Friday, June 05, 2009

Air France 447

The loss of AF 447 is a real puzzler. In the modern aviation era -- which I take to be roughly since the advent of the 707 -- and particularly the last twenty years, the sudden destruction of an airliner while in cruise flight is vanishingly rare.

What is worse in this case is that almost all the clues lie under a couple miles of water, and might never be found. However, that does not mean there is absolutely nothing to go on.

Watts up with that? posted an excellent detailed meteorological analysis of the conditions along 447's route of flight. It paints a pretty compelling picture that violent weather was the ultimate cause of 447's demise.

However, while it goes some way towards what, since the author is a meteorologist, not a pilot, there isn't a whole lot of discussion about the other "w" word: why did the four year old A330, one of the most modern airliners in the world, get in that position? After all, thunderstorms are not rare phenomena. For professional pilots, thunderstorm avoidance comes with the territory, particularly during the summer.

Obviously, one must detect to avoid. To that end, there are essentially two systems available to a pilot: the Mark One Mod Zero eyeball, and radar.

In many cases, visual acquisition works fine. During the day, cumulonimbus clouds are distinct, and at night the lightning can be seen for a hundred miles. Unless, that is, the thunderstorms are embedded in more widespread cloud. In that event, our eyes need help.

Enter radar. On modern airplanes, the radar is an amazing piece of kit. Among other things, it will overlay a color display (colors correlated to precipitation intensity) on the navigation display, and can, in the takeoff and landing phases, predict windshear.

Which, on the face of it, makes AFR 447's run-in with severe weather even harder to understand.

Time to dig a little deeper.

To visualize what the radar can see, imagine looking at an airplane from the side. The radar beam emanates in the shape of an isosceles triangle, with the apex at the airplane's nose, and the base at the range setting of the navigation display. The apex's included angle is roughly 10 degrees, centered around the pilot-set tilt angle, which is pitch stabilized (i.e., within platform limits, the angle is with respect to the horizon, not the airplane's horizontal axis; that means pitch changes do not affect the display). The triangle sweeps 90 degrees either side of the nose, providing a panoramic view of the weather ahead.

What about that tilt angle? That depends upon altitude. The closer to the ground, the greater the tilt angle, up to about 5 degrees, to reduce ground clutter. At cruise altitude, on a typical day, the tilt angle is roughly zero to -0.5 degrees. That means the lower edge of the beam hits the ground about 80 miles from the airplane's nose. Beyond that there may be some ground returns, but it is easy to distinguish them from weather returns, because only weather will appear within 80 miles.

You might begin to notice a limitation here. Just inside eighty miles, low altitude precipitation will not show, because it is just below the beam. The closer to the airplane, the higher weather can be, and still remain below that sweeping triangle, until a thunderstorm directly under the airplane will be outside the radar's field of view.

However, since an airplane will cover that 80 miles in about ten minutes, that means a thunderstorm would have to climb nearly explosively from low altitude in that time in order for its effects to reach a plane at 35,000 feet.

Which they can do. A self respecting thunderstorm probably has as much energy as a middling nuclear weapon, and can, once it starts developing, climb at 6,000 feet per minute. Put in more ordinary terms, that is a vertical velocity of about 70 miles per hour. The airplane I fly climbs better than just about anything that isn't an F-whatever. Lightly loaded, it can manage about 8,000 feet per minute at lower altitudes.

So it is entirely possible, although not common, for a burgeoning thunderstorm to climb fast enough to smack an airplane from below, all the while remaining just out of view.

That is why, if in an area with convective activity*, we will shorten the display range to get a more detailed picture, as well as essentially eliminating the possibility of getting schwacked from below.

There is also a less obvious radar limitation. What we see on the display is, in effect, colors that are correlated with the ratio of returned to transmitted energy. The more precipitation, the greater the ratio, and the color will change from light green through yellow to red. However, if there is enough precipitation, then no radar energy gets through that to anything behind it: sufficiently intense storms act like an impenetrable wall.

The other aspect to consider is aircraft performance. At cruise, we operate inside a fairly small envelope, sometimes referred to as the "coffin corner". Ten to twenty knots faster, and we hit maximum mach; twenty or so knots slower, and we run out of buffet margin. Additionally, typical cruise at about Mach 0.8 translates into a turn diameter of as much as 25 miles.

So, if an airplane runs quickly rising air, akin to suddenly going downhill, airspeed can suddenly increase beyond max mach. Conversely, passing through sinking air means the airplane must "climb" just to stay level, and may not have sufficient thrust to maintain altitude and airspeed.

Turn radius just complicates matters. When it comes to thunderstorms, pilots always have to have a "bolt hole", which has to be pretty big at altitude.

Finally, there is the A330 itself.** It is a full fly-by-wire airplane which, among other things, incorporates flight envelope protection. It won't let the pilot go to slow, or too fast, or bank too steeply, or let the pitch get out of hand. If things go wrong, it will also go into manual reversion which, if my memory of that simulator session serves, is very unpleasant.

