Why Bolts Come Loose: Two Kinds of Loosening and How to Choose a Locking Method

1. The short answer: "the bolt came loose" is two different failures

When the assembly line reports that "the bolt came loose", the sentence that usually follows is "let's add a spring washer". But if nobody has established which kind of loosening it is, that step is quite likely money wasted.

Threaded fasteners loosen by two distinct mechanisms:

Both failure modes end with "it no longer clamps", but in one case the material is deforming and in the other the part is turning. Adding an elastic element to a rotational loosening problem, or applying threadlocker to a relaxation problem, are textbook examples of the wrong medicine.

The two mechanisms are taken apart below, and a shop-floor troubleshooting sequence is given at the end.

2. Why it does not loosen under static load: self-locking and the 0.8T criterion

An ordinary thread is inherently stable under static load, for two reasons:

• Self-locking of the thread pair: the lead angle ψ is smaller than the equivalent friction angle ρ′ of the thread pair, so the inclined plane cannot be pushed back.
• Friction at the bearing face: after tightening, friction still exists between the nut (or bolt head) and the clamped part or washer.

That gives a practical static criterion: under preload Q_p, the resisting torque that prevents the nut from rotating is approximately

T′ ≈ 0.8 T

where T is the tightening torque. The meaning is plain: for a joint tightened to torque T, as long as the external loosening torque applied at the nut or bolt head does not exceed 0.8T, the joint will not loosen.

But this criterion has a precondition — the friction coefficient is stable. Under fluctuating load, vibration and shock, the friction coefficient of the thread pair collapses, the friction resistance at the thread flanks and the bearing face momentarily disappears, and the formula immediately becomes invalid: the thread pair no longer satisfies the self-locking condition, microscopic relative sliding appears, and the nut begins to rotate. Repeat that often enough and the joint is loose.

So keep one sentence in mind: self-locking only holds under static load. Once vibration enters the picture, the textbook self-locking condition no longer governs.

3. The real culprit is transverse slip, not axial vibration

"The bolt was shaken loose" is only half right. What matters is the direction of the vibration.

Loosening driven by transverse slip happens in four steps, none of them difficult to follow:

• Step one: the friction is "diverted". Under a transverse force, all the friction at the thread flanks and at the nut (bolt head) bearing face is diverted to resisting the sliding.
• Step two: for that instant there is effectively no friction. The friction reserve is exhausted and the thread pair is, in practice, frictionless.
• Step three: the preload itself pushes the nut back. The stretched bolt stores energy, and with nothing left to resist it, it drives the nut back down the thread helix — the lead angle is small, but the force pushing it is large.
• Step four: accumulation. The nut rotates by a tiny angle, clamp force drops a little, and the next cycle repeats the whole sequence.

A single slip event typically produces less than 0.01° of rotation — completely invisible to the eye. Repeat it a few hundred times and the clamp force reaches zero.

This mechanism was set out systematically as early as 1969 and was later standardised as a transverse vibration test method: DIN 65151 (international counterpart ISO 16130; the rail vehicle world also has DIN 25201-4). It is still regarded as the most severe vibration exposure that can be applied to a threaded joint.

4. Why loosening is self-reinforcing

There is one critical characteristic here: loosening does not progress at a constant rate — the looser it gets, the faster it gets loose.

The chain runs like this. The threshold for partial slip depends on the available friction capacity μ·F_i (μ being the friction coefficient and F_i the current preload). When the transverse force exceeds that threshold, the contact zone switches from static to kinetic friction, the bolt rotates a small step, and the preload F_i drops a little. But as soon as F_i drops, the threshold μ·F_i for the next cycle is lower and slip is easier — every cycle lays the ground for the next one.

The result is the shape of the preload decay curve: flat first, then steep. For the first few dozen cycles almost nothing appears to change, then it steps downward, and finally it collapses over a very short interval.

Two very practical corollaries follow:

• "It's been fine for six months" is not evidence that no locking feature is needed. A joint that survives 500 cycles will often fail suddenly within the next 50.
• "Well lubricated means less likely to loosen" is wrong. Lubrication does lower the friction coefficient and lets you reach a higher preload at the same torque; but the lower the friction coefficient, the lower the threshold μ·F_i against slip, so the resistance to loosening gets worse. Lubrication helps tightening and hurts locking. As a rough comparison of common conditions: a phosphated and oiled steel thread pair has a total friction coefficient of roughly 0.10–0.12; dry, untreated, unlubricated steel is roughly 0.16–0.20.

5. Seven factors that accelerate loosening

6. Locking methods fall into three families

By principle, every method belongs to one of three classes:

Two design moves also deserve priority: use a fine thread (smaller lead angle, better self-locking), and combine two locking methods of different principles so that they complement each other.

7. Comparison table of common locking elements

One rule of discipline: all evaluations of locking performance are relative and comparative. Conclusions depend heavily on fastener size, material, coating, manufacturing and installation quality, and locking ability may differ in different environments or operating conditions. Any test result is a design reference only; before production use, verify on a vibration rig against the actual service condition.

8. First decide whether a locking feature is needed at all

Locking elements are not standard equipment. The judgement turns on the ratio of clamp length L to nominal diameter d, plus the load direction:

Why are short joints the most dangerous? Introduce one concept — the critical slip distance: the relative displacement of the joint faces needed for the bolt head or nut to lift off the bearing face and start being driven in rotation. In a joint with a short clamp length this value may be less than 0.1 mm; under the same load a bolt with a long clamp length may need more than 0.5 mm. The reason was already mentioned in section 5: a long bolt bends with the joint instead of sliding at the interface.

So what really determines the outcome is not only the element under the nut: hole clearance, clamp-length ratio, joint stiffness and preload all matter at least as much as the locking element itself.

