Structural Engineering & Statics
where the load goes
Connections

The joint that carries nothing until it slips

Tighten the bolts hard enough and the plates are clamped together with a force nothing applied. The joint then carries shear by friction, the bolts are in tension and not in shear at all, and the load path has nothing in common with the joint it looks identical to.

Assumes The hole that goes oval, and the one that tears to the edge and The connection is not a point, and every diagram on this site says it is.

16 min read 7 figures The load must go somewhereWhich failure arrives first

Two bolted joints, photographed side by side, are indistinguishable. Same plates, same holes, same bolts, same spanner marks. One of them carries its load by the bolts pressing against the sides of their holes. The other carries it by friction between the plates, with the bolts carrying no shear whatever.

The difference is how hard the bolts were tightened, and it changes the load path completely.

A preloaded joint, before and after it slipsTwo preloaded bolts at 137 kN each, on one friction face at μ = 0.5. The joint carries 137 kN by friction with the bolts in tension and not in shear at all; past that it slips into bearing and carries 188 kN with the bolts now in shear. Two different mechanisms, one joint.00.511.522.533.544.55050100150200displacement, mmload, kNfriction 137 kNbearing 188 kNslipthe rising branch is drawn, not solved: it is elastic shear of the plates
Fig. 1 The load–displacement path of a preloaded joint: two M20 bolts at 137 kN of preload on one blasted friction face. It carries 137 kN with essentially no movement, slips, and then behaves as an ordinary bearing joint up to 188 kN. Three regimes, and no single number describes it — which is why the same connection is designed twice.

What preloading does

An ordinary bolt is tightened enough to hold the assembly together. A preloaded bolt is tightened to about seventy per cent of its own tensile strength, deliberately, and stays there.

That stretches the bolt, which clamps the plates. The clamping force is real, large and permanent: for an M20 grade 8.8 bolt it is 0.7×800×245=1370.7 \times 800 \times 245 = 137 kN, which is more than the same bolt’s shear capacity of 117.6 kN.

Now apply shear to the joint. Friction across the clamped interface resists it, with a capacity of

Fs=ks n μ FpF_s = k_s\, n\, \mu\, F_p

per friction surface — the number of bolts times the preload times the coefficient of friction, with ksk_s a factor for hole clearance. And while friction is carrying the load, the bolt is doing nothing but being stretched. There is no shear in it. It is not touching the sides of its hole, because it never moved.

That is the entire mechanism, and it is worth dwelling on how unlike a bearing joint it is. The bolt’s shear capacity is not in the calculation. The plate’s bearing capacity is not in the calculation. The hole’s clearance is not in the calculation except as a modest coefficient. What is in the calculation is a tension and a friction coefficient, and neither of those appears anywhere in the analysis of the joint sitting next to it.

The surface, which is the thing nobody controls

Friction coefficient is the only material property in the formula and it is a property of neither the bolt nor the steel. It is a property of how the plate was prepared, and it varies over a factor of two and a half.

surface μ slip resistance
blasted, no coating 0.50 137.0 kN
blasted, sprayed aluminium 0.40 109.6 kN
wire-brushed, mill scale removed 0.30 82.2 kN
untreated, as rolled 0.20 54.8 kN

Two M20 bolts, identically tightened, in an untreated joint carry 54.8 kN of slip resistance against 137 kN in a blasted one. The joint has lost sixty per cent of its capacity to a surface preparation decision, and there is nothing visible in the finished connection to say which was done.

A preloaded joint, before and after it slipsTwo preloaded bolts at 137 kN each, on one friction face at μ = 0.2. The joint carries 54.8 kN by friction with the bolts in tension and not in shear at all; past that it slips into bearing and carries 188 kN with the bolts now in shear. Two different mechanisms, one joint.00.511.522.533.544.55050100150200displacement, mmload, kNfriction 54.8 kNbearing 188 kNslipthe rising branch is drawn, not solved: it is elastic shear of the plates
Fig. 2 The same joint with the faying surfaces left as rolled. The friction plateau has dropped to 54.8 kN and the bearing capacity has not moved, so the joint slips at under a third of its ultimate load. This is the same drawing as the one at the top of the page with one number changed, and the number is about paint.

That dependence is the practical weakness of the whole method. Bolt preload can be measured, controlled and verified — by torque, by turn-of-nut, by load-indicating washers. Surface condition is verified by somebody looking at it, and it degrades between blasting and erection through weather, handling and site dirt.

Why the ultimate capacity is unchanged

Here is the result that reframes what preloading is for.

The joint above slips at 137 kN and reaches 188 kN in bearing. So preloading has not raised the connection’s ultimate capacity: it was 188 kN as a bearing joint and it is 188 kN as a preloaded one. The bolts are the same bolts and the plates are the same plates.

