Anti-Loosening for Threaded Fasteners: Choosing Between 10 Methods

1. Why this article: two real accidents

In June 2010 the Space Loos attraction at OCT East in Shenzhen collapsed, killing six people and injuring ten. The investigation found the immediate cause was fatigue fracture of a connecting bolt, and the fracture started where the threaded joint had loosened under cyclic load, concentrating stress at the thread root.

In 2007 a high-speed train derailed in the United Kingdom. Gong et al. cite the case in their review: cyclic load at the wheel interface worked a nut off its bolt, and that was one of the direct triggers of the derailment.

Both joints used ordinary bolts with spring washers. "Spring washer equals anti-loosening" is almost muscle memory in engineering offices, but since Junker published his transverse-vibration theory in 1969, five decades of testing have shown that a spring washer holds preload far worse than most people assume.

This article is based on Review on anti-loosening methods for threaded fasteners (Gong H., Liu J. H., Feng H. H., Chinese Journal of Aeronautics, 2021), together with NASA-STD-5020, ISO 16047, GB/T 3098 and GB/T 10431. It gives the real performance of ten mainstream methods in a form an engineer can check against a drawing.

2. Why threads loosen: the mechanism

Many engineers assume loosening means "the nut turned by itself". In reality it is a staged process.

2.1 Preload is the lifeblood of a joint

When a bolt is tightened the joint is compressed and the bolt is stretched. That tension is the preload F_v. As long as F_v holds, the joint cannot slide and the thread pair cannot rotate. Loosening is not the nut turning: it is preload falling below a critical value, after which the thread pair starts to rotate.

Preload is lost along two paths:

• Elastic and embedding relaxation: micro-asperities on the joint faces flatten and the contact creeps, so F_v drops 5–20 % within minutes to days.
• Rotational loosening under transverse vibration: Junker's 1969 tests showed that when the clamped plates move perpendicular to the bolt axis, friction at the thread and bearing faces momentarily disappears and the bolt unwinds itself on the helix angle. This is the most dangerous and most common mode.

2.2 The Junker test: one exam hall for every locking method

Locking performance today is judged with the Junker transverse vibration test (DIN 65151 / ISO 16130 / GB/T 10431). An eccentric cam drives two plates into relative transverse motion; the bolt passes through both, and a load cell records F_v against time.

The metric is preload retention: residual F_v divided by initial F_v after 2000 cycles. Higher is better.

2.3 Five variables that set the loosening rate

1. Initial preload F_v: higher is safer, but never beyond yield.
2. Friction coefficient μ at thread and bearing face: higher μ resists rotation, but scatter in μ makes preload unpredictable.
3. Transverse amplitude: more relative slip between plates means faster loosening.
4. Clamp length to diameter ratio L/d: a slender bolt absorbs vibration. Below L/d = 3 the joint will almost certainly loosen.
5. Joint stiffness: a stiff thick plate transmits transverse slip more readily than a thin sheet or a casting.

3. Seven conventional locking methods

3.1 Plain washer

A plain washer spreads the bearing load, protects the joint face from the rotating nut and bridges an oversized hole. It provides no locking at all. Pinelli et al. (2012) found preload retention essentially identical with and without one.

3.2 Spring washer

This is the most over-rated part in the catalogue. The idea is that the split edge bites into the bearing face and stores elastic tension. The test data disagrees:

• Dravid et al. (2012): bolts with spring washers lost preload faster than bolts with no washer, because the split edge concentrates stress and accelerates local slip.
• Pinelli et al., over 24 months of vehicle service: not one spring-washer joint retained useful preload.
• The split edge scores the joint face under repeated vibration and becomes a corrosion initiation site.

3.3 Belleville (disc spring) washer

A conical disc washer stores elastic deflection and its real job is compensating relaxation: as the joint embeds, the disc opens and keeps clamping force. Sawa's FEA work shows a clear benefit on joints with large thermal mismatch (aluminium to steel, bimetallic stacks). It does not stop rotational loosening under transverse vibration, so it suits thermal cycling (exhaust manifolds, heat exchangers) rather than continuous transverse excitation (bogies, screen decks).

3.4 Wedge-locking washer (Nord-Lock type)

The most widely adopted solution of the last 20 years in wind, rail and mining. A pair of washers: radial teeth on the outer faces, wedge ramps on the inner faces, with the wedge angle α greater than the thread helix angle β.

