Boeing And The Lessons For Cars
The Market Ticker - Commentary on The Capital Markets
Logging in or registering will improve your experience here
Main Navigation
Display list of topics
Sarah's Resources You Should See
Sarah's Blog Buy Sarah's Pictures
Full-Text Search & Archives
Legal Disclaimer

The content on this site is provided without any warranty, express or implied. All opinions expressed on this site are those of the author and may contain errors or omissions.

NO MATERIAL HERE CONSTITUTES "INVESTMENT ADVICE" NOR IS IT A RECOMMENDATION TO BUY OR SELL ANY FINANCIAL INSTRUMENT, INCLUDING BUT NOT LIMITED TO STOCKS, OPTIONS, BONDS OR FUTURES.

The author may have a position in any company or security mentioned herein. Actions you undertake as a consequence of any analysis, opinion or advertisement on this site are your sole responsibility.

Market charts, when present, used with permission of TD Ameritrade/ThinkOrSwim Inc. Neither TD Ameritrade or ThinkOrSwim have reviewed, approved or disapproved any content herein.

The Market Ticker content may be sent unmodified to lawmakers via print or electronic means or excerpted online for non-commercial purposes provided full attribution is given and the original article source is linked to. Please contact Karl Denninger for reprint permission in other media, to republish full articles, or for any commercial use (which includes any site where advertising is displayed.)

Submissions or tips on matters of economic or political interest may be sent "over the transom" to The Editor at any time. To be considered for publication your submission must include full and correct contact information and be related to an economic or political matter of the day. All submissions become the property of The Market Ticker.

Considering sending spam? Read this first.

2019-03-11 11:50 by Karl Denninger
in Technology , 162 references Ignore this thread
Boeing And The Lessons For Cars
[Comments enabled]

The crash of a 737-MAX, second of that specific aircraft make and model in the last year or so, at first blush appears to have potentially been related to over-reliance on automation.

We do not yet know for certain why the plane went down -- but it was a brand new aircraft with very few (~1200, if I read the reports correctly) hours on it.  The "black box" recorders have been found so we'll soon get something more-definitive on exactly what happened.

At first blush, however, this may have been related to what preliminary investigation says is an "angle-of-attack" piece of automation that, if it detects what it believes is an impending stall, applies "forward" pressure on the control yoke so as to pitch the nose forward.  This reduces the angle of attack and, if indeed the aircraft is about to have an aerodynamic stall, can prevent it from happening.

However, there are two assumptions here -- first, you have the altitude available to execute that maneuver and the indication of an impending stall is accurate.

There are procedures to disengage and override this system should it malfunction -- if you detect the malfunction in time, do the right thing, and have the space for the recovery.  What if the aircraft is not actually stalling and you're close to the ground, or worse, you do the wrong thing in that circumstance?

There are reports of Tesla's "autopilot" plowing into the back of fire engines or "splitting the difference" between two available lanes, ramming a barrier in the middle.  It's reasonable to assume in that case the reason it happens is that the vehicle "thinks" where it's going is clear -- and it's wrong.  Whether the sensor data is incorrect or the computer incorrectly evaluated it the result is the same -- you crash.  Ditto here, if that's what happened.

It's always possible there was a terrorist event involved here or a mechanical failure of some sort -- but it appears, at first blush, this may be the same sort of thing that took out the other aircraft.

The problem with relying on automation is that it takes attention from the human away as they become complacent.  If the automation then fails to correctly process the situation, either due to bad decision making or bad sensor input people get hurt or die from entirely-avoidable incidents.

I don't call these "accidents" because they're not.  Accidents are acts of God.  An event due to culpable negligence is not an "accident", it is an incident.  An event that occurs because a sensor returned bad data to a computer is not the fault of the computer; it executed its program.  It's the fault of the idiot who relied on a piece of machinery that had insufficient inputs to determine that one of them was giving it bad information.

Now about those self-driving "cars" you want all over the road......

