Efflux Mk. II: Circulation Control and Drooperon Testbed

[Pieliker96]

The pitching moments you describe sound about right. The pitch down is because of a larger airfoil generated pitching moment similar to what all flaps cause at small deflections, while the pitch up is because of a combination of flow separation and an interaction between the wing and tail and suggests that you're still getting flow separation on the flaps at those deflections and angles of attack. A higher aspect ratio wing with a shorter chord should avoid such large high-speed pitch coupling, while a T-tail could mostly avoid the low-speed pitch coupling but would result in a design that was vulnerable to pitch up and it might not totally eliminate it.

I was also thinking about what you were saying about increasing roll rate using the drooperons, and I think you might see something similar. Flaps (and downwards deflecting ailerons) both increase camber and effective angle of attack (by shifting the chord line) but drooperons decrease effective angle of attack while increasing camber. Therefore, you might see a small decrease in roll rate at low angles of attack by deflecting the drooperons and ailerons together. However, they should definitely help at high angles of attack where they should enhance aileron effectiveness and reduce adverse yaw.

I'm planning to use the drooperons to act opposite of the trailing edge: ex. Deflect the TE down and LE up on one wing, do the opposite on the other - this effectively changes the angle of attack of each section. It'll be effective at high speeds but will start to get problematic when the drooperon tends to stall the section by increasing AoA too far. Of course if the wing is unloaded (load factor = C_L = 0; i.e. AoA = constant regardless of airpseed) then it shouldn't matter.

For unloaded roll rate you're looking for the maximum lift coefficient difference between the two wings. When you start to offset that by adding G in the roll you will eventually run into CL_max or CL_min, where it then makes sense to decrease drooperon deflection (or even have it operate in the same direction as the TE) to keep that side from stalling and/or achieve a greater magnitude of CL at that angle of attack.

 

[telnar1236]

I'm planning to use the drooperons to act opposite of the trailing edge: ex. Deflect the TE down and LE up on one wing, do the opposite on the other - this effectively changes the angle of attack of each section. It'll be effective at high speeds but will start to get problematic when the drooperon tends to stall the section by increasing AoA too far. Of course if the wing is unloaded (load factor = C_L = 0; i.e. AoA = constant regardless of airpseed) then it shouldn't matter.

For unloaded roll rate you're looking for the maximum lift coefficient difference between the two wings. When you start to offset that by adding G in the roll you will eventually run into CL_max or CL_min, where it then makes sense to decrease drooperon deflection (or even have it operate in the same direction as the TE) to keep that side from stalling and/or achieve a greater magnitude of CL at that angle of attack.

That makes a ton of sense. I hadn't considered that you could implement drooperons like that. Just watch out for control reversal in maneuvers where you are already at a high angle of attack.

 

[L Edge]

I'm planning to use the drooperons to act opposite of the trailing edge: ex. Deflect the TE down and LE up on one wing, do the opposite on the other - this effectively changes the angle of attack of each section. It'll be effective at high speeds but will start to get problematic when the drooperon tends to stall the section by increasing AoA too far. Of course if the wing is unloaded (load factor = C_L = 0; i.e. AoA = constant regardless of airpseed) then it shouldn't matter.

For unloaded roll rate you're looking for the maximum lift coefficient difference between the two wings. When you start to offset that by adding G in the roll you will eventually run into CL_max or CL_min, where it then makes sense to decrease drooperon deflection (or even have it operate in the same direction as the TE) to keep that side from stalling and/or achieve a greater magnitude of CL at that angle of attack.

How are you doing the deflection process? (Individual or grouped) Make sure you have a way to zero out each of the servos for zero deflection and trim. PITA, but you don't want a roll on takeoff when it is suppose to be off. Faster the speed, the quicker it comes.

 

[Pieliker96]

How are you doing the deflection process? (Individual or grouped) Make sure you have a way to zero out each of the servos for zero deflection and trim. PITA, but you don't want a roll on takeoff when it is suppose to be off. Faster the speed, the quicker it comes.

