A “Toolbox” for Forensic Engineers

Failure Due to Manufacturing Faults 143
all 48 rivets, whereas the distribution would in fact be parabolic, with the
maximum loading being near the middle of the seam. This distribution could
result in the rivets near the middle, which is where the splits appeared to
have started, being exposed to a shear stress well above the average, in which
case the factor of safety may have been almost eliminated. To take into
account these two factors would be a daunting theoretical task. Furthermore,
when the panels of a stressed skin container are loaded in shear a secondary
force field develops due to buckling, the so-called diagonal cross tension field,
which imposes additional shearing force on the rivets. An alternative
approach to the problem would have been to measure strains in a real container
loaded to its full capacity and then infer from these strains what the
stresses must be. It is a technique often used to study stress distributions in
complex systems. However, the investigation did not justify this costly and
time-consuming analysis. Furthermore, it would only have supplied data for
a static container, whereas these containers split while they were being handled,
so a dynamic analysis under transient loading conditions is really
required.
Whatever theoretical modeling is undertaken, what is definitely known
is that the original design of these containers had been approved and tests
on prototypes had met the requirements of all relevant standards without
reservation. Moreover, the vast majority of these containers, presumably not
constructed with defective rivets, were still giving satisfactory service several
years after the design was introduced.
Another factor that came to light during the stressing examination was
the effect of load distribution inside the container. A single heavy item, such
as a machine tool, places a different loading on the sidewalls than the same
total load distributed evenly. If the container being handled bumps into
something, then the inertia loads also add significantly to the shear stress on
the rivets and is an obvious explanation as to why the splits occurred while
containers were being off-loaded and why all those that failed had held heavy
items of machinery. It is easy to appreciate that a container packed full of
clothing, for example, would be much less vulnerable than one with a single,
heavy item fastened to the floor near the middle. Additionally, the heaviest
part of a machine tool might be at one end, which would increase the shear
forces in the riveted seams at that end of the container.
The formal report concluded that a batch of containers had been produced
with a side-wall strength below that which was intended, because the
rivets had been set in the wrong condition, that is, without having been
solution treated as specified on the drawing. The factor of safety in the basic
design was adequate, but in the containers that split open it was reduced to
an unacceptable level because the rivets were too weak. The failures were