Calculating Bolt Preload and Clamp Force: Torque, Friction and Tightening Method

Why a torque figure is mostly a friction specification

When you tighten a bolt, the rotary motion of the head is converted into linear stretch of the bolt and compression of the parts. Only a small share of the input torque does that useful work. Torque/tensile testing shows the breakdown: about 10% of the input torque becomes preload, roughly 50% is consumed under the head or nut, and about 40% goes into the threads ¹. That split is why a bare torque number is a weak way to specify a joint. Most of what you are setting is a friction condition, not a clamp force. Friction coefficients themselves vary by around ±20%. Even a calibrated hand torque wrench accurate to ±1% ends up applying torque that deviates by ±10 to ±50% ⁴. Lubrication is the lever on that friction. A low-friction lubricant reduces friction at both interfaces and lowers the torque needed to reach the same preload ¹. The reverse is the trap: a well-lubricated zinc flake coating shifts the coefficient of friction at the thread interface. Dry-steel torque values applied to that fastener over-tension the joint ¹⁵.

From clamp force to torque: the T = K·D·P relationship

The simplified relationship between torque and preload is T = K·D·P, where T is tightening torque, K is the nut factor, D is the nominal diameter and P is the clamping load ⁸. ISO 16047 defines the same quantity as the torque coefficient, K = T/(F·d). It standardises how clamp force, torque coefficient, total friction, thread friction and bearing friction are measured, with test machine accuracy of ±2% ²³. K is where the thread friction and the underhead bearing friction are collapsed into one number. That is why the same bolt needs a different torque under a different lubrication condition: the geometry term D·P is fixed by the joint, and K carries the friction. A supplier-published worked example uses K = 0.20 nominal with bounds of 0.18 and 0.22 ¹⁷. Treat K as a design input you must justify, not a constant. The worked example is explicit that the K range, wrench tolerance, relaxation, proof strength and external load all remain inputs that have to match the real joint and assembly process ¹⁷. If you cannot source a K for your exact coating and lubricant, the honest route is a torque/tensile test on the production combination rather than a table value.

Setting the target clamp force as a fraction of proof load

The target preload is normally expressed as a percentage of the fastener's proof load. The percentage is chosen by joint type rather than by bolt size. One source gives 75% of proof load for reusable joints and 90% for permanent joints, with clamping load deviation held to ±3% ⁸. Another states that most fasteners need torque to reach 65–80% of proof load ¹⁶. A worked M12 example uses 70% ¹⁷. Read those together and the practical band is 65–90%, with 70–75% as a common nominal target and 90% reserved for joints that will not be reopened ^{8, 16, 17}. The sources do not settle a single mandatory percentage; the value depends on joint type, loading and the applicable standard ^{8, 16, 17}. Start from the proof load, not the yield strength. Where a proof load value is not given, one source calculates it as 92% of yield strength ⁸. The proof load is the load the fastener can carry without permanent set. That is why it is the reference for a preload target rather than the ultimate strength.

Worked examples: M12 and M6

For an M12 × 1.75 thread with an assumed proof strength of 600 MPa, the tensile stress area works out to 84.266 mm², giving a proof load of 50.560 kN. At a 70% target, nominal preload is 35.392 kN, and with K = 0.20 the nominal torque is 84.940 N·m ¹⁷. That single torque figure hides a wide preload band. Using the K bounds of 0.18 and 0.22 with a ±5% torque tolerance, the same 84.940 N·m produces an initial preload anywhere from 30.566 kN to 41.291 kN ¹⁷. The example is clear that this does not turn 84.94 N·m into a released torque specification ¹⁷. For a lubricated M6 steel bolt, the arithmetic is shorter. With a proof load of about 700 MPa and a tensile stress area of about 20 mm², proof load is 14,000 N. At a 75% target the preload is 10,500 N, and the torque needed is roughly 7–8 N·m ¹⁶. Note that the M6 figure is stated for the lubricated condition — the same bolt dry would need more torque for the same clamp force.

What friction and lubrication assumption to write on the drawing

Because K carries the friction, the lubrication state is part of the specification, not a shop-floor detail. If you write a torque value derived for dry steel and the fastener arrives lubricated, you over-tension the joint ¹⁵. The same logic runs the other way: a lubricant reduces the torque needed to reach the target preload. A dry-derived torque is simply the wrong number for a coated part ¹. Coating callouts can leave this uncontrolled. In an ISO 10683 zinc flake designation, the absence of the 'L' character means the coating contains no lubricant component. The friction coefficient is uncontrolled and the torque-tension relationship is unpredictable ⁷. If you need a predictable relationship, specify the '-L' variant or add a separate lubricant specification ⁷. So the drawing should carry three things together: the target preload or clamp force, the torque derived from a stated K, and the lubrication or coating condition that K assumes. Where the joint is sensitive, a torque/tensile test on the actual fastener, coating and lubricant combination is the way to confirm the K you used ²³.

