search

convexengineering / SPaircraft

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

Edits to code post draft submission #97

Closed mayork closed 7 years ago

mayork commented 7 years ago

@1ozturkbe this will be for any edits we might want to make from here on out

mayork commented 7 years ago

all i did was tighten the HT lift coefficient constraint so it is less sketchy and more defensible in the spajoa paper

1ozturkbe commented 7 years ago

Cool. I fixed the bending typo. Unfortunately I used a different material yield stress somewhere than I should have.

whoburg commented 7 years ago

I know you guys are on this but note the test failure -- please don't merge until tests pass

mayork commented 7 years ago

Tests should just be failing due to file name changes

mayork commented 7 years ago

Oh shoot this is a different branch than the one with renames...for some reason only macys failed...odd...

mayork commented 7 years ago

@1ozturkbe i'm going to rename and files and stuff to try and align this with pk_journal. It'll take some time, for now just checkout the last passing commit until a commit starts passing tests.

mayork commented 7 years ago

@whoburg @1ozturkbe is there any interest in including the high fidelity landing gear model in the AIAA Journal paper model? I don't think it would be hard for me to integrate it since I just did for the spajournal model.

1ozturkbe commented 7 years ago

hmm yeah that sounds good. It's probably good not to 'backtrack' to a lower fidelity model anyways. Good call.

mayork commented 7 years ago

If @whoburg agrees with you i'll add it in.

I personally think we probably should to avoid the "back track"

mayork commented 7 years ago

also @1ozturkbe I completed all file renames and folder creations. All that is left now is the standardized wing box.

1ozturkbe commented 7 years ago

The wingbox... that is a little tricky. the HT has something like 6-8 variables that we would need to consistently handle in VT and wing and would be just dummy variables. If we are ok with this, I can take care of this.

1ozturkbe commented 7 years ago

Actually, this may actually help in the case of a wing-engined aircraft, since the constraints are similar. But for the rest, it would just be making sure the variables don't hit the bounds.

mayork commented 7 years ago

@1ozturkbe if we have the variables in the model and they aren't used it isn't an issue. I'm going to proceed and make the change myself (you focus on writing while you have time)...also I'm already on a "code rampage" lol

1ozturkbe commented 7 years ago

K sounds good. Keep up the momentum!

mayork commented 7 years ago

@1ozturkbe pull this and see if you can figure out the import error when you try and import anything from SP_aircraft.py. I swear it was working a few commits ago and nothing seems to have changed.

mayork commented 7 years ago

@1ozturkbe i think that the wing box is now centralized...however...still have that import error

mayork commented 7 years ago

@bqpd i'm getting an import error on this branch (spajoa_edits) and I can't figure out what is happening. Can you take a look?

whoburg commented 7 years ago

I agree with having landing gear in the model but I'm confused -- I thought there was just one model. Do we have two and if so why? I thought the second paper was going to reference the models from the first paper and add the engine. Is that not correct?

Let's try to avoid duplicating a bunch of code, that will be a huge pain later.

mayork commented 7 years ago

@whoburg two models exist only in the sense that the paper I'm leading has the engine integrated and the full flight profile where as the paper Philippe is leading doesn't not have those two models. Everything else in the two models is identical with the exception of inclusion of the landing gear. Hopefully that makes sense.

mayork commented 7 years ago

@bqpd @1ozturkbe I figured out the import error. I never added __init__.py to the repo and then accidentally deleted my local copy.

whoburg commented 7 years ago

ok and where exactly is the code living for each of the two papers?

I hope they share a bunch of imports?

mayork commented 7 years ago

@whoburg the code for the fully system paper we submitted to you is the current master. The PR #93 is the code on master minus the full flight profile and engine plus file renaming and reorganization done by Philippe as well as a few minor tweaks to the underlying equations. This PR is the full system code with the same file renaming and organization and underling constraints as #93.

mayork commented 7 years ago

alright @1ozturkbe i think the code rampage is over now

mayork commented 7 years ago

@1ozturkbe I integrated the landing gear model for just the optimal 737 and it makes us burn like 8% less fuel lol...will you take a look at the solution? everything looked reasonable, not sure why the fractional landing gear weight comes out so low. Maybe we need to adjust some of the substitutions?