Any attempt to unwind lifts the bolt over the ramp. Because α > β, the axial force needed to climb the ramp exceeds the axial force released by unwinding, so the harder it tries to loosen, the tighter it gets.

Misiek (2017) and Lai's FEA both confirm preload retention above 80 % in a standard Junker test. Drawbacks: high cost (a pair costs 20–50× a spring washer), flatness sensitivity on large sizes, and unsuitability where hole clearance is excessive.

3.5 Serrated washer

Internal or external teeth bite into the bearing face for mechanical lock. Effective on soft materials (aluminium, plastics, thin sheet) where the teeth can embed; on hardened or high-strength steel the teeth wear flat and locking collapses. It also damages the joint face badly, so avoid it on painted or anodised parts.

3.6 Double nut

Two nuts pressed against each other generate extra normal force in the thread pair. Nine out of ten double-nut joints on the shop floor are assembled in the wrong order.

The correct sequence (the NASA-STD-5020 practice):

1. Torque the lower (power) nut and hold it with a wrench;
2. Run the upper (slave) nut down against it until the faces seat;
3. Tighten both against each other until the axial force between them reaches 30–50 % of yield.

Zhang et al. (2019) measured more than three times the preload retention for the correct sequence. With the wrong sequence a double nut performs like a single nut.

3.7 Nylon-insert and all-metal prevailing-torque nuts

A nylon collar at the top of the nut is expanded by the bolt and adds friction. Simple, removable, cheap — the default locking nut in automotive and electronics.

Its fatal weakness is temperature. Above 120 °C the nylon creeps and locking torque decays by more than half within a few hundred hours. No nylon nuts on exhaust manifolds, brakes or furnace parts.

All-metal prevailing-torque nuts (Spiralock-type triangular thread, ASTM A194 Grade C) deform the top of the nut into an ellipse, slot or crimp. They work to 250 °C but should not be reused more than 5–10 times.

4. Three newer locking methods

All three are gaining share quickly in aerospace, space and robotics.

4.1 Deformed-thread locking nut

What it looks like: the last two or three threads of the nut are changed from a 60° V to a curved root (about 30° arc). Externally it is indistinguishable from a hex nut, but it feels tighter to drive because the bolt thread is wedged rather than slid.

Why it locks: in an ordinary nut the radial clearance at the thread root lets the bolt oscillate under transverse load and walk itself loose. The deformed thread moves contact from root to crest and removes that clearance structurally. Junker retention is above 90 %, it can be reused 15 times or more, and it is the fastest-growing choice for aircraft and robot joints.

4.2 SMA shape-memory nut

Made from nickel-titanium alloy, heated above its transformation temperature, expanded over the bolt and allowed to contract. Preload is 30 % higher than a conventional nut, but cost is extreme and use is limited to critical aero-engine locations.

4.3 Threadlocker (chemical locking)

Not a mechanical device but chemical locking. Loctite 243 (medium strength) and 263 (high strength) cure anaerobically and bond the thread pair into one body.

Cheatham and Dattilio both found the highest preload retention of any method, with no damage to the joint face. Limitations: the surface must be clean (oil can cut bond strength by 70 %), 243 can be broken with hand tools at room temperature while 263 is permanent and needs 150 °C or drilling, and neither suits continuous service above 150 °C.

5. Three comparison tables

Table 1 — ten locking methods ranked by Junker performance

Source: Gong et al. 2021, Table 2. Standard Junker test (DIN 65151), 2000 transverse cycles.

Table 2 — recommended methods by application

Table 3 — cost versus performance

Relative to "spring washer + plain nut" = 1×. Multipliers rather than prices, so you can apply your own purchasing figures.

6. A seven-step selection framework

Run any drawing through these seven steps and the choice usually makes itself.

Step 1 — Is there transverse vibration?

Rock the joint by hand. If relative movement is visible (> 0.1 mm), or the application is rotating machinery, rail vehicles, road vehicles or wind turbines, eliminate spring and plain washers immediately and continue.

Step 2 — What is the temperature?

• ≤ 120 °C: nylon nuts are acceptable.
• 120–250 °C: all-metal nut, wedge washer or high-temperature threadlocker.
• > 250 °C: Belleville washer plus high-temperature threadlocker, or an Inconel all-metal nut.

Step 3 — Will it be dismantled?

• Frequent access (covers, service panels): wedge washer, all-metal nut, double nut.
• Never (structural): high-strength threadlocker.
• Occasional (annual service): medium-strength 243.

