I'm wondering about couple of well known crashes (mainly AF447 and AirAsia 8501) and thinking: is there any case where triple redundant attitude sensor failure has ever occurred? In those accidents it seems that the pilots assumed that they cannot trust the indicators and forced the airplane to stall to ground.

I understand that air speed sensors can have triple failure (e.g. AF447 had such a situation for a minute) because those depend on measuring external environment. However, the attitude indicators are fully closed system which doesn't require any external inputs. As a result, having failure of three fully independent attitude sensors at the same time looks impossible to me.

Why do commercial pilots keep ignoring the attitude indicators in case other instruments are assumed to have faults?

  • 1
    $\begingroup$ Why do you say "pilots keep ignoring the attitude indicators"? The attitude does not help to prevent a stall because it does not show angle of attack. $\endgroup$
    – Bianfable
    Commented Feb 8, 2021 at 9:05
  • 2
    $\begingroup$ Moreover, basic flying techniques would have saved both of those aircraft. Massively oversimplifying: the crews in those cases were staying inside the bubble of semi-automation / assistance rather than going back to basics of power + attitude to recover. This is understandable considering the mindset that is desired for that type of flying, but frustrating to those of us looking back in hindsight. $\endgroup$
    – Arkhem
    Commented Feb 8, 2021 at 9:55
  • 1
    $\begingroup$ It also doesn't tell us anything about the relative effectiveness of those two approaches. If it is customary to rely on "semi-automation" on such flights, we have no real idea how many lives that might also have saved when "basics" would have failed. Hindsight rarely shines a light on things which did not happen. $\endgroup$
    – Dan
    Commented Feb 8, 2021 at 10:38
  • $\begingroup$ I was using the expression "pilots keep ignoring the attitude indicators" because if the plane is at e.g. 30000 ft and you're not sure if AOA sensor or speed sensors have failed, the fact that ALL attitude sensors failing simultaneously is so low that you could just try to put the plane in nearly horizontal position and wait for a bit. Even if that doesn't result in fully correct AOA it would be much better than pushing or pulling the stick in panic. A plane at 30000 ft in horizontal orientation and engines running is not going to suddenly fall from the sky. $\endgroup$ Commented Feb 8, 2021 at 11:54
  • 2
    $\begingroup$ Regarding the age old tradition of power+attitude: if pilots assumed attitude indicators had failed (only reasonable explanation for ignoring them), how do they reference their attitude during pitch black night in IMC? That's right, they don't. Fully agree on the importance of basic flying skills, but kinda hard to aviate with no airspeed, AoA data, and not trust in attitude indication. $\endgroup$
    – Jpe61
    Commented Mar 13, 2021 at 5:03

2 Answers 2


Yes! More or less, it depends on your definition of failure (see later).

As you said, the probability that three Inertial Reference Units (IRUs) fail is extremely rare. This is because IRUs are self-contained, internal to the aircraft, and independents. Instead, air data probes are subject to ice and external factors (for instance, bird strikes), thus more subject to faults. The IRS failure probability is so low, that in the non-neo A320/A330 the ECAM message "TRIPLE IRS FAILED" doesn't even exist (it shows all the three messages: IR1+IR2 FAULT, IR2+IR3 FAULT, IR1+IR3 FAULT).

However, yes, it happened: This is a real picture of an A330-220 (F-GRSQ) flying at FL240 with all IRUs failed in 2014:

failed IRs

According to the BEA report of the F-GRSQ incident (published in 2018), only 2 triple IRU failures incidents in Airbus aircraft happened since 1999: this one and another one in 2002.

In both cases the reasons of the triple failure are unclear. However, the BEA report suggests that the triple failure was triggered by erroneous pilots actions responding to a single IRU failure. In the F-GRSQ the possible explanation is that the pilots identified IR2 drifting too much and decided to switch it to ATT mode (degraded mode, providing only attitude), but they erroneously switched IR3 to ATT. Hours later, when IR2 drifted too much, Autopilot and FBW computer detected the data inconsistency and started to consider IR2 as failed, while IR3 was in ATT mode (and then not used by AP/FBW). Probably, at this point, by following a non-standard procedure, the pilots tried to restart all the IRs to regain the position on Navigation Display, including the working IR1, which then lost alignment and caused the triple failure condition. Unfortunately, the lack of CVR recordings didn't allow the verification of this theory, but no malfunction was found in IR1 or IR3.

