embedment relaxation recovery
Embedment relaxation and how to recover from it
By ISOKLAMP Engineering, Inc. Editorial Team · Updated
The loss that happens before the machine ships
Tighten a joint to spec on the assembly line. Run it for an afternoon. Measure it again. It is looser, the nut has not moved, and nothing is wrong with your torque procedure.
That is embedment relaxation, and it is the fastest-acting of the non-rotational loss mechanisms. It is also the most predictable, which makes it the easiest to design against once you accept that it will happen.
Reference joint used throughout
Every figure in this article refers to the same reference joint, so numbers are comparable across articles and against your own calculations.
| Parameter | Value |
|---|---|
| Bolt | M16 × 2,0, property class 10.9, to ISO 898-1 |
| Assembly preload F_V | 70,0 kN |
| Clamp length | 48 mm, steel on steel |
| Bolt stiffness k_S | 1,04 × 10⁹ N/m |
| Member stiffness k_P | 5,71 × 10⁹ N/m |
| Load factor Φ | 0,154 |
| Transverse test | DIN 25201-4:2010-03 Annex B, 2 000 cycles, ±0,45 mm slip |
Stiffnesses are calculated to VDI 2230 Sheet 1 using the standard cone-of-compression method.
What is physically happening
No machined surface is flat. Every face carries asperities, peaks at the scale of the surface roughness. When you tighten a bolt, the entire clamp load initially bears on the tips of those peaks, so local contact stress vastly exceeds the nominal bearing stress. The peaks yield plastically and flatten until enough contact area has developed to carry the load elastically.
Each flattened interface makes the stack shorter. The bolt relaxes accordingly.
Per-interface budgets
A bolted joint has more interfaces than people count. An M16 bolt through two plates with a washer under the head and one under the nut has six load-bearing interfaces.
| Interface condition | Embedment per interface |
|---|---|
| Ground or lapped steel, Ra < 0,4 µm | 1 to 3 µm |
| Machined steel, Ra 0,8 to 1,6 µm | 3 to 8 µm |
| As-rolled or shot-blasted steel | 6 to 15 µm |
| Zinc-flake or PTFE coated | 8 to 20 µm |
| Stainless against stainless | 5 to 12 µm |
| Cast iron | 8 to 18 µm |
| Aluminium members | 10 to 25 µm |
Sum across every interface in the load path.
Worked example. M16 class 10.9 through two as-rolled steel plates, plain washers under head and nut:
| Interface | Budget |
|---|---|
| Under bolt head to washer | 4 µm |
| Washer to plate 1 | 8 µm |
| Plate 1 to plate 2 (faying) | 10 µm |
| Plate 2 to washer | 8 µm |
| Washer to nut face | 4 µm |
| Thread flanks | 5 µm |
| Total | 39 µm |
At 0,884 kN per micrometre for the reference joint, 39 µm costs 34,5 kN. The joint arrives at service at 35,5 kN against an assembly preload of 70,0 kN. That is 50,7 percent residual, and it fails the DIN 25201-4 Annex B criterion of 80 percent before a single hour of duty.
This is not an unusual joint. It is an ordinary one.
Why the standard answers only defer the problem
Retorquing after run-in is the textbook answer and it does work, once. Retorque the joint after the first shift and you restore the preload. The difficulty is that retorquing re-establishes contact on fresh asperity geometry and starts a second, smaller embedment cycle, and that on a plant with tens of thousands of joints the labour is the dominant cost rather than the parts.
Hardened washers raise the bearing surface hardness and reduce embedment at two of the six interfaces. Real and worth doing. It does not touch the faying surfaces or the thread flanks.
Better surface finish works and is the most effective preventive measure available. Going from as-rolled to machined can halve the budget. It also costs money at every interface and is often not available on castings, weldments or bought-in components.
Longer bolts reduce k_S so each micrometre costs less clamp force. Good engineering, rarely enough on its own, and usually no room.
None of these recover the loss. They reduce it or defer it.
Recovery instead of prevention
The alternative is to accept the embedment and put the length back as it happens.
