Imagine you put up a very long domino toppling construction, with lots
of tunnels, bridges, levers and turns... If one domino falls by
accident anywhere, half of your work may be lost. One way to avoid this
is to remove some of the dominoes. You only put them back for the final
and intended exhibit. That way, if the dominoes start falling anywhere,
only a small part of the construction will be lost, because the
movement was not transmitted to the rest.
Determine the boundaries
At what distance should the dominoes be placed from each other?
One way is to test a given distance. If it works OK, you keep that
distance. Doing so will bring problems. It will work often but it may
ruin your efforts on the long term. A far better attitude is to test
several distances. you will then notice that above a given distance you
start having problems and below another distance you start having
problems too. Choose a distance in-between...
Don't always use the median value
Say 5 millimeters and 30 millimeters appear to be the boundary values.
Then you might think that (5 + 30) / 2 = 18 is the optimal
value. It often is not. Rather, the optimal value will be found using
the logarithms of the values. In this case, the logarithm of 5 is 1.61
and the logarithm of 30 is 3.40. The median value of these logarithms
is (1.61 + 3.40) / 2 = 2.505. Then exponentiate back that value of
2.505 and you get a value of 12. There is a fair chance that 12 is a
saver value than 18.
Widen the boundaries
Suppose you have to choose between two types of dominoes. They're of
the same size... Well determine their boundary values. Say type 1 has
boundary values of 4 and 35 millimeters while type 2 has boundary
values of 7 and 29 millimeters. There is no doubt that type 1 has less
chance to fail you. The same way, when you conceive something, try to
make choices that increase the boundary values for whatever parameters
you are dealing with. Then you get a better chance that in adverse
situations, or simply trough aging, the device won't fail.
Do like the others
Copying what other people do works often. If in doubt, just copy. If
you're sure they're wrong, experiment and try to make out who's right.
Do never criticize the common approach just because you believe it's
not optimal. Quite often, in order to do better than the common
approach, you will have to invent a completely new system.
Do not confuse monster and production
A "monster" is a device you build to experiment something, possibly to
show that a given mechanism makes sense. For example in Electronics,
when building monsters we just love to use tantalum condensators. They
are cheap and very little, they can have huge capacities and altogether
filter efficiently high frequencies. They're just perfect... Except for
one thing: one of the main sources of failures in devices are those
same tantalum condensator, because they sometimes don't resist to being
used for months or years in a row. So, when conceiving the production
devices, we *must* replace the tantalum condensators with more
conventional ones. This implies to use bigger ones and sometimes even
to use two condensators to replace the one tantalum condensator (one
for the big capacity, one for the high frequencies). This is whatthe
prototyping is meant for... The last prototype should be a viable
production device.
Decrease the safety factor
Suppose you build a bridge. It has to withstand a maximum possible load
of cars and trucks of 1,000 metric tons. You compute it, so that it can
support 1,001 tons... Eight months after its opening, the bridge
collapses, under a load estimated at 200 tons... You ponder on the
event and you come to the conclusion that you should have built a
bridge capable to support 5,000 tons. That's what engineers call a
safety factor. A safety factor of 5 means that if the bridge has to
resist to 1,000 tons, you build it for 5,000 tons. That way you have a
fair chance that it survives to a load of 1,000 tons. This way of doing
has well served mankind since ages. But it may not always be optimal.
Suppose for example that the above bridge weighted by itself about
10,000 tons. It collapsed under a supplement of 200 tons... It may well
be that those 200 tons did almost not contribute to the collapse and
that the bridge would have collapsed anyway, even under zero load, a
few months later, under its own weight. Maybe the bridge is
ill-conceived... or ill-built... Maybe that a bridge,
with the same shape and size but weighting 50,000 tons in order to be 5
times stronger, would have collapsed even earlier... Secondly, a
heavier bridge will be stiffer, hence it may resist less to
earthquakes. Thirdly, the stronger bridge may have a frequency of
resonance that matches some of the frequencies it encounters naturally.
Fourth but not least, by spending five times more concrete and steel on
that bridge, you prevent the country from building a bridge somewhere
else. Even if your stronger bridge does the job correctly, maybe it
wastes valuable resources that could have been used elsewhere. So, the
point is that you should better try to understand how your bridge
behaves and what the possible threats against it are, rather than
boldly using a strong safety factor. You have to study how twists and
sideway forces act on the bridge and how it transmits those forces and
possible shock waves. You can implement techniques to
detect weak points in the construction. Say one pillar was made with
bad concrete, well you let such things be diagnosed, you take the bad
pillar down and you replace it by another one... Once the bridge is in
service, you can use detectors to monitor its behavior. Should it bend
too much or should abnormal cracks be heard, you can then remedy to the
problem, before any accident occurs. This way, some aerospace devices
achieve safety factors of a few
more than 1. Some even have safety factors below 1, thanks to the use
of actuators that counterbalance the adverse forces and shock waves,
dynamically.
Decrease the amount of components
Some engineers conceive electronic circuits or cars the same way that a
lady decorates her interior. They just add whatever seems useful or
advised. Protections, regulators, enhancements, dampers and other such
perfectisations. If you see such a radio or audio amplifier that is
made of a sea of transistors and other pieces, you can be sure that it
will fail soon and that the sound will be of poor quality. Madman Muntz
was reputed for finding ways to decrease the amount of electronic
components in the devices his factories made. This yielded for a
example a television set that was the most affordable of its time, that
consumed less power, heated less and had less failures. There were some
drawbacks, like less sensitivity when far from the emitter, but overall
this allowed many more people to own a good television set. The
ultimate refinement is to make individual pieces cumulate different
functions. With a cleverly placed resistor, a single transistor can
become both protected from overheating, get a more linear response, be
less sensitive to power source variations and degrade gracefully. The
distortions produced by this simplified assembly can be less disturbing
than the noise produced by a complex assembly that is supposed to
produce no distortion. Decreasing the weight, volume, power consumption
and price of the system, helps to implement the system redundantly,
which can lead to a high level of security.
