trying to spin

[de profundis]

Hey guys,

I was wondering whether basic and crucial aerodynamic behavior of aircraft is applicable in FlightGear.
I mean stalls, spins, critical AoAs, realistic crabbing and slipping on crosswind landings, correctly simulated behavior in the “area of reverse command” etc.

I got here, searching the forum for answers and I think this is the right thread.

Even though, I know zero about FDMs (except that there are two kinds mostly used in FG), I had realized that simulating conditions near or outside aircrafts’ envelopes would require some specific coding of their FDM. From what you are discussing here, this coding seems very complicated.

The reason I am posting here is to provide a link to a very informative and interactive course on the basic aerodynamic concepts (Essential Aerodynamics: Stalls, Spins, and Safety scroll down that page and you'll see it), from the Air Safety Foundation of AOPA USA, hoping that it can assist your effort in better simulating these factors. Creating an account there is free and easy and you neither have to be a certified pilot nor a USA citizen. I am amazed from the quality and abundance of educative information they freely provide

I will do some testing with the aircrafts you propose (c172 + tweaked FDM by dany93) and I might be able to provide some feedback. Imo though, the focus should be in improving the “simpler” and essential behaviors for a more realistic everyday experience, like crabs, slips, mushes etc. and then consider the more complicated aerodynamics of spins.

many thanks for your work
Michalis
ps: check this out too...

[Johan G]

ps: check this out too...

That was a very interesting thought. Focus on early stall prevention through early and regular drills vs. training a few times doing full stall (or even spin) recovery during pilot training. It have not really crossed my mind before, but as I see it early and regular early prevention drills would be way better than to have pilots raise the nose and ignore stall horn the few times stall is approached during pilot training.

The way I see it it is all about drill. In bad weather and with aircraft lining up behind you to avoid going to their alternates you should not have to think when the aircraft starts to buffet or the stall horn sounds. Just follow the drill as if it where a reflex.

[dany93]

....I mean stalls, spins, critical AoAs, realistic crabbing and slipping on crosswind landings, correctly simulated behavior in the “area of reverse command” etc.

Imo though, the focus should be in improving the “simpler” and essential behaviors for a more realistic everyday experience, like crabs, slips, mushes etc. and then consider the more complicated aerodynamics of spins.

Hi Michalis,

Thanks for your links.
Glad to see your interest in this subject.
I have exactly your way of thinking. I've applied it to DR400 JSBSim as well as I could with my knowledge (see below) but, in a first step, I didn't dare to make important modifications to C172P (taking into account the history and importance of this aircraft in FG).

However, tests after tests, I've been led to try doing some refinements on my C172P's FDM version (which may be pushed if considered as valuable):
- Forward-slip bank angle (increased),
- cross-wind landing (for 20 kt): side-slip bank angle against wind, decrabbing (by crossing the commands),
- slip-skid ball response (more sensitive).
- deletion of some actions of which I didn't see the justification.

The question is that these behaviors are often partly qualitative, subjective. Difficult to be sure if an improvement is truly closer to reality. Still, I (IMHO and currently...) consider these as improvements. Waiting for feedbacks, hopefully...

As you are really interested, here is my current version of the FDM for C172P (including all that above + stall and spin). C172P_FDM_20140116 . [EDIT] Rather look for the last one in this thread [/EDIT]
For stall and spin particularly, I'm not still truly satisfied for now by some points that I cannot improve, and I'm permanently hesitating between different compromises. But, as I've already said, my script for stall and spin in rather basic and, anyway, that's evolution.

Also if you like, I've applied these principles for a longer time to DR400 JSBSim (my signature, Patrouille de France hangar). But rather take my last version DR400_20140116, still improved from my point of view. Take Robin DR 400-180 (JSBSim). Rembrandt doesn't matter here. Don't bother for the messages in the console, due to dds, it works well.

Needless to say that I'm very interested if you give your feedback.

Thanks and regards,

Dany

[hvengel]

Hey guys,

I was wondering whether basic and crucial aerodynamic behavior of aircraft is applicable in FlightGear.
I mean stalls, spins, critical AoAs, realistic crabbing and slipping on crosswind landings, correctly simulated behavior in the “area of reverse command” etc.