What changed is the displacement. Below 137 kN the preloaded joint does not move at all, where the bearing joint takes up 2 mm of hole clearance and then ovalises. That is the whole product.

Which means the design question is not “is this joint strong enough?” but “does this joint need to not move?”, and the answer is a serviceability judgement rather than a strength one. The cases where it is yes:

  • joints subject to load reversal, where a joint that slips one way then slips back, repeatedly, and works the holes oval;
  • joints in fatigue-loaded structures, where slip introduces a stress concentration and movement at exactly the detail that is already the fatigue-critical one;
  • bracing that has to control sway, where the frame’s stiffness depends on the brace’s, and a brace with 2 mm of slip at each end has far less than its section suggests;
  • anything where the movement itself is the problem — machinery alignment, crane rails, connections carrying calibrated instruments.

In every one of them the argument is about stiffness rather than strength, which puts this essay squarely in the same territory as stiffness is not strength and for exactly the same reason: two limit states, and the one that governs is not the one the connection is named after.

How a force gets into a bolt, and how anybody knows

There is a practical problem hiding behind the number 137 kN: it is a tension in a steel bar, and nobody can see it.

Four methods are used to put it there and each measures something other than the thing it wants.

Torque control. Tighten to a specified torque and infer the tension. The relation between the two is T≈kdFpT \approx k d F_p with kk a nut factor around 0.2, and the trouble is that kk depends on thread condition, lubrication and surface finish. Scatter of ±30% on the achieved preload is normal, which is why torque control is the least favoured method and why the code values assume it.

Turn-of-nut. Bring the assembly to a snug fit and then turn the nut a specified further fraction — typically a half or three-quarters of a turn. This measures extension rather than torque, and extension is much more directly related to tension because it is a length divided by the bolt’s own stiffness. Better scatter, and it depends on the snug condition having been reached.

Load-indicating washers. A washer with raised protrusions that flatten at a known clamping force, leaving a measurable gap. This measures the clamping force, which is the thing wanted, and it can be verified after the fact with a feeler gauge — which is its real advantage: an inspector can check a finished joint.

Direct tension measurement, with an instrumented bolt. Accurate, expensive, and used for verification rather than for production.

The ordering of those four by accuracy is also their ordering by cost, and the design values in codes are written for the middle of the range. What the list makes clear is that preload is a construction quantity rather than a design one: the calculation assumes a force that somebody on site has to deliver, and the delivery method has more scatter than any material property in this entire field.

A preloaded joint, before and after it slipsTwo preloaded bolts at 137 kN each, on one friction face at μ = 0.3. The joint carries 82.2 kN by friction with the bolts in tension and not in shear at all; past that it slips into bearing and carries 188 kN with the bolts now in shear. Two different mechanisms, one joint.00.511.522.533.544.55050100150200displacement, mmload, kNfriction 82.2 kNbearing 188 kNslipthe rising branch is drawn, not solved: it is elastic shear of the plates
Fig. 3 The middle of the surface range, at μ = 0.3 — a wire-brushed face rather than a blasted one. Slip resistance falls to 82.2 kN, so the joint now slips at 44% of its ultimate capacity instead of 73%. The bolts, the plates and the preload are identical to the figure at the top of this page.

The bolt that is doing two things at once

A preloaded bolt is already at seventy per cent of its tensile capacity before the structure is loaded. If the connection then carries external tension as well as shear — as a preloaded end plate does — the two have to be reconciled.

The reconciliation is not the obvious addition. Applying external tension to a preloaded assembly does not add its full value to the bolt, because most of it goes into relieving the compression in the clamped plates rather than into stretching the bolt further. The plates are far stiffer than the bolt, so the bolt picks up only a small fraction — typically 10 to 20 per cent — of the applied tension until the plates separate entirely.

That is a genuinely useful property and it is the reason preloaded bolts perform so well in fatigue: the range of stress in the bolt is a small fraction of the range applied to the connection, and fatigue is governed by range. A preloaded bolt in a fluctuating tension joint can see a tenth of the stress range an ordinary bolt in the same joint would see.

What external tension does reduce, directly and fully, is the clamping force — and therefore the slip resistance. So the slip check for a joint carrying both is written with the preload reduced by the applied tension, which is the one place the two effects add up in the unhelpful direction.

What the preload does over time

Preload is a stress that was imposed rather than applied, which means it is subject to everything that erodes an imposed stress — and this site has an essay about precisely that.

Three mechanisms take preload away:

Embedment. The surfaces under the bolt head and nut, and the faying surfaces, are not perfectly flat. Under 137 kN of clamping they flatten locally, the bolt shortens slightly, and the tension drops. Most of this happens in the first hours and it is the reason bolts are sometimes retightened.