Go to responses (registration required to post)
 



 
Comments.......
User: Not logged on
Login Register Top Blog Top Blog Topics FAQ
Showing Page 1 of 4  First1234Last
User Info Boeing And The Lessons For Cars in forum [Market-Ticker]
Maynard
Posts: 1028
Incept: 2007-11-27

gig harbor wa
Report This As A Bad Post Add To Your Ignored User List
I have no dogs in this fight anymore but I would ground every one of them till they figure out what happened to both those birds.
Maybe they are related (hardware/software). Maybe its training, IDK. Still ground them.

As to automated driving..thats insane. I don't really like being in the passenger seat because I have no control.
Azengrcat
Posts: 453
Incept: 2010-05-31

Report This As A Bad Post Add To Your Ignored User List
Yikes, thats gonna hurt the Garbage Equity Capital Aviation Services revenue stream a bit.

Go to planespotters and count the number of lsf GECAS in the remark column.
Vernonb
Posts: 2359
Incept: 2009-06-03

East of Sheol
Report This As A Bad Post Add To Your Ignored User List
They want to totally automate airflight yet can't even get the traction control in a vehicle to understand WHY the tires are slipping. That is usually the first thing a human realizes in a situation.

Does the human respond correctly? It depends on training.

Now take a machine. It realizes something is not right. Without proper understanding of the cause it trys to respond with canned action within its sensor limits. Some of those responses may be fatal in a given situation.

Canned digital responses for a myriad of real world analog problems without the understanding of why. That's a disaster in the making.

Makes me wonder what happens if my car starts to lose altitude. /sarc off


----------
"Mass intelligence does not mean intelligent masses."
Tickerguy
Posts: 156505
Incept: 2007-06-26
A True American Patriot!
Report This As A Bad Post Add To Your Ignored User List
Quote:
They want to totally automate airflight yet can't even get the traction control in a vehicle to understand WHY the tires are slipping. That is usually the first thing a human realizes in a situation.

Does the human respond correctly? It depends on training.

Yep.

A couple months ago I was in Michigan driving on unplowed streets where the intersecting (more-major) street was plowed. The only way through that intersection IF YOU STOPPED (as you should, since there WAS a stop sign!) was to intentionally mildly spin the drive wheels through the plowed-up snow at the entrance to the intersection.

TCS on my car refused to allow it! Fortunately there's an "off" switch and shutting it off quickly allowed me to be on my way with no drama at all.

Without said "off" switch I would have been unable to proceed - literally, I would have been stuck, unable to come off idle at all in the forward direction and, since I got somewhat into the mire, I might have been unable to reverse out of it either.

The vehicle DID NOT KNOW that (1) I was in heavy snow, (2) I was traveling at an impossibly-slow speed to have a significant collision due to wheelspin and (3) that I was INTENTIONALLY commanding throttle intending to produce a mild amount of wheelspin due to differential traction which was not only reasonable it was THE ONLY way to proceed having complied with the STOP sign.

----------
Winding it down.

Dvanderp71
Posts: 658
Incept: 2007-08-05

Amsterdam
Report This As A Bad Post Add To Your Ignored User List
Funny example of the Traction control in cars. I have two independent examples of people who fried the clutch in their car trying to fight the TCS on snow.
That is what happens if people are unaware of systems in their vehicles.
In last years crash of Lion air this is what seems to have gone wrong. What I read out of it is that Boeing installed a system called MCAS in their MAX version to avoid re-certification of pilots. 737 pilots including the Lion air pilots, were not made aware of this new system and its workings. As such they also could not know or figure out in time to disable it when it engaged because of a faulty sensor reading. In itself also very bad that a single bad sensor can cause this.
It would be pretty bad if this accident has the same cause - also reflected in Boeing's price drop today.
Elkad
Posts: 536
Incept: 2009-09-04

Report This As A Bad Post Add To Your Ignored User List
IIRC correctly from the other crash.

If you are stalling (or the sensors think you are), the autopilot pushes the nose down. On every prior version, if you grabbed the yoke, it stopped fighting you.

