Page 268 - Handbook of Materials Failure Analysis

264 CHAPTER 11 The reliability design of mechanical system

Load (external stress, internal stress)
Component F = f (t)
(material)

Failure mode : malfunction
Failure mechanism : fracture
Failure site
FIGURE 11.5
Two elements of failure mechanics, stress, and materials.

Additionally, there are two other issues involved in establishing verification
specifications. First of all, we think that new products can check the combinations
of specifications used for similar products. This is a misunderstanding. Potential
problems inherent in new products cannot be identified using old specifications.
The new product incorporates innovative structures, new materials, and different
software for upgrading performance and decreasing cost. These cannot be adequately
tested using existing specifications. In terms of reliability, adopting new structures
may alter failure mechanisms or introduce new ones, as two elements of failure
mechanics in Figure 11.5, stress and materials, have changed.
By using the previous specifications, new failure mechanisms are not easy to be
identified for products that have the design modifications. In addition to updating the
specifications, we should also consider what new testing might be effective. For
example, is it possible to apply the test specifications for the Boeing 777 fuselage
made of aluminum alloys than to the Boeing 787 Dreamliner fuselage, which incor-
porates new materials, like CFRP? Obviously, we know that the previous test specs
would be improper.
The other issue is that reliability quantitative specifications include estimating
item lifetime. Accident caused by material degradation during use could tarnish
the company’s reputation. But most people consider this task beyond the scope of
possibility. Generally, obtaining quantitative results in reliability analysis is very dif-
ficult. Reliability specialist Patrick O’Connor wrote in Practical Reliability Engi-
neering that there are basically three kinds of situations—small, moderately large,
and very large number of factors and interactions [9]. A small number of factors
can be predicted with the physical model. A very large number can be predicted with
statistical models. Predictive power diminishes, however, in the case of a moderately
large number of factors pertaining to reliability.
Reliability prediction is a necessary task to be undertaken. Let’s look at a product
in the standpoint of reliability problems. We know that there are a few sites in