A torque wrench measures how hard you twisted. It says nothing about how hard the joint is clamping. Those two numbers are related, but loosely, and the gap between them is where most bolted joint failures begin.
Consider two M10 class 8.8 bolts taken from the same carton and tightened to 50 N·m with the same wrench. One may end up holding 25 kN of preload and the other close to 20 kN, purely because one had a drier thread or a rougher washer face. The steel did not change. The torque did not change. What changed was friction, and friction is what decides how much of your wrench reading survives as clamp force.
That is the whole argument for treating a torque table as an estimate rather than an instruction, and it is worth unpacking properly before anyone copies a torque value from a catalogue into a purchase specification.
Where the torque actually goes
When a bolt is tightened, the applied torque has to overcome three things: the inclined plane of the thread, the sliding friction on the thread flanks, and the sliding friction under the bearing face of the head or nut.
Take an M10 class 8.8 bolt, dry-ish assembly, tightened to 75% of its proof load, which is 25.2 kN. Working through the Kellermann–Klein relations that ISO 16047 is built on, with a bearing face diameter of 13.3 mm and a friction coefficient of 0.15 on both interfaces, the torque comes to about 51 N·m. Of that figure, only around 12% is doing the useful work of stretching the shank. Thread friction takes roughly 39%, and the under-head contact takes the remaining 49%.
Push the friction coefficient up to 0.20 and the picture shifts further: the same 25.2 kN now costs about 66 N·m, with bearing friction alone passing 50% of the total. Drop friction to 0.12, as a well-lubricated joint might show, and 42 N·m is enough.
| Friction coefficient μ | Torque for 25.2 kN (M10, 8.8) | Helix work | Thread friction | Under-head friction | Implied K |
| 0.12 | 41.9 N·m | 14% | 38% | 48% | 0.166 |
| 0.15 | 50.9 N·m | 12% | 39% | 49% | 0.202 |
| 0.20 | 65.9 N·m | 9% | 40% | 51% | 0.261 |
Read the last column first. Nothing about the bolt changed across those three rows: same diameter, same property class, same proof load. Only the surface condition changed, and it moved the torque needed by more than 50%. Any table that lists a single torque for an M10 bolt without naming the friction condition is really listing one possible answer.
K is measured, not assumed
The torque coefficient K condenses all of this into one number through the familiar relation T = K · F · d. ISO 16047 defines K as T/(F·d) and sets out how to determine it: the clamp force at the measurement point is normally 75% of the proof load of the part under test, unless something else is agreed.
Two limits in that standard matter commercially. K is only valid for the friction condition, the diameter and the geometric proportions it was measured on. It does not scale from M10 to M24, and it certainly does not transfer from an electroplated bolt to a hot-dip galvanized one. A supplier who quotes K without saying what it was measured on has quoted a number, not a property.
| Surface condition | Typical K range | Note |
| Plain, lightly oiled | 0.14 – 0.18 | Depends heavily on mill scale and storage |
| Electroplated zinc, dry as supplied | 0.18 – 0.22 | Most catalogue tables assume this |
| Hot-dip galvanized (ISO 10684) | 0.22 – 0.30 | Rougher, thicker coating; also needs clearance allowance |
| PTFE or MoS₂-based coating | 0.10 – 0.16 | Coater-specific; ask for test data per lot |
| Zinc-flake (Dacromet-type) | 0.14 – 0.20 | Process variant and lubricant additive both matter |
Those ranges are field and supplier values, not standardised constants. They are useful for budgeting and hopeless as a basis for a safety-critical joint calculation. If the joint matters, the K value belongs in a test report from an ISO 16047 rig, run on your coating and your lubrication, at your diameter.
Preload has a ceiling, and it is set by the standard
Behind every torque figure sits a permissible clamp force. For metric steel bolts that ceiling comes from ISO 898-1, which fixes minimum tensile strength, nominal proof stress and the stress area the load is spread over.
The stress area is not the shank area. ISO 898-1 defines it on the mean of the pitch and minor diameters, which gives A_s = (π/4)(d − 0.938194P)². For M10 coarse that is 57.99 mm², against 78.5 mm² of plain shank, so roughly three-quarters of the shank cross-section carries the load. Proof load is then A_s multiplied by the nominal proof stress: 580 MPa for class 8.8 up to M16, 830 MPa for 10.9, 970 MPa for 12.9.
