Specifying flatness, straightness and twist after processing

A drawing that says "bar to be flat" has specified nothing. The word is used loosely to cover three geometrically distinct characteristics with different causes, different measurement methods and different achievable values. Arguments at incoming inspection are usually arguments about which one was meant.

Flatness is a property of a surface: every point must lie between two parallel planes separated by the tolerance value. It says nothing about the direction of the error.

Straightness is a property of a line, an edge or the axis of the bar. On rectangular busbar it splits into two. Bow is curvature in the thickness direction, so the bar rocks when laid on its face. Camber is curvature in the width direction, so the bar looks like a banana when laid flat. A bar can have plenty of one and none of the other.

Twist is rotation of the cross-section about the longitudinal axis, so the two ends are not in the same plane. It is the only one of the three you cannot see by laying the bar on a table and looking along it, and it is the one most often discovered at assembly.

None of them means anything without a reference length. "Camber 2 mm" tells you nothing. "Camber 2 mm maximum over any 1000 mm" is a specification.

What the mill already gave you

Before blaming your own process, look at what the material arrives with. ASTM B187/B187M, the specification covering copper bus bar, rod and shapes, sets straightness in Table 10 as a maximum depth of arc measured over a stated portion of the length:

• Bar generally: 1/4 in [6 mm] over 60 in [1500 mm]
• Half-hard rectangular bar, 3 to 15 mm thick and 50 to 150 mm wide: 1/8 in [3 mm] over 96 in [2400 mm]
• Rod: 1/2 in [13 mm] over 120 in [3000 mm]
• Shapes: 1/2 in [13 mm] over 72 in [1800 mm]

The method matters as much as the value. The standard says to place the bar on a level table so the departure from straightness lies horizontal, and to measure the maximum depth of arc with a steel scale and straight edge to the nearest 1/32 in [0.8 mm]. Measuring horizontally removes gravity from the result. A 3 m length of 100 × 10 mm copper weighs 27 kg and will sag several millimetres under its own weight if supported at the ends and measured vertically, and that sag is not a property of the bar.

Three millimetres of camber over 2.4 m as supplied is a great deal more than most assemblers expect. If your finished-part tolerance is tighter than your incoming tolerance, the process has to remove error rather than preserve it, and that has to be designed in.

How punching produces bow and camber

Punching does not remove material neutrally. It leaves a residual stress field around every hole, and the sum of those fields moves the bar.

Two mechanisms act. Around the hole itself, material is drawn inward and downward as the punch enters, which puts the entry-side region into compression, and material is pushed out through the exit face during fracture, which puts that region into tension. When the tensile term dominates, the bar bows. This is local: a heavily punched region will bow while an unpunched region of the same bar stays flat, which is why bow often appears as a kink rather than a smooth arc.

The second mechanism is the stroke itself. As the punch drives down, the bar deflects into the die opening. The top surface goes into compression and the bottom into tension, and the plastic component of that does not recover.

Four things make it worse, and all four are controllable. Hole density sets the accumulated stress and comes from the drawing. Thickness raises rollover and force, and force is what deforms the surrounding material. Tool condition matters most in practice: a dull punch needs materially more force than a sharp one and produces more rollover, which is the usual explanation for a bar that suddenly starts coming off the machine bowed when it never used to. And clearance, tight, raises force and raises bow. That last one is the same variable that drives tool life, and it is treated in punch-to-die clearance for copper and aluminium busbar.

Camber, as distinct from bow, comes from asymmetry in the width direction. A hole pattern running down one edge of a wide bar puts the stressed zone off the neutral axis and the bar curves in plan. Patterns symmetrical about the centreline largely avoid it, which is one of the quieter arguments for the standard hole positions in DIN 43673-1.

Two countermeasures beyond sharp tools and correct clearance. Control stripper or presser-foot pressure: it holds the bar flat during the cut, but on soft copper excessive pressure will itself bow the bar, and backing it off is sometimes the fix. And sequence the holes to distribute the work rather than completing one end before starting the other, which the CNC will do for you if the nesting is set up for it.

How shearing produces twist

A guillotine shear does not cut across the whole width at once. The upper blade sits at a rake angle so the cut progresses from one side to the other, which is what keeps the force manageable. That progression applies a couple to the offcut, and the offcut rotates.

Rake angle is the dominant variable. A larger rake reduces cutting force and increases twist; a smaller rake does the opposite. Where the rake is adjustable, the correct setting is the smallest angle that will still make the cut, and using the same angle for 6 mm and 16 mm bar guarantees excess twist on the thin one.

Aspect ratio decides whether the twist is visible. Shear manufacturers commonly cite a minimum width of 10 to 15 times material thickness before a cut strip will lie flat. Busbar cropped to length is wide relative to its thickness, so the twist shows up as a small wind at the sheared end rather than a corkscrew. On narrow bar at 20 × 10 mm the ratio is 2

and the twist is real.

