The Short Answer
A bolt does not unwind itself because a thread slowly turns under tension. It comes loose because the parts it clamps slide against each other, and the sliding lets the nut rotate a few degrees at a time. GB/T 10431-2008 §1 confines its method to transverse vibration, and §2 describes the loosening it measures as the result of alternating transverse displacement between two clamped metal plates. That is the bolt self-loosening test most suppliers mean when they say a fastener passed a Junker test.
Three points follow:
- The test covers one loading path. Lateral slip at the joint interface, under one preload, one surface condition and one amplitude.
- The result is a curve, not a verdict. §1 states that a result is only a comparative evaluation under specific conditions and carries no absolute quantitative index. A pass means slower decay than the reference, nothing more.
- Four boundaries limit what a pass covers. Loading path, preload and surface state, temperature and creep, and service life.
The rest of this page takes those statements apart: why transverse slip drives loosening, what the angle arithmetic settles and what it does not, how the two common standards differ, and how to read a report that arrives with a pass line and little else.
Why Transverse Slip, Not Axial Vibration
Consider a joint under axial tension. The bolt stretches, the plates compress, and the clamp force rises and falls with the load. The nut does not move relative to the bolt, because both sit on the same load path and no force pushes the thread flanks sideways.
GB/T 10431-2008 §2 describes a different arrangement. The fastener clamps two metal plates, and the machine forces one plate to move alternately against the other in the plane of the joint. The plates slip, the clamp force falls, and the standard treats that fall as the measure of locking performance. The mechanism in §2 never invokes vibration along the bolt axis. The displacement that does the work is within the joint plane.
A transverse load starts slip when it exceeds the friction the clamp force can supply, written as μF. Below that threshold the joint is rigid and nothing happens however long it vibrates. Above it, the plates slide and the thread is loaded in a direction it was not designed to resist. Preload appears in every loosening discussion because it sets the force at which slip begins.
The method is usually credited to Junker’s 1969 work on self-loosening under vibration. The test that carries his name is the one codified in GB/T 10431-2008 and, on the American side, in MIL-STD-1312-7A.
The Loosening Chain
Four stages, in order.
1. Slip starts at the interface. Once the transverse load passes μF, the plates move relative to each other. The motion is small: the amplitudes in Annex B Table B.1 sit at ±1.2 mm for M12 and never exceed ±2.0 mm across M5 to M24.
2. The thread flanks slide. Slip at the interface is transmitted to the thread contact. In a static joint both the thread contact and the bearing face sit in static friction. Once the plates move, the contacts are forced to slide in alternate half-cycles, and static friction no longer applies to whichever contact is moving.
3. The helix stops balancing. Preload pushes the thread along the helix and friction holds it back. Remove the friction at one contact for half a cycle and the tangential component along the helix is no longer opposed. The nut rotates through a small angle in the loosening direction.
4. Clamp force falls in steps. Each cycle removes a little preload. The analysis clause, §6.2.2, plots the ratio of residual to initial clamp force against cycle count on a logarithmic axis. That plot records a per-cycle process rather than one failure event. At the end of the chain §2 states the limit case: the clamp force can be lost completely.
Stage 2 carries the argument. While every contact stays in static friction, the thread is self-locking and no amount of load turns it. Loosening requires something that forces the contacts to slide. Transverse displacement does that; axial loading does not.
The Angle Check That Does Not Settle It
The static argument runs on two angles, and it is worth doing the arithmetic before trusting either one.
Three assumptions sit in that block. The friction coefficient is an assumption, not a measurement; 0.15 is a common dry-steel illustration, and plated or lubricated threads differ. The pitch diameter comes from the basic profile and ignores tolerance. The cos 30° term refers the friction force to the normal of a 60° flank.
The result: at μ = 0.15 the friction angle is 8.53°, nearly three times the 2.94° lead angle. For the thread to unwind under a purely axial load the effective friction coefficient would have to fall below 0.044, which a dry or lightly oiled steel thread does not do. A static joint with proper preload is genuinely self-locking.
So the angle sum cannot explain field loosening, and it cannot certify a joint either. What it explains is why the acceptance method is a dynamic test: the static criterion stops being valid once the contact is made to slide. Thread pitch moves the lead angle, and that comparison belongs to a separate discussion. What pitch does not change is the shape of the argument: a static angle check cannot certify a joint that is being made to slide.
