Fatigue: Why Most Parts Really Break

The load that never lets up

Guides 1 to 3 in this rung all told a story about one big moment — a single overload that pushes a crack past its critical length and it runs. But here is the uncomfortable truth that surprises most beginners: the majority of parts that fail in service never see that one big moment at all. They break under loads they carried safely thousands of times before, brought down slowly by a load that simply kept coming back. This is fatigue, and by common estimate it is behind 80 to 90 percent of all mechanical service failures — the broken axles, cracked wings, snapped bolts, and split crankshafts of the real world.

You can feel the whole idea in your fingers. Take a steel paperclip and pull on it hard — it barely stretches; you cannot come close to snapping it by pulling. Now bend it back and forth. After maybe ten or fifteen cycles it breaks cleanly, and the spot even feels warm. Each individual bend puts a stress on the outer fibre that is well below what a single pull would need to break it, yet the repetition does what no single load could. That is the signature of fatigue: failure under cyclic loading at a stress far below the static strength. The static strength number you trust from a tensile test — the yield and ultimate — is simply the wrong number to design by when the load cycles.

Counting cycles: the S-N curve and the endurance limit

To design against fatigue you first have to describe the load honestly. A cyclic load swings between a maximum and a minimum stress, and we split that swing into two numbers: the mean stress sigma_m = (sigma_max + sigma_min)/2, the steady level the load hovers around, and the stress amplitude sigma_a = (sigma_max - sigma_min)/2, how far it swings up and down. A rotating axle bends the same way each turn, so a point on its surface goes from full tension to full compression and back — a fully reversed cycle with sigma_m = 0. The amplitude sigma_a is what drives the crack, and more tensile mean stress makes fatigue worse.

Now run the classic experiment, first done systematically by August Wöhler on railway axles in the 1860s. Take many identical specimens, cycle each at a chosen stress amplitude until it breaks, and count the cycles to failure N. Plot amplitude on the vertical axis against N (on a log scale) on the horizontal, and you get the S-N curve — the master chart of fatigue. It slopes down: the higher the amplitude, the fewer cycles a part survives. Push the amplitude near the ultimate tensile strength and it fails in a handful of cycles; drop it low and it lasts millions.

S-N CURVE   (stress amplitude vs cycles to failure N, log scale on N)

  stress
  amplitude
    ^
    |*
    | \  * STEEL (and other BCC metals)
    |  \__
    |     \____
    |          \_______  knee, then FLAT ->  ENDURANCE LIMIT
    |                  ==========================  safe forever below this
    |   *
    |    \  ALUMINIUM / COPPER (FCC)
    |     \___
    |         \_____ keeps sloping DOWN: NO true limit
    +--------------------------------------------------> N
       10^3    10^4    10^5    10^6    10^7    10^8

  Read it: pick an amplitude -> the curve gives the cycles you get.

The S-N curve. Many steels flatten into a horizontal endurance limit — below it, in principle, the part never fatigues. Aluminium and copper have no such shelf; the line keeps falling, so a 'fatigue strength' is quoted at a fixed life such as 10^7 or 5x10^8 cycles.

Here is the single most useful feature. For many steels the S-N curve flattens out below some amplitude into a horizontal shelf: cycle below that level and the part appears to survive forever. That threshold is the endurance limit (or fatigue limit), and it is often roughly 0.35 to 0.5 of the ultimate tensile strength. So a steel with a UTS of 600 MPa might have an endurance limit near 250 MPa — keep the amplitude under 250 MPa and, in a clean lab, it can run past 10^8 cycles unharmed. This is why a well-designed steel crankshaft can spin for the life of an engine.

How a fatigue crack is born and grows

A fatigue failure has three acts, and understanding them ties this whole rung together. Act one is initiation. Even a smooth part is not perfectly smooth: the surface has scratches, machining marks, inclusions, or just the natural roughness where slip bands poke out. Cyclic loading works these into a tiny crack, almost always at the surface, and almost always at a stress concentration — a sharp fillet, a keyway, a bolt thread, a corner. A part with a mirror finish and generous rounded corners resists initiation; a part with a sharp inside corner is quietly manufacturing its own crack.

