Bolted joint loosening and locking

A bolt that is loose after six months in service was almost certainly tightened correctly. That is the uncomfortable part. Something took the clamp force away afterwards — and there are two quite different things it could have been, with two quite different cures.

A close view of a rusted bracket joint on a vehicle underbody with a nut backed off several turns and a witness mark of clean metal where the washer has migrated, alongside a second joint in the same bracket still tight, and a torque wrench resting against the chassis rail. — The witness ring shows where the washer used to sit. This nut did not lose its preload; it turned.

Two mechanisms, not one

A two-part diagram distinguishing preload loss from rotational self-loosening. On the left, a bolted stack is shown before and after, with surface asperities and a paint layer flattened under load, the stack shortened slightly and the bolt stretch reduced, labelled as embedment and creep with a note that nothing has rotated. Beside it a graph shows clamp force falling steeply in the first hours then levelling off. On the right, the self-loosening mechanism is drawn as a sequence of four frames, with the joint faces slipping transversely, the thread friction momentarily released, the helix angle driving a small backward rotation, and the nut returning with slightly less engagement. A graph beneath shows clamp force descending in a staircase against number of transverse cycles. An inset shows a thread as an inclined plane with the helix angle and the friction force marked. — Nothing turns in the left-hand case; the stack simply got shorter. On the right the nut really is unwinding, one slip cycle at a time.

	Preload loss	Self-loosening
Does anything rotate?	No	Yes
Cause	Embedment, creep, thermal relaxation	Transverse slip at the faces
Timescale	Mostly the first hours	Cycles, so miles
Cure	Joint design, longer bolt, fewer soft layers	Preload, then a locking device
Evidence	Low residual torque, marks unmoved	Witness marks moved, nut backed off

Preload loss is the joint getting shorter. Surface asperities flatten under enormous local pressure, paint layers squash, gaskets creep. The bolt has not turned; it has simply been allowed to relax. Most of it happens quickly, and it is worst where there are many soft or painted interfaces under a short bolt.

Self-loosening is the nut genuinely unwinding, and its driver is not what most people assume.

What Junker actually showed

A thread is an inclined plane held in place by friction. Gerhard Junker’s transverse vibration test established two things that changed fastener practice:

Transverse displacement at the joint face — not axial vibration — is what drives self-loosening. Sideways slip momentarily releases the friction holding the helix, and the helix angle does the rest.
Once the faces slip, a conventional split spring washer contributes essentially nothing.

The first cure is preload, not hardware

If the joint faces never slip, the loosening mechanism is never triggered at all. So the most effective single measure against self-loosening is a high, correctly achieved preload — which is a tightening control question, covered in threaded fasteners and torque.

A correctly preloaded joint with no locking device routinely outlasts an underpreloaded joint with one. Reaching for a lock washer before fixing the tightening is treating the symptom.

Locking devices, and which mechanism each addresses

Read the two right-hand columns together. Most devices address rotation only, and a device that has been flattened is not addressing anything.

Device	Stops rotation	Stops preload loss	Reusable
Prevailing torque nut (polymer insert, deformed thread)	Yes	No	Limited
Wedge locking washers (cam pair)	Yes	No	Yes
Thread locking adhesive	Yes	No	Controlled process
Serrated flange nut	Yes, into the face	No	Marks the surface
Split spring washer	No	No	—
Longer bolt, fewer soft layers	—	Yes	—

Prevailing torque nuts put friction on the thread itself, so torque is needed to turn them even at zero clamp load. That is the point: if preload is entirely lost, the fastener still does not fall out. It does not prevent the loss.

Wedge locking washers use a cam angle greater than the thread helix angle. A nut trying to rotate backwards would have to climb the cams and stretch the bolt further, which it cannot. This is why they survive the Junker test — the mechanism does not rely on residual spring force.

Thermal relaxation

A joint that runs hot loses clamp force for two reasons at once: differential expansion between the bolt and what it clamps changes the stretch, and at temperature both materials creep under sustained load.

Neither is a tightening error, so neither is fixed by tightening harder — and retightening hot then leaves it overloaded cold. The answer is in the design: compatible materials, and a longer, softer bolt that absorbs the movement with a smaller change in load. The same reasoning appears in pipe supports and expansion, where the movement is larger and the argument identical.

Diagnosing a loose joint

The two mechanisms leave different evidence, and it is worth looking before deciding:

Witness marks — a clean ring where the washer sat, offset from where it is now, means it rotated.
Residual breakaway torque with the marks still aligned means preload was lost without rotation.
Bright rubbed patches on the joint faces mean the faces were slipping, which points to insufficient preload whatever else was fitted.
Fretting debris at the interface says the same thing more loudly.

What to take away

Preload loss and self-loosening are different failures. Identify which one you have.
Embedment and creep shorten the stack; nothing rotates, and most of it happens early.
Transverse slip, not axial vibration, drives self-loosening.
The split spring washer does not work, and the Junker test showed it.
The best defence is a high, correct preload so the faces never slip.
Prevailing torque nuts resist rotation at zero clamp load — they do not restore clamp load.
Wedge washers work because the cam angle exceeds the thread helix angle.
No locking device fixes embedment. That is a joint design problem.
Thermal relaxation is design, not assembly — use a longer bolt and compatible materials.
Witness marks tell you which mechanism you are looking at.