Most failed bolted joints do not fail because the bolt was too weak. They fail because the bolt was not tight enough. Insufficient preload is the single most common root cause of fatigue cracking, loosening, and gasket leaks in mechanical assemblies — and the fix is rarely a bigger fastener.
This article walks through what preload actually does, how torque converts to preload (badly), and why the difference between a dry and lubricated thread can swing your clamp load by 4x.
What Preload Does
When you tighten a bolt, it stretches like a stiff spring. The clamped parts (washers, flange faces, gaskets) compress in series. The bolt carries tension; the joint carries compression. As long as the external load on the joint is less than the preload, almost none of that external load is felt by the bolt itself — the joint just unloads slightly while the bolt stays at constant tension.
That is the whole game. A properly preloaded joint behaves like one stiff part. An under-preloaded joint lets the bolt see every cycle of external load and fails by fatigue in surprisingly few cycles.
The Joint Diagram
If you plot bolt force vs. joint deflection, the bolt is a soft spring (low stiffness, kb) and the clamped parts are a stiff spring (high stiffness, kj). External load is shared between the two in proportion to their stiffnesses. The ratio of external load that the bolt sees is:
Φ = kb / (kb + kj)
For a typical steel-on-steel joint, Φ is around 0.1–0.2 — meaning the bolt only feels 10–20% of the external load. That is the magic of preload. Lose it, and the bolt sees 100% of the cyclic load.
Turning Torque Into Tension
The relationship engineers use is the short-form torque equation:
T = K · F · D
- T — applied torque
- F — bolt preload (tension)
- D — nominal bolt diameter
- K — the dimensionless “nut factor” or torque coefficient
This equation hides every difficult thing about bolt tightening inside K. Friction at the threads, friction under the nut face, thread pitch, helix angle, surface finish, plating, lubricant — all of it disappears into a single empirical number.
The K-Factor Range Is Brutal
Published K-factor values for the same nominal bolt grade:
| Condition | K (typical) | Range |
|---|---|---|
| Plain steel, dry, as-received | 0.20 | 0.17–0.23 |
| Zinc plated, dry | 0.22 | 0.18–0.27 |
| Cadmium plated | 0.16 | 0.14–0.20 |
| Lightly oiled | 0.15 | 0.13–0.18 |
| Moly disulfide (MoS2) lubricant | 0.10 | 0.08–0.12 |
| PTFE / waxed | 0.08 | 0.06–0.10 |
Read that table carefully. A plain dry bolt at K=0.20 vs. the same bolt with anti-seize at K=0.10 will produce twice the preload at the same torque. If a print calls for 100 ft-lb on a 1/2″ bolt and the assembler reaches for the anti-seize bucket, that fastener will be tensioned to 200% of intent — likely past yield.
Conversely, if a torque spec was developed on lubricated bolts and the shop runs out of lube, the same torque produces half the preload — and the joint fails by separation under the first dynamic load.
How Much Preload Do You Actually Want?
Standard industry practice is to preload to 75% of proof load:
- Proof load is the load below which the bolt shows zero measurable permanent set — roughly 90–93% of yield for SAE J429 / ISO 898 grades.
- 75% of proof gives you margin for tightening scatter, gasket relaxation, and external load.
- For reused bolts, drop to 60% of proof — used threads have damaged engagement.
- For critical fatigue-loaded joints (engine head bolts, structural splices, lifting eyes), some specs go to 90% of proof using angle-of-turn or stretch measurement instead of torque.
Why Torque-Only Tightening Is Unreliable
Even when K is well-controlled, torque-only tightening has roughly ±25% scatter on actual preload. The reason: roughly 50% of input torque overcomes underhead friction, 40% overcomes thread friction, and only 10% actually stretches the bolt. Small variations in friction swing the preload dramatically.
Better methods, in order of accuracy:
- Torque-only: ±25% on preload. Industry default.
- Torque-then-angle: ±15%. Snug to a small torque, then turn a specified angle. Used in automotive head bolts.
- Bolt elongation measurement: ±5–10%. Ultrasonic or micrometer. Used in critical pressure vessels and turbomachinery.
- Hydraulic stud tensioner: ±3–5%. Direct stretch via hydraulic ram. Used in large flange bolts, wind turbine towers.
Running The Numbers
To work through a torque spec on a real bolt — including grade, plating, lubrication, and preload target — use the FicientDesign Bolt Torque Calculator. It pulls UNC/UNF and metric thread data, lets you select a K-factor by surface condition, and outputs torque at any percentage of proof load.
If you are working pressure flanges, jump directly to the ASME flange bolt load pages for class-specific gasket seating and operating bolt loads. For structural connections, pair the calculator with the Section Properties Calculator when designing the connecting elements themselves.
Three Rules to Live By
1. Specify the lubrication, not just the torque.
2. If you do not know K within ±10%, you do not know preload within ±25%.
3. Loose bolts kill more joints than overloaded bolts. Preload is your friend.
A torque spec without a lubrication callout is incomplete. A torque spec with a lubrication callout but no controlled lubricant in the shop is wishful thinking. The fastener is the cheapest part of the assembly; the consequences of getting its preload wrong are not.