Aerodynamic effects of autostruts

[renhanxue]

I was playing around with a simple plane with infinite fuel enabled, in order to test the effects of dihedral and high vs low wing, but I was getting some extraordinarily weird results. I was expecting the high wing dihedral arrangement to counteract sideslip disturbances and make the plane not tend to bank into level turns, but that wasn't what was happening at all. This was my test craft:

It is, as far as I can tell, completely symmetrical around the longitudinal axis. Reaction wheels were disabled and I was flying without SAS. Ignore the radial parachutes on the bottom, I was trying to balance out the high wings to neutralize torque from the engine thrust axis being off center mass, but they have no effect on the test and I just removed them.

After taking off and trimming pitch neutral, I was getting a very noticeable and annoying roll tendency to the left, and I just couldn't figure out why. I could trim it out, sure, but where was it coming from? On my actual planes I was planning to fly with SAS on (for autotrim, mostly) and I find that making the plane as stable as possible without SAS on helps a lot with making it fly well. After investigating if there was asymmetric drag or lift going on somewhere and scratching my head for a bit I discovered this little bit of weirdness: if the wings are set to autostrut to heaviest part (the engine), the plane consistently rolls to the left. If they are set to autostrut to the root or grandparent part (the cockpit in both cases), the roll tendency is much less pronounced but still there. With autostruts disabled, the wing configuration does what I expected, and the plane is quite well-behaved in level turns - just bank it over 20-30° in either direction and it will quickly stabilize in a clean turn with no sideslip by itself with no input other than the pitch trim, and it has a weak tendency to bank out of the turn and return to wings level regardless of turn direction. Rigid attachment of the wings seems to have a similar effect of making the plane tend towards to rolling left, but the effect is quite weak and it could easily be a mistake on my part (my testing methodology is hardly all that scientific and the testing environment isn't exactly controlled).

What is going on here? Why does this happen? Surely autostruts don't actually have drag (even if they did, I don't think that would cause this). I've seen some discussion of some old symmetry joint rigidity bug that could have similar effects but I couldn't find it on the bug tracker.

Here is the craft file for the test aircraft (with autostruts and rigid attachment off) in case you want to play around with it yourself. To reproduce my test conditions, take off (it is pretty nose heavy at low speed but you have more than enough pitch authority to counteract that) and get into stable level flight, without SAS, with nose pointed at horizon. Adjust pitch trim so that the plane maintains level flight. I did my tests at quite low AoA - around half a degree. Bank the aircraft left or right and observe how it behaves in a turn. Now, while you're in such a turn, autostrut the wings to the heaviest part and observe the roll tendency. It's more noticeable in a right hand turn because then the high-wing dihedral tendency to return to wings level combines with the tendency to roll to the left. It is also more noticeable at higher speeds.

Edited January 17, 2017 by renhanxue

[fourfa]

Very interesting.  I'm quite sure that it has nothing to do with aero drag, but about tiny deflections of the wing from aero forces - which can change when tweaking the autostruts.  However I don't have any data to back that up right now.

I had a really persistent problem in 1.1.3 where my SSTO cargo plane would fly great until delivering the payload and reloading the scene back to the plane, then would have a pronounced roll on re-entry.  If you never, ever switched away from the plane, no problem.  What I now realize was happening - the landing gear autostruts (the only kind present in 1.1.3) were strutting to heaviest part (something in the payload usually), then on reload were changing to whatever the heaviest remaining part was and causing some weird asymmetry.  If you never switched, it never reloaded and recalculated.  That's my theory anyway - seems to have gone away in 1.2 but I bet there are echoes of it in there

Edited January 17, 2017 by fourfa

[renhanxue]

Further observations: if you grab the offset gizmo and move the wings laterally inwards towards the fuselage so far that they end up swapping places and the "root" of the wing is now the tip (but on the other side of the aircraft), like so:

Then the roll tendency inverts and the plane wants to roll right instead. I think the effect is somewhat less pronounced, though.

As an aside, the elevons work fine in this scenario and still get the roll direction right (even if you don't reattach them after offsetting the wings), but for some reason the roll authority is considerably decreased. There might be something weird going on with the moment arm calculations, I guess?

Another - tangentially related - question: in the AeroGUI, next to the pitch, roll and heading readouts, there is a readout in parentheses that seems to be related to the rate of change in some way. Two numbers are shown, one labelled R and one A. What exactly do they mean?

[OHara]

When I try the craft you linked in KSP version 1.2.2 it is very nearly neutral in roll, returning to wings level with about a 100-second time-constant.  In version 1.1.3 a few short months ago this stability seemed impossible.  The code cleanup leading to version 1.2.0 seemed to help many such details in KSP.

Even when I auto-strut to heaviest, it rolls only 0.5°/s (roughly).  Auto-strut asymmetry was also much worse in version 1.1.3.

Auto-struts are still relatively new, and were added to solve different problems, such as keeping rovers in still cargo bays.  Given the way you approached this aircraft, I think you will get more out of KSP if you leave auto-strutting turned off.

The aero-GUI rate readouts behave as if they were rotation rates, in °/s, with the axes of rotation being aircraft-Relative and Absolute (I'm just guessing words that match the 'A' and 'R' labels and which match the behavior).   Next to heading for example, you see yaw rate relative to the aircraft yaw axis, and then heading-change-rate relative to Kerbin.  Pull your airplane into a 90° bank and pull the stick back hard, and you'll see what I mean.

[renhanxue]

I've been away from the game for a while but I've had it and played it for years and it's certainly better now than it has ever been before, make no mistake about that.

The original purpose of this small scale experiment was really to try to improve the way I design spaceplanes, since I noticed odd roll tendencies just like in the thread you linked and was wondering what caused it. The original spaceplane design I was investigating uses pretty heavy wing-mounted fuel/engine nacelles which I figured could cause weird things if there was some joint flexing going on, so I turned on autostruts thinking "hey, it can't hurt". Then I figured I should experiment at a smaller scale to make sure I actually understood the effects of wing positioning and angling in KSP (consulting the literature - that is to say, digging in forum threads - was inconclusive), and the rest of the story is in the OP.

The actual spaceplane testbed consistently rolls about 0.3°/s or so towards the left with all autostrutting (other than the landing gear's forced one) turned off. Not too bad, but maddening when I couldn't understand why.

Either way, judging by the thread you linked it does seem like this is a known issue that has been around for a while, so that does settle the question in the OP. I guess I should try KJR to see if that improves things - I was thinking I really wouldn't need that with this new nice autostrut feature, which sure seemed to help a lot with stabilizing space stations for example.

Absolute and relative does seem like reasonable explanations for the labels on the rotation rate numbers, so thank you for clarifying that too.

[AeroGav]

These days, i find these phantom rolling moments worse on small aircraft.   On my larger designs, with multiple wing segments ,  it all seems to average out.  It seems that which part the wing attaches to is absolutely crucial.

This had a constant 2 or 3 degree per second roll to the left until i took the wings off and reattached them.  Tail cone connector not a rigid enough part for wing mounting?

This was stable in the previous version, but started to roll in the current release.  The wings were attached via the engine nacelles.  After hooking them directly to the fuselage, the roll went away.

Meanwhile this seaplane ssto ,  with slightly more aerodynamic surfaces, runs straight as an arrow to 40km+.