Final Report: April 2018 737 high speed aborted TO

[Icarus2001]

If anyone would like to see some numbers to compare fatalities from RTOs above V1 vs loss of control immediately after takeoff, I would be more than happy to pull them up and compile them for your viewing pleasure.

I would love to see the numbers, with their source data.

[Chris2303]

Incident: EAT Leipzig A306 at Brussels on Nov 26th 2020, rejected takeoff above V1 due to difficulties becoming airborne

Incident: EAT Leipzig A306 at Brussels on Nov 26th 2020, rejected takeoff above V1 due to difficulties becoming airborne

[jmmoric]

https://havarikommissionen.dk/luftfa...2019/2019-234/

Think we can find a number of incidents where the decision to abort late in a departure was the correct choice.

[Momoe]

Way beyond V1 and VR - airborne

[vilas]

These are just exceptions that only prove the rule.
That you must not reject after V1 is not disputable at all. It's as bad as saying A few get the multimillion dollars lottery so every one should come out of school and put all that money in buying lotto tickets. Never know!

[Beamr]

Some may find this an interesting read: https://code7700.com/pdfs/nlr_reject...f_after_v1.pdf

As a background: The complete sample encompassed 135 high speed rejected takeoff accidents and serious incidents. In 90% of these cases the aircraft could not be stopped on the runway. The statistical results are presented for the period 1980-1993 and 1994-2008 separately.

Based on the study while 100% of pilots believe that the RTO is the correct choice to be made on the occasion, in hindsight only ~32% of the RTO decisions were correct ones. However, ~44% are clearly bad decisions, leaving ~24% of the cases unclear.

Take what you will out of it.

[olster]

Reluctantly I will join the fray. The incident in question was not at night time. The weather was fine (no thunderstorms). To reject a take off after V1 is wrong as per both main manufacturers guidance, backed up by industry and regulation. As already pointed out, rto beyond V1 in the simulator will end in repeat, retest or fail. Why? Statistically go minded has been proven the safer option as per Boeing FCOM / FCTM / QRH guidance material. But hey, what do they know compared to the armchair experts who have already proven that they are unaware of the industry definition of V1. It is not decision speed as previously understood from the 90s.

The report has been translated from another language and there are some ambiguous issues through translation. The accident report does not say that the decision was correct. National air accident reports are compelled not to apportion blame. It was the captain’s view post event. To reject post V1 and even then not conduct the rto properly is not role model performance even though apparently it is offensive to call it incompetent. Go figure.

I am only contributing in order to correct all the misleading rubbish that appears here. In the unlikely event of a current younger B737 pilot reading this and debating even in their own mind whether it is acceptable to reject post V1 because of an incident in an HS748 at Stansted 30 years ago.Dangerous and misleading. I have got 25 years of flying B737s and been a TRE for 27.Also flown out of KTM many times. Sorry for the willy waving but I just want to emphasise that the report describes an incorrect manoeuvre. I do not want to hang out the PIC to dry; we all make mistakes and we can learn from them. Cheers.

[alf5071h]

olster, et al, your concerns reflect mechanistic modern industry, the expectation of unambiguous situations, decisions which can examined without bias, and only one 'right' outcome vice being 'wrong' in so many ways.
This is not the real world; operations are uncertain, and as much as we try to remove uncertainty with rules and procedures, the unachievable ideal remains embedded in our imagination.
^^^
Re the 748; context is everything. The good outcome - an innovative violation, providing opportunity to lean.
Alternatively a poor outcome, 'blame' - because the LP fuel cock was not turned off in the fire drill ('read the report'), opportunity to learn.
^^^
'Risk is the amount of uncertainty we have to manage'.
Intelligence:- the ability to manage uncertainty; to learn from experience.
… an ability to adapt to the environment. People who are intelligent can learn, reason, solve problems and make decisions that fit their real-life circumstances.
… is something that you can change through life. It is constantly updated by your interactions with your environment.
Real world problems:-
Are for high stakes, sometimes life-changing ones
Are emotionally arousing, to the point that emotions often cloud people’s better judgement
Are highly context-driven, requiring people to balance many conflicting interests
Lack a single “correct” answer
Lack any indication that there even is a problem; or else, the nature of the problem is unclear
Need a collective solution, often by people with different backgrounds and interests
Offer only vague paths to a solution, or seemingly no good paths at all
Unfold and need to be solved over long periods of time (but time is limited)
Make it hard to figure out what information is needed or where that information is to be found
Come riddled with numerous bits of false or misleading information, sometimes deliberately posed to make a valid solution more difficult
Solving such problems requires a mixture of creative, analytical, practical and wisdom-based skills – the foundation of the notion of adaptive intelligence.
Adaptive intelligence is relevant to solving complex problems in the real world. It consists of four main skill sets: creative thinking, analytical thinking, practical thinking and wisdom (the quality of having experience, knowledge, and good judgement).
An extract from New Scientist - Essential Guide to AI.

