iler, of which there are many in use throughout the country.

Plain cylinder boilers are, as a rule, provided with mud drums located
near the back end. As a rule, also, these boilers are set in pairs over a
single furnace, and the mud drum extends across beneath, and is connected
to both, and one end projects through the setting wall at the side. Our
illustrations show a typical arrangement of this kind. Fig. 1 shows a
transverse section of the boilers and setting, while Fig. 2 shows a
longitudinal section of the same. It is a favorite method to connect the
feed pipe, F, to the end of the mud drum which projects through the wall,
and here the feed water is introduced, whether hot or cold; and there is
really not so much difference after all between the two, for no matter
_how_ effective a heater may be, the temperature to which it can raise
water passing through is quite low compared with the temperature of the
water in the boiler due to a steam pressure of say eighty pounds per
square inch. The difference in the effect produced by feeding hot or cold
water at the wrong place is one of degree, not of kind.

When a boiler is under steam of say eighty pounds per square inch, the
body of water in it will have a temperature of about 324 degrees Fahr.,
and the shell plates will necessarily be somewhat hotter, especially on
the bottom (just _how_ much hotter will depend entirely upon the quantity
of scale or sediment present). Now introduce a large volume of cold water
through an opening in the bottom, and what becomes of it? Does it rise at
once, and become mixed with the large body of water in the boiler? By no
means. It _cannot_ rise until it has become heated, for there is a great
difference between the specific gravity of water at 60 deg., or even 212 deg.
Fahr., and water at 324 deg.. Consequently, it "hugs" the bottom of the
boiler, and flows toward the _front_ end, or hottest portion of the
shell. Now let us examine the effect which it produces.

We know that wrought iron expands or contracts about 1 part in 150,000
for each degree that its temperature is raised or lowered. This is
equivalent to a stress of _one ton_ per square inch of section for every
15 degrees. That is, suppose we fix a piece of iron, a strip of
boilerplate, for instance, 1/4 of an inch thick and 4 inches wide, at a
temperature of 92 degrees Fahr., between a pair of immovable clamps.
Then, if we reduce the temperature of the bar under experiment t