737 'Tailstrike' Caused By Typo On a Tablet (arstechnica.com)

An anonymous reader writes: In August of last year, a Boeing 737 operated by Qantas experienced a tailstrike while taking off — the thrust wasn't great enough for the tail to clear the runway, so it clipped the ground. The investigation into the incident (PDF) has finally been completed, and it found the cause of the accident: the co-pilot accidentally entered the wrong plane weight data into the iPad used to make calculations about the takeoff thrust. "First, when working out the plane's takeoff weight on a notepad, the captain forgot to carry the "1," resulting in an erroneous weight of 66,400kg rather than 76,400kg. Second, the co-pilot made a "transposition error" when carrying out the same calculation on the Qantas on-board performance tool (OPT)—an iPad app for calculating takeoff speed, amongst other things. "Transposition error" is an investigatory euphemism for "he accidentally hit 6 on the keyboard rather than 7." This caused the problem: "For a weight of 76,400kg and temperature of 35C, the engine thrust should've been set at 93.1 percent with a takeoff speed of 157 knots; instead, due to the errors, the thrust was set to 88.4 percent and takeoff speed was 146 knots."

Comments

Data data everywhere and not a drop to think

[burtosis]

It still boggles my mind how we live in the Information Age and this data was not automatically uploaded and calculated. I'm not saying it dosent require a human to sign off on, but it's mildly insane it isn't all automatically calculated and simply checked.

Re:Data data everywhere and not a drop to think

[Anonymous Coward]

I don't know for sure, but I believe the pilots are supposed to *only* use the certified avionics systems in the aircraft, but they hate the godawful UIs in those things so they use iPads even though they're really, really, really not supposed to.

Re:Data data everywhere and not a drop to think

[kuzb]

Probably because boeing would charge another $750,000 for that feature.

Re:Data data everywhere and not a drop to think

[iONiUM]

Uh, isn't the weight the plane + fuel + baggage + passengers? 3 of those are clearly easily gathered (fuel, plane, baggage) automatically. Only the 'passengers' part isn't, but I'm assuming they just use averages (they don't weight us when we get on) via # of people * average weight.

So yea, why the fuck isn't this automatically calculated and prepared? What part of this requires a human who is bad at math and able to make mistakes on simple data input?

Re:Data data everywhere and not a drop to think

[kelemvor4]

It still boggles my mind how we live in the Information Age and this data was not automatically uploaded and calculated. I'm not saying it dosent require a human to sign off on, but it's mildly insane it isn't all automatically calculated and simply checked.

Is there some reason they shouldn't just use 100% thrust at takeoff and make sure the cargo being carried was less than the maximum capacity?

Re:Data data everywhere and not a drop to think

[es330td]

Is there some reason they shouldn't just use 100% thrust at takeoff and make sure the cargo being carried was less than the maximum capacity?

These engines are optimized for certain turbine speeds. By staying within the recommended ranges they reduce wear & tear and improve efficiency. In addition, max acceleration is harder on the passengers. Think of it like driving a car. Do you accelerate from every stop by pressing the gas pedal all the way to the floor or do you match how much gas you give it to the driving conditions and who and how many people are in your car?

Re:Data data everywhere and not a drop to think

[Impy the Impiuos Imp]

Well apparently both made the exact same error, for completely different reasons. So redundandy didn't help.

Really, though, why don't modern large planes all weigh themselves?

Re:Data data everywhere and not a drop to think

[jedidiah]

This seems like something the plane should be able to do all by itself. This is $50M aircraft. It should be at least as sophisticated as a hospital bed.

Re:Data data everywhere and not a drop to think

[gstoddart]

Is there some reason you don't redline your car before you pull away from a stop light?

Having pilots who don't know how to do anything other that "rev it up to full speed and let 'er rip" sounds like a terrible idea to me.

Both for maximizing fuel and not abusing the engines, doing it based on real numbers makes more sense than the equivalent of flooring it.

Re:Data data everywhere and not a drop to think

[Anonymous Coward]

And $745,000 of that is needed to comply with and get the app through DO-178C certification. The rest is pure profit.

