why bolts keep loosening
Why bolts keep loosening, and how to diagnose which mechanism you have
By ISOKLAMP Engineering, Inc. Editorial Team · Updated
Start with the evidence, not the catalogue
The usual sequence when a joint keeps failing is: fit a locking washer, then a better locking washer, then thread locker, then a bigger bolt. Sometimes it works. When it does not, it is because the fix and the failure mode were never matched, and no amount of escalation within the wrong family will help.
Ten minutes of diagnosis first will save the escalation.
Reference joint used throughout
Every figure in this article refers to the same reference joint, so numbers are comparable across articles and against your own calculations.
| Parameter | Value |
|---|---|
| Bolt | M16 × 2,0, property class 10.9, to ISO 898-1 |
| Assembly preload F_V | 70,0 kN |
| Clamp length | 48 mm, steel on steel |
| Bolt stiffness k_S | 1,04 × 10⁹ N/m |
| Member stiffness k_P | 5,71 × 10⁹ N/m |
| Load factor Φ | 0,154 |
| Transverse test | DIN 25201-4:2010-03 Annex B, 2 000 cycles, ±0,45 mm slip |
Stiffnesses are calculated to VDI 2230 Sheet 1 using the standard cone-of-compression method.
The witness mark test
Do this before buying anything.
- Bring the joint to correct assembly preload using your normal procedure.
- Paint or lacquer a single continuous line across the nut flat, the washer edge, and the member surface. Torque seal or a paint pen both work.
- Photograph it.
- Run the equipment for one normal duty cycle, or until the joint would normally have failed.
- Photograph it again and compare.
| Evidence | Mechanism | Family |
|---|---|---|
| Mark misaligned, nut has rotated | Rotational self-loosening | Mode A |
| Mark intact, joint measurably loose | Non-rotational preload loss | Mode B |
| Mark intact, joint still tight, but fails later | Slow Mode B, extend the test | Mode B |
| Mark misaligned and joint was loose beforehand | Mode B followed by Mode A | Both |
| Mark intact, joint tight, no problem found | Look elsewhere: fatigue, member, or load path | Neither |
Row four is the most common result in practice and the reason so many joints get mis-specified. Preload falls first through stack shortening. Below a threshold, the friction resisting transverse slip is no longer sufficient, the joint begins to slip, and only then does the nut rotate. The engineer arrives, sees a rotated nut, and correctly diagnoses the last event in the chain rather than the first.
The six mechanisms
Mode A: rotational
| Mechanism | Signature | Fix |
|---|---|---|
| Transverse micro-slip (Junker) | Nut rotates, mark broken, cyclic transverse load present | Wedge-lock washer, ribbed flange nut |
| Insufficient assembly preload | Rotates early, often within hours | Correct the tightening procedure first |
Mode B: non-rotational
| Mechanism | Signature | Fix |
|---|---|---|
| Embedment | Loss in first 200 cycles then stable, mark intact | Better surface finish, hardened washers, take-up |
| Coating creep | Progressive, worse hot, coated interfaces in load path | Move coating out of load path, take-up |
| Gasket or polymer relaxation | Continuous, gasket in stack, often leaks before it loosens | Take-up |
| Thermal ratcheting | Accumulates per cycle, dissimilar metals, never stabilises | Increase bearing area, or take-up |
Reading the failure interval
The time to failure is itself diagnostic.
| Interval | Likely mechanism |
|---|---|
| Hours | Assembly preload was wrong, or gross transverse overload |
| Days to a few weeks | Embedment and resin-rich layer flattening |
| One to six months | Coating creep, gasket relaxation |
| Six months to years, accelerating | Thermal ratcheting |
| Random, no pattern | Check for a cracked member or a missing spacer before anything else |
The last row matters. A joint that eats preload with no consistent interval frequently has a mechanical fault behind it, and fitting better fasteners hides it until something more expensive breaks.
