Startle time

[Genghis the Engineer]

I'm working up a project that is going to be reliant upon some assumptions about pilot response time to engine failure in an aeroplane.

Can anybody point me to any research or requirements anywhere showing pilot response times to abnormal events? I thought I remembered a figure of 1.5 seconds from engine failure to moving the collective on small helicopters as a requirement for adequate controllability, but going through CS.27 I can't seem to find anything.

All pointers gratefully received.

G

[Mad (Flt) Scientist]

Part 25 RTO actions often have a 1 second buffer added to them, but that's not just a recognition/reaction thing, the guidance also talks about 'normal variations'. It can also be the case that delay times are specified in the performance rules just to add conservatism in distances, not because the pilot reaction is expected to be that long. (And hence the perf and handling assessments for a similar failure end up with different timings)

I've seen, but can't pin down a reference as yet, numbers like 1 sec or 3 sec used as "recognition" times, based on how "in the loop" the crew are when the failure occurs, and then a further 3 seconds used as a "reaction" time to decide on, and take, a specific action. But those tend to be for "non-instinctive" reactions and recognitions - no one applies any of those as delays for something like VMCG, controlling the direction of the aircraft is instinctive and immediate. I was expecting to find those timings in either AC25.1309 or Ac25-7, but haven't tracked them down, though they may be in there and I missed them.

As a comment, I'm not sure I'd use the word "startle" in this sense, as I've seen it used in cases where it was also an inappropriate pilot reaction that was being discussed, "response" is probably a much better word, as you use in the body of the post.

[Genghis the Engineer]

Yes, I agree, startle isn't the ideal term, it was perhaps in my mind for other reasons when I typed.

Let's say "response lag".

G

[safetypee]

Genghis, following on from MFS; historically this appears to one of those areas - 'don't ask just meet the requirements', but there are few requirements, only advice, conditions, etc.

With increased engine and system reliability there is more interest in crew reaction, not necessarily time, but more in the understanding of a situation.

I recall a conversation about different failure conditions for large fan engines. Some situations the fan continues to rotate at high speed with few indications of failure, more so with fixed wing auto-rudder application; alternatively a catastrophic failure could beat the auto-rudder, requiring pilot compensation, which confusingly has to be reversed when the auto-rudder activates. Hence the issue is much wider that just the engine.

Another cautionary area - the words.

'Startle' appears more related to degraded cognition - a condition, involving a range of time span depending on the depth of impairment, individual susceptibility related to experience, training, or uniqueness of the event, and some recent studies suggest a longer term build up in an evolving situation without a major failure.

'Surprise' is often described as a mental reaction relating to a situation, with or without startle, etc, and might involve subconscious physical action.
Best avoid making words up - what is response lag …

Longtime away from regulations, but I too recall 1 sec reaction time followed by a further 1 - 3 seconds for action. e.g. autopilot deviation max rate was 4 sec in the cruise, but could be reduced to 2 sec during an approach, claiming that followup hands-on-the-stick quickened awareness.

Another example was the measured head-down to head-up transition to acquire sufficient awareness for manual landing was at least 4sec, similarly for head-down to head-up reversal, but not so for an auto-land decision - differing requirements for 'awareness' and type of decision.

Beware proposing new or novel approaches to reaction time; perhaps best enhance awareness of the failure with auto detection and alerting, guidance as to a suitable course of action, instrument displays.

RAE research was tied to particular operations - low visibility.

Other research is science based, often differing due to the subjective nature of behavioural assessment.
See https://econtent.hogrefe.com/doi/pdf/10.1027/2192-0923/a000092?download=true

and other work from the authors.
https://www.researchgate.net/publication/292947244_Fear-Potentiated_Startle_A_Review_from_an_Aviation_Perspective

Also: https://pure.tudelft.nl/portal/files/55707836/dissertation_startle.pdf
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5682572/pdf/10.1177_0018720817723428.pdf
https://pdfs.semanticscholar.org/61c0/73be673efa2f45cef0687f8f843eaefdf6a3.pdf

I am dubious of work by EASA, FAA, FSF, or those citing training solutions; much of this is highly subjective interpretation of previous research.

