There is no single fatigue-strength number that qualifies every Inconel 718 fastener. A useful fatigue result identifies the material condition, specimen or fastener geometry, surface and thread process, temperature, mean stress, loading history and failure criterion. A polished material specimen can explain behavior; it cannot, by itself, approve a threaded joint for service.
Before accepting a fatigue report, match three things: the tested part and process, the stress history, and the required outcome. A smooth-specimen result is material evidence. A representative threaded-component test is closer to the part. The joint assessment must still establish the loads that the part actually sees.
Read the test before reading the curve
Zhao and colleagues studied differently heat-treated 718 using smooth hourglass specimens under fully reversed tension–compression at 20 kHz. Fully reversed means the stress alternates equally between tension and compression: the stress ratio R = minimum stress / maximum stress is −1. Their setup also monitored specimen self-heating. [1, Section 2]
| Stress amplitude | Reported cycles | Outcome |
|---|---|---|
| 540 MPa | 1.57 × 10^9 | Fatigue failure. |
| 520 MPa | 6.13 × 10^9, run-out | Run-out; not infinite life. |
A run-out means the test ended without recording fatigue failure. These are individual observations, not a statistically qualified lower-bound curve. The material was not tested as a bolt-and-nut assembly. The paper’s treatment comparisons support the importance of condition; they do not establish an allowable stress for your threaded product. [1, Sections 4–5 and Appendix A]
An S–N curve plots cyclic stress against the number of cycles to a defined outcome. Check whether the plotted stress is amplitude, range or maximum stress, and whether it is nominal or local. Comparing unlike definitions can create a large apparent safety margin that does not exist.
A thread changes the evidence needed
A thread root concentrates stress. NASA’s Fastener Design Manual discusses fatigue at the threaded region and the benefit of compressive surface stresses introduced by cold rolling. [2, p.13] That provides a mechanism for investigating a rolled-thread route; it does not supply a universal life multiplier for 718.
For a proposed cut or rolled thread, identify the root geometry, sequence relative to aging, later surface treatment and inspection basis. The fatigue evidence must represent that supplied process. A passing GO / NO-GO result checks dimensions, not fatigue life.
Vander Voort’s actual-718-fastener study makes the inspection issue concrete: it distinguishes thread-root abnormalities, potential fatigue-initiation sites, from observed crest laps. Its sectioning and preparation precautions also matter because preparation damage can mimic manufacturing damage. [3, pp.826, 831–834] It does not provide a component S–N curve or prove a fixed rolled-thread life advantage. Agree the inspection location, preparation method and acceptance criterion; do not count every surface mark as an equivalent defect.
Worked example: small cyclic stress can coexist with high mean stress
All inputs below are fictional teaching assumptions, not a validated joint, an installation instruction or a predicted 718 service life.
Consider one bolt in a concentric, axially loaded, linear-elastic joint. Assume no bending, prying, slip or gasket nonlinearity, with the members remaining in contact.
- Installation preload F_i: 30 kN.
- Tensile stress area A_s: 80 mm².
- External separating load P: cycles from 2 to 20 kN.
- Bolt share of an added separating load, C: 0.20. Here C = k_b / (k_b + k_c), where k_b and k_c are the axial stiffnesses of the bolt and clamped members.
Using the elementary spring model described by NASA, bolt load is F_b = F_i + CP while contact is maintained. [2, pp.13–15] Force balance then gives remaining clamp force F_c = F_i − (1 − C)P.
| Quantity | Calculation | Result |
|---|---|---|
| Bolt load | 30 + 0.20 × (2 to 20) kN | 30.4–34.0 kN |
| Nominal axial stress | 30,400–34,000 N / 80 mm² | 380–425 MPa |
| Stress amplitude | (425 − 380) / 2 | 22.5 MPa |
| Mean stress; R | (425 + 380) / 2; 380 / 425 | 402.5 MPa; R ≈ 0.894 |
| Minimum clamp force | 30 − 0.80 × 20 kN | 14 kN: contact retained in this model. |
The 22.5 MPa amplitude cannot simply be compared with the paper’s 520 MPa run-out and declared safe. The mean stress and stress ratio differ substantially; the thread-root stress is not represented by the nominal stress; and the supporting evidence must represent the part, process and duty, with differences justified by an accepted component-fatigue assessment method. This arithmetic gives no fatigue safety factor or life.
Now change only the assumed preload to 10 kN. The model predicts contact loss at P = 10 / 0.80 = 12.5 kN, below the 20 kN peak. Beyond separation, the original closed-joint relationship no longer applies; do not extend the small-amplitude result across the whole cycle. Reassess the joint contact and bolt-load history first. This is why a fatigue review needs the credible preload range, including losses, rather than the assembly target alone.
Match the duty and the acceptance basis
A constant-amplitude room-temperature result does not cover arbitrary variable-amplitude loading, hot dwell or corrosive exposure. Record cycle counts and sequence, load peaks, temperature history, hold periods and environment. If heat changes preload during the duty, bring the resulting load history into the fatigue assessment; do not substitute a retained-load percentage for fatigue qualification. The preload guide covers that separate issue.
Ask what failed or ended the supporting test: crack detection, loss of stiffness, fracture or a specified run-out count. Ask how scatter, specimen count and the chosen reliability basis were handled. A single successful specimen or an average curve is not automatically the design acceptance criterion.
A buyer can make this review concrete by requesting the report’s material condition, part geometry, thread and surface route, temperature, stress definition, R ratio, frequency, cycle history, sample count and outcome definition. The design authority then decides whether it represents the drawing and service duty, needs an accepted correction, or requires different evidence.
Use the UNJ versus UN guide for thread-form and mating questions; do not change the thread system merely to chase a favorable fatigue claim. For the material condition and finished-lot specification, see how to specify 718 fasteners.
Put the fatigue requirement into the enquiry
Send the drawing revision, 718 material condition, mating components, preload basis and load/temperature history. State the governing design or component specification and any required fatigue qualification before quotations are compared. For a stud, nut and washer assembly, identify all components so that a bolt-only report is not mistaken for an assembly assessment.
BYBOLT deliveries include an EN 10204 3.1 material inspection certificate, heat-number traceability and mechanical testing, with test objects, methods, sampling and acceptance agreed in the order. Threaded products include sampled GO / NO-GO gauging. These routine records do not imply fatigue-life qualification. Any additional fatigue testing, PMI or third-party inspection must be agreed in the quotation. Submit the drawing and duty for review.