Humans must not be evaluated with mechanistic analysis after the event.
The real world is not If-Then; it is full of exceptions,
Except Pull Up, … !

[MikeSnow]

I searched the ICAO database ( https://www.icao.int/safety/iStars/P...tatistics.aspx ) for accidents with fatalities related to takeoff in the last 13 years, as the database only contained accidents since 2008. This is what I found that may be somewhat relevant to this discussion:

Crash: Avient Aviation MD11 at Shanghai on Nov 28th 2009, overran runway on takeoff - 3 fatalities

Spoiler

 

 

The crew did not properly operate the thrust levers so that the engines did not reach take off thrust. The aircraft had not reached Vr at the end of the runway and could not get airborne.
[...]
The crew members perceived something was wrong. Audibly the engine sound was weak, visibly the speed of the aircraft was low, tactically the pressure on the back of the seat was weaker than normal. Somebody within the crew, possibly on the observer seats, suggested the aircraft may be a bit heavy.
[...]
When the aircraft approached the end of the runway several options were available: reject takeoff and close the throttles, continue takeoff and push the throttle to the forward mechanical stop, continue takeoff and immediately rotate. The observer called "rotate", the captain rotated the aircraft. This shows the crew recognized the abnormal situation but did not identify the error (thrust levers not in takeoff position) in a hurry but reacted instinctively only. As the aircraft had not yet reached Vr, the aircraft could not get airborne when rotated. As verified in simulator verification the decision to rotate was the wrong decision. The simulator verification showed, that had the crew pushed the thrust levers into maximum thrust when they recognized the abnormal situation, they would have safely taken the aircraft airborne 670 meters before the end of the runway. The verification also proved, that had the crew rejected takeoff at that point, the aircraft would have stopped before the end of the runway.

Accident: Bek F100 at Almaty on Dec 27th 2019, lost height shortly after takeoff and impacted building after two tailstrikes - 13 casualties

Spoiler

 

 

The captain decided only the elevators were to be de-iced. The aircraft departed runway 05R, engines and systems were all working normally. The aircraft rotated and became airborne at 07:20:36L at 148 KIAS. The aircraft reached a maximum height of 20 feet, rolled right to a bank angle of 5 degrees, rolled left to 18 degrees left bank angle and rolled right again to 14.5 degrees losing height, the pitch dropped from 16 degrees nose up to 4 degrees nose down and went up to 14.7 degrees nose up again, the aircraft contacted ground, climbed to 6 feet AGL again, pitch reduced to 5 degrees nose up, the aircraft descended to the ground again, speed decaying reaching 129 KIAS, while on the ground the airspeed recovered again to 148 KIAS, the pitch increased 19.6 degrees, the aircraft climbed to 11 feet AGL, the gear was retracted, the aircraft lost speed again reaching 130 KIAS, pitch dropped 1.6 degrees nose down and the aircraft impacted ground a last time.
[...]
After commencing takeoff the aircraft became airborne but began to roll left and right, the left wing tip touched the runway. The aircraft sank back onto the runway. The first officer called to reject takeoff and retarded the thrust levers, the captain called "no need", advanced thrust levers and continued takeoff stating "Let's go, Let's go!". About 7 seconds later the commander instructs gear up, in the following the aircraft touched the runway 6 times with the tail, 4 times with the landing gear and one time with the wing.

