ProjectTorreyPines / FUSE.jl

FUsion Synthesis Engine
https://fuse.help
Apache License 2.0
32 stars 2 forks source link

Benchmark the Stress Calculations using the ITER case #277

Open zalzali27 opened 1 year ago

zalzali27 commented 1 year ago

Currently technology.jl,

Uses Leuer 1D solid mechanics equations to solve for radial and hoop stresses in TF coil, OH coil, and center plug.
Based on derivations in Engineering Physics Note "EPNjal17dec17_gasc_pt5_tf_oh_plug_buck" by Jim Leuer (Dec. 17, 2017)

Returns radial, hoop, axial, and Von Mises stresses for TF, OH, and plug (Pascals)
(optional) radial profiles of radial, hoop, axial, and Von Mises stresses for TF, OH, and plug (Pascals)

Generating the plot of the CS stresses using ActorStresses for the ITER case using FUSE, there is a mismatch between the two sections of the Von Mises curves at the interface as shown in the red circle, should this be resolved before proceeding to add more layers?

For reference, this plot was generated using these conditions:

act.ActorStresses.n_points = 51
dd.solid_mechanics.center_stack.plug = 1
dd.solid_mechanics.center_stack.bucked = 1
dd.solid_mechanics.center_stack.noslip = 0

CS_Stress_ITER_Bucked_Plugged.pdf Screen Shot 2023-03-29 at 1 56 36 PM

Please note that @daveweisberg highlights this issue in the solid_mechanics_GASC.pdf and notes that if the calculations are done with zero axial stress and zero void fraction, this issue disappears.

zalzali27 commented 1 year ago

@daveweisberg could you please comment if leaving the curves mismatched like that is alright, and I can use this FUSE output plot as my reference point to compare with the case of splitting the TF into two layers?

daveweisberg commented 1 year ago

This mismatch is expected if the axial stresses for the two coils are different, or if the void fractions for the two coils are different. The only requirement for the bucked CS-TF case is that the radial displacement be continuous at the interface between coils.

However, I think this case should still be checked against previous results from GASC (as found in Jim's engineering physics memos). Certain attributes of your solution don't look right to me - particularly the radial stress. A few rules of thumb to keep in mind:

1) The radial stress should be zero at the edge of a component that isn't bucked against another component. 2) Negative radial stress means compression, positive radial stress means tension (same for hoop stress).

I would recommend doing a apples-to-apples comparison with Jim's results so you can identify any discrepancies.

zalzali27 commented 1 year ago

Shown here is a comparison between FUSE, GASC and J.Leuer calculations of the von mises stress as a function of radius with ITER parameters for a CS-TF bucking case with no CS current, no voids and no axial stress. The result shows a good qualitative agreement between FUSE and GASC (& Leuer's solution) but quantitatively the stresses in the GASC result are still higher, what could attribute for this discrepancy?

Screen Shot 2023-04-03 at 2 16 36 PM

orso82 commented 1 year ago

@zalzali27 glad you got the CS stresses calculation to run.

First thing to do is to make sure that you are running with the same input parameters as GASC. Have you checked that?

zalzali27 commented 1 year ago

That's right @orso82 , an improved iteration is attached here which was obtained by adjusting the thickness of the gap, OH, and TF in dd.build to match those of GASC, but there is no agreement yet. The solid_mechanics_GASC.pdf doesn't have all the input parameters, so I'll check with @daveweisberg

Screen Shot 2023-04-03 at 3 13 51 PM

I've also plotted the three components (radial, hoops, vonmises) for four different cases of varying the OH-gap to see how the stress scale with that: Screen Shot 2023-04-03 at 2 17 13 PM

daveweisberg commented 1 year ago

@zalzali27 Based on these plots, it looks like there is a least one error in the FUSE code. Given zero axial stress and no voids, the bucked solution should have a continuous stress profile across the interface between TF and CS. The fact that you're getting discontinuous jumps in both the radial and hoop stress profiles means that the solver has a bug somewhere.

daveweisberg commented 1 year ago

@zalzali27 Ok, I've spent some time making coil stress plots with corresponding documentation of inputs and outputs. They all use the ITER case for benchmarking. These should cover 3 main configurations as well as 2 benchmarking configurations and 2 alternate configurations:

Main configurations

  1. Freestanding CS-TF with voids and axial stress
  2. Bucked CS-TF with voids and axial stress
  3. Bucked plug-CS-TF with voids and axial stress

Benchmarking configurations

  1. Freestanding CS-TF without voids and no axial stress
  2. Bucked CS-TF without voids and no axial stress
  3. Bucked plug-CS-TF without voids and no axial stress

Alternate configurations

  1. Bucked CS-TF with voids and axial stress (no slip)
  2. Bucked plug-CS-TF with voids and axial stress (no slip)

These should contain all the input parameters you would need to replicate the results. Let me know if you need more information!

ITER_buck_CS_TF_novoid_noSz.pdf ITER_buck_CS_TF_yesvoid_yesSz_noslip.pdf ITER_buck_CS_TF_yesvoid_yesSz.pdf ITER_buck_PL_CS_TF_novoid_noSz.pdf ITER_buck_PL_CS_TF_yesvoid_yesSz_noslip.pdf ITER_buck_PL_CS_TF_yesvoid_yesSz.pdf ITER_free_CS_TF_novoid_noSz.pdf ITER_free_CS_TF_yesvoid_yesSz.pdf

daveweisberg commented 1 year ago

@zalzali27 One more clarification I should probably make. The axial stress listed in the input parameters is not the actual value input into the model, rather it is an output (sorry for the confusion). This output is found by scaling the input axial stress by the void fraction coefficient (and additionally by doing the area-weighted average of the CS and TF axial stresses, in the case of the "no-slip" bucked configuration). The real input axial stresses to the model are -50.92MPa for the CS coil and 141.06MPa for the TF coil. This is the same across all configurations, since the source of axial stress is almost completely decoupled from the hoop and radial stress mechanisms.