On this one the autopilot keeps trying to push the nose down unless you manually disengage it. It would be like your car deciding you were going to get rearended and accelerating, even if you are headed for a brick wall and have your foot on the brake.
Nadavegan
Posts: 159
Incept: 2017-05-03

The South
Report This As A Bad Post Add To Your Ignored User List
Working in a large corporation, it is evident that even simple things like invoice processing can't be fully automated. That is, you still need a team of accountants to babysit transactions and make constant corrections. All automation does is allow you to make mistakes faster, because you have increased transactional volume. It is why I am extremely suspicious about any talk of automated driving.
Aquapura
Posts: 834
Incept: 2012-04-19

South of Canada
Report This As A Bad Post Add To Your Ignored User List
I don't want to jump to conclusions on any aviation incident. Full respect for air crash investigators. What I will say is that it's been over a decade since there was a hull-loss incident in commercial aviation of US based airlines. That's unprecedented and should be credited to the quality of our flight crews. Believe I read that the FO flying the Ethopian plane had more than 200 hours experience. I certainly wouldn't feel very comfortable carrying 100's of pax with just 5 weeks at the controls under my belt. Both AA & Southwest fly dozens of the same type here and I've heard nothing of any problems stateside.

Tickerguy
Posts: 156505
Incept: 2007-06-26
A True American Patriot!
Report This As A Bad Post Add To Your Ignored User List
Quote:
Funny example of the Traction control in cars. I have two independent examples of people who fried the clutch in their car trying to fight the TCS on snow.
That is what happens if people are unaware of systems in their vehicles.

Yep -- my "6" is a stick. It literally tried to kill the engine with the TCS on and had I attempted to feather the clutch I'm sure I would have burned it up in short order.

Stab TCS OFF button (which by the way requires you to come COMPLETELY off the throttle; it will NOT read the button if you're not or if the vehicle is moving) and off I went. From that point if there was ANY indication of white **** out there that button got stabbed on start (it is also NOT sticky across an engine shutdown.)

----------
Winding it down.

Tickerguy
Posts: 156505
Incept: 2007-06-26
A True American Patriot!
Report This As A Bad Post Add To Your Ignored User List
@Aquapura -
Quote:
I don't want to jump to conclusions on any aviation incident. Full respect for air crash investigators. What I will say is that it's been over a decade since there was a hull-loss incident in commercial aviation of US based airlines. That's unprecedented and should be credited to the quality of our flight crews.

Generally, me too -- but this one's different, as it's TWO hulls of the EXACT SAME make and model.

The first one I'm willing to dismiss -- **** happens.

The SECOND one of the EXACT SAME make and model, my eyebrows go up.

----------
Winding it down.
Maynard
Posts: 1028
Incept: 2007-11-27

gig harbor wa
Report This As A Bad Post Add To Your Ignored User List
apparently there are multiple witnesses that are apparently saying this
Quote:
was making a strange rattling noise and trailed smoke and debris as it swerved above a field

uncontained engine failure? wasn't there another one recently?

link https://www.reuters.com/article/us-ethio....
Thelazer
Posts: 234
Incept: 2009-05-11

Davenport, Fl
Report This As A Bad Post Add To Your Ignored User List
Lawsuits on the ready....
Flaps10
Posts: 6999
Incept: 2008-10-17

PNW
Report This As A Bad Post Add To Your Ignored User List
Reports of the Ethiopian plane trailing burning debris prior to impact.

That's not a trim problem
Tickerguy
Posts: 156505
Incept: 2007-06-26
A True American Patriot!
Report This As A Bad Post Add To Your Ignored User List
We'll see..... but do realize that we just had a small plane go down in Lake Okeechobee and had eyewitnesses claim that they saw SOMEONE SWIMMING away from the wreckage.

Nope. All pax were in the hull and dead.

Now that's pretty damned blatant -- and wrong. Eyewitless statements are, unfortunately, not very useful in these cases. Now if the FDR shows a detonated engine, well, ok......

----------
Winding it down.
Redjack
Posts: 135
Incept: 2018-01-29

Iowa
Report This As A Bad Post Add To Your Ignored User List
In my prior role, automation was my thing.

Simple process. A ribbon of material coming down the line that is then cut, split, and routed to packing machines. Simple system.

Except that it loses material on occasion. Bad photo eyes (sensors), power blips, or just a "math error". After $20 million and 90 days of work, we got it to be more reliable, but every once in a while it goes bad.

Now, I made sure that line was maintained. We fixed problems as we found them. A car will not be maintained that way.

I have moved to general management, but my opinion on auto pilots is the same. Be very careful trusting it in critical situations.
Gianmarko
Posts: 126
Incept: 2013-03-11

Bern, Switzerland
Report This As A Bad Post Add To Your Ignored User List
i am a big fan of automation. and for sure aviation is one field where automation is applicable and possible

autonomous flight is easy, i have a usd200 airplane model that will fly autonomously, and can be flown by anyone. it will recover from any attitude, and i mean any, even after being thrown in the air tail first.

however

the problem is that designers, perhaps under regulatory pressure, or pilot's union pressure, and often there is a lot of overlap as unions and regulators are run by pilots, insist in keeping deeply faulty devices in the loop: the pilots.
on top of that, they insist in having TWO pilots in the loop. did you know that in an airbus, if the right guy (the first officer) turns right and the left guy (the captain) turns left, the airplane goes straight? makes no sense to me, but thats how it is...
of course there are override buttons, and the two guys can override each other...
modern liners are very complex, and have very complex, and often badly (or not) documented and obscure failure modes, that are often very difficult to assess and remedy. many accidents happen exactly because of that. i know of many cases where the pilot went "and now what is she doing..."

also, there is something i still find absurd, which is, when the autopilot runs out of ideas, it will disengage and hand over a badly upset airplane to an half asleep pilot, perhaps at night, perhaps in the middle of the ocean (AF447)

i have an AP on my airplane. is a cheap AP but even when running out of authority, it will keep flying the airplane until i disengage it. predictable, and safe. if it disengaged suddenly at full aileron authority, it would send the airplane into 2 or 3 rolls before i could do anything. disengaging at full elevator, could result in the wings going their way...

modern liners are amazingly safe machines, and they are so perfect that the rare accidents often cant be reconducted to a pattern, and are the result of freak coincidences or occurrences.

so, in my opinion, having 40 years experience with complex IT systems, and a few decades of flying small planes, the next level of safety will be achieved when pilots will be removed from the loop, and airplanes will be flown by computers, with massive levels of redundancy. a certain burt rutan once said that the safest airplane would have a computer, a pilot and a rotweiler in the cockpit. the computer flies the airplane, the pilot feeds the rotweiler, and the rotweiler bites the pilot if he touches the controls.

the problem with "self driving cars" of course is orders of magnitude more complex than flying.

in the case of cars, the AP handing over the car to an half asleep driver tenths of seconds before impact is really something that should deserve anyone who was involved with its design to be thrown off a pier with a pair of concrete shoes on.

those who sell "self driving cars" are outright criminals.
Tickerguy
Posts: 156505
Incept: 2007-06-26
A True American Patriot!
Report This As A Bad Post Add To Your Ignored User List
Well here's an example from my home control system.

I have two "instances" of HomeDaemon-MCP here. One runs all the "stuff" in the and the second is out by the pool equipment in a weather-proof box. This avoids running signal wiring (which is subject to induced currents such as from nearly electrical storms, etc) over more than a handful of feet; this makes the design of the OVP for same (spikes destroy things, you see) much simpler and less likely to screw up.

That second system is slaved to the first BUT it will run entirely autonomously if the first disappears for some reason. It's a LOT simpler in terms of its logic tables, because it must do just one thing -- manage the pool systems. That's good, because if the circulation and chlorination are not maintained my pool turns into a breeding ground for frogs and mosquitoes in a fairly big hurry and cleaning that up pisses me off.

There are multiple sensors, valves and a VFD drive for the pump motor. There is, however, one intentionally uncovered risk -- that for "whatever reason" the pump will not have a prime and can't recover it quickly (e.g. severe water level problem, etc.) IF it happens then the result is that the graphite/ceramic seal in the pump will overheat and be destroyed. Note that this is the same risk, and has no mitigation, in a non-automated setup (e.g. where you just have a timer, or just run it all the time.)

The motor itself is protected by a slinger on the shaft; IF the seal fails and leaks the motor won't be destroyed -- it will just leak, and not all that fast either -- a few drips a minute in a severe circumstance, maybe a few gallons a day. But it will leak water until the seal is replaced.

Now I CAN put a pressure sensor on the output side of the pump, and COULD put a vacuum sensor in the input side. Given sufficient resolution this would be able to detect either a plugged inlet (which shouldn't be possible since there are two feeds and both would have to be plugged to completely stop up the suction side) OR a situation where there's no water in the pump bowl (e.g. zero pressure over some period of time on the output.)

So why don't I have those sensors in there? Two reasons:

1. They can fail like anything else, and give bad data. If the pump shuts down due to a bad sensor and I'm not home then I get the frogs. That's bad.

2. The sensors are more expensive to install than a new seal is -- by quite a bit. Further, the seals are wear items; they have to be changed every year or two anyway. Therefore IF I have such a failure the seal price is reasonably cheap and the damage is limited the seal (the motor could care less; it has its own fan cooling and running with no load on it it won't draw much current or generate much heat either.) An accurate current sensor on the VFD that can talk to the CPU (which I could also put in and would detect either a dry pump or a jammed rotor before the VFD's drive system tripped) is even MORE money (by quite a lot.)

3. If I'm paying attention and look at the cameras once in a while and I'm gone a severe problem (e.g. pipe break and loss of enough water to lower the level where it will happen) is visible, and I can shut it off manually. I still am out a seal in that instance but I stop consuming power.

So my judgment is that the sensor isn't worth it. Nobody's going to die and the damage if the "bad thing" happens costs me about $20 and a half-hour of time to change it out. Never mind that I can't do anything about it if I'm not home when it happens anyway until I return.

Now if the situation was different in terms of risks, costs, etc I'd see things differently. If someone was going to *die* I'd see things REAL differently.

But, quite quickly, I then need to cover redundancy -- if it's important enough to make SURE it works correctly then I need to somehow implement the logic in such a fashion that if it fails it does so in a way that doesn't kill people or destroy property.

Can robots do this? Sometimes. But not always, and the odds of them being successful go down a lot as the complexity goes up. Anything the robot didn't get programmed to foresee it likely cannot handle AT ALL; so-called "inference engines" are bull****; they infer nothing, really, that is beyond their trained data set.

People do that sort of thing quite well.

Robots are incapable of it.

----------
Winding it down.

Redjack
Posts: 135
Incept: 2018-01-29

Iowa
Report This As A Bad Post Add To Your Ignored User List
MG.

I used to work in a plant that had pumps that feed 600 psi liquid to an reverse osmosis plant (mechanical evaporator). When a pump runs dry, it will destroy itself and possibly spontaneously disassemble. We tried pressure sensors, gauges, etc. The best solution was a pop off valve and a setting on the VFD for under load (below so many amps, the drive shuts down).

The sensors were checked on a monthly PM. They STILL failed, turning my nice expensive RO system into a massive pile of goo that I (along with my pet engineer) had to go clean out. The pop offs did work, but they also would get stuck sometimes and either vent or allow higher pressure.

I love automation. Made a career out of it. But after 20 years I have come to the opinion you need a person near by for when the PID can't catch up, or when God decides to prove that He is a better designer than you. At my current job, they are looking at automating a very dangerous process. My input was that there had better be an "Oh Crap" button at the consul for when the automation fails before we burn the plant down and make the news. Again.
Kfell
Posts: 36
Incept: 2014-09-09

New Hampshire
Report This As A Bad Post Add To Your Ignored User List
In my 44 years of driving in the snow, the biggest danger is the jackass driving the other vehicle ESPECIALLY the plow truck drivers.
Gianmarko
Posts: 126
Incept: 2013-03-11

Bern, Switzerland
Report This As A Bad Post Add To Your Ignored User List
Quote:
I have come to the opinion you need a person near by for when the PID can't catch up, or when God decides to prove that He is a better designer than you

totally agree
the problem is that they insist in putting pilots instead of techies in the control room, and expect them to understand systems that are beyond them.
plus the industry requires thousands of new pilots per year, and they look for people who can fly airplanes rather than people able to handle the computers that fly the airplanes.

at some stage they either remove completely pilots and put in the cockpit engineers who are specialized in automation, or at least include such engineer in the loop, somehow.

currently, there are two folks in the cockpit who spend 99% of their time bored to death, and when the SHTF, they are supposed to immediately spring into action and save the day. a lot of times, they disastrously mess things up doing what they should avoid: try to hand fly the airplane.

add to that that most pilots only saw emergencies in the simulator.


dont get me wrong, overal commercial flight is incredibly safe, a lot safer than before automation was introduced, but somehow people is reluctant to accept that now the most common cause of accidents is pilots unable to deal with extremely complex airplanes.

probably the best and most cost effective solution would be to accept that zero accidents is clearly impossible.


Analog
Posts: 1628
Incept: 2010-12-29

arkansas ozarks
Report This As A Bad Post Add To Your Ignored User List
Coming out of the Nuke Industry I detest excess automation and reliance on it.

Once you start down the road of believing in a computer program, common sense leaves.
That's why i believe in a simple and solid analog control system with a good computer monitoring it and reporting to the human in charge of the machine when it needs attention.

Declare a stall by computation from AOA and airspeed sensors that can stick or get a grasshopper in the pitot tube?
That's bad enough, but to give that computer program control over the elevator is insanity.

1947 Bonanza had better stall detection than that - a small vane atop the wing that commences to flutter when airflow separates .
That's direct indication of the physical phenomenon, not some programmer's guess.
The vane operates a switch that turns on a distinct sounding alarm .

K.I.S.S.

Save your sophisticated computers for the training simulator and compute there from first principles.

see my signature

a.

----------
Never trust a computer with anything important.
Tickerguy
Posts: 156505
Incept: 2007-06-26
A True American Patriot!
Report This As A Bad Post Add To Your Ignored User List
Indeed.

Add the complacency impact of not having to actually fly the damn thing (or drive the damn thing) and thus the loss of contextual awareness (e.g. "how is it feeling?" .vs. what you USUALLY feel) and you're asking for a disaster when you dump control back to the human when the machine goes*****-up after you've deprived said human of ALL of the contextual sensory inputs for the last "X" period of time.

----------
Winding it down.

Village-idjit
Posts: 661
Incept: 2010-06-09

B'ham WA
Report This As A Bad Post Add To Your Ignored User List
Agreed Analog and Karl, besides flying the airplane and "staying in the loop" a relatively new problem or issue (since inexpensive cockpit GPS systems have come about) is what the FAA is calling the "curse of the magenta line."

Who needs pilotage, navigational landmarks, VOR, ADF or dead reckoning when you can just plug departure point and destination into that handy cockpit NAV unit, and simply follow the magenta line to where you want to go. No need to look up airspace restrictions, or even have a paper chart, as that magic display will take you there.
Wifi
Posts: 1178
Incept: 2013-02-13

Seagrove Beach
Report This As A Bad Post Add To Your Ignored User List

----------
Hurricane Evacuation Plan
1.Grab Beer
2.Run Like Hell
Login Register Top Blog Top Blog Topics FAQ
Showing Page 1 of 4  First1234Last