Crash: NOAR L410 near Recife on Jul 13th 2011, lost height - 16 fatalities

Spoiler

 

 

The captain briefed before engine start, that in case they had an engine failure above V1 and the landing gear not yet retracted, considering the length of the runway of 3100 meters, they might still abort takeoff and put the aircraft down onto the runway again. The aircraft was accelerating for takeoff from Recife's runway 18 when the left hand engine failed and lost power just at rotation. The first officer called to reject takeoff three times, the commander responded there was insufficient room to stop the aircraft and instructed to retract the gear. The first officer complied with that instruction only 50 seconds later after the captain had repeated the instruction for the 4th time. The captain instructed the first officer to request a return to runway 30 three times without realizing the runway layout of Recife had runways 18 and 36. The first officer advised ATC they would return to runway 36. The captain instructed the first officer to feather the left hand propeller, the propeller had already feathered however.

Automatic warnings "Do not sink!" and "Too low, Terrain" sounded. The captain again requested the left hand propeller to be feathered, the first officer responded the propeller was already feathered. The first officer requested to return to the aerodrome, the commander responded he was already turning.

The stall warning activated, the first officer requested to lower the nose, GPWS sounded "Do not sink!" and "Too low! Terrain". The first officer called "120 feet", the stall warning activated again. The first officer again requested to lower the nose and asked whether the captain wanted to land on the beach which met a categoric no. The first officer radioed they would land on the beach, the captain again denied, the first officer argued they would hit the top of the buildings, the stall warning activated continuously for 19 seconds, "Too Low! Terrain" and "Do not Sink!" sounded and the aircraft impacted the beach.

Crash: Ababeel IL76 at Khartoum on June 30th 2008, hit ground immediately after takeoff - 4 fatalities

Spoiler

 

 

The Sudanese CAA stated in their preliminary factual report, that the crew attempted to reject takeoff and hit the brakes, but it was too late and the crew pulled the airplane up below the required takeoff speed. The airplane was seen nose high after liftoff, then started a sharp left turn without gaining height and continued banking to the left crossing the North East Part of the Green Square, then lost height until impact with the ground. The impact occured with a nose down attitude of 30 degrees and a 60 degrees left bank.

[alf5071h]

Mike, the examples are not relevant because of either prior mistakes; i.e. as per 'spoiler'; mis-set thrust, no flap, icing check, which should have been identified before commencing takeoff, or because of system failure after take off.
These do not related to situation assessment and decision making at a critical point during takeoff.

[excrab]

I apologise, Olster, but the first part of your post makes no sense to me. You say the incident wasn’t at night, but the report states that it happened at 1621z, or 2205 LT, so it most definitely was night, being about 3 1/2 hours after sunset. You also say the weather was fine, with no thunderstorms, but we can’t draw that information from the report. It states that both crew members were aware of the bad weather in Kathmandu, and the latest METAR given in the report was at 0820, eight hours before the incident occurred, so totally pointless as far as the report is concerned. Even if they had included a weather report at the time just because it’s fine at VNKT doesn’t necessarily meant it’s fine in the valley between GURAS and the KTM. So I don’t see how we can discount the fact that the PIC may have had bad weather in his mind when he decided to reject, just from that report.

You also mention the HS748 crash at Stansted. This has, as you say, no relevance to the accident at Kathmandu, as at Stansted they didn’t reject above V1, they got airborne and then landed straight ahead on the remaining runway. However I think it is very relevant to the subject of simulator training, emergency drills and decision making. Had they continued into the air and carried out the standard drills and requested vectors back to land at Stansted they would have been airborne probably for fifteen minutes, and had the fire not gone out maybe the wing would have failed structurally. Maybe it wouldn’t, no one will ever know, but what we do know is that as a result of the Captains decision that night everyone walked away from it, and the AAIB report stated that the decision was sensible in the circumstances. Personally I think that every pilot be they young or old, flying 737 or any other type should be aware of that accident, and think outside the box. It’s really easy to say that in the simulator rejecting after V1 will require a retest, so we should never do it. But when we are accelerating down the runway, passing 150 kts and above V1 and suddenly something happens that makes us think that the aircraft is unsafe or unable to fly should we never stop? In a 737 on a 2000 metre runway probably not. But on a 4000 metre runway when we’ve got 2500 metres of it still in front of us, why not ? Real life isn’t the simulator. In real life we can’t be certain of the outcome of a scenario because we know what failures we’ve programmed, what speed we’ve armed the engine failure at, whether the fire will extinguish on the first bottle, or the second, or not at all ?

