Timber Connection Design

Where the work is

Timber design has a reputation for being simple. Spans are short, loads are light, and most calculations come down to a handful of formulas about flexure and shear. The reputation is mostly true - until you reach the connections. That is where the work is.

Wood is an awkward material to fasten. It is strong along the grain and weak across it, with the across-grain tensile strength about ten times lower than along the grain. It swells and shrinks with the surrounding air. It cracks easily when split across the grain and crushes plastically when pressed along it. A timber connection has to handle all of these properties at once, and a small detailing decision - an end distance two diameters short, a steel plate that prevents shrinkage, an over-tightened bolt - can be the difference between a ductile, yielding joint and a sudden split.

This sheet covers how Eurocode 5 frames that problem and the principles that show up in every connection you will design.

Three properties of wood that complicate everything

Three things distinguish timber from steel and concrete in connection design.

The first is anisotropy - wood's strength depends on which direction you pull it. Parallel to the grain, common softwoods reach 20-30 N/mm² in tension. Perpendicular to the grain - the direction a fastener tends to push against - the same timber might fail at 0.4 N/mm². That is roughly fifty times weaker. No other common structural material is this directional, and because connections almost always involve loads at varied angles, anisotropy is not something you can design around. It has to be designed for.

The second is moisture. Timber comes off the saw at 30-40% moisture content. By the time it is installed, it is typically at 12-18%. After ten years in a heated indoor space, it might settle at 8%. Each percentage point of moisture loss means dimensional shrinkage - most across the grain, almost none along it. A glulam column 600 mm wide when installed loses several millimetres of width over its lifetime. Connections need to allow for this movement, or they will crack the timber instead.

The third is selective ductility. Timber crushes plastically when loaded perpendicular to the grain - the embedment failure mode. It splits without warning when pulled in tension across the grain. A well-designed connection lives in the ductile zone; a poorly-designed one ends up in the brittle zone. The whole geometry of EC5 connection design exists to push the failure mode into the ductile region.

The Johansen insight

Until the 1940s, timber connections were designed empirically - from tables of allowable stresses derived from testing, with little theoretical framework for predicting which failure mode would govern. KW Johansen, a Danish engineer, changed that by applying plastic yield theory to dowel-type fasteners.

His idea was to treat the fastener as an elastic-plastic beam embedded in the timber. Under load, the timber crushes locally and the fastener bends. Failure happens when the timber's bearing capacity is exhausted and the fastener has formed enough plastic hinges to move freely as a mechanism. Johansen worked out the possible mechanisms - timber crushing on one side, on both sides, with one hinge in the fastener, with two hinges - and showed that the actual failure mode is whichever mechanism gives the lowest collapse load.

The modern EC5 tables formalise this same idea. Each row in those tables is a Johansen mechanism. The design capacity of a fastener per shear plane is the smallest value across the rows.

This matters for design because each mechanism has a different character. The pure-bearing modes (no fastener yielding) give the highest capacity, but they rely entirely on timber strength - they can be brittle. The mechanisms with fastener yielding involve the steel absorbing energy through plastic bending - these are the ductile modes. Good detailing pushes the design toward the ductile modes. This is why slender fasteners (long, thin, in dense timber) usually outperform short stubby ones for the same diameter - the slender ones use the steel's ductility, the stubby ones do not.

The embedment strength that underpins the analysis is given by:

f_{h,0,k} = 0.082 (1 - 0.01 d) ρ_k

For loads at an angle to the grain, embedment reduces via the Hankinson formula:

f_{h,α,k} = f_{h,0,k} / (k_{90} sin²α + cos²α)

These two equations, together with the fastener's yield moment, are most of what you need to evaluate any Johansen mechanism by hand.

Geometry and the splitting trap

The EC5 spacing tables look fussy until you realise that each rule prevents a specific failure mode the Johansen analysis does not capture.

End distance prevents the timber from being torn out longitudinally - a tear-out failure where the wood ahead of a loaded fastener shears free. Loaded ends require larger end distances than unloaded ones because the load is driving the fastener towards the end. For bolts in solid timber, a loaded end needs the larger of 7d or 80 mm.

Edge distance prevents the timber from splitting at the edge. When a fastener is loaded perpendicular to the grain near a free edge, the wood between the fastener and the edge can split off. This is one of the most dangerous failure modes in timber connections because it is brittle and gives no warning. The required edge distance scales with the load angle - the more perpendicular the load, the more clearance needed.

Centre-to-centre spacing prevents fasteners from interfering with each other. If two fasteners are too close, the splitting cracks each one causes can join up, and the timber between them peels off as a single block. The required spacing depends on whether you are spacing along the grain (where bolt-induced cracks are restrained by surrounding fibres) or across it (where they are not).

The full spacing geometry sits in EC5 Tab 8.4, but the underlying logic is straightforward. Every rule corresponds to a brittle splitting mode the Johansen analysis does not see, and every minimum is the distance below which that mode begins to dominate the capacity.

