New example cases and HyShot II validation harness (#87)#105
Conversation
- 01_sod_shock_tube: validates numerical scheme options (WENO5/O1, HLLC/LF, double-flux) against the analytical Sod solution - 02_isentropic_nozzle: validates geometry and area-change handling against quasi-1D isentropic nozzle theory - 04_jicf_heat_release_profile: JICF heat release profile case (in progress — reveals areas for solver improvement) - 06_autoignition_vitiated_flow: autoignition in vitiated flow (in progress — reveals areas for solver improvement)
Adds validation.py, which reproduces the HyShot II wall pressure and heat-flux comparisons (fuel-on, fuel-off, and equivalence-ratio sweep) against experimental measurements and 3D RANS (CTR) results, computing L1/L-infinity error metrics. Includes the reference-data CSVs and an example README following the numbered-example pattern. The validation figures are reproduced from StanShock result CSVs. Live end-to-end simulation of the FPV+schedule case is pending JICF fixes (mesh dependence and injector-origin non-finite values).
|
Hey @jjhans , sorry I didn't see these pull requests until just now! I don't get email notifications unless you tag me or assign me as a reviewer. I'll try to look through them and leave a few comments on each one. One general comment: Don't forget to run |
| @@ -0,0 +1,79 @@ | |||
| Experiment (Center Point),,Experiment (Upper Bound),,Experiment (Lower Bound),,Inert-HeafFlux-Body-MenterSST, | |||
There was a problem hiding this comment.
Typo: "HeafFlux"
Not a big fan of having a label for every other column, then a second label of X + Y for each. It was a little difficult to understand and takes extra effort to parse using standard tools like np.loadtxt. I would instead suggest the following changes:
- Drop "experiment" and move experimental data into their own files, try to avoid empty data points.
- "Center" should be "Mean."
- Only keep one header row, don't worry about repeating (e.g.
Mean X,Mean Y,Upper X,Upper Y,...). - Where possible, collect x-locations into a single column. E.g. experiment lower and upper have the same X coordinates, and they're almost the same as the mean X coordinates, so just have an "X Location [mm]" column followed by each Y data column.
- Variable names should not unnecessarily repeat information that's already in the file name, i.e. "Inert" "Body", and (if only one model included) "MenterSST".
- You can easily collect the following:
- Experimental mean, lower, and upper pressures for the inert and reacting cases (1 X + 6 Y columns) for Body and Cowl (2 files)
- Same, but for heat flux (2 files, 7 columns each)
- RANS equivalents for the above, named columns for different cases and turbulence models where appropriate (e.g. "React (Spalart-Allm.)", "Inert (Menter-SST)", "React 1/8C No Walls (Menter-SST)"). Don't strictly need to include "Pressure" or "Heat Flux" as it's in the name of the file. I'm open to other suggestions on how to improve this one.
- I'm not quite sure what the EQ file contains, so I'll have to come back to it later.
|
|
||
| import matplotlib.pyplot as plt | ||
| import numpy as np | ||
| import pandas as pd |
There was a problem hiding this comment.
I would prefer to avoid bringing Pandas in as a dependency just for example cases like this. It's a big library, and most of what you're doing with it here should be pretty easy to do using NumPy. If you use np.genfromtxt with names=True, it will create a structured array with column names taken from the header.
| def load_stanshock_result(path: str | Path) -> StanShockResult: | ||
| """Load a StanShock CSVWriter result file. | ||
|
|
||
| Expects the columns emitted by ``CSVWriter``: ``x [m]``, ``pressure [Pa]``, |
There was a problem hiding this comment.
My output data doesn't include the units, not sure if it was generated using a different version of CSVWriter than yours.
2 participants
Opens the pull request for
87-new-example-cases(issue #87), including the HyShot II updates. Per the plan, issue #87 stays open — the purpose of this PR is to pin down the remaining changes.What's included
04_jicf_heat_release_profile,06_autoignition_vitiated_flow).examples/hyshot_ii/):validation.pyreproduces the wall pressure and heat-flux comparisons — fuel-on (reacting), fuel-off (inert), and an equivalence-ratio sweep — against experimental measurements and 3D RANS (CTR) results, computing L1 / L-infinity error metrics. Includes the reference-data CSVs (reference_data/) and a README following the numbered-example pattern.Verified
validation.pyreproduces the three reference figures (fuel-on, fuel-off, phi-sweep) from StanShock result CSVs, matching the prior MATLAB reference plots.Hyshot2InterfaceFPV case constructs and initializes correctly on theAsymmetricBoxgeometry (200 + 6 ghost cells, valid time step).Known remaining work (to pin down)
The FPV + schedule case does not yet advance to completion on the
AsymmetricBoxgeometry:Z_3D: divide-by-zero at the injector origin (x_cl = 0) insrc/stanshock/models/jicf.py(~L291/307/308/383/385) poisons the 3D mixture-fraction array, so constructing the cubicRegularGridInterpolatorfails withValueError: RHS must contain only finite numbers. Tied to the JICF dead-code / cleanup work.Z_3Dgeneration takes ~10 min (69 axial slices at ~10 s each).hyshot_ii_frc.py) is not yet functional (WIP).Once #98 (and the injector-origin cleanup) land,
hyshot_ii_jic.pycan be wired to run the live simulation and feedvalidation.pyend-to-end.Relates to #95, #98, #104.