Act two is propagation. Once a crack exists, each load cycle opens its tip slightly, blunts it, and closes it again, advancing it by a minuscule step. This leaves the tell-tale fingerprint of fatigue: striations, a ladder of fine parallel lines on the fracture surface where, in the stable regime, each rung is roughly one load cycle. Zoom out and you see beach marks — broad clamshell ripples that mark stretches of faster or slower growth, or rest periods, each ripple containing thousands of striations. This is the crux of fractography we met in guide 1: an investigator can literally count cycles and trace the crack back to the exact point where it began.

Act three is final fracture — and this is where guide 3 comes crashing back in. The fatigue crack grows longer cycle by cycle, and remember that the stress intensity factor K at the tip rises as the crack lengthens. The moment the crack reaches its critical size — when K finally equals the material's fracture toughness K_IC — the slow story ends and the part snaps in one instant of fast fracture, ductile or brittle depending on the material. So a fatigue surface has two zones you can see with the naked eye: a smooth, beach-marked region where the crack crept, and a rough, dull final-overload region where it ran. The size of that smooth region tells you how long the part was cracked before anyone noticed.

Designing against fatigue

Because fatigue starts at the surface and at stress raisers, the whole art of fighting it is about the surface and the geometry — not about picking a stronger alloy and hoping. In fact a stronger, harder steel is often more fatigue-sensitive to a sharp notch, not less, because it has less ductility to blunt a starting crack. The moves below cost little and buy enormous life.

Kill the stress raisers. Replace sharp inside corners with generous fillets, round off keyways and holes, and smooth every abrupt change of section — this is the single biggest lever, because it directly lowers the local stress where a crack would start.
Polish the surface. A rough, machined finish is a field of tiny notches; a fine, polished surface removes the easiest initiation sites and can raise the endurance limit substantially.
Put the surface into compression. Shot peening (blasting the surface with tiny beads) or carburising leaves a residual compressive stress in the skin, a form of surface treatment. Since a crack must be pulled open in tension to grow, a compressed skin fights every cycle — this is the same trick that makes tempered glass strong, and it is worth a large fraction of the fatigue life on real shafts and springs.
Inspect for the crack you cannot prevent. Because a fatigue crack grows for most of the part's life, scheduled non-destructive testing (ultrasonic, eddy-current, dye-penetrant) can catch it while it is still small and retire the part before the final fracture — the backbone of aircraft maintenance.

One honest warning to design by: fatigue and corrosion gang up. In a corrosive environment the endurance-limit shelf can vanish, and a crack that would have stalled instead keeps growing as the freshly exposed metal at its tip is chemically attacked — corrosion fatigue. It is a cousin of stress-corrosion cracking but driven by cyclic load rather than a steady one. This is why marine and chemical-plant components are designed with far more conservatism than the clean-lab S-N curve alone would suggest.

The quiet majority — and what comes next

Fatigue earns its reputation as the number-one cause of mechanical failure because almost everything in engineering moves, spins, pressurises, or vibrates. Every rotation of a shaft, every landing of an aircraft, every pressurisation of a cabin, every wave that lifts a ship is one more cycle on the counter. The de Havilland Comet — the world's first jet airliner — was lost twice in 1954 when fatigue cracks grew from stress concentrations near its cabin openings under repeated pressurisation, and it rewrote how aircraft are designed and tested. The failures that make headlines are usually not a mystery material giving way; they are a known material, quietly counting cycles, at a corner someone did not round.

Step back and notice the shared through-line of this whole failure rung. Every mode — brittle fracture, fatigue, and the creep still to come — begins somewhere small and grows: a flaw, a notch, a surface scratch, one cycle too many. Fatigue is the purest example, a critical crack assembled patiently out of ordinary loads. Design is the discipline of respecting that small beginning: round the corners, polish the skin, compress the surface, keep the amplitude under the limit, inspect, and divide by an honest safety factor.

Fatigue is failure that needs many cycles but no heat. The final guide of this rung, creep, is its mirror image: failure that needs no cycles at all, just a steady load and enough heat and time. A turbine blade under constant pull at red heat slowly stretches and eventually ruptures — and there, surprisingly, the grain boundaries that strengthened our metals at room temperature turn traitor and become the weak paths. That twist is where guide 5 begins.