Custom objective function (e.g. 'DRAG + LIFT') #1478
Conversation
|
This pull request fixes 2 alerts when merging 4d51419 into 9efe761 - view on LGTM.com fixed alerts:
|
|
This pull request fixes 2 alerts when merging d16a78e into 9efe761 - view on LGTM.com fixed alerts:
|
|
Me: eats gingerbread for 3 weeks |
There was a problem hiding this comment.
Thanks a lot for this @pcarruscag 💐 I've been looking at the CFX Expression Language with envy for quite some time now. And this is a good push into that direction (although not quite as mighty as the CFX CEL yet ;) ). The possibility to specify a per-surface output is really the 🍒 on the 🍰 (i.e. The cherries on the strawberry cake 👍 )
Good job
| string("For multiple surfaces per objective, either use one objective or list the objective multiple times.\n") + | ||
| string("For multiple objectives per marker either use one marker or list the marker multiple times.\n")+ | ||
| string("Similar rules apply for multi-objective optimization using OPT_OBJECTIVE rather than OBJECTIVE_FUNCTION."), | ||
| SU2_MPI::Error("When using more than one OBJECTIVE_FUNCTION, MARKER_MONITORING must be the same length or length 1.\n" |
There was a problem hiding this comment.
I guess this string( wrapper can be removed in quite a few places in the CConfig.cpp
| [submodule "externals/mel"] | ||
| path = externals/mel | ||
| url = https://github.com/pcarruscag/MEL.git |
There was a problem hiding this comment.
Again, nice to have this modularized away (also with a compatible license 😉 )
| TOTAL_PRESSURE_LOSS = 39, | ||
| KINETIC_ENERGY_LOSS = 40, | ||
| TOTAL_EFFICIENCY = 41, | ||
| TOTAL_STATIC_EFFICIENCY = 42, |
There was a problem hiding this comment.
I guess these are never implemented?
There was a problem hiding this comment.
Or they were dropped in v7
| if (config->GetKind_Species_Model() != SPECIES_MODEL::NONE) { | ||
| /// DESCRIPTION: Average Species | ||
| for (unsigned short iVar = 0; iVar < config->GetnSpecies(); iVar++) { |
| % Here we define how the custom objective is computed from other outputs. For | ||
| % example, force in the z direction (computed for all MARKER_MONITORING and part | ||
| % of AERO_COEFF) plus the absolute value of massflow across the second surface | ||
| % ([1]) in MARKER_ANALYZE, scaled by a factor of 1000. It is also possible to | ||
| % use "per surface" values from MARKER_MONITORING (use the dry-run mode to see | ||
| % the names of available outputs, e.g. SU2_CFD -d lam_flatplate.cfg). | ||
| % For multizone problems the CUSTOM_OBJFUNC should be defined for each zone | ||
| % individually (with the outputs of that zone), the total for the problem is | ||
| % the sum over zones, see disc_adj_fsi/Airfoil_2d. | ||
| CUSTOM_OBJFUNC= '1e3 * (FORCE_Z + fabs(SURFACE_MASSFLOW[1]))' |
There was a problem hiding this comment.
🏅 Absolutely amazing! Being able to specify boundary markers is great... of course bonus points would be available for being able to specify the marker-tag instead of a number.
There was a problem hiding this comment.
That is a bit tricky, having first to deduce if it should come from marker analyse or marker monitoring.
But the eventual user defined functions with mass flow averages and so on would use lists of marker names.
Co-authored-by: TobiKattmann <31306376+TobiKattmann@users.noreply.github.com>
|
This pull request fixes 2 alerts when merging 14dabda into cdf780a - view on LGTM.com fixed alerts:
|
Since gcc11 some headers (including limts) are used much less in the standard library headers. See https://stackoverflow.com/questions/4798936/numeric-limits-was-not-declared-in-this-scope-no-matching-function-for-call-t
|
This pull request fixes 2 alerts when merging 130e2dd into cdf780a - view on LGTM.com fixed alerts:
|
|
This pull request fixes 2 alerts when merging 1ea8b02 into 85b2b12 - view on LGTM.com fixed alerts:
|
Proposed Changes
Allows specifying an objective function via a math expression using the history outputs and typical math functions.
General cleanup of dead code and output fixes.
PR Checklist