Bend Cracking and Flange Drift on 5052-H32 Aluminium Enclosures

The two defects have different physics. Do not pool them into a single 'the die is worn' story, because that leads to a tooling fix that will not stop the cracks.

Bend cracking - the suspect set

• Geometry: inside radius 1.0 mm on 1.5 mm sheet, so r/t = 0.67. Published forming guidance for 5052-H32 puts the minimum 90° inside radius at roughly 1t, so the part is already outside normal practice before any wear is considered.
• Grain direction: bending with the bend line parallel to the rolling direction puts the outer-fibre strain across the grain - the least ductile direction of a rolled sheet. Because the blanks are split roughly 50/50 between orientations, the existing 60,000-stroke history already contains a natural experiment: sort the scrap by orientation before spending any money.
• Lubricant: the switch to a cheaper water-based product three weeks ago coincides with the rise from 2% to 7–9%. A thinner or less EP-additised film raises friction at the die radius and moves the bend line.
• Material: the mill certificate describes the coil, not the blank after slitting, and H32 is commonly substituted with H34. Ask for Vickers microhardness and a simple 90° bend test on strip from each coil in stock.

Flange drift to 12.12 mm This is a springback and tool-geometry problem, not a cracking problem. Two sharpenings change the die radius and the punch penetration depth; a thinner lubricant film raises friction and springback; coil-to-coil thickness variation scales the result directly.

48-hour measurement plan

1. 30 consecutive parts per orientation, measured for flange height on a height gauge and for cracks at 10× magnification (record crack length, not just pass/fail).
2. Die radius and punch penetration measured against the tool drawing before and after the last sharpening.
3. Sheet thickness across the coil width and hardness per coil.
4. Lubricant viscosity and film coverage at the die entry. Then re-run 100 parts with the old lubricant only, holding everything else constant - that single change isolates the most recent variable.

Radius change 1.0 → 1.5 mm The simple outer-fibre strain estimate is ε = t / (2r + t): at r = 1.0 this gives 1.5 / (2.0 + 1.5) ≈ 43%; at r = 1.5 it gives 1.5 / (3.0 + 1.5) ≈ 33%. That is roughly a 23% reduction in the elongation demanded of the outer fibre.

Caveat for the report: the simple formula overstates the absolute strain, because the neutral axis shifts towards the inside of the bend. So do not present 43% as if it should be compared with the 12% tensile elongation of H32 - that comparison will be challenged. The valid message is the relative improvement and the move to r/t = 1.0, which brings the part back inside normal practice.

Cost and control: this is a die insert change and a change to a customer-controlled drawing. It must go through a formal drawing change request or a written deviation approved by the customer's design centre - not 'deal with it later'.

Rotating the blanks 90° Technically correct: with the bend line perpendicular to the rolling direction, the outer-fibre strain runs along the grain, which is the higher-ductility direction. The penalty is strip nesting - expect a change in material utilisation, possibly a different coil width and a new strip layout. Check the scrap data first: if the two orientations already perform differently, this change alone may solve the cracks without touching the radius.

Validation before submission

1. A 2³ factorial on the press: radius (1.0 / 1.5) × orientation (parallel / perpendicular) × lubricant (old / new), 30 consecutive parts per cell, run in randomised order.
2. Responses: cracks per 100 parts at 10× magnification, and flange height with its Cpk.
3. Select the cell that meets both requirements at the lowest cost, then run 300 parts to establish capability on flange height.
4. Present the customer with the data, the drawing change request and the revised control plan together - a change request without capability evidence will simply be rejected.

Position: 100% visual inspection is containment, not control, and cannot be offered as conformance.

Why it fails as a control:

• Visual inspection under a bench lamp detects through-cracks. It does not reliably detect incipient cracks, and it does nothing at all about flange height, which is already at 12.12 mm against 12.00 ± 0.10 mm. That is nonconforming product, and sorting cannot make nonconforming product conforming.
• There is no gauge R&R on the visual check, so you cannot state an escape rate or defend the method.
• Sorting without a written method, acceptance criteria, qualified and tested inspectors (with periodic re-qualification against boundary samples) and records is not a defensible containment. It is a claim.
• PPAP requires capability evidence from a validated process, not a sorting result.

What you can legitimately do with one week of stock cover:

1. Keep the press running only while output can be quarantined and clearly identified by lot.
2. Move immediately to the trial setting (radius, orientation, lubricant) selected from the designed experiment.
3. Run containment to a written standard, with 10× magnification, boundary samples, a dimensional check on every tray, and a signed record per lot.

What the customer must be told before the source inspection:

• That a dimensional nonconformance exists on flange height for material produced since the second sharpening, with quantities and traceability.
• The containment method and evidence, the root cause, and the corrective-action timeline.
• A request for a written deviation or concession for the affected lots.

Do not ship and hope. Concealing a known nonconformance in order to get through a customer inspection is misrepresentation, and it is indefensible whether or not the customer ever discovers it.