I got here, searching the forum for answers and I think this is the right thread.

Even though, I know zero about FDMs (except that there are two kinds mostly used in FG), I had realized that simulating conditions near or outside aircrafts’ envelopes would require some specific coding of their FDM. From what you are discussing here, this coding seems very complicated.

...

I will do some testing with the aircrafts you propose (c172 + tweaked FDM by dany93) and I might be able to provide some feedback. Imo though, the focus should be in improving the “simpler” and essential behaviors for a more realistic everyday experience, like crabs, slips, mushes etc. and then consider the more complicated aerodynamics of spins.

many thanks for your work
Michalis...

Yes all of the flight envelope including those parts that are outside of the normal operational envelope (stalls and spins) are import at least in most aircraft. There maybe exceptions such as highly automated modern airliners that are designed to actually prevent stalls. But these systems are not fool proof so even there modeling stall/spin behavior probably needs to be done to consider the FDM to be "complete" ("complete" is in quotes because there will almost always be room for improvement in the FDM no matter how much work has been done to it).

A JSBSim FDM with enough aero data does everything including symmetric (IE. no yaw - mushing) stalls nicely. Getting these things to behave correctly is basically a matter of getting the various drag, lift and moment data correct in the FDM. This is a big task but it can be done and there are a number of JSBSim aircraft where this work has been done to a level to get this behavior to be at least plausible. Once the aero data in the FDM is up to snuff things like crabs, slips, mushes (IE. symmetric stalls) and so on just work and all of the better JSBSim aircraft do these things in at least a plausible way.

Things become much more complicated when it comes to spin behavior since the JSBSim engine simply does not provide direct support for modeling this part of the flight envelope. So we need to back into it through the back door by adding code to create the forces involved in spin behavior to the FDM. Adding the code needed to make an aircraft with docile spin characteristics, like the C172, spin in a plausible way is not difficult. You can see this by looking at the code dany93 added to the C172 FDM which is only about 30 lines of added code give or take. Unfortunately this is not a general solution to the spin problem but it does give some very strong hints about what needs to be done to get a general solution. The P-51D spin code started out being based on the C172 spin code and after a lot of experimentation additional factors involved in spin modeling were uncovered along with how these factors need to be tuned to get various spin characteristics (for example how much flat tendency does the spin have, how fast does it spin, how fast does the spin develop, which wing - if any - will tend to drop first...). The P-51D spin code like the C172 code uses hard coded values in look up tables to tune the spin characteristics and because of this is very difficult to tune because it is more complex (IE. tuning is needed in a number of different functions and these functions all interact with each other). But with tuning it can be made to behave in a very broad range of ways all the way from very docile to extremely dangerous.

The problem of course is that at this point I am the only one who has any idea how the P-51D spin code actually works. What I want to do is to extract the spin code and then generalize it so that it can be tuned by setting a handful of parameters (fewer than a dozen) and then documenting what those parameters are and how they work so that others can add tuned spin behavior to their JSBSim aircraft in a fairly simple way. I am hoping that this generalized spin code can be added to an FDM and tuned to have plausible spin behavior in a few hours (4 to 8 hours) by someone who is not familiar with it. I am going to try to get a first cut at extracting and generalizing the spin code this week end.

As dany93 pointed out earlier in this thread only after the basic FDM is working correctly should spin behavior be modeled since the spin behavior needs to be configured to the specific lift and drag profile of the aircraft and this is only known once the basic FDM is working correctly. Spinning is definitely an advanced FDM feature for JSBSim FDMs.

[shock.r]

If you wanna spin, start up your A380 and take it to around 20000" and go for it. It looks amazing and I love playing it back in Fly-By view.

[dany93]

Hi hvengel,

(for information, sorry if that's obvious for you but it's not used in your P51D)
Just in case your 'stall and spin code' would seem too simplistic for you .