Two other shear faults produce distortion rather than twist. Blade gap set too wide lets the material roll and fold over the lower blade before it fractures, giving a large burr and a bowed part. Dull blades crush before they cut, driving stress into the piece and bowing it downward after separation. Both are diagnosed the same way, by looking at the cut face for excessive rollover and a short burnish band. The shearing process overview covers the geometry.

How bending distorts

Bending is intentional deformation. The unintentional part is what it does around the bend.

On a flatwise bend the outer fibre stretches and the inner compresses, and because the two are not equal the flange picks up a slight camber and the section goes marginally hollow across its width near the bend. Springback compounds it: the bar relaxes after the punch retracts, and uncompensated springback leaves the included angle wrong and the flange proud.

The specific failure to watch for is a hole close to a bend tangent. Material around the hole is drawn into the bend, the hole goes oval and its position moves. Keep clearance holes at least two thicknesses clear of the tangent, and where the drawing forces one nearer, punch after bending rather than before.

Measuring it properly

Support the bar at three points on a flat reference, a granite surface plate or a machine table verified with a straightedge, rather than along its whole length on an unknown surface.

For flatness of a face, lay a straightedge across the surface in several orientations and read the gap with feeler gauges, or sweep a dial indicator on a height gauge. Record the worst reading and the length over which it was taken.

For bow and camber, use the ASTM B187 method: bar on a level table with the departure from straightness lying horizontally, straightedge along the edge under test, maximum depth of arc read with feeler gauges. Record which of the two directions you measured. Reporting "bow" when you measured camber is the most common inspection error on busbar.

For twist, clamp one end flat on the surface plate and measure the gap under each of the two corners at the free end. Twist is the difference between those gaps divided by the bar width, reported as an angle or as millimetres per metre. State the length: 0.5 mm of twist over 300 mm and 0.5 mm over 3000 mm are not comparable.

Where any of these is an acceptance test, the reference length and the support condition belong in the specification alongside the number.

What is realistically achievable

The figures below are indicative planning values rather than guarantees, and they assume sharp tooling in good condition. Machine specifications quoted are published values.

Angle accuracy on bending is a separate specification and should not be confused with flatness. A servo bending machine holding ±0.1° on the bend angle says nothing about whether the flange is flat, and both need to be on the drawing if both matter.

Where the stack-up bites

Take a three-phase stacked run: four bars of 100 × 10 mm copper on insulated supports, 2 m between joints, bolted at each end to the adjacent section.

Each bar carries some camber, and the direction is random from bar to bar. At the ASTM limit that is around 2.5 mm over 2 m. Two bars whose camber opposes leave a gap of several millimetres at the joint face. The bolts will close it, because copper is compliant and an M12 bolt has no difficulty bending a 10 mm bar. What has happened is that the joint now carries a bending preload, contact pressure across the face is no longer uniform, and the effective contact area is smaller than the drawn area.

That is a temperature-rise problem rather than a cosmetic one. Joint resistance is set by real contact area, and under IEC 61439 the limits are 105 K rise for bare copper busbars and 70 K at terminals intended for connection of external conductors. A joint bolted through a bow starts life with less contact area than the design assumed and loses more as it thermally cycles. If the design is already close to its ampacity limit at the joint, that margin was not there to spend.

Hole position stacks on top. A cambered bar carries its hole pattern on a curve, so even a perfectly punched pattern is displaced relative to a straight mating bar. The bolts go through, because clearance holes are generous. The joint is pre-stressed anyway. Thickness tolerance stacks in the third direction: four bars per phase in a laminated stack, each at the low end of tolerance, changes total stack height and therefore clamping bolt length and spacer heights. Do that arithmetic once per standard configuration rather than on site.

When straightening is warranted

Straighten when residual bow or twist exceeds what the assembly can absorb without pre-stressing a joint, or when joint face flatness is itself out of specification. Not to make bars look better in the rack.

The order of operations is the part most often got wrong. Level the stock before punching. Levelling a punched bar redistributes stress around holes that already exist, can distort holes near the bar edge, and on hard-temper material can leave a new residual stress field that reappears after the next thermal cycle. Roller levelling redistributes stress; it does not delete it.

That makes straightening a last resort in a sensible process route. Start with levelled stock, keep punch tooling sharp and correctly clearanced on the punching and shearing machine, sequence the hole pattern so the work is distributed, take the rake angle down on the shear where the machine allows, and compensate springback on the bending station so the flange lands where the drawing says. Bars produced that way rarely need straightening.

Where it is required, specify it as a process step with its own acceptance measurement rather than leaving it to a mallet on the assembly bench, which is where a great deal of it currently happens and where none of it is recorded.