Why Nuts and Bolts Self-Loosen — and What Actually Stops It
Every engineer has seen it: a bolt that was torqued to spec yesterday is finger-loose today, after a machine ran through one shift. The instinct is to blame the torque wrench — but the wrench was never the problem. Preloaded fasteners self-loosen because of a physical mechanism first documented by Gerhard Junker in 1969, and most of the "anti-loosening" devices on the market either do not stop it or make it worse.
The short answer: transverse (side-to-side) vibration is the dominant cause of self-loosening — not axial vibration, not poor torque, not a cheap nut. Once transverse slip occurs between the nut and the bolt thread, friction breaks down and the bolt unwinds in a matter of cycles.
What Junker found in 1969
Junker's landmark experiment used a test rig that applies controlled transverse displacement to a preloaded bolt while continuously measuring the remaining preload. The result was unambiguous: under transverse vibration of just ±0.5 mm, a plain hex bolt with a standard nut lost 90% of its preload within 500 cycles. Under axial vibration — the kind most engineers assume causes loosening — the same bolt barely moved.
This flipped the industry's understanding. Loosening is not caused by the bolt "bouncing" out of the hole. It is caused by the clamped joint slipping sideways, which generates relative motion between the nut face and the thread flanks. Once that slip exceeds the friction threshold, the nut unwinds like a miniature ratchet — one fraction of a turn per cycle, until the joint is effectively unclamped.
Preload decay under transverse vibration (typical Junker curves)
| Fastener configuration | Cycles to 50% preload loss | Final preload after 1000 cycles | Verdict |
|---|---|---|---|
| Plain hex bolt + standard nut | ~100 | ~5% (nearly zero) | Fails rapidly |
| Hex bolt + spring lock washer | ~80 | ~3% | Worse than plain |
| Hex bolt + nylon insert nut (NYLOC) | ~500 | ~40% | Delays, does not stop |
| Hex bolt + prevailing torque all-metal nut | ~800 | ~55% | Better, still loses |
| Hex bolt + pre-applied thread locker (medium strength) | >10,000 (no loss) | >95% | True lock |
| Torx Plus bolt + prevailing torque nut + thread locker | >10,000 (no loss) | >98% | Best-in-class |
The three mistakes that make loosening worse
1. Spring lock washers accelerate loosening. This is the counter-intuitive finding from Junker's data. A split spring washer introduces a sharp point contact under the bolt head, which actually reduces the bearing friction — the very friction that resists unwinding. In Junker tests, bolts with spring washers lose preload faster than bolts with no washer at all. The spring washer should be retired from vibrating assemblies.
2. Nylon insert nuts are not "lock nuts." A NYLOC-style nut adds prevailing torque (the resistance you feel when spinning it on by hand), but it does not prevent transverse slip. It delays the onset of loosening — the preload decays over hundreds of cycles instead of tens — but it does not hold. After a thousand cycles, 40–60% of the preload is gone. Nylon inserts also creep at temperatures above 120°C, which eliminates them from engine and exhaust applications.
3. Relaxation is a separate failure mode from self-loosening. Even without vibration, preload drops in the first 24 hours by 10–20% because of embedding: microscopic surface roughness on the threads and clamped faces flattens under load, and the bolt stretches slightly less. This is called relaxation, not loosening. It is normal — and it means you should either re-torque after the first run-in, or over-torque by 10–15% to compensate.
What actually stops self-loosening
The only methods that survive Junker transverse vibration testing at high preload are:
Thread locking compounds (pre-applied or applied at assembly): A medium-strength anaerobic adhesive (e.g. Loctite 243 class) cures between the thread flanks and locks the nut against rotation independent of friction. It is the only method that holds >95% preload after 10,000 cycles. Toprecision supplies screws pre-coated with nylon, polyester or anaerobic adhesive in standard colors (blue, green, red) for automated assembly.
Direct tension indicators / torque-plus-angle: Not a locking device, but a tightening method that removes the friction variable. Torque-only tightening scatters ±25% on dry joints; torque-plus-angle delivers consistent preload regardless of surface finish, so the joint starts with enough clamp force to resist slip.
Bearing-surface friction control: Increasing friction under the bolt head and nut face (e.g. serrated flange bolts, wedge-locking washers) raises the torque required to unwind. This is a secondary defense — it buys cycles but does not replace thread locker for high-vibration assemblies.
Toprecision manufactures pre-coated thread-locking screws (nylon patch, polyester patch, full-bond anaerobic coating) in M2–M16 sizes, plus prevailing torque nuts and serrated flange bolts for NEV chassis, railway and heavy-equipment applications. Send the vibration profile and clamp-load requirement — we will recommend the locking method, not just the catalog option. Request a locking-solution consult with Junker test data and preload decay curves.