Re:Data data everywhere and not a drop to think

[SuricouRaven]

You shouldn't even need to calculate it - the plane can measure weight itsself. There are only three points of contact with the ground, easy to mount a strain gauge in the landing gear. It might not be precise enough, but it'll be fine for sanity checking - if the measured and calculate weights differ by more than the margin of error, plane refuses to bring the engines above taxi thrust.

Re:Data data everywhere and not a drop to think

[Tawnos]

I can think of a few ways this could be done - integrated into the pressure monitoring of the wheels and shocks, using a rolling weigh system like semi trucks do, or similarly-constructed systems. It's hard, but not impossible.

Re:Data data everywhere and not a drop to think

[Goldenhawk]

I'm a flight test engineer that works on large (passenger-class) aircraft. We do tests related to this issue.

It actually has a lot to do with takeoff safety, ironically enough. If you lose an engine at high power, the airplane will try to yaw (turn left or right) because the engine(s) on the other side are still producing thrust. At lower speeds, with less aerodynamic forces, the rudder is not capable of keeping the airplane in a straight line. So there's a speed called "Vmcg" - "velocity minimum control ground" - below which you MUST pull back the power on the good engines to avoid going off the side of the runway (you're going to have to stop the takeoff). There's also a speed called Vmca, the airborne minimum control speed (you will start to yaw out of control).

So with less power on all the engines, there is less asymmetry possible in the event of a failure. With reduced takeoff thrust, you don't need as much rudder at any given airspeed, so your Vmcg and Vmca are both lower.

This is important for takeoff because if you have a lot of runway available, you can use it by taking longer to accelerate (by having lower thrust). As a consequence, your risk in the event of an engine failure is reduced - you won't head off into the grass if it happens on the ground, and you'll be assured of sufficient control authority if it happens in the air.

So when an airplane manufacturer builds the "takeoff performance charts", these Vmc speeds heavily factor into the takeoff planning.

Now, in this tail strike mishap, the lower weight caused the iPad to compute TOO LOW a speed. Lifting off too slow takes more nose-up (pitch) angle; lift goes up as a linear function of pitch angle; lift has to equal weight to go flying. Because of the reduced takeoff thrust, they were already planning to use most of the runway to accelerate - which put them into a corner; they were too slow to take off at the normal pitch angle, but were out of room to stop. So they pulled up until the airplane started flying - which means they pulled up high enough that the tail hit the ground (just barely in this case).

Re:Data data everywhere and not a drop to think

[jbwolfe]

As long as the device and software (the iPad and app in this case) has been vetted by the regulating agency, it is acceptable to use and may actually be required. I'm unsure of the interface to which you are referring, but takeoff performance calculations are not integrated into any large transport aircraft that I have ever flown. The FMC can only calculate data accurately if its given the correct inputs- stuff like ZFW, CG, flap setting, reduced thrust setting, etc. GIGO

At my airline, takeoff data calculations are centralized (acquired through datalink) rather than carried onboard, but still require those variables.

Re:Data data everywhere and not a drop to think

[ShanghaiBill]

Ok, Einstein. Provide model numbers of the sensors capable of doing what you claim "are clearly easily gathered".

The landing gear use hydraulic and pneumatic systems, and there are already sensors that measure that pressure. The weight of the plane is a simple linear function of the pressure of the fluids in the landing gear.

Re:Data data everywhere and not a drop to think

[Obfuscant]

I wish I had mod points today. Excellent answer.

Add in the safety factor that a heavy aircraft is better lifting off further down the runway so that the following departures and arrivals of lighter aircraft are less likely to encounter wake turbulence. As a small plane jockey, you always want to take off or land before the point the previous heavy aircraft took off, or after the point the previous heavy aircraft landed.

Re: Data data everywhere and not a drop to think

[bill_mcgonigle]

hey, now, this is just an airliner, not a fancy tractor trailer.

Re:Data data everywhere and not a drop to think

[sunking2]

The real question is why is he doing the math by hand. Computers are pretty good at that.

Re:Data data everywhere and not a drop to think

[jbwolfe]

To add to this and in regard to the model flown here, The -800 and-900 versions are stretched and have higher MTOG. The end result is since the landing gear are just as short, there is even less margin for error with tail strikes. This is true both for takeoff and landing and results in higher rotation speeds as well as higher approach speeds. In some cases, this results in weight restrictions due solely to the needed increased speeds.

