Busbar twisting: geometry, limits and tooling

A twist rotates the bar about its own longitudinal axis without changing the direction it travels. It is the operation you use when the run is going the right way but the wide face is in the wrong plane: a vertical riser feeding a breaker whose lugs sit horizontally, a stack of phases that has to fan out to different terminal orientations, a drop into a device that was mounted at 90° to the busbar chamber.

Most shops run it on rule of thumb, and the rules of thumb in circulation disagree with each other by a factor of three. They disagree for a reason, and the reason is calculable.

Twist length as a function of width

Published guidance on the length of bar to allocate for a 90° twist spans a wide band. Switchgear installation practice puts the twist length at a minimum of twice the bar width. Machine builders commonly specify 2.5 times the width as a floor, with one published example giving a 50 mm twist length for a 20 mm wide bar. Other fabrication guidance quotes 2.5 to 5 times the width. Manual bench-vice practice, at the low end, quotes 1.5 to 2 times the width, but attaches a limit of 6 mm thickness and 60 mm width to that figure.

Those numbers reconcile once you look at the strain. The corner of the section sits roughly half a bar width from the axis of twist. Rotating it through an angle θ over a length L imposes a surface shear strain of:

γ = (w/2) · θ / L

which for a 90° twist reduces to γ = π·w / (4L). Converting to an equivalent uniaxial strain for comparison against elongation data, ε ≈ γ/√3:

The formula ignores thickness, which is legitimate for busbar proportions: on a 100 × 10 mm bar the true corner radius is 50.25 mm against the 50 mm the formula assumes.

Now the guidance makes sense. At 2.5 × width the corner sees around 18% equivalent strain, which soft and half-hard copper absorb without difficulty. At 1.5 × width it sees 30%, which is survivable only in thin narrow soft bar, exactly the envelope the bench-vice rule attaches to itself. At 5 × width the corner sees 9%, which is comfortable even in hard temper.

A working rule: 2.5 × width for R220 and R240, 4 × width for R290, and never below 2 × width whatever the temper. Confirm on a first article, because the EN 13601 elongation figures are minima and delivered material usually beats them.

Note the direction, because it runs opposite to bending: the longer the twist, the gentler the strain. Twist length is a design allowance rather than a machine setting, so if the route has no straight length to give up, the twist is not available at all.

Torque, and why twisting beats bending edgewise

The plastic torque needed to twist a rectangular section is approximately τy·(w·t²/2 − t³/6), which is dominated by t² and grows only linearly with width. The plastic moment needed to bend the same section edgewise is σy·t·w²/4, which grows with the square of width.

Taking half-hard copper at 180 N/mm² proof strength:

The ratio is close to 0.87·w/t across the whole range. Every millimetre of extra width makes the edgewise bend disproportionately harder while making the twist only proportionally harder, and the wide thin sections where edgewise bending is worst are precisely where the advantage is largest.

That is the quantitative case for the twist-and-flat-bend alternative discussed in edgewise versus flatwise busbar bending. Twist the bar 90°, then make the direction change as an ordinary flatwise bend where thickness governs the strain instead of width. Two easy operations replace one difficult one.

Warping, edge thinning and what happens to the section

A circular shaft under torsion keeps its cross-section plane. A rectangular one does not. Under Saint-Venant torsion, points on a rectangular section displace axially in a saddle pattern, with the largest displacement at the corners. That is warping, and it is unavoidable in a non-circular section.

Two practical consequences follow.

The twisted zone is not dimensionally identical to the parent bar. The section dishes slightly, the corners lead or lag the mid-face, and measured thickness at the edges of the twisted region comes out marginally under nominal. On normal busbar proportions the reduction is small, in the order of a few per cent, but it is a local resistance rise and therefore a candidate hot spot. Check it against your temperature-rise margin; the busbar ampacity calculator gives the headroom the run has before the reduction matters.