The plan is to get the plane up in the air, get to some cruise throttle setting in steady level flight, then do a step response stick-neutral aileron input for one full revolution. It won't be automated in any way, just flown by the pilot. Roll rate and airspeed will be averaged over that period. From there, varying the cruise throttle and the stick deflection direction should produce the dataset we need.

Here's an example of the body rates of a full-scale deflection aileron step input: right, then left. I'll have to come up with acceptance / rejection criteria based on g-loading during the maneuver (which should be ~0)

 

[Pieliker96]

We've got data!



I plotted roll rates vs. Airspeed for all cases in the log files where the plane was in-flight and the stick was commanding full deflection either way. Any residual g-loading which reduces roll rate doesn't matter in this case because it simply moves the data further inside the bounded area - i.e. we're concerned about the edges of the data, not the middle. Each maneuver was stopped by 2 full rotations.
The maximum roll rate looks linear with airspeed - i.e, once a roll is fully developed, the airplane will take the same distance along the flight path to complete a revolution regardless of speed. In that way the airplane acts similar to a screw: no matter how fast or slow you drive it in, it'll always turn the same number of times before bottoming out. The max roll rate is approximated by those two lines, whose slopes are in terms of degrees per meter. This plane takes 12 meters to complete a roll once developed to the maximum rate, regardless of the airspeed.

The drooperons, acting opposite to the ailerons with 70% dual rates, did not cause an appreciable change in the maximum roll rates with the standard camber setting. I'll be re-testing this at the reduced-camber setting to see if they will with a more symmetrical profile. Beyond that I could also try increasing and decreasing drooperon deflection.




The low-camber data shows a max normal coefficient of 0.66, or 0.10 less than the standard camber configuration. Initial maximum roll rates were observed to be higher with this camber setting, north of 800deg/sec. Also present in this chart is a brief post-stall excursion I had, similar to the snap-roll on the second flight attempt. I pulled too hard and it rapidly diverged from controlled flight, completing two full rotations in yaw on the order of 500deg/sec before I regained control.




Finally, here's envelope data for the LE drooped and the flaps at full. Max normal coefficient is a whopping 1.33 as compared to the standard camber's 0.75 - Stall speed is reduced from 13 m/s to just under 10. My pitch trim settings are close to dialled in for the flaps now, as is my adverse-yaw mix to the rudder. The flaps are also massively draggy - the power required to keep the plane moving forwards goes way up.

Next flights will focus on testing the drooperons for roll rate at the reduced camber setting, as well as testing the flaps without the leading edge droop against the existing data. I'll also have to replace the other flaperon servo, since it inexplicably died somewhere between final approach of flight 16 and a post-flight circulation control demo.

 

[Pieliker96]

I've discontinued drooperon testing because I think I can get more valuable data by moving to the circulation control part of the experiment.

Here's the roll envelope as it stands. I tested two different drooperon mixes, one in either direction, against baseline in the standard camber wing configuration. The results seem the same regardless of drooperon mix, with an outlier or two from snap-rolls. It's possible they could have a greater effect when the wing is loaded - i.e. maybe they could prevent one wing from stalling and therefore maintain some lift coefficient differential between the two. However the practical test case for this looks like something leading up to where I've experienced snap rolls with in earlier flights, so I doubt the ability to generate useful data here. Testing at different baseline camber settings may also yield a difference - or not, the only way to know is to test.



While the drooperon mixes may be quantitatively the same in maximum roll rate, they are not in piloting. I personally found the negative mix (i.e. moving both leading and trailing edge in the same direction, both up or both down, altering camber between each wing as opposed to angle of attack) to generate a significant pitch-up moment which I needed to counter, leading to one of the aforementioned snap-rolls which can be seen in the data as a spike to -1000deg/sec. Standard ailerons felt fine, and standard drooperons felt a tad different but were manageable with no significant pitch trim changes in the roll.