This is what a helical take-up does. The ISK-16 carries 0,50 mm of reserve against a 39 µm budget, a reserve factor of 12,8. Because the take-up is actuated by micro-slip and embedment is fastest exactly when the joint is first loaded, the mechanism is most active precisely when the loss is fastest.
The predicted curve for the reference joint shows the interaction clearly:
| Cycles | ISOKLAMP CFR | Wedge-lock pair | Plain washer |
|---|---|---|---|
| 0 | 100,0 % | 100,0 % | 100,0 % |
| 50 | 97,8 % | 96,4 % | 88,1 % |
| 120 | 96,9 % | 94,8 % | 71,3 % |
| 300 | 98,1 % | 94,0 % | 48,6 % |
| 800 | 99,1 % | 93,5 % | 26,4 % |
| 2 000 | 99,4 % | 93,1 % | 12,8 % |
ISOKLAMP figures are predicted by finite-element analysis of the ISK-16 geometry coupled to VDI 2230 Sheet 1 mechanics with a plastic asperity model. Baselines are calibrated against published data, sourced on the test data page.
The wedge-lock washer curve settles at 93,1 percent and stays there. It arrested rotation successfully, which is what it is for, and it lost 6,9 percent to embedment that it has no means of recovering. The CFR curve dips to 96,9 percent while embedment outruns take-up, then climbs as the embedment rate collapses and the take-up rate does not.
Designing with an embedment budget
- List every interface in the load path, including thread flanks.
- Assign a budget per interface from the table above, using the worse of the two mating surfaces.
- Sum.
- Multiply by your joint's clamp force loss per micrometre, from Φ = k_S/(k_S + k_P).
- Subtract from assembly preload.
- Check the result against the minimum clamp force your design requires, not against the assembly preload.
Step 6 is where most joint designs are actually decided, and where most of them are accidentally already failing.
Full method comparison on locking methods compared. Related mechanisms in preload loss. Derivation in the technical report.
Frequently asked questions
What is embedment relaxation in a bolted joint?
Embedment relaxation is the plastic flattening of surface asperities at the faying, underhead, nut-bearing and thread interfaces of a bolted joint. Local contact stress on the asperity tips greatly exceeds nominal bearing stress at assembly, so the peaks yield until enough contact area develops to carry the load elastically. Each flattened interface shortens the stack and the bolt relaxes.
How much preload does embedment cost?
Between 3 and 15 micrometres per interface depending on surface condition, across typically six interfaces. A representative M16 joint through as-rolled steel plates budgets 39 micrometres in total, which at 0,884 kN per micrometre costs 34,5 kN from a 70 kN assembly preload. That is 50,7 percent residual before any service duty, well below the DIN 25201-4 Annex B criterion of 80 percent.
How long does embedment relaxation take?
It is largely complete within the first 200 load cycles, which is why joints measured after a single shift often read substantially lower than they did at assembly. The rate collapses after that, which is exactly why a take-up mechanism can overtake it: embedment slows, take-up does not.
Does retorquing fix embedment relaxation?
It restores the preload once. Retorquing re-establishes contact on fresh asperity geometry and starts a second, smaller embedment cycle, so it defers rather than eliminates. On a plant with tens of thousands of joints the recurring labour cost usually dominates the parts cost, which is the practical argument for designing the loss out instead.
Can embedment loss be recovered rather than prevented?
Yes. A helical take-up carrying a reserve of clamp length restores the lost length as embedment occurs. The ISK-16 carries 0,50 mm against a typical 39 micrometre budget, a reserve factor of 12,8. Because the take-up is actuated by micro-slip and embedment is fastest when the joint is first loaded, the mechanism is most active exactly when the loss is fastest.
Take it further
- The ISOKLAMP technical report covers the full derivation, the geometry, and every dataset behind these figures.
- Full residual clamp force dataset gives the residual clamp force numbers for ten securing methods, with sources.
- Contact sales puts you in touch with the engineering team. Bring us a joint that keeps failing and we will run a VDI 2230 Sheet 1 analysis on it.
Engineering questions go to engineering@isoklamp.com. An engineer answers, not a form.