ASTM E466

What the test does

A machined metal specimen with a reduced central section is held at both ends in an axially aligned load train and cycled between a fixed maximum and a fixed minimum force. Loading is periodic and usually sinusoidal, and the amplitude does not change from one cycle to the next. Cycling continues until the specimen meets the failure criterion declared beforehand, normally separation, or until a predetermined number of cycles has passed without it. Nothing is measured on the specimen meanwhile: the machine holds the amplitude and counts.

What it measures, and why it matters

A single test yields one number — cycles endured at a stated stress amplitude and mean stress — and alone it means little. Value comes from a set run at several stress levels, from which the stress-life curve is drawn. A designer reads an allowable stress off that curve at a chosen design life, or the stress below which no failure appears within the cycle count the laboratory can afford. Comparing curves is how a change in melt practice, surface finish or heat treatment is shown to matter, and notched specimens run beside plain ones show how sensitive an alloy is to stress concentration. Because strains stay predominantly elastic, the practice describes long-life behaviour only.

Specimen

Three families are used: round bar with the test section blended tangentially into the ends, a continuous-radius or hourglass bar, and flat coupons for sheet and plate. The reduced section is proportioned from the specimen rather than fixed absolutely — roughly twice the test-section diameter for tangentially blended bars, more than three times the minimum diameter for continuous-radius bars — and the gripped ends carry considerably more area, so damage accumulates where it is being measured. Those proportions come from a superseded edition and should be checked against the current text.

Surface condition is part of the specimen, not an incidental of its manufacture: machining marks, grinding burn and preparation residual stress shift life further than most compositional differences. No conditioning period applies, and the number of specimens follows from the statistical plan for the curve rather than a count set here; statistical analysis of the results goes to ASTM E739.

What the machine must be capable of

No force capacity is prescribed; an accuracy demand is prescribed instead. The force amplitude must be held close to target at the test frequency, verified dynamically rather than by static calibration, because a load cell and control loop exact at rest can fall short at speed. Capacity follows the specimen — tens of kilonewtons at peak for a stout bar in a high-strength alloy, a few for thin sheet coupons — and the frame is chosen well above peak cyclic force, since amplitude control degrades near full scale. Alignment is the second hard requirement, verified with a strain-gauged specimen rather than assumed: the practice caps the bending strain the load train may impose.

The usual working band is about 0.01 to 100 Hz, across which fatigue strength is generally unaffected by rate for most metallic engineering materials, though localised yielding can heat the specimen and shift the result. No run-out is fixed; one million to ten million cycles is the common laboratory choice.

No extensometer is fitted. Round specimens take threaded-end or button-head axial adapters and sheet takes hydraulic wedge grips; either way the load train must transmit the cycle without backlash, since a joint that unloads and re-seats corrupts the waveform and frets the specimen where it is held. Testing is in laboratory air at room temperature, recorded and reported.

What goes wrong in practice

Failure inside the gripped length or at the fillet is the classic invalid result. It says the load train is misaligned, the grip section is undersized, or fretting has started a crack under the jaws; the life recorded belongs to the grip arrangement rather than the material, and the test is discarded.

Bending from an imperfect load train is the quiet version of the same fault. A little superimposed bending adds a mean stress on one side that nobody accounted for, and the whole curve sits low with no individual result looking wrong.

Self-heating catches laboratories that push frequency to finish a long run: results from a warm specimen are not comparable with cold ones at the same nominal stress. And scatter is routinely mistaken for a material difference — two or three specimens at one stress level cannot separate a real improvement from ordinary fatigue variability.

Related and equivalent standards

ISO 1099 is the nearest counterpart, covering force-controlled constant-amplitude axial fatigue of metals; intent is close, but specimen and reporting details differ.

The costly confusion is with ASTM E606, which controls strain, needs an extensometer on the specimen and addresses short lives in the plastic regime, whereas this practice controls force, fits no extensometer and stays elastic. Supporting documents carry the rest: dynamic force verification to ASTM E467, alignment to ASTM E1012, data presentation to ASTM E468, statistics to ASTM E739.