How GB/T 10431-2008 Runs the Test
The apparatus clamps the specimen between two metal plates and imposes an alternating transverse displacement; §3.2 requires a sinusoidal waveform. Conditions are fixed so that runs can be set beside each other: ambient temperature and pressure in a clean environment (§4.1), lubricant on the thread and bearing surfaces (§4.2), a frequency taken from service or set to 12.5 Hz when unspecified (§4.3), and a free-state amplitude chosen so the bolt is not sheared (§4.3). Preload follows the test requirement or the specimen’s working clamp force (§4.4). Ten specimens per group are suggested, with 15 more when scatter in vibration behaviour or installation torque is large (§4.5).
| Parameter | Setting in GB/T 10431-2008 | Clause |
|---|---|---|
| Waveform | sinusoidal | §3.2 |
| Frequency | service frequency, or 12.5 Hz when undetermined | §4.3 |
| Amplitude | free state with clamp force at zero; must not shear the bolt | §4.3 |
| Preload | test requirement or the working clamp force | §4.4 |
| Specimens per group | 10, plus 15 when scatter is large | §4.5 |
| Tightening | torque wrench to a set torque or clamp force; both recorded | §5.3 |
| Stop condition | clamp force falls to a preset value, or a set cycle count | §5.4 |
| Evaluation | residual to initial clamp force plotted against cycles | §6.1, §6.2.2 |
The tightening step reaches the clamp force through torque, and the conversion between the two runs through the K-factor (torque and preload K-factor calculation). A report that states one without the other cannot be reproduced.
Annex B is informative, not mandatory. It recommends a preload and a free-state amplitude for class 12.9 bolts from M5 to M24. Divide each preload by Rm times As at 1200 MPa and the ratio holds across the table:
| Size | As (mm²) | Recommended preload (N) | Free-state amplitude (mm) | Preload ÷ (1200 MPa × As) |
|---|---|---|---|---|
| M6 | 20.12 | 9,750 | ±0.6 | 40.4% |
| M8 | 36.61 | 17,750 | ±0.8 | 40.4% |
| M10 | 57.99 | 28,150 | ±1.0 | 40.5% |
| M12 | 84.27 | 40,900 | ±1.2 | 40.4% |
| M16 | 156.67 | 76,000 | ±1.6 | 40.4% |
| M20 | 244.79 | 119,000 | ±1.9 | 40.5% |
| M24 | 352.50 | 170,000 | ±2.0 | 40.2% |
The ratio stays between 40.2% and 40.5%. The recommended test preload is a fixed fraction of Rm times As, about 0.40, not a per-size engineering decision. The stress area comes from the basic profile: As = (π/4) × (d − 0.9382P)².
That fraction matters when it is set against a design preload. GB/T 32076.2-2015 §4 defines the design preload as Fp = 0.7 × Rm × As. For M12 at class 12.9, with As = 84.27 mm²:
- Fp = 0.7 × 1200 × 84.27 = 70,787 N, or 70.8 kN
- Annex B test preload = 40,900 N
- ratio = 40,900 / 70,787 = 57.8%
A run at Annex B numbers is a run at 58% of the design preload for that bolt. On class 8.8 the same 40.9 kN is 86.7% of Fp, which is 47.2 kN at the nominal Rm of 800 MPa (GB/T 3098.1-2010 §7.1, Table 3). The gap depends on the property class bought, and the torque that produces a given preload at each class is tabulated in the bolt torque chart. §4.4 allows the laboratory to set preload to the working clamp force instead, so ask which value was used before accepting the result.
