ISO 7438

What the test does

A piece of metal is bent, in one continuous movement, until it reaches a prescribed condition — a stated bend angle, legs parallel at a set spacing, or legs in direct contact. Three devices are recognised. The commonest sits the test piece across two supports and drives a former into the middle. The second replaces the supports with a V-block, so the piece folds into the vee as the former descends. The third clamps one end and wraps the free length round a former. The load is then removed and the outside of the bend inspected.

What it measures, and why it matters

Nothing is plotted and nothing is calculated. The output is the state of the convex surface after bending: either it meets what the product standard demands, or — where the product standard is silent — it is judged on the absence of visible macro-cracks.

The value of that binary answer is that it interrogates the surface and near-surface structure, where forming operations actually fail. A plate that passes a tensile test comfortably can still split on the brake if it carries rolled-in laps, a decarburised skin or heavy segregation bands, because the tensile coupon shares strain along a gauge length while the bend concentrates it into the outer fibre. Bend results are an acceptance gate for delivery condition, forming behaviour and heat treatment, not design data.

Specimen

The method is written around rectangular test pieces, and the rules below apply to them. Their edges are rounded, the permitted radius scaling with thickness, because a square-cut or burred edge is a crack starter that has nothing to do with the material. Width follows the product where the product is narrow and is otherwise taken from a defined band, so pieces from wide plate are cut to a standard width. Thickness is normally the product thickness; above a threshold the piece may be machined down on one side only, and the unmachined face goes on the tension side so the as-supplied surface is the one under test. Testing is at 10 °C to 35 °C, or 23 °C ± 5 °C where controlled conditions are called for — one of the few genuinely specified temperature bands among bend methods. Piece count comes from the product standard.

What the machine must be capable of

The standard has nothing to say about the frame. Every dimensional clause governs the jig; the press behind it is the laboratory's own choice, sized to the heaviest section it must fold. Stroke deserves the closer look, since bending to parallel or contacting legs eats far more travel than bending to a stated angle.

The geometry is what the standard controls. Support span is not a setting but an output: the standard derives it from former diameter and test-piece thickness, and it must hold through the bend. Fall below what that relation yields and the supports begin to hold the ends instead — the piece is drawn rather than folded, and the strain history is no longer a bend. V-block edges are radiused in proportion to specimen thickness, within a defined range. Former diameter is the product standard's to set, not the laboratory's, and the nearest one to hand is a different test.

No numeric bend rate appears in the publicly readable clauses, and no force accuracy class is prescribed: the method controls fixture geometry, not frame metrology, so a load reading is a process indication rather than a result. No extensometer is used.

What goes wrong in practice

Span drift is the classic. Rollers free to move, or supports set closer than the geometry allows, clamp the ends and quietly change the deformation mode; the piece looks bent and the strain history is wrong.

Former misalignment is the second. A former that is not centred between the supports produces an asymmetric bend, and cracks then appear off the apex where nobody is looking for them.

Edge condition is the third, and the most argued over. Rounding the edges is a requirement here, not good practice, and a piece that arrives at the jig square-cut or flame-cut has not met the method; what it then does at the bend is not evidence about the steel. Related to it: where thickness has been reduced by machining, putting the machined side in tension tests a surface the product does not have, and usually flatters it.

Related and equivalent standards

ASTM E290 is the counterpart and gives the same kind of visual verdict, but the two differ in the detail: E290 leaves span, former size and ambient conditions to the product specification, while ISO 7438 fixes the span relation and states a temperature band. ISO 7438 also organises the work by device, E290 by degree of constraint, including free bending and bend-and-flatten. EN ISO 7438:2020 is the same text adopted in Europe. Where a product standard names one of the two, its geometry governs.