LAND3 disconnects before flare on heavy windshear/crosswind

[kekstrom 1]

Is there some logic on the 737-800 ngx fail-operational land3 autoland that makes it disconnect if there is too much drift from the ils on the final or for example coming in too fast/slow due wind shear or in other cases making the approach unstable?

I played around yesterday putting some heavy weather to my sim using Activesky 2012. I set up crosswin of 25-30kts, heavy windshear and turbulence just to see how the autoland handles it. Though a bumpy ride, the approach went pretty good but about 200ft agl there was heavy change in windspeed increasing my ias from desired 142 to 158, soon after the LAND3 mode disconnected and I ended up flying the plane manually..

 

Any explanation for this? I always thought that after LAND3 has engaged its (almost) impossible to get it disconnected the only method being pulling the A/P disconnect bar or pushing TO/GA switch..

 

Thanks!

[scandinavian13 5,192]

Ever take a moment to look at the listed limitations?

[killkenny1 22]

I don't think it's allowed to make an autoland during such bad weather conditions. Bad visibility yes, but strong (cross)winds, don't think so...

[MED1473 18]

Hi all,

yeah those parameters were definitelly outside the limitations of the autopilot, therefore erratic behaviour is not only possible but expected. The limitations may vary slightly from operator to operator but are generally ....

 

Headwind 20kts

Crosswind 15kts

Tailwind 10kts

cheers, Matteo Capocefalo, MED1473

[kekstrom 1]

Yup, I know and knew that the conditions were out of limits, though I remeber the crosswind limit being 20kts so its not too much beyond limits. Luckily its a sim so trying anything is allowed:) Still I am not sure that the autopilot should disconnect when it deviates from the ILS beam or target app speed? Erratic behaviour yes, thats understandable when the winds throw you all over the place but what logic would make the autopilot disconnect once it has commited to the land3 mode?

 

Any thoughts?

[G-CIVA 62]

Yup, I know and knew that the conditions were out of limits, though I remeber the crosswind limit being 20kts so its not too much beyond limits. Still I am not sure that the autopilot should disconnect when it deviates from the ILS beam or target app speed? Erratic behaviour yes, thats understandable when the winds throw you all over the place but what logic would make the autopilot disconnect once it has commited to the land3 mode?

 

Any thoughts?

 

 

I dont think you are getting it.

 

The Autopilot does not have some inbuilt logic saying "I must land now because I am in Autoland Mode" & it is also not some mythical amazing super computer.

 

You exceeded its design limits by asking too much of it & at that point it (the Autopilot) ran out of zeros & ones & it told you it was leaving Dodge City by disconnecting & leaving you in control.

 

The Autoland function of the autopilot is designed to land the aircraft (or get it to DH dependant on capability) in low visibility; it is not designed to assure a safe & accurate landing in extreme scenarios like the one you have painted.

 

The designers (of the autopilot & aircraft systems) would be expecting YOU THE PILOT to manually land the aircraft under such conditions, or, alternatively, if the visibility prevented you from doing so they would have assumed that YOU THE PILOT would have had the common dog to examine the conditions & well beforehand use your super computer (brain) & decide to divert & land somewhere else where the WX conditions were more favourable.

 

Simple as.

 

PS - After all - imagine the furore of the PMDG Team had modified things a tad to allow you to land the aircraft in a 50kt crosswind using the Autopilot & Autothrottle? That definately would show less logic than a 2 channel autopilot that suddenly runs out of zeros & ones.

[kekstrom 1]

OK G-CIVA i hear you.

I just don't think you understood my questions. Of course I would not expect the autoland to land the airplane smoothly on the centerline in a middle of a hurricane. I read over and over my initial question and I cant see why you have come to that conclusion.

 

I'll try to be more specific. Computers are based on algorithms and are expected to do as they are programmed. So my question is: What are the exact variables (crosswind, turbulence, ILS deviation) that make the autoland mode disconnect on the 737-800? I couldn't find anything on the FCOM or FCTM and when basically EVERY single variable about aircraft performance etc. has been documented in the manuals, I am thinking that the autopilot should NOT disconnect in autoland mode due to exeeded weather conditions.