The slip nobody models

Most engineers, the first time they design a moment-resisting timber connection, treat it as rigid. They model the structure with rigid joints, distribute moments to the connections, design the fasteners for the calculated demands, and move on.

Then the structure gets built and deflects more than the model predicted.

Timber moment connections are semi-rigid, not rigid. Three things contribute to joint slip.

The first is hole tolerance. Bolts are typically fitted into holes 1-2 mm oversized to allow installation. Until the load is high enough to push the bolt against the hole edge, there is no resistance - the joint just slides. For a moment connection where the lever arm between the bolts and the group centroid is, say, 200 mm, 1 mm of bearing slip translates to 0.005 radians of rotation. A connected member 4 metres long swings 20 mm at its far end. That is a lot of unmodelled deflection.

The second is embedment yield. Even after the bolt has come into bearing, the timber under the bolt crushes - elastically at first, then plastically once the bearing capacity is reached. This is not slip in the rigid-body sense, but it adds rotation under load.

The third is the fastener bending. Single-shear connections in particular let the fastener bend noticeably under load, and this contributes both rotation and lateral displacement at the joint.

EC5 §7.1 gives slip moduli in tables (K_ser for serviceability, K_u for ultimate) that approximate the elastic part of this behaviour. Where joint deformation matters - usually in unbraced frames or long-span moment-resisting beams - these moduli need to be in the model. If you treat both the connection and the structure as rigid, the deflection you get on site will surprise you.

A related caveat: over-tightening bolts at installation does not help. It crushes the timber under the washer, locally reduces stiffness, and does nothing to keep the bolt tight as the timber dries. Tighten firm-and-snug, then revisit at twelve months.

Moisture, the slow disaster

If you set the timber moisture to 18% in your model and assume it stays there, you have made an assumption the timber will ignore. Timber takes up and releases water continuously, equilibrating with the surrounding air. The dimensional consequences are negligible along the grain but significant tangentially and radially - most softwoods change width by about 0.25% for every 1% change in moisture content in the tangential direction.

This matters for connections in two ways.

First, restrained timber cracks. If you fasten steel side plates on both faces of a glulam beam - top and bottom continuous, with bolts going through - the steel does not move as the beam dries, but the beam wants to. The bolts pin the timber against shrinkage and the section splits horizontally. The standard remedy is to detail the plates so the timber can shrink: separate top and bottom plates, slotted holes oriented along the shrinkage direction, or central plates set into slots in the timber rather than wrapping around it.

Second, bolts loosen as the timber shrinks around them. This is the source of the standard 12-month re-tightening specification. If a connection is in a service-class-2 environment and is not accessible for re-tightening, you are designing for a state where clamping force has already reduced - or moving to a different fastener type. Glued-in rods, for example, do not have this problem.

End grain is the worst case for moisture effects because it dries fastest. Connections at the ends of members carry two penalties - they are closer to the surface, and the wood there is drying faster than the bulk material. Most splitting failures in timber connections happen near member ends.

Single shear, double shear, and the geometry that decides

Single shear or double shear is rarely a choice. A beam slotted into a column with a central steel plate naturally produces double shear. A beam attached to a column face with a side plate naturally produces single shear. The structural designer rarely gets to pick the type - the question is what geometry the layout forces on you, not which type is better.

Where there is a choice, double shear has two structural advantages: each fastener carries more load (failure has to occur at two shear planes at once), and the symmetry tends to reduce eccentricity-induced moments. Single shear has the practical advantages of being faster to detail and easier to construct.

EC5 separates the two cases. Tab 8.2 covers single-shear connections; Tab 8.3 covers double shear. The mechanisms are conceptually similar - timber crushing, fastener yielding, combinations of the two - but the geometry of each shear plane and the symmetry of the load path produce different governing equations.

Fire

Connections in fire have the same problem as the rest of the timber structure, only worse. The timber chars from the outside in, reducing the effective section. The fasteners - particularly steel ones, which conduct heat well - heat up quickly, lose strength, and conduct heat into the timber around them. The combination is bad enough that exposed metal connections often set the fire-resistance time limit for an entire structure.

Two protection strategies are common. Encapsulation - plasterboard, timber-based panels, intumescent coatings - slows the temperature rise at the fasteners. Concealed detailing buries the steel away from the fire by recessing bolt heads, using internal plates, or specifying glued-in rod systems where the steel is entirely embedded in timber. For long-duration fire ratings on exposed timber structures, glued-in rods are often the only option that works.

EC5 Part 1-2 sets out the calculation framework. The general approach is to design the connection for the reduced section after the appropriate char depth, with reduction factors applied to fastener strength to allow for the elevated temperature.

Pulling it together

A well-designed timber connection has a clear load path, fails in a ductile way under overload, leaves room for the timber to move, and acknowledges that the joint is not really rigid. Eurocode 5 turns these concerns into geometry tables and mechanism equations, but the rules make more sense once the underlying problems are visible.

Most poor-quality timber connections fail because the designer treated them as small steel connections and forgot that wood has its own ideas.