Vibration testing shakes things randomly
Why a swept tone is the wrong test, and what replaced it.
A modern vibration test does not step slowly through one frequency at a time the way an older test would, it delivers a broad, statistically random mix of many frequencies all at once, because that scrambled, simultaneous mix is a far closer match to how a real road, a real aircraft hold or a real factory floor actually shakes anything sitting on top of it.
What is actually happening
An older style of vibration test, still useful for certain narrow diagnostic questions, works by sweeping slowly through a range of frequencies one at a time, holding at each frequency briefly before moving on to the next, so that a part's response to a single, pure tone of vibration can be measured cleanly at every point along the sweep. This kind of test is genuinely good at one specific thing, finding exactly which frequency makes a given structure resonate most strongly, since sweeping past that frequency produces an unmistakable spike in the measured response that is easy to spot and easy to record, and a well-run sweep can pin that frequency down precisely enough to design a fix around it, whether that fix is a stiffening rib, an added mass, or simply a decision to keep the part away from wherever it will actually meet that exact frequency in service. What it does not do, no matter how carefully the sweep is run, is recreate what actually happens to a part riding in the back of a truck or strapped inside an aircraft's hold, because a real transport journey never delivers one clean frequency at a time, it delivers dozens of different frequencies simultaneously, layered on top of each other in a shifting, statistically random combination that never repeats itself exactly from one moment to the next. A test built from one frequency at a time, however thoroughly it covers the whole range in sequence, is structurally incapable of reproducing what a part actually experiences when several of those frequencies arrive together and interact.
The washing-machine-spin-cycle comparison
A washing machine's drum spinning fast with one heavy, bunched-up towel caught at a single point on its rim produces a vibration that settles into one clear, repeating rhythm, strong enough to walk the whole machine across the floor and steady enough that a person standing nearby can predict almost exactly when the next thump is coming. The same machine loaded normally with a full mixed wash, a dozen items of different weight and shape distributed unevenly around the drum, produces a shake that never settles into anything so clean, jittering unpredictably from one moment to the next because every item is contributing its own small, independently timed push to the drum's motion at once, none of them lining up neatly with any of the others. The single towel's steady thump is the swept sine test, a clean, isolated, predictable response to one clear cause. The mixed wash's jittery, unpredictable shake is the random vibration test, and it is the mixed wash, not the single towel, that resembles what a real transport journey actually does to whatever it is carrying.
Why a single resonance is not the only real risk
A structure rarely has just one frequency at which it resonates, most real parts and assemblies have several, sometimes many, each tied to a different mode of flexing or twisting built into the part's own shape and mass distribution. A swept sine test finds each of those resonances one at a time, in isolation, and reports how severely the part responds to each individually, which is genuinely useful information on its own. What it cannot show is what happens when several of those resonances are excited together at the same time, the way a genuinely random, broadband vibration input naturally does, since two or more modes responding simultaneously can combine in ways that are not simply the sum of testing each one separately, one mode's motion sometimes feeding energy into another rather than the two staying politely independent of each other. A part that passed a swept sine test comfortably at every individual frequency checked can still fail a random vibration test run across the same overall frequency range, purely because the random test is the only one of the two that ever actually asked several of the part's resonances to respond at once. This is the specific reason random vibration testing became the standard for anything destined for real transport or field service, once the equipment existed to generate a genuinely broadband, statistically controlled shake rather than only a clean, single-frequency one, and a swept sine test was demoted from the primary qualification method to a diagnostic tool used mainly to identify exactly which frequency a random test failure is actually coming from.
One figure worth keeping in mind
A structure whose several resonant frequencies sit close together can respond to genuinely random, simultaneous vibration far more severely, sometimes by several times over, than the simple sum of how it responds to each of those resonances tested one at a time by an older swept sine method, because random testing lets the separate responses combine and reinforce each other rather than measuring each in polite isolation.
What this does not explain
Passing a well-run random vibration test is strong evidence a part will survive the mechanical shaking of a real transport journey, but it is not, on its own, a complete substitute for the actual journey, since a laboratory shaker table cannot easily reproduce every other condition riding alongside real transport vibration at the same time, the temperature swings inside an unheated trailer, the humidity of a shipping container that has crossed an ocean, or the sustained static load of everything stacked on top of a part for the whole trip. A random vibration test answers the vibration question thoroughly and honestly, which is exactly the question the swept sine test it replaced could only ever answer partially, but it still leaves the combined effect of vibration arriving together with heat, moisture and static load as a separate question a shaker table alone cannot settle by itself.