Most tightening torque goes into friction — why preload differs at the same torque
Why a bolt tightened to the specified torque can still break or work loose. Where the torque goes, and how much preload changes with friction, with a worked example and graph.
"We tightened to the specified torque, and the bolt broke." We hear this from time to time, and also "We tightened to the specified torque, and it came loose." Both have one thing in common: torque alone cannot fix the preload. Between torque and preload there is friction, and friction takes most of the torque.
The purpose of tightening is preload
A bolt clamps parts like a stretched spring. That clamping force is the preload. With enough preload the parts do not slip or open, and the joint resists vibration loosening. Torque is the turning force applied with a wrench to create that preload; "how many N·m?" really means "how much preload?".
Worked example conditions: M5×0.8, A2-70, pan head with an assumed under-head contact diameter of 8.4 mm (KS B 1023 annex dk minimum), clearance hole 5.5 mm, thread friction μG = under-head friction μK, tool scatter excluded (VDI 2230 formula)
Where the torque goes
Wrench torque splits three ways. Under the conditions above with a friction coefficient of 0.30:
| Where it goes | Share (μ 0.30) |
|---|---|
| ■ Climbing the thread helix — the part that creates preload | 6.6% |
| ■ Thread friction between external and internal threads | 40.0% |
| ■ Under-head friction between head and part | 53.5% |
Here about 93% of the torque goes into friction. The split varies with size, head type and friction, but friction always takes most of it, so a small change in friction moves the share that reaches preload a great deal. (The three terms of torque = preload × (0.16P + 0.58·d₂·μG + μK·Dkm/2).)
Same torque, different preload
The same fact seen from two directions:
| Friction coefficient | Torque for 2,500 N preload | Preload calculated at 6.68 N·m |
|---|---|---|
| 0.30 | 4.88 N·m | 3,425 N |
| 0.50 | 7.91 N·m | 2,111 N |
Conditions above, tool scatter excluded. All preloads in the table are within this model's permissible assembly preload under both friction conditions (3,854 N at μ 0.30, 2,865 N at μ 0.50).
With higher friction the same preload needs about 1.6 times the torque. On the shop floor, though, the other direction matters more: operators tighten to a set torque. When the torque wrench clicks, tightening ends — so high-friction parts finish with less preload and low-friction parts with more, even though both feel equally tight.
6.68 N·m is the torque set in this example for dry stainless (friction 0.30–0.50). Applied as it stands to a lubricated part (friction around 0.10), the same formula gives 9,084 N, far beyond the permissible assembly preload of 5,194 N — a region where the bolt may stretch permanently or fail.
If friction is higher than expected, preload falls short and the joint may loosen under vibration or slip. If friction is lower than expected, preload overshoots and the bolt may stretch or fail. When setting a torque, the friction condition assumed (finish, lubrication, surface) matters as much as the number.
Two friction conditions that can lead to failure during assembly
We are sometimes asked, "Did it break because friction was high and preload was low?" Higher friction actually gives less preload at the same torque. Two typical friction-related paths to failure during assembly have opposite causes; not every failure is one of the two.
| ① Friction lower than expected | ② Friction abnormally high, e.g. galling | |
|---|---|---|
| What happens | Preload rises at the same torque, and the combination of tension and torsion during tightening can exceed the permissible level | The torque–preload relation assuming a constant friction coefficient no longer applies; further tightening can cause torsional failure or thread damage |
| Typical cause | Lubricant or rust-preventive oil, change of finish — the left region of the graph | Galling (seizure) of stainless threads |
| Feel when tightening | Tightens normally, then fails | Suddenly stiff or binding part way |
| After failure | Signs of the bolt stretching | Torn or seized threads |
Higher friction within the defined range is already in the calculation: the maximum allowable torque includes the case where high thread friction raises torsion. Galling lies outside the calculation's assumptions, so it cannot be prevented by calculation; galling itself has to be prevented. The fracture surface gives some indication, but an accurate judgement needs the fracture surface itself.
What we check with you
When we hear "it broke at the specified torque", we look at both the product and the assembly conditions. On the assembly side:
- Has the lubrication or surface changed? (①) — lubricant, rust-preventive oil, or a change of finish or lot changes friction.
- Has the tool changed? (①) — power and impact tools can apply more torque than set; do not reuse a torque set for a manual torque wrench.