Also of note regarding Vmcg, if Vr (reject) is lower than Vmcg, you're overweight for takeoff. Virtually never a factor for a modern aircraft.

Re:Data data everywhere and not a drop to think

[MouseR]

The iPad is not at fault here. Pilot did simple math and forgot to carry over a "1". There's no carry over when you let a software add.

It's a whole system failure: paper being handed out to be hand-computer and then the number punched into a iPad for final trust numbers which are then entered in the avionics system.

A frickin piezo on the landing gears would have done the trick.

Re:Data data everywhere and not a drop to think

[Andy Dodd]

I've never seen a flight where they were threatening bumping that they weren't gate-checking bags because every seat was full and there wasn't enough room for carryons.

(Not weight reduction here - bag was still on the plane, just in cargo instead of overhead bins in the passenger compartment.)

Re:Data data everywhere and not a drop to think

[Andy Dodd]

The real story behind $10,000 hammers you hear people joke about in defense & aerospace:

It's a $500 hammer because it's made from premium materials and is designed to handle things most hammers in civilian use would never see, plus $9,500 worth of government-mandated paperwork proving you didn't rip off the government.

Re:Data data everywhere and not a drop to think

[leathered]

I'm only a private pilot but you also need to take into account temperature, air pressure (QNH), headwind component, the amount of runway you have to work with as well as weight to calculate takeoff performance.

Re:Data data everywhere and not a drop to think

[CohibaVancouver]

Interestingly enough, most commercial planes can't measure how much fuel they're carrying

Of course they can. Here are some examples of the gauges on a 737 -

http://www.b737.org.uk/fuel.ht...

The Gimli Glider ran out of fuel because the gauges weren't working and they messed up the manual calculations.

Re: Data data everywhere and not a drop to think

[michelcolman]

The day they can get tire pressure sensors, temperature sensors, fuel gauges etcetera to work reliably without failing or giving incorrect indications every now and then, maybe they can try measuring weight. Technology moves slowly in aviation.

Re:Data data everywhere and not a drop to think

[michelcolman]

Because you need more speed if the weight is higher. And you may also need more thrust if the runway isn't long enough.

Re:Data data everywhere and not a drop to think

[Anonymous Coward]

My step-brother is an onsite supervisor helping to build the new AP1000 reactors at the Vogtle Electric Generating Plant.

He's constantly complaining about the paperwork, how every nut and bolt is documented, and how people are slow to sign-off on paperwork because they're afraid of the buck stopping with them if something breaks.

And yet _his_ projects (the boilers) are way ahead of schedule. Their the only ones ahead of schedule, in fact. Why? Because even though he complains about the red tape and meddling of regulators, he's actually incredibly smart, very hardworking, and he manages to whip the crews he supervises into shape. He doesn't half-ass anything, but he's courageous enough to make the hard decisions, including extra work, that keeps things on time and under budget.

Government (as well as corporate, for similar reasons) projects are expensive not because of paperwork and red tape, but because _real_ people drag their feet. And because they will use paperwork and similar frictions as an excuse for poor performance, rather than treating it as any other challenge one might encounter.

I couldn't even imagine the amount of graft and corner-cutting that would occur without the paperwork holding people's feet to the fire.

Most people are lazy most of the time. Most people don't take pride in their work most of the time. Civilization works because a minority of people prefer getting the job done to whinging, even if they complain almost as much as the next guy.

Re:Data data everywhere and not a drop to think

[Deadstick]

Because the aircraft can't measure its own weight.

Some of the larger ones can. They have strain gages on the landing gear axles.

(It would seem simpler to put pressure gages in the oleo struts, but static friction in the struts corrupts that.)

Re:Data data everywhere and not a drop to think

[DarkOx]

I am not a pilot so I don't know but I am curious. In all the old war novels I read they just run the throttles to "take off power" and go.

Is there any reason you don't just slam the throttles open on a modern jetliner until you are off the ground? Why do anything other than make sure you are not 'to heavy' to make it before you run out of runway? What not just 'punch it'.