Would I expect it to land perfectly? Off course not. Of course I would expect it to crash the aircraft off the runway, stall on approach or put the aircraft in some other unwanted situation if I would not manual take over because it just cant handle the drifting winds etc. But to disconnect by itself? If it is PROGAMMED to do so, then I understad it completely but what are the criteria? You say that "The Autopilot does not have some inbuilt logic saying "I must land now because I am in Autoland Mode"". Well the autopilot has programmed logic to follow the glidepath, maintain speed, flare and rollout on toucdown etc. because thats how its programmed. No computer has a AI that says, "Ok I cant handle these kind of winds, I think ill just disconnect" if it is not programmed to do so.

 

Get my point?

 

Now, if there's somebody with solid facts and knowledge about this matter I would appreciate very much if somebody educated me on this. Thank you in advance!

 

 

I dont think you are getting it.

 

The Autopilot does not have some inbuilt logic saying "I must land now because I am in Autoland Mode" & it is also not some mythical amazing super computer.

 

You exceeded its design limits by asking too much of it & at that point it (the Autopilot) ran out of zeros & ones & it told you it was leaving Dodge City by disconnecting & leaving you in control.

 

The Autoland function of the autopilot is designed to land the aircraft (or get it to DH dependant on capability) in low visibility; it is not designed to assure a safe & accurate landing in extreme scenarios like the one you have painted.

 

The designers (of the autopilot & aircraft systems) would be expecting YOU THE PILOT to manually land the aircraft under such conditions, or, alternatively, if the visibility prevented you from doing so they would have assumed that YOU THE PILOT would have had the common dog to examine the conditions & well beforehand use your super computer (brain) & decide to divert & land somewhere else where the WX conditions were more favourable.

 

Simple as.

 

PS - After all - imagine the furore of the PMDG Team had modified things a tad to allow you to land the aircraft in a 50kt crosswind using the Autopilot & Autothrottle? That definately would show less logic than a 2 channel autopilot that suddenly runs out of zeros & ones.

[MED1473 18]

Hi again,

if you're looking for a specific reason for this behaviour it may be connected to the unstabilized approach resulting from the wx conditions. The ap disconnects if the lateral or vertical deviation fron the ils signal exceeds a certain degree (I'm sure its in the manuual somehere but I can't rember where). When on autoland the ap logic assumes that you are on a stabilized approach an if this is not he case, because as you get near the threshold the ils beam becomes much narrower, you might exceed the limits and the ap gets unhooked from the beam. Mat

[G-CIVA 62]

& .....

 

Drive a vehicle or an airplane betwixt these beams & the LOC & G/S receivers on your aircraft & bingo!

 

In addition to what is said above the Radio Altitude & air data sensors feed a number of queues into the PITCH Mode & Autothrottle logic respectively.... create instabilties & variations beyond preset limits & the two channels within the autopilot take a quick vote ....

 

Like them the left lobe & right lobe of my brain are doing this right now ... with far too many zeros & ones to account for I have disconnected.

[Fabo 213]

Also: Why disconnect? So you have a very real, very present whoop-whoop style annunciation that this is not going to happen - and you dont have to wonder all day if you will land OK or stall and die in marginal conditions.

[Houghton11 12]

What are the exact variables (crosswind, turbulence, ILS deviation) that make the autoland mode disconnect on the 737-800? I couldn't find anything on the FCOM or FCTM and when basically EVERY single variable about aircraft performance etc. has been documented in the manuals, I am thinking that the autopilot should NOT disconnect in autoland mode due to exeeded weather conditions.

 

You are partially right in a round-a-bout way. The autopilot doesn't have a magical connection to the weather gods and decides "Ohh...this airport's METAR is saying there's a 35 knot cross wind...I'm just gonna disconnect now." There is a much more complex...but strangely simple explanation for how autopilots and basically any sort of automation system on aircraft behave, when it comes to redundancy and disconnection events.

 

There are numerous factors and computations involved. Right now if I showed your comment to an 'E' guy (what we call an avionics tech up here in Canader) he'd be punching the screen :lol: . First of all, lot's of things are not listed in the flight manuals...which is what is provided with the NGX. In fact the 'pilot' documentation is quite narrow in it's scope and rightly so. Everything is there that they need to know.

 

Now, to answer your question, there are many things that will set off an autopilot. First, I want you to look at this diagram of an example of an aircraft autopilot system.

