Threaded fasteners and torque

A bolt is not a pin and it is not a rivet. It is a spring that has been deliberately stretched, and everything useful about a bolted joint comes from that stretch. Get it wrong and the joint either loosens over months or fails on the first load — and in both cases the bolt will look perfectly fine.

An assembly operator holding a DC electric torque tool on a bolt in an engine mounting bracket, the tool's LED status ring lit green, with a rack of graded bolts and a torque transducer and its readout on the bench alongside. — The tool reports that it delivered the torque it was asked for. What that became inside the joint is a separate question.

What the bolt is actually doing

In a well designed joint the bolt does not carry the load. It clamps the two faces together hard enough that friction between those faces carries it.

A diagram of a bolted joint as a spring system. A bolt through two clamped plates is shown in section, with the bolt drawn as a tension spring stretched between head and nut and the clamped material drawn as a compression spring around it. Arrows show the clamp force at the interface and a shear load being carried by friction between the faces rather than by the bolt shank. Beside it, a joint diagram plots force against deflection, with the bolt line rising and the joint line descending, meeting at the preload point, and an external load shown dividing between an increase in bolt tension and a reduction in clamp force. A second panel compares a short stubby bolt and a long slender bolt of the same diameter, each losing the same small amount of length to embedment, with the short bolt losing a large fraction of its preload and the long bolt a small fraction. — Bolt and joint are two springs in opposition. A long bolt is a softer spring, which is why it keeps its preload when a short one has lost it.

That has three consequences that catch people out:

Preload matters more than bolt strength in most joints. A strong bolt at low preload gives a joint that moves.
An external load does not all go into the bolt. It divides between raising bolt tension and reducing clamp force, in proportion to the two stiffnesses.
A longer bolt is a better bolt, other things equal. It is a softer spring, so the small permanent losses — embedment, gasket creep, paint settling — eat a smaller fraction of the stretch. Short stubby bolts are the ones that go slack.

Property classes

The marking on the head is not a grade in a vague sense. It is two numbers with exact definitions, set by ISO 898-1.

Class	Tensile strength	Yield	Typical use
8.8	800 MPa	640 MPa (80 %)	General structural and chassis work
10.9	1000 MPa	900 MPa (90 %)	Suspension, engine mounts, brackets
12.9	1200 MPa	1080 MPa (90 %)	Highly stressed, compact joints

The first number is the nominal tensile strength in hundreds of megapascals. The second is the yield-to-tensile ratio in tenths — so 10.9 yields at 90 % of 1000 MPa.

Where the torque goes

This is the part that surprises everyone the first time they see the numbers.

A torque budget diagram. A horizontal bar representing the total applied torque is divided into three parts, roughly half labelled as friction under the bolt head bearing face, about a third labelled as friction in the threads, and a small remaining portion labelled as useful work stretching the bolt. Beside it, a scatter band shows achieved preload against applied torque for three conditions, dry, oiled and zinc flake coated, with the three bands widely separated and each band itself broad, illustrating that the same torque gives very different clamp loads. A third panel shows a torque against angle curve, rising steeply through the snug region, then a long straight elastic slope, then flattening into a yield plateau, with marks indicating a torque control target, a torque plus angle target and a yield control point at the start of the plateau. — Only a small slice of the applied torque becomes clamp load. The rest is spent overcoming friction that nobody measured.

Typically about half the applied torque is consumed by friction under the bearing face, about a third in the threads, and only the remaining 10 to 15 per cent actually stretches the bolt.

Three ways to control the tightening

Method	What it controls	Scatter in preload	Where used
Torque	Applied torque only	Wide, ±25 % or worse	Non-critical joints
Torque + angle	Rotation past a snug point	Much tighter	Safety-related joints
Torque to yield	The bolt’s own yield plateau	Tightest	Head bolts, main bearings

Angle control works because once the joint is snug, further rotation corresponds to bolt elongation through the thread pitch — geometry, not friction. Measuring the angle measures the stretch, more or less directly.

Torque to yield goes one step further and tightens into the plateau where the bolt starts to yield. The bolt itself now sets the limit, so friction barely matters. The bolt has been permanently stretched, which is exactly why cylinder head and main bearing bolts are replaced, never reused.

Threads in aluminium

A nut can be chosen to outmatch the bolt. A tapped hole cannot — it is whatever the parent material happens to be.

In steel, thread engagement of roughly 1 × diameter is usually enough.
In aluminium, roughly 2 × diameter, because the material is weaker.
Where the depth is not available, a steel insert carries the thread instead.

Applying the steel torque figure to a tapped aluminium hole strips it. The weaker material always sets the limit, and the fastener catalogue does not know what it has been screwed into. The same logic governs the fits and engagement lengths in fits, limits and clearances.

Practical points that decide whether a joint holds

Hardened washers spread the bearing load and give a consistent friction surface. Substituting a plain washer changes both.
Clean, undamaged threads. A damaged thread reads as high torque long before the joint is tight, so the tool signs off a joint with no preload in it.
Tightening sequence on multi-bolt joints — cross pattern, in stages — because each bolt relieves its neighbours as the joint settles.
Grip length, so the plain shank rather than the thread sits in the shear plane where shear is unavoidable.

What happens to that preload over the following months is a separate subject, covered in bolted joint loosening and locking.

What to take away

A bolt is a stretched spring; the clamp force it generates is what carries the load.
An external load splits between bolt tension and lost clamp force.
A longer bolt loses proportionally less preload to settlement.
8.8, 10.9, 12.9 — tensile strength in hundreds of MPa, then yield ratio in tenths.
Nut class is chosen so the bolt breaks before the thread strips.
Only 10 to 15 % of applied torque becomes preload; friction takes the rest.
A torque figure without a friction condition is an incomplete specification.
Angle control measures stretch through the thread pitch, so friction stops mattering.
Torque to yield bolts are single use. Replace them.
Tapped aluminium needs about twice the engagement of steel, or an insert.