Key takeaways
Sine sweep tells you where the structure is weak; random vibration tells you how much margin is left. Choosing the wrong one does not produce a conservative result - it produces an invalid one. When both are run, the sweep must come before random and dwell, otherwise what you read is the structure after degradation.
For a given part, sine and random are not a looser and a stricter version of the same test - they are two different examinations. Choosing the wrong one does not make the conclusion conservative, it makes it invalid: using a random spectrum to demonstrate that a bracket's resonance is acceptable, or a sine dwell to demonstrate solder-joint endurance, means measuring with the wrong ruler. Three questions settle it: is the excitation narrowband or broadband, is the subject resonance dwell or accumulated damage, and is the criterion "find the resonance" or "still working afterwards".
The most common enquiry is "which is stricter, sine or random, and can we run only one". The question is aimed at the wrong target, so no answer exists. Set out what each path answers, and the selection follows. GB/T 28046 and ISO 16750 present both methods side by side in their mechanical load sections; neither implies that one replaces the other.
1. Put the idea of "looser versus stricter" aside
Sine is single-frequency excitation: frequency controlled, magnitude controlled, repeatable. What it yields is structural information - which frequency amplifies the response, which location cracks first.
Random is broadband excitation with energy distributed by a spectrum. What it yields is cumulative information - how much margin is left at the connecting interfaces after running in a broadband environment.
One answers "where is it weak", the other answers "how much longer will it last". The first is diagnostic, the second is a qualification. There is no stricter one; there is only the one that matches your question.
How the three stress types - sine, random and shock - divide the failure modes between them has already been set out in mechanical load conditions. This article does not repeat it; it covers how to choose, how to sequence and how to read.
2. Three questions that fix the direction
First: is the excitation source narrowband or broadband? Engine orders, road input and vehicle-level road spectra are broadband. A single-speed rotating component, the fixed fundamental of a pump or fan, or one known disturbance source is narrowband.
Second: are you answering "where is the resonance" or "how long will it last"? The first needs a sweep, the second needs time at a spectrum.
Third: at which layer does the criterion sit? Structural integrity, functional performance, or parameter drift still within an allowed band - each implies a different run and different monitoring.
| Object and concern | Preferred method | Reason | If chosen wrongly |
|---|---|---|---|
| --- | --- | --- | --- |
| Brackets, displays, long cantilevers; resonance frequency and amplification | Sine sweep with resonance dwell | Point-by-point single frequency exposes the peak and its amplification | Random only: the peak is flattened by spectral energy and the resonance is invisible |
| Solder joints, pins, BGA; accumulated damage | Random vibration | Broadband energy at once, close to the real transfer spectrum | Sine only: energy sits at a few frequencies, accumulated damage is never exercised |
| Parts mounted near a single-speed rotating component | Fixed-frequency dwell as the main step | The excitation itself is narrowband | Running a broadband spectrum dilutes the energy and under-tests the part |
| Relays, electrolytic capacitors, heat sinks - heavy and overhanging | Mechanical shock, alongside both vibration types | Inertial force dominates, not fatigue | Substituting vibration never delivers the inertial load |
| General parts with no known excitation source | Low-level sweep first, then the agreed spectrum | Characterise the structure before fixing the qualification spectrum | Running the spectrum directly leaves you without knowing where the weak point is |
The "preferred" column is a directional judgement derived from excitation source and object of concern, not a rule. Magnitudes, spectra, sweep ranges and dwell times all follow the applicable standard revision and the customer technical agreement; no values are given here.
3. When both are run, the order cannot be reversed
The usual combination is a low-level sine sweep to locate resonance, a fixed-frequency dwell at that resonance, then random per the agreed spectrum. Each step produces something the others do not, so none can replace another.
There is a hard reason for the order: the sweep is a medical check, dwell and random are the loaded run. The check must come before the loaded run, otherwise what you read is the body after the run.
Run random first and sweep afterwards, and the resonance frequency you read may already have shifted lower - fasteners loosened, interfaces worn, material softened. You then cannot tell whether that frequency is the original state or the degraded one, and the report cannot attribute anything.
The reverse trap is just as real: "low level" in a low-level sweep must be low enough to cause no damage. Set it too high and the first sweep is itself a qualification, so every later conclusion rests on a sample that has already been disturbed.
German OEM projects commonly reference the mechanical load clauses of the LV 124 system, whose magnitudes and spectra differ from the GB/T series. Mixing the two systems in one outline means the sequence cannot be reproduced even when it is correct. This fails more often than the choice of method does.
4. Five items a report must show
| Item | If missing |
|---|---|
| --- | --- |
| Control point position and count | A control point on the fixture versus on the sample can differ by an order of magnitude; without the position the data cannot be reproduced |
| Sweep rate, or spectrum shape and overall RMS | "Per the standard" alone cannot be reproduced after a revision or in another laboratory |
| Dwell time and dwell criterion | Without both, the record only says the part was shaken |
| Axes and mounting attitude | Whether all three axes were run and whether the vehicle attitude was reproduced sets the validity range of the conclusion |
| Online monitoring items and sampling | Intermittent opens and contact-resistance jumps during the test are invisible in a static re-check |
The fifth is the one most often dropped. The sample is unpowered and unmonitored, it powers up afterwards, and the report says "functionally normal" - while the part may have opened dozens of times during the run. Such conclusions are overturned at vehicle road-test stage. How functional status is graded is covered in functional status A to E; the monitoring items must be fixed at outline stage.
Fixing and observation for flexible parts such as harnesses is a separate topic in harness vibration testing. Before commissioning, the vibration test capability page is worth checking for table size, force and axes against your sample.
5. How to read a sweep curve
A resonance shifting lower usually means stiffness degradation. Loosened fasteners, fretting wear at interfaces, cracks and heat-softened material all push frequency down. It is the most informative single change - provided the control points and mounting were identical both times.
A lower peak does not mean improvement. It may be higher damping, or a sensor that is no longer properly bonded, or a slack clamp. When the peak drops, check the measurement point before drawing a conclusion.
Amplification sets how destructive the dwell is. At the same input level, a location with high amplification accumulates fatigue far faster, so dwell time should be set by the weakest point, not by an average.
Beyond the curve, watch for non-linearity during the sweep: a peak that no longer scales linearly with level, or a frequency that drifts with level. If it appears, the structure has entered a non-linear region and any spectrum conversion based on linear assumptions no longer holds.
6. A few plain words
On selection, asking "which is stricter" never produces an answer; asking "what is my excitation source and what am I judging" does.
When TMC LAB takes on a vibration commission, the nature of the excitation, the mounting location and the acceptance items are confirmed first, then the combination of sweep, dwell, random and shock is decided and written into the test outline for customer confirmation before any slot is booked. Magnitudes and spectra follow the standard text and the technical agreement; no values are preset at outline stage.
If the vehicle mounting attitude, tightening torque and harness routing can be sent with the samples, locating resonance and fixing the dwell frequency becomes far more straightforward. Scheduling is arranged per project; you can state your expected delivery date when submitting samples, and we will confirm a feasible plan against the current schedule and report progress at each milestone.