 

 

This is just an example of the single channel of an autopilot system. In this case, this simplified diagram shows all the back and forth going on between a typical Flight Control Computer and everything else. For your situation, I want you to take notice of the numerous Spoiler Position Sensors, Airleron Position Senors and LVDTs (Linear Variable Differential Transformers). These units are sending (in this case) analogue signals back to the FCC. These signals are converted to digital signals, monitored, and constantly compared. Roll attitude, roll acceleration, inputs from the inertial reference system, flight management computer, air data computer, and mode control panel are being compared to these other signals from the position sensors and LVDTs.

 

Now, let's begin to look at your situation exclusively. In this case, you are flying a fully automated landing, in which 3 FCC's are providing dual channel (pitch and roll) outputs to 2 independent sets of aileron, spoiler, and elevator (stabilizer) autopilot servos. There is a lot going on. First, each FCC is constantly monitoring each other's signals and comparing them. Now this is where things get wonky between fail-passive and fail-operational differences but lets just move on here. In addition, each FCC is sending, receiving, comparing, and monitoring it's own outputs and inputs from the position sensors and LVDT's....which the other 2 autopilots are also monitoring. Confusing enough eh ;)?. Back to your situation. There is a set of disengage logic that is applicable for every autopilot mode. In the Autoland case, there is very strict disengage logic present. It can be from VOR course needles not copacetic, to loss of power. In your case, (and this is why I explained the monitoring of position sensors and such in detail), things are not jiving. First of all, you've got an unstable nasty crosswind. The FCC senses that it is getting blown of the ILS horizontal course. This FCC is trying to let say...roll right to crab into the wind a bit to correct. It sends the signal to the actuator. It knows the actuator is working and moving because the LVDT says it is. Then it knows the aileron is in the position it wants to be because the signal from the aileron position sensor jives with the signal it sent in the first place. Now, it's comparing this movement to the roll acceleration and roll attitude....Hmm...wow that wind is really blowin...The plane isn't doing what the FCC wanted it to...okay..then...process repeats and the autopilot tries to correct again. Okay..we've just about reach the end of the aileron travel...the plane's starting to react then...BAAAM!

 

Ohh now that heavy turbulence has kicked in (that you set up). The aileron's near the end of it's travel, the FCC was still trying to get the aircraft in the correct attitude, and now it has sensed (through numerous inputs) that the aircraft has begun a violent movement opposite of what it wanted to to before! Ahh but it hasn't given up yet...it still has a couple tricks up its sleeve. Now it's going to try and fix this turbulence. It does what it did before...looks at the inputs, determines an output, monitors output, then compared all this to the other 2 FCC's that are also sending their own independent outputs to their own independent actuators! Nope...the plane isn't doing what the FCC wanted it to do. In your case, it most likely found itself in a situation where the control surface was at the end of it`s travel, the plane was not going where the autopilot wanted it to do and the other three FCC's (which it`s still monitoring) were saying the same things. It`s tried and failed. There's nothing more the autopilot has been programmed to do. The Roll Channel monitoring logic clicks in...the events and signals are compared to a magical digital table of 0s and 1s which say the control movement and surface sensor position does not match aircraft acceleration and attitude...it compared this logic to what the other 3 FCC's are saying then comes to the same conclusion....F*** IT.

 

Viola....your disconnect.

[rocketfs 2]

And still no one here was able to answer the (valid) question: "When will it disconnect?"

 

-> How much off the glideslope, LLZ, speed,...?

 

 

Bert Van Bulck

[richardnobes14 13]

It will disconnect at the point when the conditions become too difficult for it to process, by which time you should have manual control anyway.

[Houghton11 12]

And still no one here was able to answer the (valid) question: "When will it disconnect?"

 

-> How much off the glideslope, LLZ, speed,...?

 

It's a lot more convoluted than just how many feet one is off the LOC etc..etc...The internal logic to an exact Autopilot behaviour is something deep-deep inside specific engineering documents.

 

Below is an example of both Pitch and Roll mode disconnect tabled from an aircraft autopilot system. Not the NG though.

 

Pitch Disconnect:

 

Roll Disconnect:

 

As you can see, the specifics are kind of there, but things like "selected IRU longitudinal acceleration not valid" is kind of what I broke down in my previous post.