747megatop From United States of America, joined May 2007, 316 posts, RR: 0 Posted (2 months 3 weeks 3 days ago) and read 2639 times:
I was reading an article on Flightglobal where a waypoint data entry error on a 777 led to it's rapid descent when approaching MEL. The crew luckily recovered and did a low approach into MEL and landed safely
This brings me to one question - How come sophisticated software such as one found on modern airplanes cannot do a simple validation on data entered and prevent invalid entries? Is there something i am missing or don't understand?
From what the article describes one shudders to think what would have happened if the crew hadn't recovered in time; perhaps another SFO OZ like incident if they hadn't taken corrective action? Bravo to the crew for recovering and a safe landing.
francoflier From France, joined Oct 2001, 3514 posts, RR: 11 Reply 1, posted (2 months 3 weeks 2 days 17 hours ago) and read 2480 times:
It would seem the crew somehow entered an improper VNAV constraint when entering the VFR approach to the runway and sequencing it to the FMC route.
As they were following the LNAV/VNAV profile, the set up the MDA in the MCP and let VNAV take care of the descent profile. VNAV then tried to 'dive' down to make the 'wrong' alitude constraint.
I don't know exactly how that constraint could have been entered, though it would seem the crew confused the Runway point and the centerline extension point created by the FMC and entered the altitude constraint on the wrong one.
It doesn't say how high the aircraft was when it started to descend and how long the crew took to react, but I'm guessing it was pretty low already. That's where it pays to always double guess what LNAV/VNAV are doing so as to be able to revert to more basic modes if anything's amiss.
Don't overestimate how smart FMCs are. Your smartphone has 20 times the computing power of a 777 FMC.
Additionally, every line of code has to be carefully considered as any programmed command could have unintended consequences.
But then it doesn't need to be foolproof. FMCs are meant to be a convenience, a workload reducing tool. Pilots aren't supposed to let it take first stage unchecked. Other Autopilot modes are always available, or they can even actually fly the thing, for a change...
Looks like I picked the wrong week to quit posting...
Max Q From United States of America, joined May 2001, 3828 posts, RR: 18 Reply 2, posted (2 months 3 weeks 2 days 14 hours ago) and read 2427 times:
Quoting 747megatop (Thread starter): From what the article describes one shudders to think what would have happened if the crew hadn't recovered in time; perhaps another SFO OZ like incident if they hadn't taken corrective action?
Well, no need to shudder, your fears would be somewhat more realistic if this was a crew that allowed the automation to fly them without keeping an eye on things and intervening if necessary (Like Asiana)
These things can happen, a competent crew will do exactly what this one did. Disengage the Autopilot, correct the approach and fly it manually, as they did.
An alternative would have been to go around but their response was a very reasonable and correct one.
Automation errors occur, that's why there are PILOTS in the cockpit to correct them.
I have intervened manually on more than one occasion myself for various different reasons. The number one priority is always:
FLY THE AIRCRAFT.
The best contribution to safety is a competent Pilot.
Starlionblue From Greenland, joined Feb 2004, 16360 posts, RR: 66 Reply 3, posted (2 months 3 weeks 2 days 13 hours ago) and read 2419 times:
It is an often repeated anecdote that the most frequently uttered phrase in a modern cockpit is, "what's it doing now?"
Contrary to common belief, airline pilots are not infallible gods of the sky. Well, ok they are but they are also human beings. Programming errors happen all the time, a bit like how you press the wrong icon on your iPhone or click the wrong button on your TV remote. That's why you keep checklists and human brains handy in the cockpit. Even a well trained crew makes errors perhaps once in a thousand actions. (Pilots without much experience are closer to one error per hundred actions). By errors here I don't mean flying off course or below glide path. I mean selecting the wrong mode or twisting the HDG knob the wrong way, or as in this case entering wrong information in the FMS.
