Originally Posted by
vilas
5. A few failures of automation is not enough evidence against automation just as one Sully or Al Haynes doesn't make a summer. There are any number of fully serviceable Aircraft crashed through inadequate piloting.
I worked at airbus incident analysis and little system malfunction were commonplace.
Aircraft descending below minimum descent altitude, initiating the first sid turn on the wrong side, unable to switch to land then flare mode, etc...
Originally Posted by
vilas
Actually the question should have been how to acquire/retain manual skill. That's what all the discussion is about.
Disagree on that point.
Airbus could very well design a cockpit full of automation but that would still leave a possibility to fly raw data.
This is impossible, so the slow erosion of manual skills is almost unavoidable, even if we would all be well aware of the problem and willing to practise raw data flying, in most cases (RNAV) we just can't, due to a design choice.
Originally Posted by
sonicbum
Regarding the last bit on how to acquire/retain manual skill I believe we all know the answer is : training, and more specifically SIM training. But SIM training is expensive and, a
It is almost impossible. A correct amount of sim training would be one session per month. So multiply by about 5 or 6 the required sim capacity. Take that hangar that your airline uses to stuff a few sims into, and imagine building 5 others, recruiting tens of TRIs, etc.. Impossible.
Plus, a sim remains a sim. Since it represents the aircraft, you still can't fly raw data RNAVs in it. Since it does not match perfectly the aircraft, the training is not as realistic.
Flying raw data manual departures and approaches on a regular basis on the line, to me, is the only option to maintain a high standard of manual skill.
Originally Posted by
Uplinker
...
Re instruments, I agree, indications are too small for a given deviation or are badly designed, and this is why raw data flying is such a challenge. I remember finding NDB tracking in a PA28 to be very difficult because the NDB needle was on one dial and the heading bug was on another, and there was no bug for the NDB track - you had to remember what it was. One had to continually compare the needle with the heading instrument, and parallax and misreading could occur. Also, you were flying an aircraft that never stayed where you put it, so you were busy hand flying and continually correcting the aircraft and tracking a non bugged NDB needle, it was quite common for the NDB to drift more than 5 degrees out. When I later flew the Dash 8, you could overlay the NDB needle, the NDB track bug and the compass rose all on one instrument and suddenly NDB tracking was a piece of piss ! Instead of having to remember the NDB track and look at a different instrument to read what the heading was and then go back to the NDB instrument, all you had to do was glance at the one composite overlay. You did not even have to read any values, you could see at a glance if the NDB needle was under the track bug, and if it was just one degree out one side, you clicked the AP heading bug towards it by one degree. Really easy.
I have always found LOC and G/S displays to be too limited. By the time you can see a deviation, it is quite a large error. With my engineer's hat on, I would redesign the display so the markers were in two halves. One half would move as they currently do, the other half would move over its whole range of the display for say 1/4 of a degree LOC or 50' G/S - much more sensitive and a large movement for a small deviation - so you would be able to see a deviation before it became too big. The other half would display as it does now. A lot of the time the sensitive marker would be pegged on one extreme or the other, but when you had captured the LOC and G/S they would come off the stops, and a perfect ILS would see all the bugs centred.
Yes ! You explain very well how a simple design choice leads to an easier flying.
Two other examples :
When flying an DME arc with an old DME like this one :
https://s7d2.scene7.com/is/image/hon...detail-470x290
You have the DME distance, but also the very precise and precious indication of DME speed !
So to fly a DME during flight school, on an aircraft equipped with a DME, I used the TAS/200 to initiate the turn towards the DME, and then I did not worry about the wind, the angles, the calculations, all this, I just kept the DME speed around zero.
Let's say after the initial interception I was 0.1nm away. I would set 10kt convergence until it displayed the correct DME distance, at which time I set the DME speed to zero by diverging a little.
Then, the DME speed increased slowly, and when it reached a limit that I chose like 5kt, I converged until I read 5kt on the other side. At some point after several corrections like this one, the DME distance which was stuck on the correct DME distance could deviate by 0.1nm. I corrected this and continued.
This allowed me to make almost perfect DME arcs, remaining always within +/-0.1nm. And I would have been always within 0.05nm if I had an indication of the distance of that precision.
Whereas even the best other students struggled to remain within +/-0.3nm.
I could very well see when they would deviate, thanks to the DME speed. A DME speed deviation of about 30kt outwards would appear. At 0.1nm they would reduce the DME speed to 20kt. But by the time they reached 0.2nm deviation, their DME speed is now 25kt outwards.. They apply a larger correction and now have 10kt DME speed outwards, increasing. They reach 0.3nm deviation at 15kt outwards. They apply the same large correction and they are now at 0kt outwards, but increasing outwards. They can reach 0.4nm at 5kt outwards, and only then with another correction they can finally converge ! Now the risk is huge to overcorrect.
When I gave them the technique they aced the DME arcs, problems like above described never happened anymore and instructors were amazed.
Then, we changed airplanes, and I could no longer use this technique. Because DME speed was not given anymore !
So I had to find another technique. Add 90° to the bearing and that is my desired route.
Also works very well. I did an almost complete DME circle in a glider like this, around my home gliding airfield, with just my GPS watch.
A computer can analyse in real time the deviation, the derivative of the deviation, and it's derivative's derivative !
A human cannot derive a derivative from a needle, at least not as well as a computer.
Regarding the ILS, I practise them raw data on the airbus.
The difficulty is that the green diamond should give an indication of the LOC's derivative : right of course, deviate to the right, and so on.
But since there is a few degrees imprecision on the diamond, especially after a long flight, this does not work. The computer can compute the LOC's derivative, but I have no indication of it. So a deviation trend must start before I can detect it, and this still does not give me the heading at wich the derivative is zero.
Instead of your double indicator, I would rather think of a very sensible "LOC trend" or "glide trend", as is done with the speed trend, displayed on the side of the diamond.
Then you would just set the LOC trend to zero, and if there was a deviation, you could see much faster if it is growing or reducing, which would also help the PM's job.