I have been searching for a long time how to use 'stall-hyst-norm' (found in jsbsim Internal Properties). My opinion was (is) that it can improve realism by delaying the stall recovery (E.G. I've often seen rapid bank-angle oscillations by too rapid recovery of AoA critical angle, or I feel stall recovery is too easy, too rapid at least with my simple code).

'stall-hyst-norm' is implemented and set by inserting such lines in the FDM (inside the <aerodynamics> tags, DEG or RAD units), see C172P or SenecaII.

Code: Select all

        <hysteresis_limits unit="RAD">
            <min>0.09</min>
            <max>0.297</max>
        </hysteresis_limits>

(otherwise, don't know for the use of <alphalimits>)
Important, from my point of view. Still to observe (for me) and to try how to use it well. At first tests, I find a true efficiency on hardness to recover from stall and spin with my code.

From the feedback(s?), I feel there is not too many people interested in this stall and spin behavior...

[hvengel]

Yes I saw this property when fishing around the property tree the other day while working on my spin code. I have no idea what it is or how to use it and I was unable to find any documentation about it other than it being listed as a property in the JSBSim manual and on some FG wiki pages.

I agree with your assessment concerning spin recovery in JSBSim aircraft being too rapid/simple when using your simpler spin code.

The challenge with this is to get it so that the development and decay (for lack of better words) of the spin can be configured to cover a broad range of spin behaviors. One of the issues is that JSBSim without any added spin code will only spin, even a little, with extreme control inputs and will recover instantly when the controls are relaxed. This is probably OK as a starting point since many GA aircraft are very docile and are not far from the default JSBSim behavior. What is missing is a way to make stall/spin less docile all the way up to unrecoverable spin behavior for aircraft that are not safe to spin. As you pointed out, self sustaining spins (IE. spins that continue with the controls relaxed) are not possible with your simplified code because the yaw and roll forces needed to sustain the spin go away as soon as the controls are relaxed. My spin code actually has added functions that create these roll and yaw forces in a controlled way and allow for the spin characteristics to be tuned. There is actually a feedback loop that can be tuned to get any degree of sustain to the spin all the way from one that quickly self damps (very docile) to one that is divergent (IE, unrecoverable) and anything in between. This should also result in default JSBSim stall/spin behavior if all of the configuration parameters are set to minimum values so it should be possible to include the spin code without changing default FDM stall behavior.

I have my spin code extracted from the P-51D and am in the process of generalizing and parameterzing it. What I have so far looks good but there is still lots of stuff that is P-51D specific that needs to be generalized and also needs to be parametized to make it accessible and easy to configure for other FDMs. I am thinking that by early next week I should have this (mostly) generalized so that it could be used with any JSBSim FDM or at least a solid first cut that could be tested in other FDMs.

I am also thinking that when this gets close that I will put together a wiki page with the code and documentation on how to add it to a FDM and configure it. The P-51D can be used as an example of how this can be done but it might also be good to have example code for a more docile aircraft with a simpler FDM like the C172 so that people can see how to apply this to the more common aircraft types.

In the mean time I would like to know more about about hysteresis_limits and stall-hyst-norm. Perhaps one of the JSBSim guys can pipe in and explain what these do and how they might be used.

[ludomotico]

From the feedback(s?), I feel there is not too many people interested in this stall and spin behavior...

Yes, please. There is lack of knowledge about how to participate in a technical discussion, not lack of interest in stall and spin.

Thank very much for your efforts.

[dany93]

In the mean time I would like to know more about about hysteresis_limits and stall-hyst-norm. Perhaps one of the JSBSim guys can pipe in and explain what these do and how they might be used.

What I have understood at this point:
By setting, say

Code: Select all

        <hysteresis_limits unit="DEG">
            <min>5.0</min>
            <max>17</max>
        </hysteresis_limits>

stall-hyst-norm, normally at 0, switches to 1 at 17 deg (alpha-deg or corresponding alpha-rad value, different from alpha-wing-rad if wing incidence is not zero).
Later on, it returns back to 0 only when alpha-deg comes down to 5 deg. Close to the usual hysteresis behavior (typically iron magnetization, with magnetic field h corresponding to alpha angle for us), but the delay is not dependent on the max value attained by the variable alpha. A kind of 'rectangular' hysteresis loop, more simple but already good for aerodynamics (linear or turbulent) regimes.