Relaxation. The bolt is a steel bar held at constant strain at seventy per cent of its yield, which is precisely the configuration in which stress leaks away. At ambient temperature the effect is small over a structure’s life; at elevated temperature it is not, and preloaded connections in fire are treated with care.

Coating creep. If the faying surfaces are painted, the paint is a polymer under 137 kN of compression, and polymers creep. This is the reason the surface treatments allowed in preloaded joints are a short list.

The clamping force is therefore not a constant and the design values allow for its decay. What is worth carrying is the shape of the problem: an imposed force decays and an applied force does not, and preload is the only force in this entire field that was imposed.

Two checks on one joint

The consequence of all this is that a preloaded connection carries two calculations that look at different things.

At the serviceability limit state, the check is slip: does the joint stay where it is under working load? Slip resistance against the serviceability load, using the surface’s friction coefficient and the preload after decay.

At the ultimate limit state, the check is bearing and bolt shear — the joint is assumed to have slipped, and the calculation is the same one an ordinary bearing joint gets. Both plies checked, net section checked, block shear checked.

The joint has to pass both, and the two are not simply related: a joint can be comfortable in slip and marginal in bearing, or the reverse.

Bearing and tear-out against end distanceA 20 mm bolt in a 10 mm plate. Below 165 mm of end distance the bolt tears a channel out to the end and the capacity is proportional to that distance; above it the plate crushes in front of the bolt and the end distance stops mattering. At 40 mm the capacity is 52.12 kN and the mode is tear-out.020406080100120140160180200050100150200end distance, mmbearing capacity, kN40 mm → 52.12 kNcorner at 165 mmplate crushesbolt tears out
Fig. 4 The second of the two checks, which is the whole of the analysis for an ordinary joint and the ultimate half of the analysis for a preloaded one. Preloading does not remove this check; it adds a check in front of it. A designer who treats slip resistance as the connection’s capacity has done the serviceability calculation and left the ultimate one undone.

The two curves, side by side

It is worth putting the two joint types on one page, because the difference is not in the height of anything.

A preloaded joint, before and after it slipsTwo preloaded bolts at 137 kN each, on two friction faces at μ = 0.5. The joint carries 274 kN by friction with the bolts in tension and not in shear at all; past that it slips into bearing and carries 188 kN with the bolts now in shear. Two different mechanisms, one joint.00.511.522.533.544.55050100150200250300displacement, mmload, kNfriction 274 kNbearing 188 kNslipthe rising branch is drawn, not solved: it is elastic shear of the plates
Fig. 5 The same bolts at the same preload in a double-lap joint: two friction surfaces rather than one, so 274 kN of slip resistance against 188 kN of bearing. This joint never slips before it fails — the friction plateau is above the bearing capacity — and it is the only arrangement here in which the connection’s ultimate capacity and its slip capacity are not two separate numbers.

That is the exception rather than the design intent. A joint detailed so that friction outlasts bearing is a joint whose ultimate failure is the plates rather than the interface, and the arithmetic that produced it — two faces rather than one — was a geometric decision rather than a structural one.

Where the friction faces are counted

One arithmetic detail that catches people. The slip resistance is per friction surface, not per bolt.

A single-lap joint — two plates, bolts through both — has one interface and one friction surface. A double-lap joint — a central plate between two covers — has two, and the same bolts at the same preload deliver twice the slip resistance: 274 kN rather than 137.

That is a doubling for no extra bolt and no extra preload, and it happens to coincide with double shear doubling the bearing capacity too. So the double-lap arrangement is better on both limit states at once, which is unusual enough to be worth reaching for whenever the geometry allows.

Why this is not the default

Given that a preloaded joint is stiffer, better in fatigue and better under reversal, the obvious question is why every connection is not one.

The answer is cost, and it is worth being specific about where the cost sits, because it is not in the bolt. A grade 8.8 bolt costs what it costs whether it is snug-tightened or preloaded. The cost is in:

  • surface preparation, which means blasting the faying surfaces and then protecting them until erection;
  • controlled tightening, which is slower than a rattle gun and needs a method and a record;
  • inspection, because a preload that was not achieved is invisible and the connection looks identical either way;
  • restrictions on what can be in the joint — no unqualified paint on the faying surfaces, and limits on packing.

None of those is large individually and together they are a real multiple on the cost of an ordinary bolted connection, repeated across every joint in a building.

So the honest design position is that preloading is a targeted measure. Use it where movement is the problem, on the connections where it is the problem, and use ordinary bolts everywhere else. A structure with every joint preloaded is one where somebody has answered a serviceability question with a blanket instead of a calculation.