Choosing a tightening method to match the joint's criticality

Torque control is friction-dominated, and that shows up as scatter in the preload actually achieved. Torque-controlled tightening is quoted at ±25–40% accuracy, influenced by friction and thread conditions ⁵; other sources put the same method at ±25–30% ^{25 4}. Torque-angle control tightens to an angle past snug, which removes most of the friction contribution, and is quoted at ±10–15% ^{5, 25}. Direct bolt stretch measurement removes friction from the measurement entirely and is quoted at ±3–5% ²⁵. If your joint can tolerate the torque method's scatter, it is the cheapest and fastest option and is suited to non-critical connections ⁵. Where that scatter is unacceptable, torque-angle or direct stretch measurement is the specification to write ^{5, 25}. The sources do not give a volume or cost threshold at which you should step up a method — the decision rests on the consequence of a preload error, not on a published break-even. One condition makes torque scatter worse than the headline figure: rusty bolts or stainless steel fasteners prone to seizing can show scatter higher than the typical ±30% ⁴. If your joint uses either, budget more margin or move to a method that does not depend on friction.

Checks before you commit the specification

These are the practical conditions that decide whether the method you picked can actually be executed on the line.

• Confirm the tooling and skill the chosen method needs: torque control is low cost and easy, angle control is moderate cost and moderate ease, stretch control is high cost and hard ¹³.
• Check that the site can support the method — matched assemblies, correct tools, power supply and trained personnel are prerequisites for fast repeatable pretensioned installation ²⁷.
• For safety-critical high-strength assemblies, allow a 24–48 hour relaxation period after initial tightening before the final torque check ¹².
• Decide whether you need torque-curve monitoring or elongation measurement; digital torque wrenches with real-time curves and laser elongation measurement are offered as ways to improve preload consistency ¹¹.
• Verify the K value against the actual coating and lubricant combination rather than a generic table, since the sources provide no tabulated K set per lubrication condition ²³.

Where the sources disagree

Two disputes matter for the specification. Neither is resolved by the sources, so both are things to verify by test or by calculation rather than accept from a table.

Where the torque goes: the 50/40/10 split, K-factor ranges, and the 75%-of-yield target

The short-form equation T = K·D·F only becomes useful once three questions are answered: how much of the input torque actually becomes preload, what K value matches the friction condition of the joint, and what fraction of bolt strength the target preload should be. The sources behind this section give concrete numbers for all three, and they largely agree — with conditions attached.

Torque control is inefficient by physics, not by tooling. Nord-Lock's bolting analysis reports that the torque breakdown 'usually shows that only 10% of the input torque is useful', with underhead friction consuming about 50% and thread friction about 40% of the input. Engineers Universe frames the same split as 'torque energy → useful tension ~10% (typ.)' and notes the split varies by joint; the same source puts typical preload scatter at ±25–30% for a plain K-factor torque spec even with a calibrated wrench. That scatter band, not the wrench reading, is the real accuracy of torque control — and it is why friction condition, surface finish, and lubricant state dominate whether a torque table lands on target.

The nut factor K condenses all of that friction into one number. Engineers Universe gives clean, lubricated threads roughly K = 0.10–0.15 and dry, corroded, or reused fasteners roughly 0.20–0.30 or more; MechCodex states a typical span of 0.10–0.25 and quantifies the sensitivity: a change in K from 0.18 to 0.20 is an 11% preload shift at the same torque. Lubrication and coatings push K down. Five Flute notes that thread lubrication produces 'more bolt elongation and preload for a given torque', and EPC Land warns that PTFE or zinc-flake coatings have much lower friction than bare steel, so applying bare-steel torque values to a coated bolt can over-tension and potentially yield it — use the coating manufacturer's K-factor. Maitenics states the rule plainly: a torque value from one lubrication condition does not produce the same preload under another.

The target preload is usually set as a percentage of the bolt's yield or proof strength, and 75% of yield is the commonly cited working point. MechaniCalc formalises the input as F_PL = %yld · Sty · At — preload as a percent of tensile yield times the tensile stress area — while AMESWeb uses the same structure against proof load (Fi = At × Sp × PPS/100). A worked example from EngiCalcHub for an M10×1.5 class 8.8 bolt lands at a recommended preload of 27.8 kN, exactly 75% of its 37.1 kN yield load, leaving a 25% margin that absorbs the combined tensile-plus-torsional stress during tightening. Engineers Universe describes the workflow in reverse: the chosen percentage produces the force F fed into T = K·D·F, which is why an error in K propagates straight into the achieved preload.

The corpus gives K ranges by broad condition (lubricated vs dry) but no verified per-finish or per-coating K tables beyond the qualitative coating statements cited; the classic '50/40/10' split is stated as a typical example by one manufacturer source (Nord-Lock), with a second source (Engineers Universe) presenting the same ~10% useful-torque figure as a varying typical value rather than a fixed rule.

FAQ

Related pages

• Bolts product page
• How to Determine Tightening Torque and Choose a Tightening Method for a Bolted Joint
• Installing a Bolt to a Specified Torque: Procedure, Tooling and Verification
• Bolt Stretch vs Torque Control for Preload Accuracy: What the Evidence Settles
• Proof Load vs Yield and Tensile Strength in Bolt Selection: What the Evidence Settles