[BalusKaptan]

I'm unaware of where this is published but it most likely is somewhere.
With regards to the terms 'Startle' and 'Response', these for practical use are two separate events. It can best be seen in a simulator session. An example, you have 3 abnormals programmed, an engine failure somewhere, a hydraulic failure and an electrical failure. Crews know these are coming, it's part of the continuous 3-year cycle of licence renewal, but not the specific failure nor the order of occurrence and so have an overall level of preparedness and foresight. If the Checker then goes off-script and throws in, for example, an explosive decompression you get the Startle factor (that's the What The F**k moment) which is then followed by the reaction (that's the time for the brain to engage the correct recall response prior to activating it).
Though my example is a bit crude we have come to understand over many years that this is about as close as you can come to simulating the real thing as possible. Startle factor can be quite significant in real life, line flying, as everything and anything is possible and you cannot be keyed up for an (unknown) event permanently. Aircraft annunciators, warnings, cautions etc are designed to rapidly concentrate the crew on the actual problem to reduce this Startle factor as much as possible so as to get to the Response phase.

[Pilot DAR]

G2,

In FAA AC 27-1B, Look for AC 27.79, (b) (6) (i), Landing Gear Loads, High Hover (around page B-40). It specifies a one second delay between recognizing the engine failure, and lowering the collective. I've done the test a few times, that's a long second!

Otherwise, I'm not aware of a similar standard for engine failures on airplanes, though AC 23-8B would be a good place to look. I have had long discussions with Transport Canada Aircraft Certification about the required delays for pilots to recognize electrical system failures, and we have agreed on some benchmarks, depending on the nature of the annunciation. That presumes a failure which does not make itself apparent though...

[Genghis the Engineer]

All of you thanks, might be useful if I explain a bit more of what I'm doing.

I'm revisiting the good old Turnback, not as a certification point, but a possible research project. My question to myself is how it might be possible to optimise flying the maneouvre, My proposed approach is to create a performance model, look to what additional aeroplane performance data may be needed to make that work and probably obtain that through flight test (for example height lost during a max aileron zero-thrust rolling reversal is not readily available data!), use that model to predict the best way (bank angles, heading changes, speeds) to fly the turnback - then validate that in flight test. Early days at the moment, I'm on lockdown so playing with research project ideas.

However, it is clear that the interval between engine failure (let's assume catastrophic and instantaneous, as that's worst case) and pilot response won't be zero, and will be significant in the turnback performance as the aeroplane is bleeding energy for as long as it's left in the climb attitude without thrust. For the time being, my model will I think leave this as a variable so that I can test its important (not actually got to coding it yet), but it would be good to have an idea from some research or research-supported-regulation (ha!) I can use as a starting point.

G

[safetypee]

Genghis, a challenging project.
There appears to be two main aspects; a quantitative analysis of aircraft performance, using definable variables.

Then there is a qualitative part relating to human performance; you seek to 'fix' this with a recognition time. For a theoretical analysis a representative value can be chosen - an assumption.

However, any practical analysis would depend on further assumptions.
Given engine failure recognition (establishing safe flight), there would be a further decision to turn-back or not. This requires different, wider understanding, with significant demand on sense-making and delay before manoeuvring.

Recalling my basic military training - never, ever turn-back; greater experience questioned when would a turn-back situation become a forced landing at the departure airfield. An astute instructor answered this by declaring IMC, requiring ejecting.

Another recollection; a reputable regulator posed a special condition on a multi-engine commercial aircraft, to demonstrate an all-engines inoperative landing - any event encountered in their country. A mix of flight demonstration with reduced power and piloted simulation defined the performance boundary in order to reach an airport. Discussing the approach and landing, the aircraft was capable, the tp declined further tests on the basis that the outcome would be so subjective it was meaningless.

Finally - James Reason; what ever assumption you choose, it will invariably be wrong.

Good luck.

[Mad (Flt) Scientist]

The timings I was thinking of come from a Flight Controls Harmonization Working Group draft AC (possibly some earlier doc, but they are in this at least) that can be found in the recommendation letter from TAEIG to ARAC in May 2001, on the FAA website at this link below. Delay times are discussed on page 45 of the pdf, section (e)(1)(iii) of 'Enclosure 2':

https://www.faa.gov/regulations_policies/rulemaking/committees/documents/media/taefch-fcs-8261998.pdf

That has the 1 second for recognition, and 1 or 3 seconds for reaction statements.

I don't think that draft AC was ever formally issued - it ran into the FAA's diversion of rulemaking resources post 9/11. There is an FAA Policy (or similar document) which allows any draft doc submitted to ARAC to be used as guidance, with the usual "unless we say you can't do that" caveat of course. So it's almost as good as a 'real' AC.