MIL-STD-1312-7A vs GB/T 10431
The two methods share a purpose and differ in almost every parameter.
| Aspect | GB/T 10431-2008 | MIL-STD-1312-7A |
|---|---|---|
| Source of relative motion | two clamped plates displaced in the joint plane (§2) | a test cylinder held by the fastener, traversing slots in the vibrating fixture (§4.1.2, §5.1) |
| Waveform and frequency | sinusoidal; service frequency or 12.5 Hz (§3.2, §4.3) | sinusoidal; 1,750 to 1,800 cycles per minute, about 29.2 to 30.0 Hz (§4.1.1) |
| Amplitude | free-state amplitude, bolt must not shear (§4.3); Annex B gives ±0.6 to ±2.0 mm | 0.450 ± 0.015 in peak to peak (§4.1.1), which is 11.43 mm |
| Preload | test requirement or working clamp force (§4.4); Annex B recommendations | seating torque specified in the acquisition document (§5.1) |
| Specimen hold | clamped plates and fixtures per Annex A | cylinder assembly with washers added so the axis stays perpendicular to the vibration (figure 1 notes); slot clearance criteria for other fixtures (§4.1.2.1) |
| Failure criterion | clamp-force decay curve; result is comparative (§1, §2) | inspection of components after the run; if rotation is the failure measure, parts are scribed to show relative movement (§5.1) |
| Report | curve and statistical analysis (§6.1, §6.2.2) | seating torque, bake time, reuses, duration, frequency, inspection results (§6.1) |
The amplitude comparison needs a caveat. GB/T 10431 measures free-state amplitude with the clamp force at zero, while MIL-STD-1312-7A states a peak-to-peak figure for the vibrating fixture. Taking 11.43 mm against the 2.4 mm peak-to-peak equivalent of ±1.2 mm gives a factor of about 4.8, and frequency roughly doubles. Read that as an order-of-magnitude difference, not an equivalence, because the two numbers are defined differently.
Neither standard carries a universal pass value. MIL-STD-1312-7A §5.1 takes the seating torque from the acquisition document and vibrates for the required time; the report records that duration (§6.1). GB/T 10431-2008 §1 rules out an absolute index. In both cases the parameter set comes from the buyer or the test request, which means a bare word such as “passed” carries no information by itself.
What a Pass Proves, and What It Does Not
A pass says that, inside one parameter set, this assembly lost clamp force more slowly than whatever it was compared against. §1 states that this is all the result can support. Four boundaries follow.
1. One loading path. The method imposes lateral slip. A joint loaded axially, in torsion, or in bending does not reproduce that motion, and a pass on lateral slip says nothing about the other paths.
2. One preload and one surface state. §4.2 requires lubricant on the thread and bearing surfaces. Preload, plating, roughness and whether the parts were reused all change the friction condition at both contacts. A report produced with a different lubricant or a different coating describes a different joint, even with the same part number.
3. No temperature, no creep, no embedding loss. §4.1 sets room temperature, normal pressure and a clean environment. Zinc and aluminium creep under sustained load at elevated temperature, coatings and paint films compress, clamped polymers relax. Each of those is a preload loss that the curve never sees, because the test neither runs hot nor runs long enough to produce it.
4. No service life. §5.4 stops at a preset force or a set cycle count, and the fallback frequency is 12.5 Hz. One hour at 12.5 Hz is 45,000 cycles. A machine running at 30 Hz for eight hours a day over 250 working days accumulates about 216 million cycles. A run that ends at 45,000 cycles compares designs; it does not predict that machine.
Locking Methods and Their Failure Boundaries
Locking hardware is usually classified by what it does to the joint, and the classes behave differently once slip starts.
| Class | Method | Mechanism | What it changes in the test | Failure boundary |
|---|---|---|---|---|
| Friction | Nylon-insert nut | An elastic insert bears on the thread and adds a prevailing torque | Opposes the small back-rotations while the insert stays elastic | The insert creeps at elevated temperature; prevailing torque is not clamp force |
| Friction | All-metal prevailing torque nut | A deformed thread section creates interference with the bolt thread | Same opposition, with metal-to-metal contact | Galling, and torque falls after a few assembly cycles |
| Mechanical | Cotter pin with castle nut | A pin through the nut and a cross-hole blocks rotation | No relative rotation while the pin is intact | Needs a drilled bolt; cannot stop preload loss from embedding, so the joint can lose clamp force with the nut locked |
| Mechanical | Double nut | A second nut jammed against the first stores opposing preload | Holds while the jam preload survives | Transverse slip can relieve the jam, after which the pair acts as one nut |
| Mechanical | Wedge-locking washer pair | Cam faces steeper than the lead angle must climb before the nut can turn | Blocks rotation at the washer interface rather than at the thread | Needs clean, hard bearing faces; reused pairs wear the face |
| Adhesive | Threadlocker | Cured polymer fills the thread clearance and carries shear | Acts as a bond until the bond breaks | Temperature limit, cure and contamination; disassembly breaks it; it does not correct under-tightening |
| Preload | Higher preload | Raises the transverse force μF needed to start slip | Delays the start of the curve | Bounded by bolt strength, thread stripping and embedding in the clamped parts |
Preload sits in the last row because it is not a locking device. It sets the threshold at which slip begins and it is the variable most often missing from a comparison between two test reports. Two curves produced at different preloads cannot be ranked against each other, whatever the pass lines say.