Re:Data data everywhere and not a drop to think

[lgw]

Hospital beds don't usually have airfoils. You've got a wind, say, 8% of the aircraft's takeoff speed, gusting to 12%. The aircraft has wings. How does this all work again? Plus, the kind of sensors that work at this scale tend to be temperature sensitive, magnetic field sensitive, and hard to calibrate.

As usual, the stuff that /. armchair experts thought up in 5 minutes did, in fact, occur to the actual professionals designing these things. You could likely make an automatic system of some sort, but would it be more accurate than the current method, which works pretty well on average?

Re:Data data everywhere and not a drop to think

[Mass Overkiller]

Depending on the environmental conditions, going "balls to the wall" on a modern jet without computer controls would likely wipe out the engines by over torquing them and overheating them. Modern engine controls keep the engines together. Basically what happens now is that the computer is programmed for the aircraft configuration and the computer says "max power is XYZ". Then pushing the throttles all the way forward only nets you Max Power of XYZ. Some systems have an override condition, provision or button/lever, but most throttle systems on take off are "set and forget". This is to keep the engines together and keep them from over spinning, overheating or over torquing. Modern jet engines have more power than necessary for take off and climb. If you notice right after take off, the engines spin down slightly. This is to prevent over heating and increase engine reliability.

Re:Data data everywhere and not a drop to think

[Lisandro]

There's a good number of reasons for this, and they're not immediately obvious.

- Optimize fuel usage; you want to burn only as much as you need on takeoff
- Reduce engine wear as much as possible. Engine overhauls are awfully expensive for airlines.
- Don't overthrust engines in case of failure. If an engine goes out on takeoff with too much airspeed the airplane might not be able to correct yaw before running out of runway.

On small piston aircrafts, yes, you usually go full power on takeoff every time. Airlines are a very different beast though.

Why do these data need to be entered manually?

[QuietLagoon]

Where do the pilots get the data? Is it displayed to them? Is it shouted at them as they board the plane?

.
Why can't the source of the data convey the data to the tablet apps automatically? Why involve an error-prone human in the process?

Re:Why do these data need to be entered manually?

[Solandri]

Because if an automatic system fails in an "obvious" way and causes an accident, people here would be posting "why wasn't a human in the loop?"

In an ideal world, both automatic systems and people would complement each other with a person manually checking the automatic results. But in the real world, once you automate the system, people tend to get lazy and stop double-checking the automated system's calculations. NASA ran across the same problem - in the post-Challenger investigation they discovered that having multiple inspections actually decreased safety. Each inspector assumed the other was doing his job, so became more lax and sometimes didn't check things thoroughly or sometimes even skipped checks. Even in this incident, the pilot using the iPad didn't bother checking to make sure the number he entered into the iPad was the number he thought he typed. He just assumed the input method worked, despite everyone who's tried to type an email on a touch interface knowing that errors occur frequently.

So you end up with an either/or situation. Either the system has to be completely automated with the engineers trying to think up every possible scenario during design, and it'll still occasionally fail in ways you never thought of which will lead people to question why humans were left out of the loop. Or it has to be reliant on humans doing everything by hand, to ensure they take their job seriously and actually do it. Unless an automatic system has a track record showing decades of reliability (this is why computers on aircraft are usually more than a decade old), aerospace usually relies on humans.

Re:Why do these data need to be entered manually?

[jbwolfe]

Obtaining performance data requires knowing how many people and bags and where they are located, which runway is used, flap setting CG and weather data like pressure, winds and temperature. These factors, exclusive of the atmospheric conditions, are dictated by the humans operating the plane and while they could theoretically be automated, there likely isn't enough return on the invested effort to do so.

Re:Why do these data need to be entered manually?

[jbwolfe]

As a unionized airline pilot, I can assure you that your statement is completely false. I'd suggest that you visit here: http://alpa.org/ to get a better idea about the kinds of things that union has done to improve safety in the industry. Such as the following: science based duty limits, TCAS, Captains authority, security of the cockpit, hazardous cargo, safety reporting system, etc.

I suspect, sadly, that your bias against unions is an indicator that your mind is made up already.

