Friday, March 2, 2012

Mechanical Engineering


4.     Feedback in mechanical engineering

This goes back a long way, right back to ancient times, where float valves were used to regulate the speed of Greek and Roman water clocks. A float valve uses some air filled object floating on the surface to operate a valve which lets water in (or out) to restore the desired level. Float valves are still used today in toilet flush systems, where a ballcock is used  to open or close the water inlet. This (as you should know by now) is an example of negative feedback, because it keeps the water level steady. Another current use of float valves is in carburettors, though these are dying out in favour of fuel injection systems.

These ancients were pretty smart guys though, not limited to simple water clocks. The maestro seems to be one Ctesibius or Ktesibios (around 270 bc),  who invented a number of mechanical devices, including a water organ with rows of sound pipes and a keyboard; and a very fancy water clock with a number of  whistling birds and ringing bells, no less. Unfortunately, none of his work survives, but he was certainly a pioneer of feedback mechanisms.

A little later on brings us to the windmill, which was enhanced in 1745 by a blacksmith, one Edmund Lee who added a fantail to keep the face of the windmill pointing into the wind. This of course is yet another example of negative feedback. These mechanical negative feedback mechanisms are also examples of simple control systems, but purely mechanical rather than electronic, and it is the electronic form which dominates these days.

The next big step forward in mechanical control was the use of centrifugal governors. It seems their first use was in 1787, when Thomas Mead regulated the speed of rotation of a windmill by using a centrifugal pendulum.

But the best known early use was by James Watt in 1788 to regulate the speed of his steam engine. This was a big step forward in mechanical engine design, and one of the keystones of the industrial revolution.

It works by having a pair of weighted balls suspended from arms attached to a spindle, which is connected to the output shaft of the engine. The arms are linked to a valve that regulates the steam input. As the engine speeds up beyond the desired rate, the spindle rotates faster, and the flyballs are driven outwards and upwards by centrifugal force. This movement partly closes the valve, reducing the amount of steam delivered to the engine. Likewise if the engine slows too much, the valve lets in more steam to speed it up again. A beautiful example of negative feedback, keeping the speed stable.




Governors are also used in gas and steam turbines and internal-combustion engines, where the governor regulates the flow of fuel. In hydro-electric turbine generators, the governor alters the water flow by opening and closing gates and valves, in aircraft engines it varies the pitch of the propeller blades attached to the engine.

An ideal governor controls speed exactly, with no variation. This requires something that responds quickly and accurately. But things don’t always work do smoothly; there is often a time lag while the governor responds, and this can cause instability, with the system varying alternately between too fast and too slow.

People can also act as governors. When you drive a car, you exert pressure on the accelerator to keep the car at the desired speed (well, you should). If the car starts to go too fast, you ease off the pressure, and if it starts to go too slow, you increase it. A learner driver can have problems with this control, and this can lead to the classic lurching motion, which is a form of the instability mentioned. As the car speeds up, the learner’s foot is thrown back a little too much, and the car slows quickly, which throws the learner forward and the pressure is increased too much. This is a governor that is over responding. Responding too much, too little or too late is a classic problem with any feedback control system, and there are many examples of problems this can cause in life, from central heating systems to the management of the country’s economy.

Another early example of mechanical feedback was in The Great Eastern, which was one of the largest steamships of its time. It possessed a steam powered rudder with feedback mechanism designed in 1866.

By the late 20th century, mechanical feedback was giving way all round to microprocessor based electronic control systems. Systems using vacuum pressure to adjust the timing of car engines were being replaced by engine management systems. As is often the case with developing technology, the components for electronic engine management had actually been in place for quite a long time before they found their way into the average production car.

There are a couple of reasons for this, and these reasons are often not well understood by pundits trying to predict when new developments will become widespread in use. Firstly, there is the simple reason that it takes a long time for a new technology to be developed from initial prototypes to large scale mass production. There is a huge difference between making one or two, or even a few dozen working devices, and making hundreds of thousands or millions. Production engineers are quite a different breed from research and development engineers, and there is often a certain amount of healthy friction between them. The R & D (research and development) guys tend to look down on the production teams as merely putting into practice what the geniuses in R & D have produced. The production teams are somewhat scornful of the R & D guys, regarding them as living in an ivory tower, with no concept of real world problems. And there is also a big difference between getting something to work in the development department to making sure that it works ‘in the field’ where variations in components and environment can play havoc with the intentions of the designers. As indeed can the infinitely subtle ways in which those confounded people, the users, can abuse and misuse the product – a process which engineers often refer to as ‘finger trouble’. Just as teachers sometimes think that schools would be fine if it was not for the students, engineers sometimes cannot help feeling that the product would be fine without people using it.

But there is a second reason why it often takes a long time for a new development to make it into the marketplace. A new technology may have a cost/benefit advantage over the old technology, but the trouble is, the old technology keeps improving as well. Or at least, it keeps improving up to a point where it finally runs out of steam. So it can take much longer than expected for the new development to overtake the old one. In fact, sometimes it never does. It took longer than many people expected for semiconductor memories to overtake the now obsolete ‘core’ memories, back in the 1960-70’s, because people kept finding better and cheaper ways to make the old core memories. Primarily, this involved using female labour in the Far East, whose nimble fingers could make the small devices better and faster than previously expected. But later on, Intel spent a large amount of time and money developing ‘bubble’ memories, based on magnetic bubbles. These never really caught on, because semiconductor memories and floppy/hard disc technology was developing at such a pace.

Machine Tools.

I have already described how robots use feedback, but there is another kind of mechanical system that uses feedback, and that is the Machine Tool. Machine Tools are in a sense, the ancestors of robots. Whereas robots can be (but are not always) mobile, machine tools are always firmly fixed to the floor. It is a question of bringing Mohammed to the mountain, if the mountain cannot be moved. Parts to be manufactured are ferried to one or more machine tools to be punched, drilled, milled, bent, cropped or otherwise manipulated. Simple machine tools (such as a lathe) can be, and still are, operated manually. But large complex tools are now generally controlled by programs, like computers.

The program controls the sequence of operations to be performed, and as in robots, feedback can be used to detect the results of the operation so it can be controlled more precisely. For example, feedback can be used to take into account the wear and tear on the machine’s own tools. So as the tool wears, the machine can be made to apply more pressure, or to move further.

Another use of feedback is in error detection and recovery. If something goes wrong during a machine tool operation, the tool can sense it and either call for human assistance (Help!), or take recovery action itself. And in these days of ever increasing emphasis on health and safety, machine tools can also detect safety hazards and take appropriate action.


Early machine tools were controlled by programs on a variety of media – punched paper tape or cards, magnetic tape – but now generally they are driven by microprocessors with semiconductor memory. Most small ‘computers’ now are in fact microprocessors which are built into other equipment – cars, washing machines, Televisons, DVD players, as well as machine tools. My car has, I believe, about 6 microprocessors hidden away somewhere, one of which just controls the ‘crash detection’ system. Microprocessors like these are called ‘embedded’, because they are included in the product, but hidden away from the user. Which is just as well, otherwise there would be even more finger trouble.

No comments:

Post a Comment