Thermal cutoffs (TCO's) take many forms, but those commonly found in consumer products are small devices used to prevent overheating. Like fuses, they are typically "one-shot", non-resettable devices which open upon reaching a particular temperature. They are usually used as backups to thermostats or other temperature-regulating components.
A common design for thermal cutoffs, with some of the failure modes, is shown below.
This type of device uses two springs, one heavy spring which holds the contacts (see Star Contact on the diagram) closed, and one light spring (trip spring) which opens the contacts when the compression on the heavy spring is released. An organic pellet serves to hold the heavy spring compressed, and the melting point of the pellet determines the temperature rating of the TCO.
A number of compounds are used for the pellets, and they do not all behave the same. Several have been found to cause problems. In some cases, choosing a TCO with a temperature rating too close to the operating temperature of the appliance will result in pellet shrinkage over time. This may cause slow opening, in which arcing damage to the contacts is possible. If the pellet sublimates, and the material condenses in the wrong place, the device may operate incorrectly. Because of this last phenomenon, the temperature gradient across the TCO must be correct, typically the pellet at the coolest end.
I had the opportunity to study several models of TCO in-depth. I examined 5 of one product which had been involved in "incidents" (anything from scorched countertops to fires). It was obvious that, in at least some of these units, the TCO had failed to open, and in two of them, the TCO looked like it had been hit by a torch. However, since the units had been severely overheated, this was not considered surprising.
Then a recall was issued on the product, and the manufacturer agreed to divert 1000 of these recalled units for close examination. We examined over 500 of these. We removed the TCO's, mounted them in series connections, and radiographed them, noting the state of each pellet. We then temperature-cycled the TCO's repeatedly (while carrying normal operating current), and radiographed them again to look for pellet shrinkage. Finally, each TCO was individually heated rapidly as if it were in an overheating appliance. We did get a number of failure-to-open cases (in excess of 1%), including a couple which developed that "hit with a torch" appearance during the FTO incident: the excess heat is produced by the TCO itself.
Working with our metallurgy department, I developed a means for back-filling TCOs with a clear plastic, and sectioning them to examine their internal state at various stages of pellet strinkage, or after FTO or a fire. The diagram above is based on those observations.
The TCO is a great product, but not fool-proof. The recalled product in question had turned out to be unexpectedly durable, and the failures were showing up at twice the expected life. Bimetallic thermostats are not totally reliable either. A manufacturer should not be surprised if a thermostat fails after 100,000 cycles or less, and from my studies it appears that there's about a 50% chance that the thermostat will fail closed. If that happens, you need something like a TCO. It is likely that the thermostat is the life-limiting component of many heating appliances, and in that case, a 1% FTO rate of the TCO is definitely going to cause serious consequences when you have hundreds of thousands, maybe millions of the products in use. Many manufacturers now use two TCOs, two different temperature ratings, in an attempt to avoid this problem. This becomes even more important when the product is intended to run unattended.
One other observation of the study surprised me: in spite of a fairly high rate of "nuisance trips" in products that had been in service for about a decade, in none of the units examined did we find the TCO bypassed! I found plenty of evidence that people had opened the units to find the problem, and I found other types of repairs, but when they found an open TCO, they seemed to just give up. I've personally known one electrician who was not above bypassing a TCO, so it surprised me quite a bit that the general public seems fairly well-behaved in this department.
There is one class of product I wish had a TCO, the "protected" fluorescent lamp ballast, basically a current-limiting coil. Ballasts frequently fail after repeated cycling as the wire insulation degrades, causing shorting of adjacent turns. The units I have encountered contain a thermostat. When deterioration begins to cause excessive heat, the thermostat opens, and the lamp goes out. The trouble is, in a few minutes it comes back on. I investigated one fire in which a ballast continued to cycle on and off for several days over a holiday break, culminating in a fire that was quite costly. There's no point in a self-resetting thermostat in a component that usually overheats only when failing. A TCO, or a TCO plus a fuse, would be a better solution.