Why do commercial pilots keep ignoring the attitude indicators in case other instruments are assumed to have faults?

I don't think that pilots thought it was an IR problem in the sort of incidents you mentioned.

  • 5
    $\begingroup$ Great answer! I would call this pilot induced failure of triple redundant attitude indicator. $\endgroup$ Commented Mar 15, 2021 at 13:51

After occurrences such as AF447, industry study has concluded that

  • the "startle factor" comes into play during these 'Upset' scenarios.
  • Formal study of aerodynamic theory is, for most pilots, a thing from the distant past. There develops a reluctance to make simple aerodynamic interpretation.
  • Apart from this gawk factor there's a bit of "this can't be happening', because of the super safe technology we have become programmed to believe in. (and statistics mostly prove this is true.)
  • The fact that the airplane reached the stage of approach to/stall indicates a severe loss of situational awareness, and a late recognition of such a fact can fluster the pilot, rendering her/him unable to make the required decisions nor take recovery action. In the AF 447 case, the 2 FO's were bouncing around in severe turbulence, and icing, with ECAM messages galore and without the PIC in the cockpit.

Typical Upset scenarios are not caused by failures of the IRS/attitude reference and modern technology and system architecture do not generally give unreliable info, they either work, or they don't, not anything in between.

Let's assume the case of triple IRS failure, it's surprising that we haven't referred to the 'standby' flight instruments. These include attitude and heading from independent sources and are DC driven (= will continue working on loss of all AC elec power, ie batteries only). The point being made is that it is very rare to have no attitude data. Procedures are based on comparing the LH and RH PFDs to the STBY, and taking the 2 that match to be correct and "switching out" the other.

The conclusion is that ICAO has mandated UPRT, Upset Prevention and Recovery Training, and this is probably the only way to escalate the problem sufficiently in the minds of pilots. Pilots have been lulled into a strange state of ennui, due to the genuine overall improvement in safety due advances in technology and procedures.

  • $\begingroup$ TL;DR: pilot training is not adequate and as a result, modern planes fly by magic according to an average pilot. When the magic fails, the pilot goes in panic mode. $\endgroup$ Commented Mar 15, 2021 at 13:47
  • $\begingroup$ Clarke's First Law: Any sufficiently advanced technology is indistinguishable from magic $\endgroup$ Commented Mar 15, 2021 at 13:47
  • $\begingroup$ Just wait until we have a large population of self-driving cars on the road, "piloted" by drivers who were never trained well in the first place (I'm looking at you, US driver training standards), and now haven't the slightest clue what to do when their autopilot fails. $\endgroup$
    – FreeMan
    Commented Mar 16, 2021 at 18:24
  • $\begingroup$ @FreeMan I'm assuming that will be a different paradigm, one I haven't even thought of thinking about so far . . . . but come to think of it, that autopilot can fail too. They'll probably be able to achieve a lower mortality rate as compared to human driven cars just like the much quoted probability of death by aircrash is lower than that of a pedestrian crossing the road. $\endgroup$
    – skipper44
    Commented Mar 16, 2021 at 19:38
  • 1
    $\begingroup$ My point is that despite the incredible capabilities of autopilot, regulations require highly trained pilots and frequent retraining of pilots in the cockpit of every aircraft. Drivers (in the US, at least) are not highly trained in the first place (I know it's much better in other countries) and there are zero refresher courses required. Ever. Autopilot will fail, be it in an aircraft or road car. In the plane, there's someone who can take over. Average Joe in his car who hasn't actually driven in 4 years won't have a clue what to do... $\endgroup$
    – FreeMan
    Commented Mar 17, 2021 at 11:42

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .