When working in aerospace, I at one time was the supervisor of an electronics packaging group. Packaging in this case meant the mechanical design of electronic equipment —structurally, thermally, chemically, whatever— and the group was tasked with figuring out how to cut the weight from electronics subsystems and improve integration of the system. When first asked to do this, I was reluctant, because electronics never were my main love. But what I realized more and more clearly as I worked on the job, is that although solid state electronics components themselves are surprisingly reliable, (once they have gone through a bit of usage) in use they fail mechanically, often because of mechanical inputs, either from the environment or from people assembling, testing, and maintaining them. Even if this is not the case, if electronic devices are expected to do physical work, they need not only an input (which can be from an electrical device), but an actuator which allows them to effect the physical world. Such actuators are often a combination of electrical and mechanical (solenoids, electric motors, etc.) and exhibit traditional wear and anger at contaminating particles, and require such things as lubrication and controlled loads and temperatures. So it was an excellent assignment, because we could work on challenging mechanical problems that neither the electronic nor the traditional mechanical engineers were spending much time on.
I also became much more aware of the fact that given enough complexity and usage, electro-mechanical systems will fail. And there is a good possibility that they will fail at some point if they are not designed with the humans who produce and use them in mind, because not only do things break, but people (and even robots) screw up. I have many memories of examples of this—a flight spacecraft, built and assembled in clean rooms, showing clear signs of one of the technicians having eaten a sandwich containing tomatoes over one of the electronic packages—a contract technician at the Cape actually putting two male connectors together enough to blow up a science instrument on another flight vehicle (thought to be an impossibility).
I mention this, because I suspect that the seemingly miraculous success of digital devices, both in performance and profit, has caused us to give less attention to this. Perhaps it is not only because electrical engineers would just as soon not work on mechanical problems and vice versa, but also because technology has been changing so fast that many devices are essentially thrown away in a relatively short time (Apple hopes it is a year).
We hear more and more about autonomous cars. But they certainly have not been in use long enough to have good data on their behavior when operated and maintained over long period of time by typical people. I tend to keep my automobiles for 10 to 15 years and 150,000 to 200,000 miles. I expect to have minor problems with them in that time period and buy them replacement parts, but not to have a failure capable of killing me. And so far they have done that, partially due to the many years of experience and testing we have invested in them.
How many autonomous cars have been run in traffic for even 200,000 miles? One of the last two “incidences” my wife and I have had with cars was due to a neighbor backing rapidly out of a tree-and-fence lined driveway into my innocent wife, who was driving down the adjoining street. If the cars were autonomous, would its sensors have been able to detect the other car? Which one would have stopped? Both of them? The other was due to an almost new front tire delaminating while in heavy traffic moving rapidly on a very high bridge. This required a bit of quick decision making with little data. Neither of us had experienced such a thing. We had to decide when and where to get off the road, which was difficult because there was no turn off lane, and neither of us was sure what was happening to the car and whether it would continue in its present bumpy and veering mode, or go into a more catastrophic one. I am told that autonomous cars will be better than we are, but would they know what to do in uncommon failure modes such as this?
As another example of electrical failure over time, my previous pick-up was mechanically both reliable and fairly rugged, but as it acquired miles, it began to have failures in electrical components such as coil packs and solenoids. By the time I gave it to my brother’s farm after 125,000 miles, I had given up on many of its push-button operated features. The buttons would push, but the features wouldn’t operate. To make matters worse, accessing the faulty actuators was sometimes a nightmare (the rear-most coil pack and spark plug, the window and door-lock actuators), with the result that I looked hard to find a replacement vehicle with a minimum of such things. I succeeded and love my new one (even though it is too big) because it has very clever long-used devices such as cranks on the windows and keys to lock and unlock doors.
And then, of course, there are the interactions with electrical systems that grossly overestimate the capability of humans. An example was the almost catastrophic Three Mile Island nuclear plant shut-down many years ago—the report of the study committee (the Kemeney Report) portrays a very believable situation in which operators trained only in normal conditions (the way we learn to drive) made all the wrong responses when behavior became abnormal. Would computers have done better? Only if all failure modes could have been quantified and predicted.
Such mechanical weaknesses in electrical systems can be found at all levels. Stanford University has many buildings and large conference and class rooms, in which people who not only work at Stanford, but come from all over the world, give talks. I am one of them, and the room scheduling is done by a combination of a large computer and people who represent individual fiefdoms (and no, they do not always coordinate or agree), so I am often finding myself in rooms I am not familiar with. These days, of course, people tend to use large numbers (often too many) graphic images, usually in some format such as Power Point, and usually stored in their powerful portable computer, because these rooms tend to be equipped with built-in projectors, but not computers. The rooms do tend to be equipped with what are called smart panels, which typically contain the necessary electronics, buttons, switches, and lights to integrate one’s portable computer, microphones, lights,or whatever with the room. they are certainly simple compared to one's computer and the equipment in the room.
But my impression from using them, is that these panels are all different, usually without good instructions on how to use them (instructions are usually written by IT people who assume they shouldn’t be necessary and the rooms do not come equipped with these IT people) and unlabeled wires, (often coming from under the rug or other mysterious sources). Some screens are controlled by a switch, and others automatically when the projector is switched on, window covers may or may not be synchronized with the projector, and if one’s computer is on the odium, it often blocks the buttons, switches, and lights.
If I have never used it, I routinely visit the room I will use the day before my talk, and typically spend 15 minutes to a half hour figuring it out, sometimes failing and looking for someone who uses it often. Even if I have used it, I arrive at my talk ridiculously early because undoubtedly some one else will have used the room and left it in a different state than it was in the day before, especially if they have been able to physically access the projector. People still don’t seem to realize how the computer and the projector interact, so they fiddle with the manual projector controls, which of course are also not standardized. I often end up with several VIP types trying to help me figure out the AV system (often to their consternation with little success), but I guess it introduces an interactive and informal atmosphere to the session.
I used to complain mightily when the standard slide projector (the Kodak Carousel) was almost universal. The remote was often missing, slides could drop out if the retaining ring was not securely fastened to the drum, etc. etc. etc. But at least the projectors varied little from year to year and did not attempt to interact with all other equipment in the room, and after one learned their foibles, one could show slides almost anywhere. I almost want them back, because I have a sneaky feeling that such things as “smart panels” will not be standardized in my lifetime, and of course, they are not smart at all.