Fatigue is not weakness

Why a part fails at a load it has already carried many times.

A part failing under a load it has already carried thousands of times without incident has not become weaker with age or use in the ordinary sense, it has been accumulating microscopic damage on every single cycle since the very first one, invisible and undetectable right up until that accumulated damage finally connects into a crack large enough to break the part outright.

The mechanism behind fatigue failure

A part loaded well within its rated strength survives any single application of that load with no visible harm at all, and a strength calculation checked once against the largest expected load correctly predicts that survival. What that same calculation does not capture is that repeated loading, even at levels far below the material's rated strength, causes a slow, cumulative process at the microscopic scale, tiny imperfections in the material growing by an infinitesimal amount with every single cycle of load and release, called fatigue. Each individual cycle's contribution is too small to measure, let alone notice, but the damage does not heal or reset between cycles, it accumulates, and after enough repetitions a crack finally forms and begins growing large enough to be a genuine structural threat, at which point the part can fail suddenly under a load it has shrugged off uneventfully thousands of times before.

The light-switch comparison

Flicking an ordinary household light switch takes almost no effort and leaves no visible mark on the mechanism, and the millionth flick of that same switch, performed with the identical light touch as the very first one, still feels exactly the same to the finger doing it. Yet switches do eventually wear out and fail, since the internal contacts and spring mechanism accumulate real, physical wear with every single operation regardless of how gently each flick was performed, wear too small to notice on any one flick and impossible to ignore once enough of them have happened. Nobody blames a failed light switch on the final flick being unusually hard, everyone understands, correctly, that the failure was earned gradually across every one of the flicks that came before it, which is precisely the same logic fatigue asks an engineer to apply to a structural part loaded and unloaded thousands or millions of times over its service life. The part that finally cracks under an ordinary load has, in a very real sense, been failing quietly since the day it first entered service. The failure, when it finally comes, rarely looks dramatic either, a switch that used to click crisply starts to feel a little softer or a little stickier for weeks beforehand, the spring's temper gradually giving way and the contacts slowly pitting from thousands of tiny sparks, small warning signs available to anyone paying attention long before the mechanism actually stops working.

Why a single strength check misses this entirely

A strength calculation asks one question, will the part survive this specific load applied once, and answers it correctly for exactly that scenario. It has no mechanism built into it for asking a completely different question, will the part survive this load applied and removed repeatedly, ten thousand times, a million times, across its actual working life, since that second question depends on cumulative damage a single static calculation was never designed to track. A part can pass its strength check with a comfortable margin and still fail well within its intended service life purely from fatigue, which is exactly why fatigue is analysed as its own separate discipline entirely, with its own dedicated testing and its own dedicated calculations, rather than being folded into an ordinary strength check as an afterthought. Establishing that separate answer usually means physically cycling real test parts, or a representative sample of the material itself, tens of thousands of times on a dedicated rig built purely to apply and remove load repeatedly, since no calculation alone can substitute for actually watching how many repetitions a given design genuinely tolerates before it cracks.

The one number worth remembering

A part loaded to only a modest fraction of its rated strength on every single cycle can still fail after a large but entirely finite number of repetitions, sometimes in the thousands, sometimes in the millions depending on the material and the load, a number that a strength calculation focused purely on the largest single load a part will ever see has no way of predicting at all, since that calculation was never asking about repetition in the first place.

Where some materials genuinely stop accumulating damage

Not every material behaves identically under repeated loading. Many steels show a genuine endurance limit, a load level below which the material appears able to survive essentially unlimited cycles without ever accumulating fatigue damage in any meaningful sense, which is why a correctly sized steel part kept below that threshold can, in practice, be treated as safe from fatigue for the whole of its service life. Aluminium and many other common structural materials show no equivalent floor, continuing to accumulate at least some fatigue damage at almost any load above zero, only more slowly the lighter that load is, which means a design in aluminium can never fully retire the question of cycle count the way a correctly designed steel part sometimes can, only push the eventual failure further out by keeping the repeated load as low as the design can reasonably manage. That endurance limit is also easy to lose in practice even in steel, since a single overload cycle that briefly exceeds the threshold can leave behind microscopic damage the material never had a chance to accumulate at lower loads, which is why a steel part is only genuinely exempt from the cycle-count question if every load it ever sees, including rare overload events, stays reliably below that floor.

What this changes in practice

Designing against fatigue means asking not only how large the worst single load will be, but how many times a more ordinary, everyday load will actually repeat across the part's intended service life, and sizing the part against that repeated, cumulative demand rather than only against its single worst moment. A part deleted or thinned down purely because it comfortably passed a one-time strength check, without ever being asked how it would fare after the tens of thousands of ordinary cycles its actual job would demand of it, is carrying a risk that will not show up in testing, in inspection, or in any single measurement, only in service, and often without warning, at which point the honest question is never whether the part was weak, it is how many cycles it was actually asked to survive against how many it was ever checked for.