1ozturkbe commented 7 years ago

whaaat? Ok, will check.

mayork commented 7 years ago

@1ozturkbe how would you feel about right before submitting the AIAA Journal paper updating all results and merging this branch to master?

1ozturkbe commented 7 years ago

I like the idea. Note that I made a bunch of variable name changes while we were writing the other paper that I will need to import both into the model and our paper.

mayork commented 7 years ago

@1ozturkbe the variable name changes are just in the appendix right?

mayork commented 7 years ago

If so i'm not that worried about fixing that before submission. we can tweak the variable names later

mayork commented 7 years ago

oh btw @1ozturkbe all the weird LG stuff is fixed. results are looking pretty sweet.

1ozturkbe commented 7 years ago

@mayork it just that spajoa-edits needs to be updated with the new variable names. I can do this later (once I'm done with final edits on Philippe's paper).

mayork commented 7 years ago

@1ozturkbe we also need to recreate the following

I think that's all the code changes that aren't accounted for. I'd like to get these done by Thursday

mayork commented 7 years ago

@1ozturkbe solution time has decreased by a factor of 3. I think it's these dumb dxlead and dxtrail things. I vote we take it out, what do you think?

mayork commented 7 years ago

@1ozturkbe never mind I think it might be the landing gear...

mayork commented 7 years ago

@1ozturkbe even master is running slow

mayork commented 7 years ago

@bqpd this is relatively urgent and baffling...our code seems to be running like 3x slower and we have no idea why. This entire branch and the master are doing it. Thoughts? Has mosek slowed down or something? I checked out old commits and am getting the slower speed. The only change I know of is that I renewed my mosek license

@1ozturkbe @whoburg

UPDATE: I really think it is the individual GP iterations that are running slower...

mayork commented 7 years ago

@1ozturkbe how about we just cut the 777 from the paper. The model has never been that stable and doesn't add that much to the paper since it isn't too different from the 737. I think the issue probabaly has to do with increased sweeps, speed, etc. degrading some fits

1ozturkbe commented 7 years ago

@mayork I think we need to get to the bottom of this, and I would rather not cut the 777 section. The big reason to keep it is because it gives people proof that we can scale our model. I do think that the trailing and leading edge constraints may be causing slowdown, because they add many vectorized SP constraints. I will remove those, and perhaps add something a little more sophisticated to calculate the CG of the VT, such as another area centroid calculation. But I do think it is important to have the 777 in there.

mayork commented 7 years ago

@1ozturkbe it's the Dxlead and Dxtrail causing the slowdown just check out the commits prior to that and check the speed, it's the same.

There's else wrong with the 777 model

1ozturkbe commented 7 years ago

I guess there is no longer an issue with those constraints; I think this is because we no longer substitute the wing location like we used to in Philippe's old version of the code.

mayork commented 7 years ago

@bqpd the GP solves each are taking a bit over 0.67 seconds, any idea what is going on here?

@mjburton11 have any of your codes slowed down?

mayork commented 7 years ago

@1ozturkbe how does VSP look?

1ozturkbe commented 7 years ago

Awesome. I'm fixing it up now, with the new variables that we have added. Realized there were issues with splitting of constraints between D8 and conventional configs for the LE/TE.

mayork commented 7 years ago

@1ozturkbe 777 works. is there anything else left to do?

1ozturkbe commented 7 years ago

@mayork merge when you are happy!

mayork commented 7 years ago

@1ozturkbe lets hold off and merge right before submitting the paper just in case anything else crops up (hopefully it wont!)

1ozturkbe commented 7 years ago

K, I'd you prefer. Tbh, I don't see why we don't just merge to master and edit master. At this point, we have all the spajoa edits in, so to me the rest are modeling updates.

mayork commented 7 years ago

noting that we are generating the paper's first submission solution info with commit https://github.com/hoburg/SPaircraft/pull/97/commits/ab23af470a7f4b53ee855c14f2a956b0163a526b

mayork commented 7 years ago

@1ozturkbe main is broken in all the subsystem models. Since the model is fully integrated can we just delete all the non-functioning code we used to run the subsystems one by one?

1ozturkbe commented 7 years ago

Hmm, in a perfect world I would like to get them up and running, but we both have to move on to better things. Feel free to delete.