello!

I am testing the streamwise periodic (SP) adjoint solver and found that the Gradient Validation is spot on for PRESSURE_DROP based SP solver, however, with MASSFLOW based SP solver the gradients are off by more than twice the value. @TobiKattmann any clue why that might be the case ?

I am currently implementing an additional term in SP solver which needs to be updated as pressure drop is being updated in MASSFLOW SP.

Thanks,
Nitish

[TobiKattmann]

Are you missing code from this PR by any chance #1536 (merged into develop on Feb 26 and part of 7.3.1. master version)

There we added an 'additional adjoint adjoint equation' for pressure-drop as it is also a fixed point equation just as the flow solution and therefore needs to be handled alike.

[NAnand-TUD]

Yes, I am up to date with the develop branch so I am certain that I am including that the new adjoint equation is being accounted for. I see that there is AD::Register for PressureDrop in the version I am using.

To be double sure I also ran on the commit which includes your updates from 26th Feb, however, still no validation for the MassFlow-based SP solver. Can you kindly share your config file or point me to the one in the TestCases that you used for validation? Thanks!

[NAnand-TUD]

@TobiKattmann: I am running some verification for the Turbulent case now, and I find it rather difficult to make a comparison. What test case did you use for the verification of the constant heat-flux problem?

PS: By verification: I am comparing a 30 in-line circular fin simulation with that of the SWP solution.

[TobiKattmann]

I used my standard 2D unit cell setup. It is of course easiest to use a constant massflow in the periodic setup, and then either also constant massflow for the non-periodic simulation but I went for a block-profile with a velocity inlet where of course with fixed density and area the prescribed velocity regulates the domain's massflow. Otherwise there was really nothing special to it. I simply used the downstream profiles at the ''same" periodic location as you see below and that worked well enough for me:

I am not sure whether I get you concern correctly though

Edit: and for pressure and temperature i simply zero-ed one location for each profile such that one only compares the profiles

[NAnand-TUD]

Yes, you got my question right. Periodic BCs were to blame there... Once I switched to SYM BCs the flow converged rather effortlessly... I am still trying to get verification for temperature equations. I will post the plots when I am done implementing them. Plots loading soon....

[NAnand-TUD]

Great ! All sorted. Here are my results for iso-thermal BCs...

[NAnand-TUD]

Hello @TobiKattmann,

So far the Massflow based SP solver has struggled to give consistent gradient validation. Can you tell me what Objective Function you used to validate the Gradients? Can you also share your case?

-Nitish