How to Read the Report
A report usually leads with the conclusion. The useful reading starts further down.
The first thing to read is where the curve leaves its flat segment. An assembly that holds for a long stretch and then falls steeply behaves differently from one that decays from the first cycle, even when both reach the same residual force at the stop point. The second is the spread between specimens, because §4.5 anticipates scatter and asks for 15 additional specimens when it appears. A tight group of ten curves supports a comparison; one curve does not.
What to Send for a Review
Seven items settle most of these questions: the standard and edition, the initial clamp force and how it was measured, the amplitude with its definition, the frequency and stop count, the lubricant and surface treatment, the specimen count with individual curves, and the property class and coating of the parts tested.
With those in hand the review is mechanical. Does the preload in the report match the preload in the joint, does the amplitude match the slip the joint actually sees, and does the surface state match what will be assembled. Where any of the three does not match, the report describes a different joint, and a pass line from it should not be carried across.
References
- GB/T 10431-2008 — Transverse vibration testing method for fasteners, Clauses 1, 2, 3.2, 4.1-4.5, 5.3, 5.4, 6.1, 6.2.2 and Annex B Table B.1
- MIL-STD-1312-7A — Fastener test methods, Method 7: Vibration, 1.1, 4.1.1, 4.1.2, 4.1.2.1, 5.1 and 6.1 (no online source)
- GB/T 32076.2-2015 — High-strength structural bolting assemblies for preloading, Part 2: Suitability for preloading, Clause 4
- GB/T 3098.1-2010 — Mechanical properties of fasteners: bolts, screws and studs, 7.1 Table 3
- [G. H. Junker — New criteria for self-loosening of fasteners under vibration, SAE Transactions, 1969 (no online source)]
Key Standards
Standards and clauses referenced by this article.
- Transverse vibration testing method for fasteners1, 2, 3.2, 4.1-4.5, 5.3, 5.4, 6.1, 6.2.2, Annex B Table B.1
Source for the transverse-slip loosening mechanism, the comparative-only status of the result, the test conditions, the sample size, the stop criterion, the residual-clamp-force curve, and the recommended preload and amplitude values used in the two tables.
- Fastener test methods, Method 7: Vibration1.1, 4.1.1, 4.1.2, 4.1.2.1, 5.1, 6.1
Source for the cylinder-and-slot fixture, the 1,750 to 1,800 cycles per minute frequency, the 0.450 ± 0.015 in peak-to-peak amplitude, the seating torque taken from the acquisition document, and the rotation-based failure inspection.
- High-strength structural bolting assemblies for preloading - Part 2: Suitability for preloading4
Source for the design preload definition Fp = 0.7 Rm As used in the preload comparison, and for the scope of the suitability-for-preloading test.
- Mechanical properties of fasteners - Bolts, screws and studs7.1, Table 3
Source for nominal Rm 800 MPa at class 8.8 and 1200 MPa at class 12.9, used to compute As-based preloads.
References
Sources used for fact checking and background context.
- SAMR openstd - official free full-text portal for GB standardsState Administration for Market Regulation / Standardization Administration of China
- MIL-STD-1312-7A - Fastener Test Methods, Method 7: VibrationU.S. Department of Defense
- G. H. Junker - New criteria for self-loosening of fasteners under vibration, SAE Transactions, 1969SAE
Deep Reading
More systematic selection, procurement, or inspection guides.
- Bolt Property Classes 4.8, 8.8 and 12.9
Where the Rm values used in the preload ratio come from
This page is one entry in the fastener engineering reference. The library covers standards, dimensions, materials, and failure analysis; every figure is traced back to its source.
Continue reading in the library →