Next up is the flight envelope testing with the flaps down, with and without the leading-edge droop. The data looks functionally identical, it's hard to tell a difference. The problem with not having a test pilot in an actual plane, or telemetry, is that I can't really get to particular flight conditions, or do things like detect a stall very easily. Another confound is that the plane measures it's normal acceleration, not its acceleration relative to the flight path. This means I can only directly calculate normal force, but not lift force. The two are geometrically related by angle of attack, which means they are reasonably coupled close to zero alpha, and this is why we can still see differences in the flight envelopes due to camber setting. However, when angle of attack gets large, in stalls or otherwise, the normal coefficient diverges from the lift coefficient by exceeding it. These are seen in the data as excursions well above the bounding parabola due to the fact that aerodynamic drag force is now the normal force: in the extreme case of 90 degrees angle of attack, falling straight down at terminal velocity, the lift coefficient would be zero and the normal coefficient would be one. Dynamic effects allow that peak normal coefficient to far exceed to maximum lift coefficient due to the high amounts of drag produced post-stall, which makes data analysis difficult.

My approach so far has been to fly such that visually and by my feedback loop with the controls I can use my intuition about stall to max-perform the wing while still avoiding it. There were some cases in these most recent flights where I had to put in a good bit of down-elevator and dump the nose to avoid or get out of stall, which may be those humps above the parabolas at 13 and 17 m/s. If I disregard those and squint really hard at the data I may have seen an increase in the maximum lift coefficient from drooping the leading edge, but if I'm being honest with myself if there is a difference it's masked by flawed test methodology.



A method to improve my testing of normal coefficient would be to perform max performance turns at different throttle settings and gradually increase the turn rate and decrease the radius until the aircraft stalled - the stall point may be more distinguishable then, especially if the snap-roll tendencies showed themselves. Plotting the calculated normal coefficient instead of a V-N diagram may also help make seeing patterns in the data easier, by making the cutoff between stall and normal flight a line instead of a parabola.

More down-trim was needed without the droop, which makes sense as the angle of attack of the wing was effectively increased.

Onto circulation control testing!

 

[telnar1236]

I've discontinued drooperon testing because I think I can get more valuable data by moving to the circulation control part of the experiment.

Here's the roll envelope as it stands. I tested two different drooperon mixes, one in either direction, against baseline in the standard camber wing configuration. The results seem the same regardless of drooperon mix, with an outlier or two from snap-rolls. It's possible they could have a greater effect when the wing is loaded - i.e. maybe they could prevent one wing from stalling and therefore maintain some lift coefficient differential between the two. However the practical test case for this looks like something leading up to where I've experienced snap rolls with in earlier flights, so I doubt the ability to generate useful data here. Testing at different baseline camber settings may also yield a difference - or not, the only way to know is to test.

View attachment 232023

While the drooperon mixes may be quantitatively the same in maximum roll rate, they are not in piloting. I personally found the negative mix (i.e. moving both leading and trailing edge in the same direction, both up or both down, altering camber between each wing as opposed to angle of attack) to generate a significant pitch-up moment which I needed to counter, leading to one of the aforementioned snap-rolls which can be seen in the data as a spike to -1000deg/sec. Standard ailerons felt fine, and standard drooperons felt a tad different but were manageable with no significant pitch trim changes in the roll.

Next up is the flight envelope testing with the flaps down, with and without the leading-edge droop. The data looks functionally identical, it's hard to tell a difference. The problem with not having a test pilot in an actual plane, or telemetry, is that I can't really get to particular flight conditions, or do things like detect a stall very easily. Another confound is that the plane measures it's normal acceleration, not its acceleration relative to the flight path. This means I can only directly calculate normal force, but not lift force. The two are geometrically related by angle of attack, which means they are reasonably coupled close to zero alpha, and this is why we can still see differences in the flight envelopes due to camber setting. However, when angle of attack gets large, in stalls or otherwise, the normal coefficient diverges from the lift coefficient by exceeding it. These are seen in the data as excursions well above the bounding parabola due to the fact that aerodynamic drag force is now the normal force: in the extreme case of 90 degrees angle of attack, falling straight down at terminal velocity, the lift coefficient would be zero and the normal coefficient would be one. Dynamic effects allow that peak normal coefficient to far exceed to maximum lift coefficient due to the high amounts of drag produced post-stall, which makes data analysis difficult.

