The torque from the table says little about the preload
Where the torque actually goes
A bolted joint does not hold because of tightening torque. It holds because of the preload that presses the parts together. Torque is merely a way of producing that force, and a fairly poor one.
The tightening torque splits into three parts: the thread pitch, friction on the thread flanks, and friction under the bolt head.
M_A = F_M · ( 0.16 · P + 0.58 · d2 · µ_thread + D_Km/2 · µ_head )
P pitch d2 pitch diameter
µ coefficients of friction in thread and under the head
D_Km effective friction diameter under the head
For M10 with µ = 0.14 in both interfaces the three bracket terms come to roughly 0.24, 0.73 and 0.90 millimetres. Only the first one creates preload. Put differently: about 13 percent of the torque tensions the joint, nearly 40 percent goes into the thread and almost half under the head.
So the decisive quantity depends on the two numbers you know least well. A coefficient of friction is not a material constant; it is the result of surface, coating, oil, number of tightening cycles and tightening speed.
What friction does to the preload
Two calculations for the same part, M10 property class 8.8. First the view from the table: which torque belongs to which coefficient of friction if the yield strength is to be used to 90 percent.
| Coefficient µ | Tightening torque | Preload |
|---|---|---|
| 0.08 (MoS₂, wax) | 34.8 Nm | 29,530 N |
| 0.10 (oiled) | 40.3 Nm | 28,480 N |
| 0.12 (lightly oiled) | 45.2 Nm | 27,410 N |
| 0.14 (plain, reference value) | 49.6 Nm | 26,330 N |
| 0.16 (dry, degreased) | 53.5 Nm | 25,270 N |
| 0.20 (zinc plated, dry) | 60.1 Nm | 23,230 N |
Figures from the reference page for M10 8.8, calculated to VDI 2230.
This table is usually read the wrong way round. It answers the question: which torque do I need if I know the friction. On the shop floor the question is the other way round. There the torque is given and the friction is whatever happens to be on the bolt.
Always tighten to 49.6 Nm, the table value for µ = 0.14, and this is what happens:
| Actual coefficient | Preload at 49.6 Nm | Assessment |
|---|---|---|
| 0.08 | about 42,000 N | above the yield load of roughly 37,000 N, the bolt yields |
| 0.10 | about 35,000 N | well above design intent, plastic deformation possible |
| 0.14 | 26,330 N | the design case |
| 0.16 | about 23,400 N | 11 percent below design intent |
| 0.20 | about 19,200 N | 27 percent below design intent |
Scaled linearly from the same series. The yield load follows from R_p0.2 = 640 MPa and A_S = 58 mm².
A factor of two between the best and the worst case, with the fitter doing exactly the same thing. That is why a torque figure without a friction figure is incomplete.
The tightening factor makes the scatter visible
VDI 2230 wraps this uncertainty into the tightening factor α_A, the ratio of maximum to minimum assembly preload. The design has to survive both: the part must not be overloaded at the upper value, and the joint must still hold at the lower one.
| Tightening method | α_A | What that means in practice |
|---|---|---|
| Impact wrench, uncontrolled | 2.5 … 4 | the lower preload is a quarter of the upper one |
| Torque wrench, friction estimated | 1.6 … 2.0 | the normal case in assembly and service |
| Torque wrench, friction measured | 1.4 … 1.6 | worth it from medium volumes upwards |
| Angle-controlled tightening | 1.2 … 1.4 | the yield point is passed on purpose |
| Yield-controlled tightening | 1.1 … 1.2 | the tool detects the knee in the curve |
| Bolt elongation measurement | 1.1 … 1.2 | ultrasound or a measuring pin, a special case |
Guide values to VDI 2230. The exact classification depends on lubrication and setting accuracy.
The impact wrench row is the interesting one, because it happens in nearly every workshop. Design a joint to survive α_A = 4 and you need a bolt that can take four times the required force in the favourable case without failing. Changing the tightening method is usually cheaper than changing the bolt.
Embedding: the loss after assembly
After tightening, the preload drops without anyone touching the joint. The roughness peaks in the interfaces, under the head and in the thread flanks are flattened. This is not creep and not loosening, but plastic deformation in the micrometre range, largely finished within the first hours of operation.
For an estimate you count an embedding amount per interface. Typical magnitudes at medium roughness are about 3 µm per thread interface, 3 µm per head or nut face and 2 to 3 µm per internal joint face. A stack of two plates quickly reaches 10 to 12 µm.
Whether that is a lot depends on resilience. A short, stiff bolt loses a substantial share of its preload to 10 µm of embedding; a long reduced-shank bolt barely notices. That is why waisted bolts and long clamping lengths are not a fad but the only effective answer to embedding. The second route is re-tightening, which only helps if the joint stays accessible.
What belongs on the drawing
A torque figure on its own cannot be executed. It only becomes complete with three statements.
- The torque with a tolerance, for example 50 Nm ± 10 percent. Without a tolerance it is unclear whether an impact wrench is acceptable.
- The coefficient of friction assumed, for example µ_total = 0.14. That is the value the torque belongs to, and the only way for assembly to notice a different lubrication state.
- The lubrication state, meaning plain, lightly oiled, secured with adhesive or assembled dry. A drop of oil on a bolt that was calculated dry raises the preload by around thirty percent.
With thread-locking compounds there is the added point that the adhesive itself changes the friction, and depending on the product it can go either way. The manufacturer data sheets give their own values, and those replace the table value rather than supplementing it.
When torque is the wrong quantity altogether
For safety-relevant joints, for short clamping lengths and anywhere the scatter gets expensive, another method takes over. Angle-controlled tightening runs up to a defined snug torque and then turns on by a fixed angle. Because that angle converts directly into bolt elongation, the resulting force hardly depends on friction any more. The price is that the bolt deliberately passes its yield point and must not be reused.
On large joints in plant engineering the elongation is measured directly instead, by ultrasound or with a measuring pin along the bolt axis. That is accurate but laborious, and only pays off where the consequences are correspondingly expensive.
For everything below that the torque wrench remains the tool of choice. You just have to know that it sets a force you are not measuring.
Frequently asked questions
May I oil a bolt that has a table torque?
Only if the table value applies to the same lubrication state. Common table values refer to µ = 0.14, meaning plain and lightly oiled. Add grease on top and you push the friction towards 0.10 and produce a considerably higher preload at the same torque. On high-strength bolts that can pass the yield point.
Why do torque tables from different manufacturers disagree?
Because they make different assumptions: 90 or 70 percent utilisation of the yield strength, friction 0.12 or 0.14, torsion during tightening included or not, different bearing diameters under the head. Each table is consistent in itself. They can only be compared when those boundary conditions are stated with them.
How often may a high-strength bolt be reused?
With torque-controlled tightening in the elastic range, several times, provided the thread and the bearing face are undamaged and it is re-lubricated. The friction does rise with every cycle, because the surfaces roughen. After angle-controlled or yield-controlled tightening the bolt is plastically deformed and should be replaced.
What is the minimum thread engagement and why does it depend on the material?
It is the thread length beyond which the bolt breaks before the female thread strips. Because that is a question of the shear strength of the weaker partner, it depends on the material of the female thread. In quenched and tempered steel about 1.2 times the diameter is enough; in aluminium the same bolt needs a good twice that.