None of this comes as any surprise. Weather avoidance is so common as to be essentially a routine part of the job, made far easier with color radars, and easier still on modern airplanes that overlay radar returns over the God's-eye-view navigation display.

Yet AFR 447 is gone, nonetheless.

Time to start speculating.

Despite all the technology, AFR 447 hit violent weather through one or more of:

    Sheer bad luck. Directly overflying an extremely quickly building thunderstorm that just managed to stay out of view.
    Bad luck of a different kind, mixed with bad planning. They got into a widespread area of rapidly building storms, and, thanks to that large turn circle, ran out of bolt holes.
    Failure to note that severe foreground weather was hiding stuff behind it, and thereby running out of boltholes. This problem is rare, and relatively insidious. However, all flight manuals address it in detail.
    Complacency led them to not be continuously working with the radar.

I think the most likely explanation for the encounter is running out of boltholes.

Having hit the weather, what caused the airplane to come apart? In and of itself, turbulence, no matter how extreme, probably isn't the cause. Aircraft structures are both very strong and resilient. However, at altitude, the narrow speed envelope can come into play. Strong up and down drafts can cause airspeed excursions outside the coffin corner***. That happens, and the coffin corner has built in margins. However, a sufficiently strong updraft could cause an irreconcilable problem: can't fight the updraft, because of greatly increased airspeed, yet going with the updraft could, due to temperature effects on Mach number, result in exceeding critical Mach.

This is where a little more speculation comes in. The A330's flight control system would have been trying to keep airspeed within the envelope, and if it was physically unable to do so, might have gone into manual reversion. Given the circumstances, that would likely have made the airplane impossible to fly. (IIRC, pitch control becomes an approximate kind of thing).

Alternately, they may have gone enough beyond maximum mach to reach critical mach, which would have put enough of the horizontal stabilizer into shock stall so that the airplane would have lost the stabilizer's balancing effect, and pitched sharply nose down, making the overspeed problem even worse.****

Either way, or for that matter, any other way, still leaves the fundamental problem largely unfazed: why did AFR447 fly into such severe weather?

Thunderstorms contain as much energy as an atom bomb. As strong as modern airplanes are, going through one can cause excursions beyond controlled flight, leaving pieces scattered over miles of ocean floor, and thousands to grieve.

* Satellite based weather observation is so good that I can't remember running into convective activity about which we had not been previously warned.

** I have flown the A320, which has an essentially identical flight control system. However, that was seven years ago, so my memory may let me down.

*** On descent into Narita yesterday, we ran into substantial turbulence that briefly caused airspeed to increase beyond max Mach, and which the flight control system could not handle without intervention -- briefly leveling until airspeed came back down. Not a big deal, though; just another day at the office.

**** Airplanes designed for supersonic flight move the entire horizontal stabilizer to control pitch for this reason.


Blogger Bret said...

Great article! It contains insights that I've never seen anywhere else.

Have you considered submitting it Pajamas Media?

These airplane posts give a whole new meaning to the daily "Duck".

June 05, 2009 10:38 PM  
Blogger Frankie said...

Having hit the weather, what caused the airplane to come apart?

Do we even know that it 'came apart'?

If it came apart, there would be debris floating in the water. But none has been found. Why has none been found, after 5 days of searching? It's simply not credible that all of it would have sunk.

There was a stream of messages from the aircraft which are consistent with it 'coming apart'. But there are almost certainly explanations of such messages which don't require the aircraft to have 'come apart' to generate them.

What was the last recorded position of AF447? Is it possible that it could have changed course after this (possibly as a result of hijacking/damage)?

As things stand, AF447 has simply disappeared, and any one explanation is as good as another. Break-up in flight is one highly likely explanation. But the absence of floating wreckage would appear to be ruling that out.

June 06, 2009 5:12 AM  
Blogger Hey Skipper said...


Well, you got me there -- I in fact made an assumption that is not supported by the some evidence that should be there, but we either have not found, or isn't there to be found.

But there are almost certainly explanations of such messages which don't require the aircraft to have 'come apart' to generate them.

No, but the kinds of failures which lead to the messages just stopping point most directly to in flight breakup.

The transoceanic flying I do is almost always across the Northern Pacific or North Atlantic. The technology has been continuously upgrading in the Japanese, Anchorage, Shannon, and Gander regions. A couple years ago, we were making voice position reports on HF radios. Now everything is data linked, so that those agencies have essentially as good an idea of where we are as if we were in radar contact.

I have never flown through the region AFR447 did, so I do not know that ATSU's capabilities.

As I noted in the post, I am assuming (perhaps speculating is a better word) that the simplest explanation is also the correct one. Had 447 encountered weather, they would have altered course to avoid it. Ordinarily, we request permission prior to doing so. However, there are procedures for deviating from oceanic tracks without permission. Looking at WUWT's depiction of the weather situation, they could easily have deviated by at least 60 miles. (Last night, as we were going by Hong Kong from the Philippines, some planes were diverting as much as 60 miles off track for weather).