- Is the torque from another property class? (①) — an 8.8 torque table applied to A2-70 or 4.8 bolts can overshoot. For the same M5 pan head, the calculator's set torque for A2-70 (dry) and 4.8 (electro-zinc, unlubricated) differs by about two times.
- Have stainless threads galled? (②) — stainless-on-stainless or fast tightening invites galling (more on galling).
- Is it a countersunk head? — the head geometry does not guarantee that the head carries as much as the shank (KS B 0233 scope).
So torque is set as a range
Friction is a range, not a single value. Dry stainless is VDI 2230 friction class E (0.30 or more) and scatters widely. Class E gives only the lower limit of 0.30; the upper limit of 0.50 in this example is an assumption before measurement, and if actual friction exceeds it, preload can fall below the calculated minimum. Tightening torque is therefore best read as three numbers. Below are results from the tightening torque and preload calculator; unlike the example above, they include the pan head dimension range (dk minimum and basic), independent thread and under-head friction (each 0.30–0.50) and ±6% torque scatter.
| Number | Value | Meaning |
|---|---|---|
| Maximum allowable torque | 7.08 N·m | Allowable assembly torque at the most unfavourable combination in the assumed friction range |
| Set torque | 6.68 N·m | Reduced so that a high-side torque (±6%) stays under the maximum (about 68 kgf·cm) |
| Preload range | 1,937–3,629 N | Calculated preload range under the assumed conditions — about 1.9 times |
Even with the friction range and torque scatter assumed here, the same work instruction gives calculated preload differing by nearly two times — which is why controlling dry stainless joints is difficult. To calculate other conditions yourself, use the tightening torque and preload calculator. The relation between tightening, loosening and breaking torque is covered in Tightening, loosening and breaking torque.
The numbers above are example calculations from the VDI 2230 formula and the friction ranges of the reference table, not guaranteed values. The friction ranges are not measurements of our products. Confirm important joints with the actual parts and tools. Tell us your conditions and we will review them with you. Specification enquiry
FAQ
The bolt broke at the specified torque. What is the cause?
Both the product and the assembly conditions need checking. Two friction-related paths are typical. If lubricant, rust-preventive oil or a change of finish lowers friction, preload rises at the same torque and the combined tension and torsion during tightening can exceed the permissible level. If friction becomes abnormally high, as in galling of stainless threads, further tightening can cause torsional failure or thread damage. Power or impact tools, torque values from another property class and the head strength of countersunk screws are also worth checking.
Did it break because friction was high and preload was low?
Higher friction gives less preload at the same torque. But if thread friction becomes abnormally high, as in galling, the torque–preload relation for a constant friction coefficient no longer holds and further tightening can cause torsional failure or thread damage.
What percentage of torque actually creates preload?
About 7% in the example of an M5 pan head at a friction coefficient of 0.30. The remaining 93% goes into thread friction (about 40%) and under-head friction (about 53%). The split varies with size, head type and friction, but friction always takes most of it.
Can I keep the same torque after applying lubricant?
No. Lubrication lowers friction, so the same torque gives much more preload. Using a torque set for dry stainless on lubricated parts can push the bolt beyond the permissible assembly preload, so set the torque again for the lubricated condition.
Why do some bolts tightened to the same torque come loose?
Parts with higher friction develop less preload at the same torque. With the friction range assumed here (dry stainless 0.30–0.50) and torque scatter, calculated maximum and minimum preload differ by about 1.9 times, and more if actual friction exceeds the assumed range. The bolts with less preload loosen more easily under vibration.
References
- VDI 2230 Blatt 1:2015 Systematic calculation of highly stressed bolted joints — Joints with one cylindrical bolt — tightening torque terms (pitch, thread friction, head friction), permissible assembly preload, friction classes
- Würth Industrie, DINO technical part, ch. 6 — 6.3 friction class table (p.1790)
- ISO 3506-1:2020 Fasteners — Mechanical properties of corrosion-resistant stainless steel fasteners — Part 1: Bolts, screws and studs with specified grades and property classes — minimum proof strength of A2-70
- KS B 1023:2016 Cross-recessed machine screws, Annex Table A.1 — pan head diameter dk
- KS B 0233:2000 Mechanical properties of steel bolts and screws — scope: countersunk and similar heads vs tensile and torsional requirements
Last revised 2026-10-09