In the laboratory, a prepared specimen was fitted to an apparatus,
and then subjected to repeated stressing. The nature of the
stressing was controlled, and in this case, the loading was
sinusoidal with a maximum stress 
and minimum stress
. The specimen was inspected periodically, at which time
a replicate tape was placed on the surface, and an image of any
cracks present was recorded. At the end of the experiment, the
images recorded were subjected to microscopic examination and
measurements of cracks recorded.
Data was recorded for three different specimens, at three different stress levels. The first specimen did not fail in the laboratory; the time of failure for the other two specimens was recorded.
The data consist of the lengths of cracks at the final observation point, followed by the lengths of any cracks which subsequently coalesced to form these cracks, recorded at the earlier time points.
One cycle is defined to be the time between the application of
and the next application of . It is known that
the deterioration of the specimen depends upon the number of times
it is stressed. It is possible, therefore, to simulate what might
amount to years of use in a real application, in a shorter time in
the laboratory, by just increasing the frequency of the stressing.
In this case the stressing was carried out at 30Hz. Time,
therefore is recorded as number of cycles and represented by N,
which is defined to be the number of cycles of stress undergone by
a specimen.