Closed 1ozturkbe closed 6 years ago
interesting test failures...is it fragility or different MOSEK versions?
test this please
Works on my computer... perhaps mosek versions? Let me dig into the error message. Update: Hmm, definitely not the robustness.
WOOOOHOOO ready to merge. Both multimissions run with no maxOPR/BPR optimization (True,False args) and with full opt. (False, True args), for optimal_737 and optimal_D8. I suspect it had to do with TE/LE vectorized signomial inequalities :P
@mayork would appreciate your review ASAP. Would like to merge so that we can get master up to date with all of the code clean-ups.
And as a heads-up, please do not squash merge, and just lmk when you have gone through. I would like to be able to reference the specific commits in the future.
@1ozturkbe review done. The changes looked good but I want to document any solution changes between this branch and what we have in the paper.
overall looks really good
What do you mean exactly by document? My reaction will depend on the depth of the documentation asked. I'm even skeptical about keeping the same input parameters in master for the sake of having it match the code. We do have a commit hash in the paper, so if someone wants to run exactly what we used to have they can easily do that.
Also, the modeling changes in this PR were the most minute. It was mostly to improve the cleanliness and convergence of the code.
Hey @mayork , just checked the changed against the 737 values. The weights are all within 1klbs. For the D8, our weights a little higher now (+2klbs for both dry weight and fuel weight), likely because of the weird engine debugging code that had not been removed. The 777 is way lighter now in MTOW (530klbs vs 590klbs), no idea why. Investigating...
Are there any relaxed variables for the 777?
On Nov 4, 2017, at 11:20, Berk Ozturk notifications@github.com wrote:
Hey @mayork , just checked the changed against the 737 values. The weights are all within 1klbs. For the D8, our weights a little higher now (+2klbs for both dry weight and fuel weight), likely because of the weird engine debugging code that had not been removed. The 777 is way lighter now in MTOW (530klbs vs 590klbs), no idea why. Investigating...
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
Also check the flight profile plots (the ones that compare ours to TASOPT...sometimes the 777 flight profile gets super weirded out)
On Nov 4, 2017, at 11:20, Berk Ozturk notifications@github.com wrote:
Hey @mayork , just checked the changed against the 737 values. The weights are all within 1klbs. For the D8, our weights a little higher now (+2klbs for both dry weight and fuel weight), likely because of the weird engine debugging code that had not been removed. The 777 is way lighter now in MTOW (530klbs vs 590klbs), no idea why. Investigating...
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
No relaxed variables. Added yaw rate constraint to make VT more reasonable. But here is the issue: engine is small (33% light) = VT small = CG close to wing = HT small. So the aircraft is 12.5% light. And flight profile is fine.
If the engine is small it's due either to the flight profile (time to climb constraint needs to be more restrictive, check the plot), increased gradient at TOC (reasonable since 777 is at a lower initial cruise -> should need to have better climb gradient to have some ceiling), or too low of an initial rate of climb (possibly related to #1)
My bet is it's the time to climb or climb gradient. Check the 777 TASOPT climb gradient at TOC.
On Nov 4, 2017, at 11:40, Berk Ozturk notifications@github.com wrote:
No relaxed variables. Added yaw rate constraint to make VT more reasonable. But here is the issue: engine is small (34% light) = VT small = CG close to wing = HT small. So the aircraft is 15% light.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
Hmm TOC rate is 1500ft/min, but it is flattening around 34kft, so perhaps increasing this is an idea. I think it makes sense to make it cruise higher so that the TOC rate works at 38kft. The initial rate of climb is fine. I also don't want to fudge it too much artificially.
Alright everything looks great. Once the tests pass I will merge.
fyi we are within 1% of paper for 777 as well.
Sweet, great work 👍👍
On Nov 4, 2017, at 12:04, Berk Ozturk notifications@github.com wrote:
Merged #107.
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.
Opening to be able to see all of the changes. Can discuss methods to get this working here as well.