d were released at any time, the piece would return
to its original length. However, if the load had been excessive, and
then relieved, the extensometer would no longer read exactly 2.0
in., but something more.

Soft steels "give" very quickly at the yield point. In fact, if
the testing machine is running slowly, it takes some time for the
lower head to catch up with the stretching steel. Consequently at
the yield point, the top head is suddenly but only temporarily
relieved of load, and the scale beam drops. In commercial practice,
the yield point is therefore determined by the "drop of the beam."
For more precise work the calipers are read at intervals of 500 or
1,000 lb. load, and a curve plotted from these results, a curve
which runs straight up to the elastic limit, but there bends off.

A tensile test therefore gives four properties of great usefulness:
The yield point, the ultimate strength, the elongation and the
contraction. Compression tests are seldom made, since the action
of metal in compression and in tension is closely allied, and the
designer is usually satisfied with the latter.

IMPACT TESTS

Impact tests are of considerable importance as an indication of
how a metal will perform under shock. Some engineers think that
the tensile test, which is one made under slow loading, should
therefore be supplemented by another showing what will happen if
the load is applied almost instantaneously. This test, however, has
not been standardized, and depends to a considerable extent upon
the type of machine, but more especially the size of the specimen
and the way it is "nicked." The machine is generally a swinging
heavy pendulum. It falls a certain height, strikes the sample at
the lowest point, and swings on past. The difference between the
downward and upward swing is a measure of the energy it took to
break the test piece.

FATIGUE TESTS

It has been known for fifty years that a beam or rod would fail
at a relatively low stress if only repeated often enough. It has
been found, however, that each material possesses a limiting stress,
or endurance limit, within which it is safe, no matter how often
the loading occurs. That limiting stress for all steels so far
investigated causes fracture below 10 million reversals. In other
words, a steel which will not break before 10,000,000 reversals
can confidently be expected to endure 100,000,000, and doubtless
into the billions.

About the only way to test one p