There is one more consideration that decides it in some structures regardless of cost. A preloaded joint’s slip resistance depends on a surface condition that cannot be inspected once the joint is closed. For a structure whose connections will be inaccessible for their whole life, that is a property that has to be delivered right the first time and can never be verified again — which is a different kind of risk from a bolt that is simply the wrong size, and it is one the drawings cannot record.

A preloaded joint, before and after it slipsFour preloaded bolts at 137 kN each, on one friction face at μ = 0.5. The joint carries 274 kN by friction with the bolts in tension and not in shear at all; past that it slips into bearing and carries 376 kN with the bolts now in shear. Two different mechanisms, one joint.00.511.522.533.544.550100200300400displacement, mmload, kNfriction 274 kNbearing 376 kNslipthe rising branch is drawn, not solved: it is elastic shear of the plates
Fig. 6 Four bolts rather than two. Both the friction plateau and the bearing capacity double, so the shape of the curve is unchanged and the joint still slips at the same fraction of its ultimate load. Bolt count is the one variable that moves both mechanisms together — which is why it cannot be used to buy slip resistance relative to strength.
The net section, and the path the tear takesA 200 mm plate with two holes staggered by 0 mm at a gauge of 60 mm. The straight path through one hole leaves 156 mm; the diagonal path through both leaves 156 mm after the s²/4g correction adds 0 mm back. The shorter of the two decides, at 78% of the gross section.g = 60net width 156 mm of 200the critical path crosses two holes, with s²/4g = 0 mm added back
Fig. 7 The holes, which the two limit states see differently. At serviceability the load crosses by friction and no force passes through the net section at all; at ultimate the joint has slipped into bearing and the full deduction applies. The same plate has two effective sections depending on which question is being asked.

What to take from it

Preloading changes the mechanism, not the capacity. The joint carries shear by friction with the bolts in tension, up to a slip load, and then becomes an ordinary bearing joint with an unchanged ultimate capacity.

The variable that decides slip resistance is the surface. Sixty per cent of it can be lost between a blasted face and an untreated one, and nothing about the finished connection shows which it is.

Preload is an imposed force, and imposed forces decay. Embedment, relaxation and coating creep all take it away, which is why the design value is not the tightening value.

And the joint is designed twice. Slip at serviceability, bearing at ultimate, and passing one says nothing about the other.

Where this goes next

020406080100120140160180200050100150200end distance, mmbearing capacity, kN40 mm → 52.12 kNcorner at 165 mmplate crushesbolt tears out The hole that goes oval, and the one that tears to the edge A bolt pressing on the side of its hole either crushes the plate in front of it or shoves a channel of metal out to the end. Which one happens is decided entirely by a distance that is usually set by a minimum in a table. 1 d10 d100 d2.7 yr27 yr0246810time since the strain was imposedstress, N/mm²E × strain: 9.6tensile strength: 3.5what is left: 2.32the one-line shortcut at χ = 0.8: 3.31 The strain that was imposed, and the stress that leaked away Multiply a restrained shrinkage strain by the modulus and the answer is three times the tensile strength — which predicts that every restrained concrete member ever cast has cracked. Most have not, and the reason is that the material creeps while it is being stressed. 0.60.811.21.41.61.8200.511.5span, relative to the firstthey cross heredeflection runs out at 1.40strength runs out at 1.54the limitstrengthdeflection Stiffness is not strength, and usually it is the one that governs A beam can be nowhere near failure and still be unusable, because it has moved too far. For most long-span members that limit arrives first, and stronger steel does not help at all. 100 kNe = 150centroidworst bolt 50.37 kNSix bolts · direct shear 16.67 kN eachelastic vector method The connection is not a point, and every diagram on this site says it is Every free body drawn here has joined its members at points. Real structures fail at the joints far more often than in the members, and the reason is that a joint is exactly the region the theory behind every other page explicitly excludes. 10⁴10⁵10⁶10⁷10⁸2050100200500cycles to failurestress range, N/mm²category 160category 90category 366.8e+74.3e+62.7e+5working range 70 N/mm² — the lives are marked The load that never came near failing anything A detail survives sixty-eight million cycles at a stress range that another detail in the same steel survives two hundred and seventy thousand of. The two lie a factor of two hundred and fifty apart, and the material is not on the plot anywhere.

Named alongside this one

Essays reaching for the same objects. Nobody chose these; they are what the concept index makes visible.

What links here

Every essay whose body links to this one.

The objects this essay names

Each one links to every other essay that touches it.

Bolt tensionConnectionFrictionLimit statePreloadRelaxationServiceabilitySlip resistance