This stall-hyst-norm (0 or 1 values) can then be used as a parameter for tables, to set some command sensitivity, lift coefficient,moment, etc, to hold longer in it's 'bad' condition even when the cause has ceased. Stall holds for a longer time, like turbulent regime in the real case.

Logically, the <max> value is close to the critical AoA for stall, but the lift curve and stall-hyst-norm are used independently.

I hope being clear enough.

[hvengel]

Yes much clearer. This or something like it might be useful.

My spin code is now setup to only kick in if the wing is stalled. The critical AoA parameter(s) (both positive and negative AoA) for this are set by the FDM developer(s) using standardized properties. This was the first generalization that I did to the spin code. The critical AoA setting can be anything from a set of constants to a set of perhaps complex JSBSim functions and it is up to the FDM developer to decide how this will be implemented to make this work for the FDM. For example for the P-51D I am using a set of functions that changes the critical AoA numbers based on Reynolds Number (fdm/jsbsim/aero/RE). So the critical AoA varies by about 4 degrees depending on Reynolds Number. The P-51D also uses Reynolds Number in the lift and drag tables so the lift and drag curves change shape with RE and that affects the stall AoA. Simpler FDMs that only use alpha for the lift and drag tables should be able to just use constants for this. So this setting could be used to set <max> in the hysteresis_limits settings for positive AoA stalls but it looks like this will only work for fixed critical AoA FDMs (IE. not the P-51D).

This is only available for positive AoA stalls so provides no assistance for inverted stall/spin. But it shouldn't be difficult to setup a custom JSBSim function that does something similar in the inverted (negative AoA) stall case and it should also be possible to set this up to use a variable critical AoA like would be needed for the P-51D. In fact something like this built into the spin code could prove to be very useful particularly for getting sustained spins in very docile aircraft. I will have to think about this some more. Very useful thing to have brought up. Thanks.

[hvengel]

From the feedback(s?), I feel there is not too many people interested in this stall and spin behavior...

Yes, please. There is lack of knowledge about how to participate in a technical discussion, not lack of interest in stall and spin.

Thank very much for your efforts.

Understandable. It took me many many hours of playing around with getting an FDM to spin in a realistic way to get a handle on how this could be done and what factors were involved. Not a simple subject and I think there are only two people, myself and dany93, who have any idea about how this all fits together. And even then I am sure that there are gaps and I suspect that we have no idea where or what or how big these gaps are. But I am finding that pulling the spin code out of the P-51D FDM is forcing me understand it better.

My hope is that by fully exposing the P-51D spin algorithms that we can get others looking at and experimenting with this and over time improve what we currently have by making it more flexible and/or easier to use. But this will not be possible until the algorithms are generalized enough and well enough documented that they can be plugged into other FDMs with minimal effort. This is non-trivial since the P-51D spin code has some complexity to it and was very intertwined with the FDM and hand tweaked with FDM specific settings and tables.

[Johan G]

From the feedback(s?), I feel there is not too many people interested in this stall and spin behaviour...

Yes, please... Thank very much for your efforts.

Though most of this thread is a bit beyond my current knowledge I am reading every post, and I am learn new stuff. A small eye opener to me was that the Reynolds number affects stall and spin behaviour, not just angle of attack (though at the same time I have always been somewhat sceptical to that only AoA would affect stall as it seems like an oversimplification).

Threads like this one is where I learn new things and expand my knowledge. Also stall and spin behaviour is one of those things often mentioned as things that should be improved but (obviously) only a handful of people this far can contribute to. With some help from you guys more people would perhaps be able to help improve those aspects of FlightGear aircraft in the future.

Take your time, but keep it coming, please. And thanks for your efforts so far.

[dany93]

Thanks, ludomotico and Johan G.

A small eye opener to me was that the Reynolds number affects stall and spin behaviour, not just angle of attack (though at the same time I have always been somewhat sceptical to that only AoA would affect stall as it seems like an oversimplification).

