Torque vs tension: why torque only estimates bolt clamp load
What holds a joint together is tension in the bolt, the clamp load. Torque is only an indirect way to get there, because roughly 90% of it is spent on friction. With a torque wrench, clamp load scatters by about ±25%, and NASA's design default for dry or as-received hardware is 35%.
Contents9 sections
Key takeaways
- A torque reading 'does not indicate the bolt tension directly' 1. It is converted to tension through the nut factor K.
- About 90% of input torque goes into thread and underhead friction, leaving about 10% for stretching the bolt 1.
- IFI accuracy by method: feel ±35%, torque wrench ±25%, turn of the nut ±15%, load-indicating washers ±10%, bolt elongation ±3 to 5%, strain gages ±1% 2.
- NASA-STD-5020B default preload variation for torque control: 25% lubricated, 35% non-lubricated or as-received 3.
- RCSC withdrew torque-pretension tables in 1954 after research showed ±40% variation 4.
Quick facts
| Torque lost to friction | About 90% 1; 85 to 95% 5 |
|---|---|
| Torque wrench preload scatter | ±25% 2 |
| Turn of the nut | ±15% 2 |
| Bolt elongation | ±3 to 5% 2 |
Tension is what you want
A bolted joint works by clamping. The bolt is stretched like a stiff spring, and the stretch pulls the parts together. Fastenal: applying torque turns the fastener, “this action stretches the bolt (similar to a spring) and creates a tension in the bolt. In most cases it is this tension or preload that we need” 1. If the clamping force is not correct, Fastenal notes, the joint can fail by bolt fatigue, vibration loosening, stress corrosion cracking or hydrogen embrittlement 1.
Torque is easy to measure, tension is not. So most joints are tightened to a torque, and the torque is chosen to produce the tension. Fastenal is blunt about the gap: the torque reading “does not indicate the bolt tension directly. Rather the torque reading is only an indirect indication of our desired tension” 1.
Where the torque goes
The conversion runs through T = K × D × F 1. K, the nut factor, carries all the uncertainty. It combines thread geometry, thread friction and underhead friction 1, and friction is most of the torque:
- Fastenal: about 90% of the input energy is lost overcoming friction under the head and in the threads; about 10% becomes bolt stretch 1.
- Bolt Science: usually between 85% and 95% of the applied torque overcomes friction, so “slight variations in the frictional conditions can lead to large changes in the bolt preload” 5.
Because the stretch is the small remainder, a modest change in friction from oil, plating, dirt or rust is a large change in clamp load.
Fastenal lists what changes K: hardness, materials, class of fit, plating, surface finish, cut or rolled threads, washers, the tool, tightening speed, whether the nut or bolt is turned, how many times the fastener was used, and every property of the lubricant 1. “Each fastener, even from identical lots … will give you different torque values for the same preload” 1.
How much it scatters
| Method | Preload accuracy | Relative cost |
|---|---|---|
| Feel (operator’s judgment) | ±35% | 1 |
| Torque wrench | ±25% | 1.5 |
| Turn of the nut | ±15% | 3 |
| Load-indicating washers | ±10% | 7 |
| Fastener elongation | ±3 to 5% | 15 |
| Strain gages | ±1% | 20 |
Other sources line up:
- NASA-STD-5020B. Torque control is “the most economical method” and “the least accurate for controlling preload” 3. Without test data, the default preload variation for torque control is 25% if lubricated and 35% if non-lubricated or as-received; for turn-of-nut or turn-angle 25%; for bolt stretch 10% 3.
- Fastenal. “Even perfect input torque can give a variation of preload by as much as 25 - 30 %” 1.
- RCSC. The first edition of the structural bolting specification (1951) carried a torque-to-pretension table. Research soon showed variation “as high as ±40 percent must be anticipated” unless the relationship is established for each bolt lot, diameter and component condition, so tables and equations were withdrawn in 1954 4.
What it means on a real bolt
An M12 x 1.75 class 10.9 bolt, plain dry, at the site’s 75%-of-proof target: 126 N·m (93 ft-lb) for a nominal 52.5 kN. Its proof load is 70.0 kN.
| Scatter | Clamp load range | High end vs proof load |
|---|---|---|
| ±25% (torque wrench, IFI) | 39.4 to 65.6 kN | 94% |
| ±35% (NASA dry default) | 34.1 to 70.8 kN | 101% |
At ±25% the bolt stays under proof load. At NASA’s dry default the high end just crosses it. That is the reason torque tables stop at 75% of proof 1, and the reason proof load itself is set at about 75% of yield “to prevent inadvertent yielding of the fastener through torque measurement inaccuracies” 2.
Fastenal adds a point about design: “Were most joints not massively over-designed to accommodate inaccurate tightening, simple tightening procedures could prove catastrophic” 1. Ordinary joints carry more or bigger bolts than strictly needed so torque control is good enough.
Getting closer to the tension you want
In rising order of cost and accuracy 1 2:
Control the friction
Same lubricant, same amount, same surfaces, every time. NASA notes the torque-preload relationship depends on where lubricant is applied 3. Bolt Science suggests avoiding plain washers, which can shift where the slip happens and change the friction radius 5.
Torque the nut
NASA: “Whenever possible, torque should be applied to the nut,” and the drawing should say which part is turned because the torque-preload relationship differs 3.
Account for running torque
On locking fasteners, running torque “does not generate preload”; specifying torque above measured running torque gives tighter control 3.
Torque, then turn
Snug with torque, then rotate a set angle. The added tension of a snugged bolt is proportional to the turn and “not affected by variations in friction” 1. See turn-of-nut and torque-to-yield bolts.
Measure stretch or tension directly
Bolt elongation, load-indicating washers, strain gages, ultrasonics or hydraulic tensioners 1.
Calibrate against the real hardware
For structural bolts installed by calibrated wrench, RCSC requires the wrench to be calibrated daily and whenever a component lot, lubrication or surface condition changes 4.
Tension changes after tightening
Even a perfectly set preload doesn’t stay put. Bolts tightened first in a group relax as the rest pull the joint together, and creep at heavily loaded thread and contact surfaces adds to the loss 1. RCSC cites about 6.5% pretension loss within five days for galvanized plates and bolts, against 2.5% for uncoated 4. Tightening patterns and multiple passes exist for this reason: see torque sequence.
Using the site’s numbers
Every bolt torque on this site is T = K × D × F at 75% of proof load, with K stated. Read each one as the center of a band about ±25% wide. To see the band for your own bolt, run the torque through the clamp load calculator. When an equipment maker gives a torque, it was set for their joint, and it comes first.
Sources
- 1
- 2
- 3
- 4
- 5Bolt Science Ltd, Bolt Science, Methods of Tightening Threaded FastenersChecked 21 Oct 2026
Spotted a value that doesn’t match its source? Report an error
Related
- Clamp load calculatorEstimate bolt clamp load from tightening torque with F = T ÷ (K × D). 75 ft-lb on a 1/2 in bolt with dry threads gives about 9,000 lbf (40 kN), ±25%.
- Dry vs lubricated torqueWhy oiled bolts need less torque than dry ones for the same clamp load, with nut factor K values from Fastenal, NASA and Henkel for oil, zinc, anti-seize and moly.
- Turn-of-nutTurn-of-nut per RCSC 2020 Table 8.1: 1/3 to 1 turn past snug by bolt length, tolerance +60°/−0°. Match-marking, verification and the new combined method.
- MethodologyHow Torque Reference calculates bolt torque from SAE, ISO and RCSC standards and sources vehicle torque from manufacturer documents, with every input listed.