Step 4 — What is the joint material?

• Aluminium, plastics, thin sheet: serrated washer or nylon nut; avoid plain washers that dish.
• Hardened or high-strength steel: wedge washer or threadlocker; serrations wear flat.
• Cast iron: plain washer plus prevailing-torque nut, because the porous surface needs load spreading.

Step 5 — Check L/d

Below 3 the bolt is too stiff for any mechanical device to save the joint. Add a sleeve to lengthen the clamp, use a longer bolt, or add threadlocker.

Step 6 — Budget and volume

Use Table 3: at 1–2× take threadlocker or a plain nut; at 4–12× take nylon or all-metal prevailing-torque nuts; at 30–100× take wedge washers or deformed-thread nuts; SMA nuts only for extreme aero-engine cases.

Step 7 — Prototype and verify

Run 2000 cycles to GB/T 10431 or ISO 16130 and release only above 60 % preload retention. Do not release on a supplier's catalogue curve — one in-house test is the cheapest insurance you can buy.

7. Six mistakes engineers make most often

8. FAQ: ten real engineering questions

Q1. Can spring washers be eliminated everywhere?

On static, non-vibrating, non-safety-critical joints (internal trim panels, lightly loaded covers) they are still acceptable. Anywhere with dynamic load, personnel risk or expensive downtime, replace them with wedge washers or threadlocker.

Q2. Is a thicker wedge washer better?

No. The wedge angle must be strictly greater than the helix angle and the pair must be used together — a single washer does nothing. Above M16 choose the cable-marked version for easier inspection.

Q3. How many times can a nylon nut be reused?

Industry practice is 5–10 assemblies. Each one expands the insert and reduces locking torque. Aerospace practice often scraps them after three.

Q4. 222, 243 or 263?

222 low strength, below M6, adjustable; 243 medium, removable, general use below M20; 263 high strength, permanent above M20. Always pick the lowest grade that holds — the higher grade may not come apart.

Q5. What locks into aluminium?

Aluminium threads are weak and threadlocker can pull the threads out on removal. Use a serrated washer that embeds in the soft metal, or install a wire thread insert first and then a nylon nut.

Q6. Can stainless bolts use spring washers?

No. Stainless on stainless galls: it seizes on assembly and shears on removal. Use an all-metal prevailing-torque nut or a coated washer.

Q7. Is 70 % of yield safe?

Yes. ISO 898-1 gives class 8.8 a yield of 640 MPa and recommends 60–75 % of yield as preload. Fracture risk starts above about 90 %.

Q8. What works above 300 °C?

Inconel 718 bolts, Belleville washers and MoS₂ lubrication. Threadlocker softens above 150 °C and nylon inserts fail at 120 °C.

Q9. Which locking nuts can be reused?

All-metal 5–10 times, nylon 5, deformed-thread 15, wedge washers indefinitely, threadlocker essentially once.

Q10. How do I know a bolt has loosened in service?

Paint a witness line across nut and joint during inspection; a broken line means movement. Critical joints can use bolt load indicators or torque-sensitive paint.

9. Standards and references

Chinese standards

• GB/T 3098.1 Mechanical properties of bolts, screws and studs
• GB/T 10431 Transverse vibration test method for fasteners
• GB/T 94.3 Spring washers — saddle and wave types
• GB/T 6184 Prevailing-torque hex nuts, style 1, all-metal
• GB/T 889.1 Prevailing-torque hex nuts, style 1, non-metallic insert

International standards

• ISO 898-1 Mechanical properties of fasteners
• ISO 16047 Fasteners — torque/clamp force testing
• ISO 16130 Fasteners — performance test for prevailing-torque types
• DIN 65151 Aerospace — locking performance of threaded fasteners
• NASA-STD-5020 Requirements for threaded fastening systems

Key literature

• Gong H., Liu J. H., Feng H. H. Review on anti-loosening methods for threaded fasteners. Chinese Journal of Aeronautics, 2021. DOI: 10.1016/j.cja.2020.12.038
• Junker G. H. New criteria for self-loosening of fasteners under vibration. SAE Automotive Engineering, 1969, 78:314–335.
• Dravid S., Tripathi K., Chouksey M. Experimental study of prevailing torque locking behavior of plain and spring washers. Int. J. Eng. Res., 2012.
• Pinelli M., Subramanian C., Tabor J. Experimental study of breakaway torque in service. Transport Research Record, 2012.