The Reynolds number (to which I was not so familiar either before working on FDMs, I just pretend to... ), through viscosity, density, obstacles dimensions, velocity, characterizes the way a laminar flow becomes turbulent. That is, when stall occurs for a wing.

The laminar flow is a competition between inertia and viscosity. In the example of a rapid fluid flow onto fixed obstacles, inertia makes fluid molecules willing to go on straight, whereas viscosity slows and leads them gradually starting from the obstacles, up to turbulence when they can no longer follow the shape of the obstacles.

For a given Reynolds number and a given velocity (a wing at an altitude), the AoA is an easy and good criteria for stall. The apparent oversimplification that you quote rather comes from the fact that, at a given altitude, a more loaded wing will stall at a higher speed (it will need a higher speed to hold laminar flow), but at first approximation for the same AoA. This, rather than for the same velocity like some people tend to believe. Example of an aircraft which is more loaded or the same one in a steep turn.
I think that this simple AoA criteria is justified by the fact that the critical AoA varies slowly with velocity (at a given altitude).
In fact, the Reynolds number (depending on altitude and velocity for a given wing) affects the angle of attack at which stall occurs. I would say that the critical AoA contains the Reynolds number effect.
This is translated in different lift-coefficient curves vs AoA, for different Reynolds numbers as a parameter. That is, different lift-maxis and different AoAs for the maxis.

(Just to share what I believe having understood. I hope having written no mistakes...Not sure. Please FIX ME! Or be critical)

[radi]

Hi dany93,

Since you asked for it I'm critical.

The Reynolds number ... characterizes the way a laminar flow becomes turbulent.

That is mostly correct, but your conclusion

That is, when stall occurs for a wing.

and

up to turbulence when they can no longer follow the shape of the obstacles.

is not.

Laminar/turbulent flow and attached/separated flow (flow separation == stall) are two different phenomenae, albeit interconnected.

At some point depending on Reynolds number, laminar flow becomes unstable and eventually turbulent (through a number of secondary/tertiary instabilities). So for low free-stream turbulence (i.e., the air way upstream the wing is not already turbulent), the flow over the first part (usually some centimeters, but maybe up to half the chord) of the wing will be laminar. For real-world sized wings, this point where the flow becomes unstable is reached before (sometimes way before) or at the point of minimum pressure on the airfoil (roughly where the airfoil is thickest), so eventually, every flow over airfoils of interest to us will become turbulent.

Both laminar and turbulent flows can be attached or separated. Once separated, most flows over wings (unless Re is very low) become turbulent and may re-attach or stay separated, depending on AoA.

Interestingly, turbulent flow can sustain more adverse pressure gradient than laminar flow, i.e., turbulent flow actually prevents flow separation to some degree. So for example, if you have a flat plate wing at low Re, i.e., initially laminar flow, the flow might separate at 6 degrees AoA. If you put a trip wire close to the leading edge, you force the flow to become turbulent, and then the flow separates only at 8 degrees (don't nail me on the exact numbers. Also, real, curved airfoils have much larger stall AoA).

For a given Reynolds number and a given velocity (a wing at an altitude), the AoA is an easy and good criteria for stall.

True, although Re is a more general (and relevant) measure in fluid mechanics than velocity. As you know, Re contains velocity, and much more.

a more loaded wing will stall at a higher speed (it will need a higher speed to hold laminar flow)

The more load you put on a wing, the more lift it must produce to sustain level flight. Lift is proportional to dynamic pressure (i.e., speed) and AoA. For the same speed, a more loaded wing must be at a greater AoA than a wing less loaded -- therefore, it will reach the stall AoA sooner. Laminar/turbulent flow is pretty irrelevant here, since at realistic Re, the flow over a wing close to stall will be turbulent, anyway.

but at first approximation for the same AoA. This, rather than for the same velocity like some people tend to believe.

Correct.

I would say that the critical AoA contains the Reynolds number effect.

True, for a given airfoil/wing shape, critical AoA depends only on Re and Mach number.

[dany93]

Thanks, radi, for these corrections and to have taken the time to write this additional and more precise information.
A good place to attempt to push our limits...