The Air France 447 accident demonstrated to me that Airbus hadn't really thought things through either. Pilots entered conflicting inputs and the plane averaged them out instead of giving (good, actionable) feedback.
Side question: is fly-by-wire an obviously good idea for passenger airplanes? It's ubiquitous in e.g. fighter jets because of the inherent aerodynamic instability of those platforms, making them hard or impossible ("Hopeless diamond") to fly without computer assistance. Passenger jets have different goals and are built to have stable flight -- the plane wants to fly level. My takeaway from Air France 447 was that I want more Boeing-style (linked, mechanical) controls in passenger airplanes than I want fly-by-wire. Am I off base?
In AF447 it has been established that the crew failed to follow multiple protocols, ignored warnings (including the audible "dual input") and while design / instrument improvements have been put forward, it's not clear they would have avoided that tragedy. The conflicting inputs only happened during the final seconds of descent.
Here's a good discussion on the why: https://aviation.stackexchange.com/questions/23577/why-do-ai...
The latest Boeing 787 has switched to fly-by-wire.
EDIT: an eerily straight-forward report of how the AF447 flight came down: https://www.popularmechanics.com/flight/a3115/what-really-ha...
The Boeing 777, much earlier, was the first Boeing aircraft with fly-by-wire.
Once you have electric control, you can start putting logic between pilot inputs and actual control surface movements. You can create dynamic, real-time-conditions-based safety stops that prevent pilots from doing common mistakes like pitching up too much and falling from the sky.
With direct control, it's up to pilots to figure out where the limits are and ensure that they are not crossed. With advanced fly-by-wire, pilots can resort to setting goals and it's up to the control system to figure out how far the aircraft can go (within safety limits) to meet them.
This is really difficult to get right, but saves lives when it finally works.
It's like memory safety in computer programming. If you're really good at programming and you never make mistakes, then you don't need it, but most people make mistakes from time to time and are better off when something checks that their buffers don't overflow and variables don't go uninitialized.
Which may end up doing more harm than good, as seen in Air France 447.
The AF447 didn't have automatic protections available due to sensor failure. Flight computers detected the failure and degraded flight controls into direct mode with less protections than under normal operating conditions. While flying on their own, pilots performed below their expected standard and there was not enough information available from sensors for automatic systems to save them.
Notably the net result was that when the pilots did the right thing (pitching down) and the sensors recovered, a stall warning sounded.
> While flying on their own, pilots performed below their expected standard
You're talking as though pilot performance is a constant. Notably these pilots had little experience flying "on their own", precisely because of these automated systems, and were thrown in at the deep end, having to take over flying under bad conditions.
This is to be expected. Pilots learn on their first single-engine prop trainer that stall horns are primitive mechanical devices based on airflow over the wing and warnings may trigger intermittently when the airflow is seriously disturbed. It's additive: blasting alarm indicates that something is wrong, but lack of warning doesn't mean that everything is right.
> were thrown in at the deep end
Shallow end. Like stall warnings, airspeed indicators are also known to become unreliable relatively often. It's like loss of cabin pressure, bird strike or a blown tire. Every pilot can reasonably expect it to happen to them, unlike the 737 MAX crashes that took place in completely uncharted territory that nobody knew to be afraid of or prepare for.
This makes the official animation at https://www.youtube.com/watch?v=n-hbWO0gL6g really puzzling. Speed keeps wildly jumping around throughout the event, an obvious sign that it's faulty. At one point, airspeed indication drops by 200 knots in three seconds to almost zero without any other indications supporting it. Everything else remains consistent. Nose is pitched up and stable. Altitude is dropping fast, vertical speed is negative. Stall warning is blasting, albeit intermittently, and the aircraft is experiencing a characteristic stall buffeting as turbulent airflow develops around wings. Why would any pilot look at that and give a nose up command? It goes opposite to their training and the flying instinct that every pilot should have.
If anything, the AF447 case shows that automatic protections should be developed further, to work under wider range of conditions than they currently do.
Had some other failure disoriented the crew instead (eg engine failure), and had the protections remained active, they probably would've saved the flight.
In which case surely the warning does more harm than good - if the warning is sounding you need to check, and if the warning isn't sounding you... still need to check.
> Shallow end. Like stall warnings, airspeed indicators are also known to become unreliable relatively often. It's like loss of cabin pressure, bird strike or a blown tire. Every pilot can reasonably expect it to happen to them
Sure, which is why it's something that should be practised under normal conditions, rather than something that you do only when it's icing.
> It goes opposite to their training and the flying instinct that every pilot should have.
But how is a pilot supposed to develop that instinct if they're not hand flying often?
What we don’t have data for is how many accidents would have happened if not for the systems that prevented them. But we should be able to compare incidents of aircraft with and without such systems and get an idea of the (no pun) impact fly-by-wire has had on safety.
In other words, the answer is, show me the data.
Fly by wire is about removing the weight and complexity of mechanical control systems. A pilot wouldn't likely have the strength to move the controls without fly by wire. However, I don't think that's what you are asking. Fly by wire does not inherently have to be computer assisted; it could simply translate your input to a control surface movement without interpretation. Of course, to get any kind of feedback, it is going to have to be computer generated. The question is where to draw the line.
Finally, you compared Airbus to Boeing. Both are fly by wire. The difference is, I guess, that the control yokes on Boeing are mechanically linked to each other but not on Airbus. However, from the yokes to the control surfaces is fly by wire either way. My understanding is that the difference is in philosophy of how much the computer does for you.
I hope I got all this right.
You also can do more complicated mappings of control inputs to control surface movements more easily with FBW (you can think of automatic traction control in a car as a somewhat analogous system - it uses differential braking per wheel, which the driver has no direct control over, to attempt to straighten out the path of the car and follow the driver's inputs from the steering wheel). As another comment mentioned, this has been happening in fighter jets for a long time, mostly due to how inherently aerodynamically unstable they are.
I can think of a particular passenger jet that doesn't.
At that point it's either repudiate AA's announcement, harming relations with a customer worth at least 100 aircraft, or go build the thing. And if they build it, they're stuck with a bunch of the problems with the 737, like being unable to change how the controls work.
If they could sub out the control system from mechanical cables to fly by wire, they wouldn't have had any visible issues with the Max.
This is a little bit unfair. Among other innovations, Concorde was the first production airliner with fly-by-wire controls (first flown March 1969!). The people who worked on it carried their experience to later highly successful designs such as the Airbus 300 and Airbus 320 series, which is a backbone of modern air travel.
The 737 MAX fiasco shows that Boeing still hasn't caught up with what Airbus was doing in the 1980s.