Maximum tightening torque for bolts and nuts
12 בJuly 2026
How Long Does a Chemical Anchor Take to Cure? Gel and Curing Times by Temperature
30 בAugust 2026
Pull-out Testing of Anchors with a Torque Wrench
Applying a known torque and checking that an anchor does not move is a fast, cheap field check — but it is a fault detector, not a measuring instrument. Here is the formula behind it, what it can prove, and what it cannot.
The test does not load an anchor to failure; it only confirms that the anchor survives a defined load. Used well, it catches gross installation faults. Used as proof of capacity, it misleads.
What the method actually does
Suppose you want to confirm that a particular anchor holds a pull-out load of P kilonewtons. There is no hydraulic puller on site, but there is a torque wrench.
The idea is simple. Tightening the anchor’s nut generates a tensile force in the bolt. Work out which torque produces a tensile force equal to P, apply it, and if the anchor does not move you have an indication that it holds at least P.
This is, in effect, an improvised proof load test: you are not pulling to failure, only confirming that the anchor survives a defined load.
The formula
The torque needed to reach a given pull-out load is T = K x D x P, where T is the torque in newton-metres, K is a friction coefficient, D is the bolt diameter in millimetres, and P is the tested pull-out load in kilonewtons. The units work out without any conversion: a diameter in millimetres times a load in kilonewtons gives a torque directly in newton-metres.
K expresses how much of the torque is absorbed by friction instead of becoming tension — the rougher the finish, the larger K, and the more torque is needed for the same force:
The K factor, by bolt finish
| Bolt finish | K (VDI 2230) | Notes |
|---|---|---|
| Electro-galvanized (cold) | 0.15 – 0.20 | The most common case; a reasonable default is 0.18 |
| Hot-dip galvanized | 0.20 – 0.30 | Thick, rough zinc layer; high and unpredictable friction |
| Stainless steel | 0.25 – 0.35 | Prone to galling; the highest and most variable K |
Calculator — what torque should you apply?
Common mid value: — N·m. Total uncertainty in the force: — — the K range combined with torque-wrench accuracy (±25%). Before applying the torque, check that it does not exceed the maximum torque permitted for the anchor by its ETA, or the maximum torque for the bolt in its strength class.
The stainless range alone spans 0.25 to 0.35 — a 40% spread at the same torque. That is the method’s main source of error.
A worked example
Take a chemical anchor with an electro-galvanized M12 threaded rod, to be checked against a pull-out load of 10 kN. Using K = 0.18: T = 0.18 x 12 x 10 = 21.6 N·m, so you apply about 22 N·m. If the anchor does not move, it passed — but consider what you really measured. If the true K were 0.15 (a smoother bolt), those same 22 N·m would produce about 12.2 kN, some 22% more than intended; if K were 0.20, the same torque would fall well short of 10 kN. The pass looks decisive, but the threshold behind it is blurred.
But what did we actually measure?
If the true K had been 0.15, because the bolt was smoother than assumed, the same 22 N·m would have produced 12.2 kN — 22 percent more than intended. At K = 0.20 you get 11.0 kN. And if the torque wrench itself read 25 percent low, the force actually applied could have been as little as 8.3 kN, meaning the anchor “passed” a test it was never really given.
This is the heart of the problem. The pass or fail result looks unambiguous, but the threshold behind it is blurred by roughly ±50 percent. The method does not give you a number — it gives you an indication.
Where the error comes from
| Source of error | Order of magnitude | Why |
|---|---|---|
| Estimating K | 15 – 20% | K depends on finish, lubrication, dirt, temperature and thread condition — the table gives a range, not a value |
| Torque-wrench accuracy | about 25% | A typical click wrench, not recently calibrated; even a new one is only a few percent at best |
| Friction under the nut | Variable | A dirty washer, an uneven concrete face or a warped plate all change it |
| Method of application | Variable | Jerky tightening, stopping and starting, or an off-axis pull all bias the reading |
The errors do not cancel — they accumulate. In practice the force applied to the anchor can deviate from the intended value by tens of percent in either direction, so combined accuracy is often around plus or minus 50%.
What it is good for — and what it is not
Suitable for
- In-process quality control, mainly on chemical (adhesive) anchors — this is its clearest use. The test catches an installation fault: adhesive that was not properly mixed, a hole that was not cleaned, a partial injection. A defective anchor fails well below the threshold, so even a rough test picks it up.
- A quick sample check on site — dozens of anchors an hour, with no special equipment.
- A quick check that a fixing was done at all, to find anchors that were forgotten or only hand-tightened.
Not suitable for
- Qualifying an anchor for an application — that needs a hydraulic pull tester, per EOTA TR 048 and ETAG 001.
- Structural acceptance or engineering approval — the approving authority will not accept torque-wrench results as evidence of capacity.
- Mechanical expansion anchors, especially wedge anchors — there the torque expands the anchor itself, so the formula does not apply; use the installation torque from the ETA.
- An anchor that already carries structural load — the torque you apply adds to the existing load.
How to do it properly — recommended procedure
- Set the test load P from the designer’s requirement — normally the working load or a multiple of it, not the full resistance.
- Choose K from the bolt’s actual finish, not from what was ordered. A lubricated bolt is not a dry bolt.
- Calculate T = K · D · P and cross-check it against the maximum torque permitted for the anchor (from the ETA) and for the bolt.
- Wait for the adhesive to reach full cure, according to the substrate temperature. Testing early will fail a perfectly sound anchor.
- Use a calibrated torque wrench, preferably digital. Apply the load in a slow, continuous movement, perpendicular to the axis.
- Set a sample size — usually 5%–10% of the anchors, and always the critical ones.
- Record everything: position, diameter, K, torque, pass or fail, date, and the name of the person testing.
- An anchor that fails is rejected. Increase the sample, and investigate the installation method.
Comparison with other test methods
| Method | Accuracy | Cost | Speed | What it suits |
|---|---|---|---|---|
| Torque wrench | Low (about 50%) | Low | Very fast | In-process QC; catching installation faults in chemical anchors |
| Hydraulic pull tester, to proof load | High (3-5%) | Medium | Medium | On-site approval, structural acceptance, testing to a defined load |
| Hydraulic pull tester, to failure | High | High | Slow | True capacity in an unknown substrate; suitability tests per EOTA TR 048 |
| On-Off-On method | Medium to good | Low | Medium | Estimating the real clamp force in bolt-nut joints, without knowing the friction |
Common questions
How many anchors should be tested?
There is no single answer. A common practice is 5 to 10% of the project’s anchors, and always every safety-critical anchor. If one fails, enlarge the sample. The designer decides.
An anchor failed the test — what now?
First check that the applied torque was within the allowable limits: you may simply have exceeded the allowable installation torque, or mis-estimated K. If it genuinely failed, reject it, test an enlarged sample, and investigate the installation method.
Why bother, if the accuracy is so low?
Because the faults it is meant to catch are gross. A chemical anchor with unmixed adhesive, or a hole full of dust, does not hold 20% less — it holds almost nothing. Even a test with 50% error catches that immediately. The method is exactly as accurate as its job requires, and no more.
Can the method be used on a wedge anchor (MTH / MTP)?
Not as a pull-out test. In an expansion anchor the torque drives the expansion mechanism itself, so the relation between torque and pull-out force is neither linear nor predictable. Anchors of that kind have an installation torque defined in their ETA: apply it, do not test with it.
What is the difference between K here and the 0.20 used in the torque tables?
It is exactly the same coefficient. The maximum-torque tables assume K = 0.20 as a reference value. Here we acknowledge that K is a range rather than a single number, and set out that range by bolt finish.
