A “Toolbox” for Forensic Engineers

Introduction 13
1. Inability to reproduce conditions of use.
2. The test technique is not monitored correctly.
3. The test may not adequately detect product defects.
There is of course failure to test under realistic conditions at all, the
results of previous experience being preferred. An example of such a failure
occurred where hot water seals achieved their intended purpose during intermittent
conditions of use, but failed completely during near continuous
exposure. Testing was not performed under the changed conditions, so the
failure mode was missed entirely (see Section 6.7).
1.5.1 FMEA
An important tool for considering potential and actual failure modes in
products is FMEA (failure modes and effects analysis). It is not in itself a
product test method but rather a way of assessing product defects and, if
necessary, specifying changes to product testing or suggesting new test methods.
4,5
Pareto analysis, FTA and SPC (statistical process control) are all techniques
used by design teams both for new and existing or long-standing
products. Most factories gather statistics on their product lines through
quality departments. Many inspectors are not engineers, however, and may
not recognize a particular problem as rating high in importance compared
with others. Thus surface blemishes are easy to spot on the line and are clearly
defects of a kind, but not usually serious enough to affect product safety.
Such obvious (patent) defects may be contrasted with latent defects, which
lie hidden within products or components. The study of latent defects is the
particular responsibility of the design team, who will use FMEA (failure
modes and effects analysis) to identify, classify and act upon. They may be
defects in products returned from customers, defects found internally or by
other methods such as routine testing and inspection.
A common design defect in many products, for example, is sharp corners
in products which, when stressed, raise the local stress above the failure stress
of the material concerned. Thus fracture will start at such corners, and
propagate into the interior of the sample. If the product shape is made by a
mass manufacturing route such as injection moulding or casting of a metal
alloy, it will occur in every product so that its likelihood of occurrence (L)
is 10 on a scale of 1 to 10. If the severity of the consequences of fracture is
high (say 8), and it is easy to detect (with a value of say 8), then the product
of all three factors is 640. This number is known as the risk priority number
(RPN), in effect a criticality threshold, so
RPN = (likelihood of occurrence, L) ¥ (severity, S) ¥ (detectability, D) (1.1)