My approach so far has been to fly such that visually and by my feedback loop with the controls I can use my intuition about stall to max-perform the wing while still avoiding it. There were some cases in these most recent flights where I had to put in a good bit of down-elevator and dump the nose to avoid or get out of stall, which may be those humps above the parabolas at 13 and 17 m/s. If I disregard those and squint really hard at the data I may have seen an increase in the maximum lift coefficient from drooping the leading edge, but if I'm being honest with myself if there is a difference it's masked by flawed test methodology.

View attachment 232024

A method to improve my testing of normal coefficient would be to perform max performance turns at different throttle settings and gradually increase the turn rate and decrease the radius until the aircraft stalled - the stall point may be more distinguishable then, especially if the snap-roll tendencies showed themselves. Plotting the calculated normal coefficient instead of a V-N diagram may also help make seeing patterns in the data easier, by making the cutoff between stall and normal flight a line instead of a parabola.

More down-trim was needed without the droop, which makes sense as the angle of attack of the wing was effectively increased.

Onto circulation control testing!

Do you have a way to measure acceleration in roll? Your roll rate might stay similar, but I would expect you to reach that max roll rate faster

 

[Pieliker96]

Do you have a way to measure acceleration in roll? Your roll rate might stay similar, but I would expect you to reach that max roll rate faster

Roll acceleration can be derived by simply taking the delta between roll rate data points and dividing by timestep.



Qualitatively the data looks very similar for all three test cases. The fixed leading edge case does have higher peaks than the others but it's hard to really say how significant these are. The data I used to generate these was designed to achieve maximum rate, not acceleration: for example, sometimes I would need to add in a bit of pitch to unload the wing more and increase the roll rate, which would happen well after the initial jump from full aileron deflection: it's possible higher roll accelerations could be achieved if the appropriate pitch trim was done in conjunction with the aileron step input.

Physically speaking, roll acceleration is limited by control surface actuator speed, roll inertia, and aerodynamics whereas maximum sustained roll rate is just aerodynamically constrained: the former is dynamic where the latter is quasi-steady-state. Since this data seems to mirror the roll rate data it suggests that the actuators and roll moment of inertia are consistent - which makes sense, since I didn't change the model of servos or the plane's mass distribution between tests.

Put another way: If the actuator performed the same, getting the control surfaces to the same maximum deflection in the same time, any difference in acceleration would literally be the result from a difference in roll rate slopes which would necessarily result in different roll rates once the system had come to equilibrium: the higher acceleration would be due to the same rise time to a higher roll rate. Since the maximum roll rates we saw earlier were pretty much the same, and the actuator dynamics and roll inertia were the same, we would expect to see the same pattern repeated in the acceleration data as the rate data, which we do.

 

[telnar1236]

Roll acceleration can be derived by simply taking the delta between roll rate data points and dividing by timestep.

View attachment 232036

Qualitatively the data looks very similar for all three test cases. The fixed leading edge case does have higher peaks than the others but it's hard to really say how significant these are. The data I used to generate these was designed to achieve maximum rate, not acceleration: for example, sometimes I would need to add in a bit of pitch to unload the wing more and increase the roll rate, which would happen well after the initial jump from full aileron deflection: it's possible higher roll accelerations could be achieved if the appropriate pitch trim was done in conjunction with the aileron step input.