Since errors are unavoidable, trying to eliminate them entirely is futile and counter-productive. The focus is on discovering and recovering from errors before they affect safety. As happened here. It hardly seems as if the flight was in danger. As Max Q says, the danger is when the crew does not take over and fly manually in the face of events which are obviously not going according to plan. If you don't understand what the automation is doing, disconnect and hand fly.
Even in basic instrument training in a plane far less sophisticated than a 777, my instructors impressed upon me the importance to monitor the automation (such as it was in a Cessna 172) and never hesitate to take over if I felt that events weren't going the way I felt they should. I can't tell you how many times I've engaged the autopilot and gone "whoa!" before immediately pressing the disengage button. Flying manually may be more work but I know exactly where the plane is going to go, meaning that at least the "aviate" bit is taken care of. Much better than trying to reprogram while also monitoring a deteriorating situation due to bad programming.
Quoting 747megatop (Thread starter): This brings me to one question - How come sophisticated software such as one found on modern airplanes cannot do a simple validation on data entered and prevent invalid entries?
Look at it this way: What constitutes an invalid entry? How could the FMS "know" that the entered routing isn't the desired one.
mandala499 From Indonesia, joined Aug 2001, 6417 posts, RR: 74 Reply 4, posted (2 months 3 weeks 2 days 12 hours ago) and read 2406 times:
OK, error entry is one thing, it happens all the time, and there are ways to check it (in ideal conditions of course)... but the question I got from this is, why the heck did they kept using the LNAV/VNAV from waypoint SHEED?
On LIZZI 7V arrivals, you go to SHEED at 257 heading and hit SHEED at 2500ft or above, you then make the turn to runway 34.... From SHEED to runway 34 is about 5NM direct line, make the turn and you should have about 8-9NM in a turn, and a nice 80deg turn.
My question is, why keep it on the LNAV/VNAV in a visual turn to final? It's a no-no in some of the airlines where I am, LNAV/VNAV for guidance is acceptable and advised, but not on AP. Had this occured at some Asian carrier, would some then scream "automation dependence syndrome"?
Am just baffled...
When losing situational awareness, pray Cumulus Granitus isn't nearby !
francoflier From France, joined Oct 2001, 3514 posts, RR: 11 Reply 5, posted (2 months 3 weeks 2 days 7 hours ago) and read 2319 times:
Quoting mandala499 (Reply 4): My question is, why keep it on the LNAV/VNAV in a visual turn to final? It's a no-no in some of the airlines where I am, LNAV/VNAV for guidance is acceptable and advised, but not on AP. Had this occured at some Asian carrier, would some then scream "automation dependence syndrome"?
It is perfectly acceptable to keep Robopilot on duty as long as you have proper LNAV/VNAV guidance, all the way to the MDA if you are flying an RNAV/GPS/Non-precision approach, or to the minimum A/P disengage altitude if you're doing a visual approach.
I'm sure policies differ from airline to airline, but according to Boeing (in this case), it's absolutely fine. Setting up a VFR approach on the FMC makes it calculate a 3.0 degrees angle approach to the RWY waypoint, just like any other non precision approach.
Unless, of course, you accidentally set up a ridiculously low altitude constraint on the 3 miles runway centerline fix...
What Boeing and every operator will tell you, however, is to avoid FMC programming during high workload periods such as the final stages of the descent and approach. In this case, it's better to go back to basic modes or F/D off - manual flight altogether. Lest this happens...
Looks like I picked the wrong week to quit posting...
blueflyer From United States of America, joined Jan 2006, 3478 posts, RR: 1 Reply 7, posted (2 months 3 weeks 1 day 1 hour ago) and read 1985 times:
Quoting Starlionblue (Reply 3): How could the FMS "know" that the entered routing isn't the desired one.
Most automated system can be programmed to validate user input if the result falls within (or outside) a certain range. It would take a very sophisticated system to tell you a route is wrong, but that doesn't preclude checking certain parameters from the route, such as speed of descent.