One thing to keep in mind about an inflight breakup is that it might have been something like a wing failure. In that case, very quickly the wing and the rest of the plane dissipate their forward momentum. The aircraft pieces will hit the water going straight down, and not particularly fast -- maybe with a vertical velocity of 100 mph.

So, the reason no wreckage has been found is that the airplane was in a couple large pieces, and hit the water like a couple falling leaves. There would not have been wreckage scattered all over heck and gone.

Note, for instance, that the Buffalo crash had a similar dynamic. Most of the aircraft was destroyed by fire, but the tail is completely intact.

At the moment, of course, you are right: the airplane has disappeared. However, all explanations other than inflight breakup are even more difficult to speculatively explain.

June 06, 2009 7:47 AM  
Blogger Hey Skipper said...



Have you considered submitting it Pajamas Media?

No, I hadn't. I would need to get home and break out my A320 flight manual to re-familiarize myself with the flight control system.

Which takes time ...

June 06, 2009 7:48 AM  
Blogger erp said...

Fantastic analysis.

June 06, 2009 7:53 AM  
Blogger Barry Meislin said...

It appears that Air Bus is covering up.

Check this out.

June 07, 2009 3:41 AM  
Blogger Hey Skipper said...


Perhaps it is more denial than a cover up.

I remember from flying the A320 that one of the things they did to prevent the computers crashing the airplane: The airplane has three flight control computers. IIRC, each computer is designed and built by a different company, as is the software. The thinking being that the odds of simultaneous hardware or software failure would be pretty much nil.

Of course, each FCC is three the same AOA, air data and IRU inputs.

However, that leads back to my previous post on the Buffalo accident: transport category aircraft do not display AOA. What's more, SFAIK, they do not calculate what AOA should be.

Aircraft weight, air density, indicated airspeed, and g-loading are all perfectly correlated. So, each FCC should be continuously running twelve (four parameters by three different inputs for each parameter) simultaneous equations, then comparing the calculated values against the measured values for each parameter.

If something doesn't match, it is wrong.

However, SFAIK, since the FCCs have triple redundancy for each data source, the thinking is that there will never be a simultaneous coincident failure.

The other potential problem lies here:

It is not yet known whether Air France 447, an A330, carried the troublesome variety of ADIRU. But if it did, and if the Air France plane plummeted into an uncommanded dive while traveling through a downdraft generated by storms - a common occurrence over the region of the Atlantic Ocean where the plane went down - it could have been doomed as it entered a steep dive and likely broke up.

(sound a little like what I wrote?)

That quote fails to mention another problem that I alluded to: a strong enough updraft could have put them into a position where not only did the gap between buffet speed and max mach disappear, they could actually have crossed.

Ordinarily that isn't possible, because the aircraft doesn't have enough climb performance to reach that point.

But, with a fast enough, sufficiently sustained updraft, it could.

So, the FCCs might have gotten shoved into a place the designers never thought they could get: too fast and too slow at the same time.

I have no idea what the FCCs would do, particularly since they may each have decided at once they had all gone stupid.

June 07, 2009 11:26 AM  
Blogger Gaw said...

Terrifically informative post. I'm sure that we'd all appreciate updates as the story develops...

For instance, does the discovery of wreckage 70km, I believe, from the last known location change your analysis?

June 08, 2009 7:13 AM  
Blogger Hey Skipper said...

For instance, does the discovery of wreckage 70km, I believe, from the last known location change your analysis?

Yes. According to what I just read, the distance was from the location of the airplane when it sent its last maintenance message.

It does not look like the airplane broke up.

No wind idle thrust glide distance for an A300 from 35,000 feet is right around 100 miles; probably around 80 with the engines off. That they only got around 45 miles does indicate losing control about halfway during that glide.

It would take about 8 minutes to cover 45 miles in a max range airspeed glide, power off. Which, of course, raises the question of why the airplane only made it 45 miles, when it should have gone 80.

Also, that there was no communication during that time is odd. On my airplane, with all the engine driven generators out, and just the emergency Air Driven Generator operating, we have an HF (long range) and UFH (line of site range) radios available.

From the news story I read:

Brazilian authorities have refused since the search began to release the precise coordinates of where they are looking, except to say the area lies southeast of the last jet transmission and could have indicated the pilot was trying to turn around in mid-flight and head to the islands.

So, the likelihood that the airplane encountered severe weather remains. However, my thinking that the airplane broke up does not.

If it remained in controlled flight for any amount of time, though, the engines should have been restarted, presuming they flamed out.

I talked to friend of mine who is an A330 Capt last night. He told me that Airbus just put out a bulletin reducing the rough air penetration speed.

June 08, 2009 10:00 AM  
Blogger Barry Meislin said...

Here's Wretchard's take on it.

You may enjoy some of the comments....

June 10, 2009 5:33 AM  
Blogger Barry Meislin said...

Hmm. Do you really want to fly in an A330 / A340?

June 21, 2009 1:30 AM  
Blogger Barry Meislin said...

Not exactly a good month, is it?

June 22, 2009 3:07 AM  
Blogger Barry Meislin said...

And finally, the magnum opus....

July 20, 2009 4:23 AM  

Post a Comment

<< Home