[olster]

I don’t really want to get involved in further debate on this. You are either in the industry and understand how performance works or you are not. It should appear obvious that you have to have a set of rules and understanding how go / no go decisions are made. The HS748 incident is not relevant in this context nor really any of the other examples. The general view here is that we have to have endless debate with a kind of free for all over the reject decision. Is that what we want or is it better that we have rules and regulations that determine how we perform in certain non normal situations? Statistically it has been proven beyond doubt that approaching V1 to be go minded. Both Boeing and Airbus promote this and indeed it is rubber stamped by regulation. Also both manufacturers in different ways reduce the reasons for reject in the high speed regime. Whether you like it or not rto twice at V1 plus on a recurrent check under any regulatory regime and you can hand your licence in on the way out. The rules are there for a purpose. They are not made up arbitrarily. What I don’t like is the intimation that we don’t have to follow or understand the technicalities of take off performance, what V1 actually means and we can find our inner Sully and use our self proclaimed superior skills and intellect for non normal scenarios. Dangerous stuff. I might add, the incorrectly performed rto in this instance, beyond V1 might not have had such a happily casualty free outcome at other airfields.

excrab, you are right on the time of day. There is a confusing mismatch between the metars and the take off time which I misinterpreted. The weather was relatively benign. and although nighttime my views on this incident remain unchanged. Finally, I have spent one dimensionally my career in the real world since the late 70s and understand real world dynamics. I have been @ V1 @ night @ KTM. I am empathetic to the crew and I am not here to castigate but we also have to be honest and realistic. Cheers.

[FullWings]

I agree with the above posters that it is not as simple as it sounds. When was the last time you were given a sim exercise where you had to reject after V1 in order not to crash? So, strangely enough, don’t reject after V1 in the sim!

Back in the real world, having an unexpected situation develop around V1, we don’t know in the case under discussion exactly where, leaves you in the position where rote behaviour (V1 = GO) is now at odds with cognition (can we fly?). I’m not a behavioural scientist but I don’t think you have to be one to realise that there is an awful lot of human factors here, not least the way we process information and make decisions, especially with conflicting inputs. I’m going to be controversial and say that one of the major causal factors here is a configuration warning that doesn’t appear to be inhibited at high speed - what’s the point of having it sounding if the consensus is that you should continue anyway? All it provides is an opportunity for confusion. On later aircraft, config warnings are inhibited approaching V1.

Statistically, it’s better to continue pretty much all of the time, so that’s what we train and what we expect to do. However, there is the possibility that one of us reading this thread may be dealt something truly nasty at some point in their career, like a multiple birdstrike at speed that causes multiple engine problems. Then you have to make a decision rather quickly...

[PEI_3721]

FullWings, you observe two important points.

First in simulation; either the situations are not sufficiently challenging, or that we behave differently in a simulator than we might in reality. Probably some of each.
Simulators are not reality; pilots in reality may not meet the imagined reality in SOPs.

Second, that the technology in more recent aircraft reflects what is now taught and accepted as good practice.
Investigations into incidents in older aircraft should consider these later standards. It would be impracticable to expect that older aircraft be modified to the latest standard, but at lest the errant config alert could be improved to work as designed.

[hans brinker]

olster

Sorry, but I have to disagree. In the SIM you get the same canned exercise year after year. I have 20 years of flying, and have never seen a reason to abort a fter V1, especially in the SIM. But I have landed on a runway covered in geese at night, and I struck a few in the flare. Had it been during take off there is a good chance I would have lost both engines. Dual engine failure after V1 is never trained in the SIM, to enforce the habit of aborting, because it is correct 99.9% of the time. But it is not correct 100% of the time.

