search

convexengineering / SPaircraft

Models for commercial aircraft design
http://spaircraft.readthedocs.org
25 stars 17 forks source link

run_737800 is not fully relaxed at its final solution #100

Closed bqpd closed 7 years ago

bqpd commented 7 years ago

@mayork, @1ozturkbe 

...
               | Mission/Aircraft/Fuselage/Relax
      C_{fuse} : 1                         Fuselage Weight Margin and Sensitivity
      N_{land} : 1                         Emergency landing load factor
      R_{fuse} : 1.082                     Fuselage radius
           SPR : 1                         Number of seats per row
   W''_{floor} : 1                         Floor weight/area density
...

and in general (as I'm sure you know) this solution is balanced on a knife-edge.

You may not know, though, that it gets a >100% better (and equally relaxed) solution with a relaxed-vars power of 10 instead of 20; I haven't checked if the solution is ridiculous.

1ozturkbe commented 7 years ago

hmm, we haven't run run_737800 in many months... :P Can you run run_optimal_737 instead, and see what you get?

bqpd commented 7 years ago

you guys set it up as the test...

bqpd commented 7 years ago

I'll do a quick check, but in general y'all need to update the tests; this knife-edge is the whole reason the 3x faster solves haven't been merged yet.

1ozturkbe commented 7 years ago

we are really silly @bqpd, our b. Will fix ASAP.

bqpd commented 7 years ago

what arguments should I use for run_optimal_737 ? I'm on master

1ozturkbe commented 7 years ago

it should be ('W{f{total}}', True, False).

mayork commented 7 years ago

@bqpd we've been running on the spajoa_edits branch for a while. I have the tests working there. I'll merge it right now and that will resolve any issues.

bqpd commented 7 years ago

tests are working on gpkit/master, to be clear, but on nomialmaps numerical issues push it to fail occassionally

mayork commented 7 years ago

alright @bqpd everything should be resolved on master now. The test is the optimal 737 run. I'm going to close this but feel free to reopen if you discover something is still messed up.

bqpd commented 7 years ago

Great, thanks!

mayork commented 7 years ago

@bqpd I've been running on nomial maps and haven't seen any issues with the test file. It has never once failed for me.

bqpd commented 7 years ago

Phew. It's passing here...trying Jenkins now. Hopefully this was just a run_737800 issue

1ozturkbe commented 7 years ago

yeah, and the issue with run_737800 is the Rfuse sub. Could have resolved it that way too @mayork, but I don't think we will be using that model in the short term anyways.