>At the recent Flight Safety Foundation Corporate Aviation Safety Seminar a
progress report from the Approach and Landing Accident Reduction Task Force
brought to light some rather provocative data. It seems that in the
accidents falling within this category, the aircraft is being "hand flown"
by the Captain 54% of the time. At this time in-depth exploration of this
data has not occurred.<
Larry and the Developers Group,
Is there any way the data analysis folks can assess if a flight director
was being programmed and used even though the approach was being hand
flown? I just finished initial training for the 767 and in the past few
weeks have flown at least 30 precision and 30 non-precision approaches as
pilot flying or pilot not flying in all levels of automation. I noticed
something specific about the non-precision approaches that contributed
significantly to the error rates.
Every non-precision approach includes some hand flying even if it is just
the transition to landing from the VDP or the MAP. The problem I noticed
is when the crew starts trying to program the flight director to give them
guidance while they are hand flying any portion of the non-precision
approach.
The crew's attention shifts from altimeters and vvi's to mode control
panels and flight director steering. The human performance and ergonomic
researchers have shown that it takes about 1 second to scan and assimilate
each element in our instrument cross check. As pilot hand flying, I would
still need to request the desired modes, the PNF would need to program the
request on the MCP then I would confirm the inputs of the PNF. So here is
the breakout of attention:
* I would spend 2 seconds stating my request
* PNF would spend 4 seconds programming it in the MCP
* I would spend one second confirming the programmed vertical descent
mode, one second confirming the programmed altitude, and one second
looking at the flight director.
All total, 9 seconds, 3 of which I am not looking at the vvi and altimeter,
if the sequence is done pretty efficiently. Any errors or confusion in
communicating the request or programming the MCP and the attention demands
are huge.
In the latter portions of a non-precision approach or a transition to
landing, I can inadvertently generate an additional 1000 fpm descent at the
blink of an eye. The vestibular sensory difference between 1500 fpm and
2500 fpm is almost imperceptible. I personally do not like to use the
flight director on any portion of a hand flown non-precision approach. I
believe that it generates a whole bunch of extra workload that takes you
away from the primary flight instruments.
This was not a problem on the hand flown precision approaches because you
hit one button before intercepting the localizer and you get flight
director steering even if you deselect the autopilot and/or never touch the
mode control panel again.
I have a theory that your group may find useful. In our first instrument
flying lesson, we were taught the control and performance basics. We make
an input using a control instrument then we confirm the outcome on a
performance instrument. When we add automation to the equation, the basics
change to "request, control, performance". If we use this theory to look
at our individual instrument flying tasks, we can see where our attention
management and time demands changes. In the hand flown non-precision
approach example above, we would see the following:
(Hand flown - no flight director)
CONTROL = ADI and Power (only ADI if you are using autothrottles)
PERFORMANCE = VVI and Altimeter
* Note: There is almost no lag time between the control input and the
performance verification
(Hand flown - with flight director)
REQUEST = Mode Control Panel
CONTROL = Flight Directors
PERFORMANCE = VVI and Altimeter
* Note: There is about 9 seconds lag time between request and performance
verification
I hope this helps.
Keep us posted on the progress of the group Larry. I think we would all
like to know more about the great work your group is doing!
Vince Mancuso