§1A bolt is a spring
Tightening a bolt stretches it. The stretched bolt pulls the joint faces together with a force called the preload, and that clamping force — not the bolt’s shank — is what actually holds the joint.
This is the idea the rest of the page rests on. A bolt is not a rivet or a pin that resists load in shear; it is an elastic member deliberately stretched so it acts as a stiff spring in permanent tension, squeezing the joint together. The stretch is tiny but real: an M10 bolt clamping a 50 mm grip at its proper preload of 27.8 kN extends by δ = FL/(AE) = 27 835 × 50 ÷ (58.0 × 200 000) = 0.12 mm — about the thickness of a sheet of paper. That hair of stretch is the entire clamping mechanism. Everything that follows — why preload matters (§2), why torque is a poor way to set it (§3–4), and how it is better set (§6) — is about controlling that 0.12 mm.
Contents§2Why preload matters
Preload is not a nicety — an under-tightened bolt is a fatigue failure and a loosening waiting to happen, and joint integrity depends on it being high, not moderate.
A properly preloaded joint gains three things. It resists fatigue: because the clamped members are already squeezed hard, an external tensile load mostly relieves the joint’s compression rather than adding to the bolt’s tension, so the bolt sees only a fraction of the fluctuating load — and, as the fatigue pages show, it is the alternating stress that kills. It resists loosening: high preload keeps enough thread and under-head friction to stop vibration walking the nut off. And it resists slip and separation: the clamped friction between the faces carries shear load, and the faces cannot gap and pump. Lose the preload and all three go at once — which is why the counter-intuitive rule holds that most bolted joints fail from too little tightening, not too much, and why the preload target is deliberately set high (§5).
Contents§3The torque–tension relation
Torque is used to set preload because it is easy to measure — through a simple relation whose weak link is a friction factor that is never quite known.
The nut factor K bundles all the friction and thread geometry into one number: roughly 0.20 for plain steel fasteners as-received, and about 0.15 lubricated. To reach a 27.8 kN preload in a plain M10, T = 0.20 × 0.010 × 27 835 = 55.7 N·m. Lubricate the same bolt and K falls to 0.15, so the same preload now needs only 0.15 × 0.010 × 27 835 = 41.8 N·m — a quarter less torque for identical clamping. The relation is simple, but K is not a constant: it varies with finish, plating, lubrication, surface condition, reuse and even speed of tightening, and a scatter of ±25% in K means ±25% in the preload you actually achieve. Torque is a proxy, and a loose one — which is exactly the trap of §4.
§4Where the torque actually goes
Almost all the effort of tightening is spent overcoming friction. Only a small fraction of the torque does the useful work of stretching the bolt — which is why torque is such an indirect measure of preload.
Per turn, the torque does work T × 2π while the bolt advances only one pitch p against the preload — so the useful share is p/(2πKD). For an M10 × 1.5 at K = 0.20: 1.5 ÷ (2π × 0.20 × 10) = 11.9%. Barely one-eighth of the work goes into stretching the bolt; the other 88% is burned as friction — roughly half under the turning head or nut face, and the rest in the threads. Lubricating raises the useful share to 15.9%, which is precisely why less torque then achieves the same preload (§3). The lesson is structural, not incidental: torque is mostly a measurement of friction, and preload is the small residue. A method that sidesteps friction altogether will always be more accurate (§6).
§5Setting the preload
The preload target is a high fraction of the bolt’s own strength — conventionally about 75% of its proof load — deliberately close to yield, because a slack bolt is the greater danger.
An M10 × 1.5 has a tensile stress area of 58.0 mm², and a property class 8.8 bolt yields at 640 N/mm² (§ the metric fasteners page), so its yield load is 640 × 58.0 = 37.1 kN. The standard target of 75% gives a preload of 27.8 kN — the figure used throughout this page, reached at 55.7 N·m dry. Two things are worth seeing in that number. First, the bolt is deliberately tightened to three-quarters of yield: the design intent is a hard-stretched spring, not a gentle nip. Second, the remaining quarter is the margin that absorbs the scatter in K (§3) and any external load — and it is not generous, which is why over-torquing a lubricated bolt (the §3 note) lands exactly on yield. Preload high, but know your friction.
§6Better than torque
Because torque is mostly friction, the accurate methods measure something closer to the stretch itself — angle, elongation or tension directly.
| Method | How it works | Preload scatter |
|---|---|---|
| Torque | measure the turning moment; rely on K | wide (±25% or worse) |
| Torque + angle | snug, then turn a set angle — angle ≈ stretch | much better |
| Bolt stretch | measure elongation (micrometer or ultrasonic) | very good |
| Tensioner / load-indicating | stretch hydraulically, or read an indicating washer | best |
| Each step down the list removes more of the friction guesswork. Angle control works because past the snug point a turn of the nut advances it by a known pitch, so the angle turned is the stretch — friction cancels out; it is why critical engine fasteners are specified as “torque to X, then Y degrees”. Stretch measurement reads the 0.12 mm of §1 directly. Hydraulic tensioners pull the bolt to load and run the nut down slack, bypassing torque entirely. Where preload really matters — heads, flanges, structural connections — the method rises up this table. | ||
§7Quick reference
The working core of the page on one card rack.
The point
preload = clamping force
M10 stretches ~0.12 mm
Relation
T = K · D · F
K ≈ 0.20 plain · 0.15 lubed
Efficiency
only ~12% → tension
~88% lost to friction
Target
~75% of proof
M10 8.8 → 27.8 kN @ 55.7 N·m
Trap
dry torque + oiled bolt
→ 37.1 kN = yield