So wrong data in Flight Management Controller

[Anonymous Coward]

So somebody entered bad data into the Flight Management Controller.

Happens all the time with Qantas, which has been caught out for having extremely lax protocols.

Search the ATSB database, there have been more than a dozen incidents in the past decade which by pure luck didn't result in mass casualties.

Then read the BS excuses Qantas gives. Such as "the ladder wasn't tall enough to check the engine cowling was locked".

5%

[rfengr]

Seems to me 5% should be well within the margin or error. Are they trying to save fuel by cutting it that close?

Re:5%

[avandesande]

Yes that is a ridiculously small margin.... it's not hard to imagine a series of errors adding up to 5%. What about a sudden tail wind?

Re:5%

[Anonymous Coward]

He meant a tailwind, you ignorant git. It's the velocity of air over the wings that causes lift. A sudden tailwind will reduce lift. Not knowing this just means you still have things to learn. Mouthing off about things you don't understand makes you an ignorant git.

Re:5%

[Talderas]

Headwinds are helpful for takeoff. Tailwinds are detrimental.

As if...

[CanadianMacFan]

As if the co-pilot couldn't have made the exact same mistake with a calculator or even paper.

Why is a tablet signiifcant to this story?

[JoeyRox]

They could have just as easily mistyped the weight using the on-board flight management system's keyboard.

"Transposition error"

[Nutria]

Is *not* an "off by one" error.

That's why...

[CanadianMacFan]

you always have both the pilot and co-pilot make the calculations to make sure that they both come up with the same number. That is if you can't have the plane download and measure the necessary inputs in order to calculate the values for you automatically.

Re:That's why...

[BZ]

They did both do the calculation. The pilot did the arithmetic wrong and the copilot typed in his result wrong, and the upshot was that the numbers they entered independently agreed with each other... and were both wrong.

Re:That's why...

[craighansen]

Well, to be fair, that's the official report. I'd use Occam's razor and note that if the pilot did the arithmetic wrong and the copilot simply copied the pilot's calculation instead of doing it himself, the records would also match this explanation. Occam's razor cuts to the chase, even as it also cuts to the quick.

Where is there check?

[Rinikusu]

I used to be a Ramp Agent at a major international shipping firm. We did weight and balance for the flights. We had several layers of redundancy for our numbers: Every container number and weight was rechecked by another Ramp Agent, and then once again at the gate to match with the load sheets. We realized if we put the numbers in wrong, that could result in loss of life (not to mention aircraft assets and cargo). We took this job very seriously. Once we turned that paperwork (now done via ACARS, supposedly), I would hate to think that the cockpit just fat fingered the numbers in on their end without having a secondary check. "Hey Captain, can we check the numbers real quick?" Probably take them 15-30 seconds at most since they'd be concerned with big picture numbers and not individual positions.

Re:Where is there check?

[wickerprints]

Both the captain and co-pilot did separate calculations. They just happened to make two different arithmetic errors that resulted in the same incorrect result, therefore failing to detect the error. The captain failed to carry a 1. The co-pilot type in a 6 instead of a 7.

What this incident shows is that automation or the use of computers to do calculations automatically, does not necessarily improve reliability. Independent and redundant systems are instrumental in reducing error, but basic vigilance and attention to detail frequently is the most effective means in preventing mistakes.

Re:Where is there check?

[Nidi62]

Both the captain and co-pilot did separate calculations. They just happened to make two different arithmetic errors that resulted in the same incorrect result, therefore failing to detect the error. The captain failed to carry a 1. The co-pilot type in a 6 instead of a 7.

What this incident shows is that automation or the use of computers to do calculations automatically, does not necessarily improve reliability. Independent and redundant systems are instrumental in reducing error, but basic vigilance and attention to detail frequently is the most effective means in preventing mistakes.

I may be wrong, but in my experience with commercial flights crews the pilots are trained to not only verbalize what they are doing but the other crew member is supposed to visually check and verbally repeat what is being done. So if the captain is putting in 70000lbs into the computer he should verbalize "weight 70000lbs" after which the FO should check to make sure that the entry was indeed "70000lbs" and verbally confirm so. If the weight does not match his calculations or what is in the system does not match what the CA then he should speak up. Sounds to me like this crew needs some additional work with CRM.