9. How to compare locking performance: reading a transverse vibration test report

Since transverse vibration is the only severe condition, it is also the only way to compare locking solutions. The standard procedure has two stages:

• Reference test. First assemble a joint with no locking element at all, and increase the transverse displacement from zero until it self-loosens completely within 300 ± 100 cycles. That displacement is the "effective displacement" for the test and also calibrates how severe the test environment is.
• Verification test. Replace with the joint including the locking element, test at exactly the same effective displacement and frequency, and record the preload decay curve against cycle count.

The usual evaluation metric is the percentage of preload remaining after 2000 cycles, together with a curve shape that indicates failure is "not reasonably expected to occur" even if the test were continued.

A usable report must contain at least: locking type and its installed position, test frequency, clamp length to diameter ratio, clamping method, effective transverse displacement, lubrication condition used, and the applicable diameter range.

The single most important warning: DIN 65151 / ISO 16130 define a test procedure; the standard contains no pass line. A supplier saying "certified to such-and-such standard" therefore constitutes no meaningful conclusion. What you need is four things — preload decay curve, cycle count, transverse stroke, and friction/lubrication condition. Only with those four can you compare anything.

10. The other path: what to do about preload loss (relaxation)

Beyond rotational loosening, a share of failures involves no rotation at all — the clamp force is simply being eaten away. There are three sources:

• Embedding loss. Because of roughness, waviness and form error, the thread flanks, the various bearing faces and the interfaces between clamped parts undergo local plastic deformation during tightening; under repeated external load in service part of that deformation continues to develop and the preload falls with it. This is also why "a larger washer reduces contact pressure" and "retighten after a running-in period" are two classic countermeasures.
• Stress relaxation at high temperature. As temperature rises, the lower yield strength, the proof strength at prescribed non-proportional extension and the tensile strength all fall; sustained service at elevated temperature produces stress relaxation, and stress relaxation inevitably brings clamp-force loss with it. As a rule of thumb, once the fastener operating temperature exceeds 150°C its performance should be examined carefully.
• Differences in material sensitivity. This is often overlooked: cold-worked fasteners (property classes 4.8, 5.8, 6.8) are more sensitive to stress relaxation than quenched-and-tempered or stress-relieved fasteners. Using a class 4.8 bolt on a joint that runs continuously hot carries a risk beyond simply "lower strength".

Two other conditions deserve attention: where fasteners made from lead-bearing steel are used at high temperature and the service temperature falls in the melting range of lead, liquid metal embrittlement (LME) should be considered; for material selection in high-temperature fasteners, standards such as EN 10269 and ASTM F2281 are useful references.

The countermeasure list for relaxation:

• Increase the bearing area: switch to flanged screws or nuts, or add a high-strength large washer, to bring down both contact pressure and indentation directly.
• Increase joint elasticity: use long bolts or reduced-shank bolts (elastic length k₀ > 4d) so the joint does not readily shed preload as the working load fluctuates. This is also why "preserve elasticity" and "resist transverse motion" are often applied together in engineering practice.
• Remove soft intermediate layers: thin paint films, plastic pads and soft washers all bring continuous creep loss — unless the layer is genuinely designed as an elastic element (a conical spring washer designed for a specific preload point, for example).
• Retighten where necessary: re-torque once after the joint has been in service for a while, to eliminate the initial embedding loss.

11. Shop-floor troubleshooting sequence

Given a "it came loose" case, work through the following order rather than reaching straight for a different locking part:

• Step one, look at the traces. Is there any trace of relative rotation between the nut (or bolt head) and the bearing face? Has the thread pair visibly backed out?
  Yes → rotational loosening. No, with only an annular indentation at the bearing face → most likely relaxation loosening.
• Step two, if it is rotational loosening. First confirm whether the load contains a transverse reciprocating component → check whether L/d falls in the danger zone → verify that the preload actually reached its design value → then choose the locking element (preferring one whose principle is complementary to what is already there) → verify by transverse vibration test.
• Step three, if it is relaxation loosening. First check service temperature and the condition of the contact faces → increase bearing area, increase bolt flexibility, remove soft intermediate layers → assess whether "retighten" needs to be written into the maintenance procedure.
• Step four, check the same thing in both cases. Is the preload really there? With torque-controlled tightening the preload scatter can reach ±25%, and if the actual preload is only half the design value, then "the locking was inadequate" and "it was never properly tightened" are two very different stories. See How Much Tightening Torque: Three Conventions and One Look-Up Table for the detail.

One sentence to finish with: preload is the first line of defence, resistance to transverse slip is the second, and mechanical positive locking is the last. Skipping the first two and going straight to the third usually only postpones the failure rather than eliminating it.

Sources

• Fastener Joint Design Handbook, chapter 3 "Bolted joints": loosening mechanisms, classification of locking methods, worked examples of common locking methods for bolted joints, and the uneven axial load distribution across engaged threads.
• DIN 65151, Locking characteristics of fasteners — dynamic test under transverse loading; ISO 16130 (international counterpart); DIN 25201-4. Including the two-stage reference-test/verification-test procedure and the 2000-cycle residual preload metric.
• Mechanism of self-loosening under transverse vibration: local slip theory, and the conclusion that friction resistance momentarily vanishes once the thread faces enter full sliding (first set out systematically in SAE technical paper 690055, 1969).
• VDI 2230: friction coefficient classes, the clamp-length to nominal-diameter criterion, and bearing-face pressure checks.
• Corporate technical manual: stress relaxation and high-temperature suitability, the sensitivity of cold-worked classes to stress relaxation, liquid metal embrittlement risk with lead-bearing steel; EN 10269, ASTM F2281.
• Preload scatter and the error of torque-controlled tightening: see the article "How Much Tightening Torque" on this site.