Physically speaking, roll acceleration is limited by control surface actuator speed, roll inertia, and aerodynamics whereas maximum sustained roll rate is just aerodynamically constrained: the former is dynamic where the latter is quasi-steady-state. Since this data seems to mirror the roll rate data it suggests that the actuators and roll moment of inertia are consistent - which makes sense, since I didn't change the model of servos or the plane's mass distribution between tests.

Put another way: If the actuator performed the same, getting the control surfaces to the same maximum deflection in the same time, any difference in acceleration would literally be the result from a difference in roll rate slopes which would necessarily result in different roll rates once the system had come to equilibrium: the higher acceleration would be due to the same rise time to a higher roll rate. Since the maximum roll rates we saw earlier were pretty much the same, and the actuator dynamics and roll inertia were the same, we would expect to see the same pattern repeated in the acceleration data as the rate data, which we do.

Fair enough. I would actually say that the acceleration data does show at least one useful trend and that it shows your system does have an effect. If you look at the -70% data points, the roll accelerations are overall noticeably lower than the other two configurations, maybe by as much as 10-15%. This is most obvious in the negative direction, where roll rates are roughly equivalent for all options, but the accelerations are almost universally lower. This could be coincidence or pilot induced since the difference isn't that big, but it's along the lines of what you'd expect from aerodynamics. It would also inform future tests, since it would suggest that the +70% drooperons might be too much and that you might do better with smaller deflections, or differential deflection, if possible, to emphasize the downwards moving wing (would also reduce adverse yaw and thereby improve roll rate). From the sound of it, additional tests might not be worth it, but I think you could definitely get the drooperons working.

My thinking is that your roll rate should have a linear relationship with airspeed (roll damping and aileron effectiveness both increase with the square of airspeed, so the ratio of max roll rate to airspeed should be about constant). From your plots, this appears to hold true for your design. However, your initial roll acceleration is not dependent on roll damping, so since your aircraft's moment of inertia should remain constant, your initial roll acceleration should increase with the square of airspeed. This might amplify differences in effectiveness at higher airspeed. So long as the actuator speed stays constant across the speed range, which it should, you should still see this trend. I'm not sure why it isn't present in you plot, but it could be anything from the fact that, as you said, you weren't trying to achieve maximum acceleration, to too large of a time increment when recording data, to the airspeed range being to small and the low end exhibiting non-linear dynamics, to aeroelastic effects at higher speed.

Personally, I really like dynamic data since it better separates the different forces and moments acting on a system, but it can also be a real pain to work with, so steady-state data definitely has its place as well.

I look forward to seeing the results of the circulation control tests!

 

[telnar1236]

Roll acceleration can be derived by simply taking the delta between roll rate data points and dividing by timestep.

View attachment 232036

Qualitatively the data looks very similar for all three test cases. The fixed leading edge case does have higher peaks than the others but it's hard to really say how significant these are. The data I used to generate these was designed to achieve maximum rate, not acceleration: for example, sometimes I would need to add in a bit of pitch to unload the wing more and increase the roll rate, which would happen well after the initial jump from full aileron deflection: it's possible higher roll accelerations could be achieved if the appropriate pitch trim was done in conjunction with the aileron step input.

Physically speaking, roll acceleration is limited by control surface actuator speed, roll inertia, and aerodynamics whereas maximum sustained roll rate is just aerodynamically constrained: the former is dynamic where the latter is quasi-steady-state. Since this data seems to mirror the roll rate data it suggests that the actuators and roll moment of inertia are consistent - which makes sense, since I didn't change the model of servos or the plane's mass distribution between tests.

Put another way: If the actuator performed the same, getting the control surfaces to the same maximum deflection in the same time, any difference in acceleration would literally be the result from a difference in roll rate slopes which would necessarily result in different roll rates once the system had come to equilibrium: the higher acceleration would be due to the same rise time to a higher roll rate. Since the maximum roll rates we saw earlier were pretty much the same, and the actuator dynamics and roll inertia were the same, we would expect to see the same pattern repeated in the acceleration data as the rate data, which we do.