Re:It's even worse than that now.

[Austerity Empowers]

That'll make for some angry birds.

Siri

[chubs]

"Okay. I found 5 restaurants like 'calculate takeoff thrust' near you"

Qantas

[Anonymous Coward]

On a recent flight from SYD to MEL our Qantas pilot ranted on the PA about how evil the ATSB was because they wouldn't let the plane take off, because Qantas engineering accidentally put an incorrect black box into our plane. He said that Qantas maintenance made mistakes like that all the time, and that, since it wasn't a safety issue to fly without a black box, the ATSB were just being pricks.

I will never set foot on a Qantas plane ever agin.

Re:Why not 100% on takeoff?

[jklovanc]

Not just fuel but engine wear and air frame stress which reduces the life of the aircraft.

Re:It's even worse than that now.

[sycodon]

Umm...Tail Strikes on take off are caused by over rotation.

Regardless, seems to me that pilots calculating the take off weight is anachronistic. Manufacturers can easily incorporate weight sensing devices into the gear. Or, weight sensors could be incorporated into a Wheel Chock device. There is no reason for guesstimates.

Re:It's even worse than that now.

[tripleevenfall]

"76,860 kg" autocorrected to "Settings thousand ate sexy cardiograms"

Insufficient safety margin

[gurps_npc]

Before you drive a 10,000 lb truck over a bridge, you make sure it is rated for 10,000 lbs.

That does not mean it can handle 10,000 lbs, it means it can handle MORE than 10,000 lbs - the excess is called a margin of safety. It can be expressed as a "factor of safety". If your factor of Safety is 2, then the engineer believes the bridge can handle twice the stated load.

For aircraft and spacecraft, the US government generally requires Factor of Safety of at least 1.2, but as high as 3 is common. That means that if the tire is expect to hold 10 tons, it was built to actually hold at least 12 tons. (All above info from Wikipedia)

Apparently the idiots that write these programs thought, hey, we don't need to use a reasonable Factor of Safety. They calculated they needed 93.1 units of thrust, but failed only 88.4. That means their Safety Factor was less than 1.12

If the morons had put in the expected margin of safety of 1.2, then the plane should have been able to take off even if they had only applied 77.67 units of thrust.

Clear failure of the programmer to build a sufficient margin of safety into their calculations

Re:amazing no ground scale or even strain gauges

[Idarubicin]

...but then the stupidity of taking off at less than 100% throttle to save a little bit of fuel at the expense of increasing risk is also a pretty dumb thing to do, engineering wise.

Taking off at less than 100% throttle means reduced acceleration, which reduces stress on the airframe (and passengers). It reduces wear on the engines and - more important - reduces the risk of turbine failure. It makes the aircraft easier to control (less unbalanced thrust) if it does lose an engine immediately before or after takeoff.

So...not just to save fuel.

Re:It's even worse than that now.

[michelcolman]

First of all, tail strikes on take-off are obviously the result of overrotation, but this usually happens because the pilot rotates at the wrong speed. You pull back, expecting the plane to leave the ground, but instead the plane remains on the ground while the nose keeps going up. Also, you may be running out of runway if the calculations were off, so you'll pull back regardless.

About the weight sensors: good idea, but this is aviation, where everything has to work reliably in pretty difficult environmental circumstances. Even something as simple as a proximity switch to determine whether or not the gear is down, fails from time to time. We often deal with incorrect tire pressure indications, temperature indications, etcetera. Measuring the weight of a plane with sufficient precision is quite a bit more complex than a simple tire pressure reading, so I can't see any manufacturer trusting that kind of system enough to let it determine take-off settings by itself. Maybe as an extra crosscheck for the data from the loadsheet, sure, but not as the primary source of information.

People always go "we should replace the pilots with automated systems because pilots make too many mistakes", but they have no idea how many mechanical failures we deal with as part of the routine of our job. We make mistakes, sure. But so does automation.

Re: It's even worse than that now.

[manu144x]

As a software developer I love it when people forget that computers run software written by humans who are error prone too :)