Reference
Exemplars
One in a Million - Using Exemplars to Prove Your Case
By Mark Hughes, P. Eng.
© 2001 Sintra Engineering Inc.
When fires occur as a result of a component or appliance that has failed, there is a tendency to think that if this happened once, maybe it will happen again. In the United States and to a lesser extent Canada, a great deal of effort is spent to prove that the manufacturers of a particular item are at fault for a real or perceived defect in their product. This real or perceived defect generally forms the basis of a product liability case (e.g. Firestone tires). As part of working on a product liability case, it is not unusual for a fire investigator to be asked to look at an undamaged version of the same product (i.e. an exemplar) to determine if the defect is present. This article examines the pitfalls of examining exemplars.
Exemplars.....what a bizarre word. It means an example of something. Typically if you have an appliance that causes a fire, then an exemplar is an undamaged version of the same appliance that caused the fire. The purpose of examining the exemplar is to assess if the fire was a consequence of something specific to the device that failed, or if the cause of the fire is something systemic and thus present in all units of that model type. This can be difficult to accomplish.
To understand why this is difficult to accomplish, one need only look at the problems with a number of coffee makers made in the 1980's (there is no specific reason to pick on coffee makers but I'm enjoying a good cup of java right now). These coffee makers were equipped with a single thermal cut off (TCO). This device would disconnect the power to the element if a temperature (as opposed to current) was exceeded. In the event that the thermostat failed, the TCO was supposed to shut down the coffee maker and prevent a fire. Unfortunately, TCO's sometimes fail themselves. As a consequence, if you had a thermostat that failed (they all do eventually) and a TCO that also failed, then the temperature of the element could run away and cause a fire (thermal runaway). The the manufacturer's solution was to manufacture all coffee makers with two (2) TCO's. This would allow for the failure of one, with the anticipation that it would be unlikely that the second would also fail.
What does this have to do with exemplars? The problems with TCO's are at the margins of use. Most TCO's function well and without incident. If you had a coffee maker and attempted to disable the thermostat, the TCO would in all likelihood prevent the thermal runaway of the element. Therefore testing an exemplar would show that the TCO works and that there is no problem with the coffee maker. An untrained investigator might then incorrectly assume that there is no problem with the particular make of coffee makers. Since TCO's that fail only represent a small percentage of those installed, it is unlikely that they will ever fail in the exemplars. This fundamentally illustrates the problem with exemplar testing. Beyond the specific example of TCO's most of the exemplar units should work and only rarely will one fail when tested.
Continuing my analogy, while I don't know the specific rate at which TCO's actually fail, let's use, for the sake of argument, a failure rate of 0.1%. That is, for every 1000 TCO's, one TCO would fail and allow current to flow after exceeding its critical temperature. As each coffee maker is only equipped with a single TCO, you would need to test 1000 coffee makers to find one with the same problem, a very impractical approach to testing exemplars. This can be made worse if the failure rate is an even smaller percentage, requiring an even larger number of exemplars to find a single unit that fails.
This means that in general, if you hook up an exemplar coffee maker it will work. If this is the case, why test exemplars? Product defects generally fall into two categories, occasional and condition specific.
Occasional defects relate to products that have a failure rate associated with their production, which means that of a 1000 produced a certain number will not function correctly. These occasional defects represent a quality control problem for the manufacturer of the product. Testing an exemplar will reveal the functionality of the coffee maker and show how it is generally supposed to operate. It may also reveal a general quality of manufacture. The coffee maker fires as previously described would result from occasional defects and we've already seen that testing exemplars my not reveal this type of defect. These coffee maker defects are well known, what about those that we are unsure about, will exemplar testing help determine cause?
Condition specific defects relate to products that have the potential to cause a fire or failure when components are otherwise free of manufacturing defects. These tend to relate to products that when operating under specific conditions, can no longer perform safely and will fail. These products are either under-designed or operating under severe conditions that would not normally be expected. An example of this type of failure is some GM transmissions (not to pick on GM) that can cause a fire when towing heavy loads or excessively spinning their wheels. If the specific conditions are properly recreated, an exemplar can be used to create the hazard or failure. Failures resulting from condition specific defects are much easier to recreate than failures that occur during normal use.
By understanding how the conditions of how a device failed, this can assist
the investigator in determining the merit of exemplar testing. Even if an
exemplar does not fail as expected, this does not mean that the specific
device that failed could not fail under similar circumstances.
return to Reference articles...