In fact, I would go one step further and say that the drooperons already work too (see my very scientific best fit lines that I drew in Microsoft Paint). Also, never mind what I said about the data not being parabolic. It just doesn't extend to zero, but if it did a parabola would be the best fit.

 

[Pieliker96]

I ran into a hardware limitation that required software correction: The bypass door servo isn't strong enough (and I wouldn't expect it to, the dynamic pressure is insane) to retract it at medium to high throttle settings. To counter this, I programmed a gated SR latch that only allows me to change the door's state at low throttle settings where the servo can cope with the requirements. I've just got to remember that I'll have to chop the throttle to change the door position, though that's preferable to potentially jamming the door closed and causing an engine-out.

 

[Pieliker96]

Well there's good news and... eh? news

The bad news is that the diverter door failed structurally and jammed to block the outlet on the last flight. By the time I noticed I had no thrust, I was just above the trees and too far to glide from the field. I attempted to un-stick the door by chopping the throttle, which induced a stall while trying to maintain flight above the trees - I then lost control and it went in at a 45 to 60 degree downline.



The upshot of the bad news is that it really isn't that bad. It managed to avoid the areas of marsh and standing water, and had its landing cushioned rather well by some thick Florida flora. All of the electronics work and appear undamaged, though I will have to clean dirt out of the pitot tube. The airframe fared well structurally, with the left side of the fuselage taking most of the load (peaking at 15g!) - With a repaired diverter door and a little TLC I have no doubt it'll be back in the air again.



The other good news is that I was successfully able to demonstrate an increase in the maximum normal coefficient with the circulation control system active: up to 1.54 from 1.33. The actual maximum value may be even higher as I was unable to push the wing to stall due to the large up-elevator trim requirements that came with adding power: a modified elevator, or extending the circulation control system to the elevator, may be in order.

 

[telnar1236]

Well there's good news and... eh? news

The bad news is that the diverter door failed structurally and jammed to block the outlet on the last flight. By the time I noticed I had no thrust, I was just above the trees and too far to glide from the field. I attempted to un-stick the door by chopping the throttle, which induced a stall while trying to maintain flight above the trees - I then lost control and it went in at a 45 to 60 degree downline.

View attachment 232719

The upshot of the bad news is that it really isn't that bad. It managed to avoid the areas of marsh and standing water, and had its landing cushioned rather well by some thick Florida fauna. All of the electronics work and appear undamaged, though I will have to clean dirt out of the pitot tube. The airframe fared well structurally, with the left side of the fuselage taking most of the load (peaking at 15g!) - With a repaired diverter door and a little TLC I have no doubt it'll be back in the air again.

View attachment 232718

The other good news is that I was successfully able to demonstrate an increase in the maximum normal coefficient with the circulation control system active: up to 1.54 from 1.33. The actual maximum value may be even higher as I was unable to push the wing to stall due to the large up-elevator trim requirements that came with adding power: a modified elevator, or extending the circulation control system to the elevator.

View attachment 232717

Congratulations on the successful test! It looks like you could have noticeable improvement in lift, though the data isn't totally clear. What is the qualitative difference like? Sorry about the crash; hopefully the repairs go quickly and you can get the diverter to be more reliable.

 

[Pieliker96]

Congratulations on the successful test! It looks like you could have noticeable improvement in lift, though the data isn't totally clear. What is the qualitative difference like? Sorry about the crash; hopefully the repairs go quickly and you can get the diverter to be more reliable.

Qualitatively I need a lot more aft stick. Besides that there's not much difference other than I'm basically gliding all the time - while I think the data shows the increase in max normal coefficient is there it's not as substantial as the jump between flaps and no flaps, thus hard to feel.

 

[L Edge]

Nice save. You gained some valuable info there. My experimental foamy's always look like a wrinkly old man from all the crashes before I discard it.

Next testing is to see if the diverted EDF flow can act like a aileron(circular tube with slot in it) so it can reduce weight in a commercial plane.