MountainBear, #178
All it takes is someone to put his nose up in the air, stick a number on
the problem, and put his hand in the other guys pocket and the rabble in
the crowd will give him a cheer.
That's the human condition in the 21st century. The obsession with
putting numbers on and compartmentalising everything, is a sickness of
the modern age and owes everything to the age of enlightenment, when man
started to discard religion and superstition in favour of science and
the classification of everything. As you quite rightly say, once
something has been quantified, everyone can go away and be happy in the
knowledge that due diligence has been satisfied, even though noone but
specialists in the field understand what the numbers actually mean.
In some ways, it's all gone too far, but it's not unique to aviation.
Having said that, it's often the case that the only way to get an
indication that there has been an improvement in any process is put
numbers on things via analytical methods. In aviation, as in things
like the climate debate, the change may be so small that it's down in
the noise and difficult to measure reliably anyway. Even so, the effort
is worthwhile if progress is made. The 1 in 10e5 will be a statistical
value that is based on mtbf values (also statistical) for individual
components and would be updated with data from in service components
over a multi year timescale. Obviously, the value doesn't mean that
there will be no failures until 10e5 hours. That single failure could be
in the next 5 minutes, but the figures are usually very conservative
and real world kit is often far more reliable than the figures might
suggest.
Any engineer will tell you that it's not possible to make any system
100% reliable. In many areas, it's a devil's compromise between cost,
safety and performance. The graph of cost vs improved safety probably
looks something like an exponential decay, in that you can get vast
improvement at the start of the curve, but beyond a certain point, you
could spend another 10x present cost to get any serious effect at all.
I suspect we are well down that curve in terms of civil aviation and
most likely need the analytical methods to detect anything.
An activity where you put several hundred people into an aluminium can,
together with tens of tons of fuel, then fly it at 35k feet, will always
be high risk, irrespective of how reassuring the numbers are. You are
also correct in saying that learning is high cost, though excessive
timidity in terms of risk taking can be a serious bar to progress. If
you look at the early space program in the US in the 60's, a high degree
of risk was accepted to attain great goals and was, imho, an example of
the highest aspirations of mankind, even though the initial driver for
it was arguably less than altruistic. Take big risks, make great
progress. If they had had the health and safety culture that exists now,
where nothing moves because of multi layered a** covering, the program
would never have got off the ground.
The question then becomes at what point does
the cost of the experience become more than the flying public will bear
and it's simply cheaper to automate the flight deck and get rid of the
pilots entirely.
I don't see the connection here. It seems as though you think that the
pilot's are the problem, when I would suggest that the systems are
nothing like smart enough in terms of the way they interface with the
pilot. Nor in the way that they degrade when expected to handle
something outside a strictly defined set of limits. From comms /
information theory, you achieve the lowest error rate when you match the
transmitting and receiving ends and use a low noise channel. Put simply,
if you want to talk to humans, the onus is for the system to talk the
correct language, rather than at present, where the human is expected to
adapt to the inadequacies and rigidity of the system.
Fully automated flight decks are, imho, a fantasy and will never happen
until computing has at least the same reasoning and abstract problem
solving ability as a human brain, trained in an activity and augmented
by years of experience. A lot of human processing is analog and driven
by subconscious responses, even if it is learned. Imagine trying to
model all that in a computer
. Having worked in computing and electronics
for a lifetime, I can tell you computers are not even close yet,
thankfully...
Regards,
Chris