Thermal cycling, and the failures caused by switching off
Why repeated warming and cooling breaks things that steady heat does not.
Repeatedly heating something up and letting it cool back down does more damage over time than holding it at a steady high temperature for the same total number of hours, because it is the number of times a joint expands and contracts that drives this particular kind of failure, not simply how hot the part gets or how long it stays there, which means switching a hot component off and on can be genuinely harder on it than simply leaving it running.
Why switching something off can be harder on it than leaving it running
A component held at a steady operating temperature settles into that temperature and stays there, every material inside it expanding once to whatever size that temperature calls for and then remaining at that size for as long as the temperature itself remains unchanged. Nothing about staying at a fixed temperature, however high, repeatedly stresses a joint between two different materials, since a single, one-time expansion is not the kind of repeated loading the earlier articles in this set on fatigue already established as the real driver of that particular category of failure. Switching the same component on and off instead sends it through a full expansion and contraction with every single cycle, and if that component contains a joint between two materials that expand and contract by different amounts for the same change in temperature, each on-off cycle stretches and relaxes that joint in exactly the repeated, back-and-forth way ordinary mechanical fatigue depends on, only here the cyclic force driving it is temperature itself rather than an externally applied mechanical load.
The pothole comparison
A small crack in a road surface lets water seep in during a wet spell, and if that water is still there when temperatures drop overnight, it freezes and expands, widening the crack by a small amount before thawing again the next day and letting still more water seep in, deeper this time, ready to freeze and widen the crack further the following night. No single overnight freeze is doing anything dramatic on its own, and a road held at a constant temperature below freezing all winter, with no thaw and refreeze cycle at all, would develop nowhere near the same damage even over an identical number of cold hours, since a crack that never gets the chance to fill with water again after a thaw has no fresh water left to freeze and expand. It is specifically the repeated cycling between frozen and thawed, each cycle doing a little more than the last one left behind, that turns a hairline crack into a genuine pothole, and thermal fatigue in a manufactured joint follows exactly the same pattern, with repeated on-off cycling doing damage a single sustained temperature, however extreme, simply does not.
Why materials joined together expand at different rates
Almost every real joint brings together two materials that are not identical, a metal lead soldered to a circuit board, a ceramic component bonded to a metal case, a coating applied over a substrate it does not perfectly match, and each of those materials has its own characteristic rate of thermal expansion, the amount it grows or shrinks for a given change in temperature. When two materials with different expansion rates are joined rigidly together and the whole assembly changes temperature, the two sides try to grow or shrink by different amounts and are physically prevented from doing so freely by the joint holding them together, and that mismatch shows up as real internal stress concentrated right at the interface between them. A single heating and cooling cycle produces one instance of that stress, felt once and then released once the temperature settles back to where it started, but thousands of on-off cycles over a working life apply that same stress thousands of times over, and the joint responds to that repetition exactly the way any material responds to repeated stress, by slowly accumulating the kind of microscopic, unhealing damage the earlier fatigue articles in this set already described in detail, only here every cycle of the fatigue damage traces back to a switch being flipped rather than to a mechanical load being applied.
One figure worth keeping in mind
The mismatch in expansion rate between two ordinary joined materials is often only a few parts in a million of length per degree of temperature change, a genuinely tiny per-cycle effect on its own, and yet multiplied across a real joint's dimensions and repeated over thousands of ordinary on-off cycles, that tiny mismatch accumulates into exactly the same kind of fatigue crack a much larger single mechanical overload would have caused in one dramatic event.
Why this matters in practice
A part rated for continuous operation at a stated high temperature has not automatically been rated for the very different demand of being switched on and off repeatedly, since the two load cases are testing genuinely different failure mechanisms, one governed by steady-state heat and material limits, the other governed by the total number of expansion-and-contraction cycles the joints inside it will ever be asked to survive. A datasheet's single maximum operating temperature figure answers the first question thoroughly and says nothing whatsoever about the second, which is exactly the gap a designer relying on that one figure alone can fall into without ever realising a genuinely different specification was missing. Equipment that is switched off overnight to save energy, or cycled on and off automatically by a thermostat responding to demand, is quietly accumulating thermal fatigue cycles that an identical unit left running continuously at the same peak temperature never sees, which is why a genuinely reliable design has to account for the expected number of on-off cycles over its working life as its own separate specification, distinct from the maximum operating temperature that a simpler check would otherwise have stopped at. This is also why a piece of equipment left running continuously through a whole shift, rather than switched off between short breaks in demand, can genuinely outlast an identical unit cycled on and off far more often, even though the continuously running unit spends more of its total life at its peak temperature and would look, on paper, like the harder-worked of the two.