Warping is also restrained at the clamps, because the clamp faces hold the section flat while the material between them wants to warp. Restrained warping generates axial stresses concentrated at the clamp edges, which is why cracking in twisting almost always initiates at the clamp line rather than in the middle of the twist. If your twists crack at the clamp, the answer is a longer free length or softer clamp inserts, not more torque.

The twisted zone is also work-hardened. Do not place a bend inside a twist or immediately adjacent to one. The material there has already spent a large fraction of its available ductility and it will crack at a radius that the parent bar would tolerate. Leave clear straight bar between the two operations.

Torsional springback

Twists spring back, and they spring back considerably more than bends do.

The elastic recovery on unloading is Δθ = T·L/(G·J), where T is the applied torque, L the twist length, G the shear modulus and J the torsional constant of the section. Copper's shear modulus follows from its elastic modulus at roughly 46 kN/mm². Working the arithmetic for a 100 × 10 mm half-hard bar twisted 90° over a 250 mm free length gives an elastic recovery of about 5°. The same bar twisted over 400 mm recovers about 8°. A 160 × 10 mm bar over 400 mm recovers about 7.6°.

These are first-order estimates from elastic unloading of a fully plastic section. Verify them on a first article rather than programming them from the page. What they establish reliably is the scale: torsional recovery runs to several degrees, an order of magnitude larger than the springback on a flatwise bend of the same bar.

Note also that the recovery grows with twist length while the strain falls with it. Lengthening a twist to protect the corners makes the angular compensation larger. There is no length that optimises both, so pick the length from the strain requirement and let the compensation table handle the angle. The same offset-table discipline described for springback compensation applies here, keyed on material, temper, section and free length, with twist entries kept entirely separate from bend entries.

Clamping and tooling

A twisting head is two clamps: one fixed, one rotating. The free span between them is the twist length, which means twist length is set by clamp position and is not continuously adjustable on every machine. Check the increment before you commit a design to a particular allowance.

Clamp faces must not mark the bar. Copper is soft, and the clamping pressure needed to prevent slip under full torque is high. Bronze, aluminium or hard polymer inserts are normal. Steel jaws with a serrated face will emboss the bar, and the emboss becomes a crack initiation site under the torsional strain.

One clamp must be free to move axially. A bar shortens slightly along its axis as it twists. Hold both ends rigidly and you superimpose axial tension on the torsional shear, which adds directly to the corner strain and produces edge cracking at strain levels the twist alone would have survived. A floating clamp, or a small deliberate axial compliance, removes the problem.

The section also has to be held square at both clamps. Rotational slip shortens the effective free length, which raises the strain above what you calculated, and it puts the finished angle out.

Published twist envelopes on busbar bending machinery typically sit around 5 to 10 mm thickness and 15 to 100 mm width, noticeably tighter than the same machine's flatwise bending rating. Twisting is a listed capability on both the EMAC-BB-H12 and the SMART-603CNC-S, running on the same control and the same part programme as the bending operations, so a part with both a twist and two bends is one setup rather than three. The ±0.1° figure quoted for the EMAC-BB heads is a bending accuracy specification; twist repeatability on any machine depends heavily on clamp grip and should be characterised on your own material rather than assumed from the bend figure.

Where twists belong in the process plan

Punch and cut before twisting, not after. A twisted bar will not sit flat on a punch table and cannot be located reliably against a back gauge.

Sequence twists before bends where the geometry allows, because a twisted section is harder to clamp square in a bending former than a straight one is.

Keep holes out of the twist zone entirely. A hole in a region of high torsional shear is a stress concentration in the worst possible place, and the material around it will tear before the twist completes.

Allow the strain-based twist length at the drawing stage. Retrofitting a twist into a route that was dimensioned without one is the usual reason a shop ends up attempting it at 1.5 times the width and scrapping the bar. Twisting and bending are both length-consuming operations, and the developed length has to carry both. The full process description sits under twisting.