Two utilisation levels dominate practice. Reusable joints are commonly tightened to about 75% of proof load, which leaves margin for external load and for a second or third assembly. Joints that will not be opened, and where the design has been checked for it, sometimes go to 90%. Going beyond that is a design decision, not a tightening decision, and it needs the joint calculation to back it.
| Size | A_s (mm²) | Proof load 8.8 | Torque at 0.75 F_p | Proof load 10.9 | Torque at 0.75 F_p | Proof load 12.9 | Torque at 0.75 F_p |
| M6 | 20.1 | 11.7 kN | 11 N·m | 16.7 kN | 15 N·m | 19.5 kN | 18 N·m |
| M8 | 36.6 | 21.2 kN | 25 N·m | 30.4 kN | 36 N·m | 35.5 kN | 43 N·m |
| M10 | 58.0 | 33.6 kN | 50 N·m | 48.1 kN | 72 N·m | 56.3 kN | 84 N·m |
| M12 | 84.3 | 48.9 kN | 88 N·m | 69.9 kN | 126 N·m | 81.7 kN | 147 N·m |
| M16 | 156.7 | 90.9 kN | 218 N·m | 130.0 kN | 312 N·m | 152.0 kN | 365 N·m |
| M20 | 244.8 | 146.9 kN | 441 N·m | 203.2 kN | 610 N·m | 237.5 kN | 712 N·m |
| M24 | 352.5 | 211.5 kN | 761 N·m | 292.6 kN | 1053 N·m | 341.9 kN | 1231 N·m |
Every torque figure in that table assumes K = 0.20, coarse thread, unlubricated assembly. Treat it as a starting point for a plated, dry joint and re-rate it for anything else. Note also that class 8.8 above M16 steps to a higher proof stress, 600 MPa rather than 580, because ISO 898-1 allows larger sizes a slightly different strength band.
Why the same torque gives different clamp loads
Set the wrench to 50 N·m on an M10 bolt and vary only K between 0.15 and 0.25. The resulting preload runs from 33.3 kN down to 20.0 kN, a ratio of 1.67. Hold the middle of that spread as the nominal value and the extremes sit about 25% away in each direction. That is the scatter that torque control lives with, and no amount of wrench calibration removes it, because the variation sits in the joint rather than in the tool.
Assembly speed adds its own contribution. A study published through ASTM in 2005 measured a single bolt, washer and nut combination tightened to a 28 N·m target across speeds from 25 to 700 rpm. Both thread and under-head friction fell as speed rose, and clamp load rose with them: the average difference between the slowest and fastest runs came to 6.3 kN with lubricant present and 1.2 kN unlubricated. On a small fastener that is a meaningful fraction of the target preload, produced by nothing more than how fast the nut runner turns.
Relaxation then takes a little back. Thread flanks and bearing faces embed as the joint settles, paint and coatings creep under load, and gasket materials compress. The loss is usually modest but it is never zero, which is why joints that are critical in service are often checked or re-torqued after a settling period rather than signed off at the wrench.
Methods that beat torque control
Where the preload actually matters, engineers stop relying on torque alone.
Torque-angle tightening drives the fastener to a snug torque that sets a known starting point, then rotates it through a calculated angle. Past yield the clamp force becomes insensitive to friction, so the scatter collapses. Turn-of-nut installation, familiar from structural bolting practice, is the same idea expressed in fractions of a turn and is far less dependent on a calibrated wrench than a torque value is.
Direct measurement is more precise still. Ultrasonic bolt meters read stretch from the change in sound transit time, and strain-gauged or instrumented bolts report load directly. Hydraulic tensioners skip torque altogether and pull the stud to the required load before the nut is run down, which is why they dominate on pressure vessel and large flange work.
Each method costs more at installation and saves more in service. The decision is economic: a bracket that holds a cable tray can live with torque control, and a bolted flange on a gas line cannot.
What belongs in the purchase specification
Torque tables get copied between projects far more often than they get calculated. Most of the failures that follow trace back to a specification that named a torque but not the conditions it assumed.
Four things close most of the gap. Name the coating and the lubrication state explicitly, since these set K. Ask for ISO 16047 test data with the lot on anything where preload is a design parameter, and state the diameter and friction condition the numbers apply to. Keep galvanized and plated parts on separate torque schedules, and remember that hot-dip galvanized fasteners need thread and clearance allowances under ISO 10684. And write down how the joint will actually be assembled, because a torque figure that assumes clean, dry threads means nothing if the crew greases them on site.
Shaanxi Flourish Industrial supplies hex bolts, studs, nuts and washers in classes 4.8 through 12.9, in carbon steel, alloy steel and stainless, with zinc plating, hot-dip galvanizing, PTFE and zinc-flake finishes. Each shipment carries EN 10204 3.1 or 3.2 material certificates, dimensional inspection reports and salt spray results, and hardness, tensile and yield testing is done in our own laboratory. Where a project needs independent confirmation, we arrange SGS, BV or TÜV inspection.
If you are building a torque schedule for a project, send us the joint details and the coating you intend to use. We will tell you what the fastener can carry, and what the test data behind that figure was measured on.
Standards and sources referenced: ISO 16047 (torque/clamp force testing, definitions of K and the 75% proof load measurement point), ISO 898-1 (property classes, proof stress, tensile stress area), ISO 10684 (hot-dip galvanized coatings). Torque and preload figures in the tables were calculated from the ISO 898-1 stress-area formula and the Kellermann–Klein torque relations, at stated friction coefficients and a bearing diameter of 13.3 mm for M10; they are worked examples, not catalogue values. The tightening-speed data is from "Effect of Tightening Speed on Thread and Under-Head Coefficient of Friction", ASTM Journal of ASTM International, Vol. 3, Issue 3, published online 21 December 2005 (DOI 10.1520/JAI13072).