[sheppey]

Dual engine failure after V1 is never trained in the SIM

I have often wonderd why not. I don't mean near the ground but at high altitude where double flameout in a thunderstorm has occurred. Garuda 737 for example. The drill in the QRH calls for a re-start and hopefully one engine always re-lights. Now your troubles are over - end of exercise and box ticked.

But what if you are unable to get both engines going? Boeing avoid that situation and leave it to the pilot to use good airmanship (or is that NTS 1,2,3 or 4?) On the other hand full marks to Airbus. At least their QRH or whatever it is called in that aircraft, leads you into a forced landing with all its additional hints.

I have yet to see a Boeing simulator demonstrate a dead stick landing. He wouldn't have a clue anyway. if a dead stick landing is allowed by a keen instructor it becomes a "fun" exercise like a barrel roll rather than a serious session. Thread drift apology

[Stuka Child]

Originally Posted by tdracer

Stuka, I'd like to see that.
Years ago, Boeing did a study of exactly that (aborting above V1 vs. continuing) and concluded just the opposite. The percentage of aborts above V1 which resulted in "catastrophic" outcomes was quite high (catastrophic being hull loss and/or multiple fatalities), where as cases where the aircraft was "unflyable" were extremely rare.

Now I'm not 100% sure we are thinking of the same thing, but I believe you are misremembering. What the study showed was that a high percentage of high speed RTOs were unnecessary and/or unjustified, versus the smaller percentage of correct calls. Pretty sure there weren't that many catastrophic outcomes.

There are many post V1 aborts that happen regularly. Much too many to list here. Of these, a number of instances result in overruns. Again, very long to list and I don't currently have a precise number, so I'd rather not say. A small number of these overruns result in fatalities. This is where we can actually dive into some figures.

This is the list of fatal accidents resulting from a high speed RTO, starting in 1980 (this includes aborts after being airborne, and I’m including even stuff like trying to take off from a taxiway or engaged parking brakes, taking forever to accelerate, etc.):

- JULY 1982: Philippine Air Lines 480, HS-748 at Yolo, 1 fatality out of 30
- SEPTEMBER 1982: Spantax 995, DC-10 at Malaga, 50 fatalities out of 394
- JANUARY 1986: VASP 210, Boeing 737-200 at Sao Paulo, 1 fatality out of 72
- JUNE 1986: Aeroflot, Tu-34 at Penza, 1 heart attack fatality out of 59
- JUNE 1989: Interflug 102, Il-62 at Berlin, 21 fatalities out of 113
- SEPTEMBER 1989: USAir 5050, Boeing 737-400 at New York, 2 fatalities out of 63
- APRIl 1990: Lao Aviation, An-24 at Luang Namtha, 1 fatality on the ground
- JULY 1992: China General 7552, Yak-42 at Nanjing, 107 fatalities out of 126
- SEPTEMBER 1993: Chaillotine Air Service, Falcon 10 at Besançon, 2 fatalities out of 3
- DECEMBER 1995: Chilean Military, CASA C-212 at Rancagua, 1 fatality out of ?
- JUNE 1996: Garuda 865, DC-10 at Fukuoka, 3 fatalities out of 275
- AUGUST 1998: Cubana 389, Tu-154 at Quito, 70 fatalities out of 91 + 10 on the ground
- MAY 1999: Kenyan Air Force, DHC-5 at Mandera, 1 fatality out of ?
- NOVEMBER 2000: Malu Aviation, An-32 at Luabo, 2 fatalities out of 11
- APRIL 2002: SELVA, An-32 at Popayan, 3 fatalities out of 8
- NOVEMBER 2003: Congolese Air Force, An-26 at Boende, 20 fatalities out of 24 + 13 on the ground
- JANUARY 2007: RAE 7755, Fokker 100 at Pau, 1 fatality on the ground
- JULY 2011: Missinippi Airways, Cessna Grand Caravan at Pukatawagan, 1 fatality out of 9
- MAY 2014: Private, Gulfstream IV at Bedford-Hanscom Field, 7 fatalities out of 7

