In carrying out a medical study of the effectiveness of a new treatment, the “gold standard” is the randomized double-blind, placebo-controlled study. This type of medical trial is specifically geared to human participants, whose human minds might otherwise influence the outcome.
The essence of the double-blind, placebo-controlled study is that neither the medical staff nor the patients know who's getting the treatment and who’s getting the placebo, or inactive substance. We might consider the people who receive the placebo to be the “control specimens.” Some new clinical trials are adding a third group of “specimens.” These are people who are treated like the first two groups but not given any treatment. They become the “natural history” group.
The situation is different in inert objects. Obviously, we do not have to worry about the thoughts and emotions of the damaged component having an influence on the microstructure, strength or composition that we determine based on our test protocol. What we do have to worry about is how to interpret the results and what to tell the client or our boss, as the case may be.
Those who are beginning their career as a failure analyst often think that it is enough to compare the characteristics of the “subject component” to a new part that has never been used. But we need to pay more attention to the provenance, or origin, of the compare-and-contrast part. Some people use a random current production component, but what gives us the idea that the current production is free from problems? If the subject part has been out in the field for a while, it may be that the current production is just as bad but hasn’t had time to manifest the problem. Of course, we may have a set of specifications against which we may compare both the subject and the control specimens.
When we are working on large pieces of machinery, or civil structures, we’re unlikely to get someone to pay for, or even find, an “exemplar,” or “reference,” to be “sacrificed to science.” In this case, finding an appropriate reference specimen from another area of the subject of the investigation will generally have to suffice. Just as care is required in selecting a reference part, care is required to select a reference specimen from the subject part.
If we are dealing with a high-temperature application, choosing an area that would have been expected to have similar exposure should be considered. However, even this may be difficult, especially if the “failure” is going to be repaired rather than replaced. Sometimes the analyst is confronted with the fact that the only “reference specimen” available is published data about the alloy. It is important to consider not only the specification, which is usually based on minimum characteristics, but to look for “typical” data as well. The “gut feel” that results from years of experience in engineering design and manufacturing and construction practice, will be based on the results of the “typical” behavior of material manufactured to a given specification.
Whether we are dealing with a manufactured component, a machine or a civil structure (e.g., a bridge or building), we need to remind ourselves that:
- The part doesn’t know the specification.
- Many specifications have fairly arbitrary requirements.
So, what is the answer? When available, for manufactured parts, a successful durability-tested part is the best “control” or “reference” part. If we are confronted with a failure analysis related to assembly failures of current production and the failure rate is very low, then using randomly pulled or even any available components might be acceptable.