Why the standard fix so often half-works
Locking devices address Mode A. They address it well: analysis of the reference joint under DIN 25201-4 Annex B gives 93,1 percent residual for a wedge-lock pair against 12,8 percent for a plain washer.
But a wedge-lock washer applied to a Mode B joint arrests a rotation that was going to happen anyway as a *consequence* of the preload loss. The joint stops coming apart, which looks like success, while clamp force continues to fall. What you get is a joint that no longer loosens visibly and still leaks, fretts, or fatigues.
The diagnostic shortcut: if a locking device improved matters without solving them, you have a Mode B joint. That single observation resolves most difficult cases.
Matching fix to mechanism
| Your evidence | Specify |
|---|---|
| Mark broken, no gaskets, steel on steel, constant temperature | Wedge-lock washer pair. Cheaper and correct |
| Mark broken, but a locking device already half-worked | Take-up plus rotation control |
| Mark intact, gasket in the stack | Take-up. See Hygienic equipment bolting |
| Mark intact, dissimilar metals, thermal cycles | Take-up sized for cycle count. See Busbar and battery joints |
| Mark intact, coated fasteners, elevated temperature | Move coating out of the load path, then take-up |
| Mark intact, composite member | Take-up with non-penetrating bearing faces. See No surface damage |
| No clear evidence | Instrument one joint with ultrasonic measurement before spending further |
When to stop diagnosing and measure
If the witness mark test is ambiguous after two duty cycles, instrument a representative joint. Ultrasonic bolt elongation measurement gives absolute preload with no disassembly, and a before-and-after pair separates the mechanisms conclusively. It costs one instrument hire and resolves arguments that otherwise run for months.
For a joint fitted with a reserve indicator, the equivalent information is available by eye at any time, because reserve consumption rate is a direct read on the stack shortening rate at that location. See Preload reserve indicator.
Mechanism detail on Non-rotational preload loss. Method comparison on Locking methods compared.
Frequently asked questions
How do I find out why my bolts keep loosening?
Paint a continuous witness line across the nut, washer and member at correct assembly preload, photograph it, run one normal duty cycle, and compare. A broken mark means the nut rotated and the mechanism is rotational self-loosening. An intact mark on a loose joint means the clamped stack shortened and the mechanism is non-rotational.
My locking washer helped but the joint still fails. Why?
Because the dominant mechanism is non-rotational. A locking washer arrests rotation, which in a joint losing clamp length is a consequence rather than the cause. The joint stops visibly coming apart, which looks like success, while clamp force keeps falling, so it continues to leak, fret or fatigue. Partial improvement from a locking device is the clearest single indicator of a non-rotational problem.
What does the time to failure tell me?
Hours suggests incorrect assembly preload or gross transverse overload. Days to weeks suggests embedment. One to six months suggests coating creep or gasket relaxation. Six months to years with acceleration suggests thermal ratcheting. No consistent interval frequently means a mechanical fault such as a cracked member or missing spacer, which should be ruled out before changing fasteners.
When should I use a wedge-locking washer?
When the witness mark is broken, the joint is steel on steel with no gaskets or coatings in the load path, and temperature is roughly constant. In that case rotational self-loosening is the dominant mechanism, a wedge-lock pair addresses it directly, and it costs less than any recovery device.
How do I confirm the diagnosis conclusively?
Instrument a representative joint with ultrasonic bolt elongation measurement, which gives absolute preload without disassembly. A before-and-after pair separates rotational from non-rotational loss definitively. On a joint fitted with a reserve indicator the same information is available by eye, since reserve consumption rate is a direct read on the stack shortening rate at that location.
Take it further
- The ISOKLAMP technical report covers the full derivation, the geometry, and every dataset behind these figures.
- Full residual clamp force dataset gives the residual clamp force numbers for ten securing methods, with sources.
- Design partner programme is open. Eight slots, two per sector. Bring us a joint that keeps failing and we will run a VDI 2230 Sheet 1 analysis on it.
Engineering questions go to engineering@isoklamp.com. An engineer answers, not a form.