TOTAL count: 317


So let's contrast that with loss of control accidents. Here are some fatal accidents where the crew decided to force the takeoff in an unsafe aircraft (non-exhaustive list this time, just want to give you an idea of scale):
- JANUARY 1982: Air Florida 90, Boeing 737-200 at Washington, 74 fatalities out of 79 + 4 on the ground
- DECEMBER 1985: Arrow Air 1285, DC-8 at Gander, 256 fatalities out of 256
- AUGUST 1987: Northwest 255, MD-82 at Detroit, 154 fatalities out of 155 + 2 on the ground
- AUGUST 1999: LAPA 3142, Boeing 737-200 at Buenos Aires, 63 fatalities out of 100 + 2 on the ground
- JULY 2000: Air France 4590, Concorde at Paris, 109 fatalities out of 109 + 4 on the ground
- NOVEMBER 2000: ASA Pesada, An-24 at Luanda, 57 fatalities out of 57
- SEPTEMBER 2005: Mandala 091, Boeing 737-200 at Medan, 100 fatalities out of 117 + 49 on the ground
- AUGUST 2008: Spanair 5022, MD-82 at Madrid, 154 fatalities out of 172
- MAY 2018: Cubana 972, Boeing 737-200 at Havana. 112 fatalities out of 113
- DECEMBER 2019: Bek Air 2100, Fokker 100 at Almaty, 12 fatalities out of 101

The entire number of people killed by takeoff overruns over a period of 40 years is equivalent to 2-3 LOC accidents. None of these lives should have been lost, and each and every one matters.
But if an accident is inevitable, I hope you can see which one will give you the best chance of survival.

(*main source: You can find them all in the Aviation Safety database, no matter the size of the aircraft. You just have to know how to search. The ones I was unfamiliar with I validated by going through the reports, just to make sure that they fit the criteria)

Conclusion: Yes, everyone knows we should be go-minded at V1. But that does not mean go at any cost. I see some of you saying, "oh you should never be in that situation anyway, these accidents were all preventable if they had made the right calculations, set the stab trim correctly, set the flaps correctly, etc etc.". But that matters zero! If you somehow find yourself in that situation, whether it's a wonky weight & balance, or wrong flaps setting, or ice or fire or a satanic force holding you by the gear, if you judge loss of control or structural failure are imminent, or that you cannot clear whatever obstacle you will be flying towards, you take the goddamn overrun!

[Icarus2001]

These do not related to situation assessment and decision making at a critical point during takeoff.

V1 is the decision.

It is simply impossible to "make a decision" about the flyablility of the aircraft after V1. That is why we have V1.

This crew were "lucky". Some crews are unlucky.

[alf5071h]

Why do you bother ?

I bother because I am still learning.
“You are either in the industry and understand how performance works or you are not.”
Re 'industry', a yes-no answer.
Re 'understanding', an opinion, after the event, without assurance that my opinion or that of others is correct, at that time, in that situation.
Re 'performance'; for aircraft a simple yes-no, a line on the page.
Re 'performance' relating to people; this defies explanation, too many variables, thought, context, situation, etc; thus the use of Behavioural Markers, ratings, not numbers.

Consider a jump-seat CRM check-ride for this incident. You rate the crew at the moment the aircraft stops, as a crew member; a judgement call either way - but you are going to walk away.
Consider the same scenario in a simulator; a decision for a pre-programmed event chosen by you, where the outcome is known (one way or another) and can be judged (rated) yes-no.
Consider now, penning words to Pprune, with incident report and FDR to hand; another rating.
If these ratings differ, why.
If they are the same, why; which is easier to explain … and why.
Why do I bother, because its another opportunity to ask why, hoping to improve understanding.
Learning from the misfortunes of others.

[olster]

Cheers George. I don’t know what happened to your post. The only reason I engaged at all is because the misinformation and lack of basic understanding masquerading as expertise is very dangerous and misleading. There are obviously very few professional pilots here but nevertheless everyone has an ‘opinion’. It is not up to me who contributes to these threads but it was optimum when those who knew what they were talking about posted.

Assuming you are in Australia, hope you are easing out of lockdown safely.