Common build failures: what actually goes wrong on the shop floor

Two screwups originate before heat treat and show up as failures in service.

Sharp inside corner that propagates a fatigue crack. A fatigue crack initiates at a specific cavity corner, propagates radially under thermal cycling, and the die scraps at 20-30K cycles instead of the 100K target. The root cause has two paths. The designer specced R 1.5 mm when the application needed R 3 mm and the analysis was never done. Or the print specced R 3 mm and the machinist or EDM electrode produced R 0.8 mm at the bottom of the cavity because tool reach, electrode wear, or finish-grind cleanup all undercut the nominal radius. The inspection is cavity dimensional check against print before heat treat and again before nitride. Silicone replica or optical CMM on critical radii. 3D corners need a sphere of radius, not just two intersecting 2D radii, and that is exactly where the field-cracked dies have their initiation sites. R 2 mm in two views can collapse to a sharp point at the intersection. Verify the as-built radius, not the print radius.

Stress relief skipped between rough and finish machining. The die warps unpredictably during heat treat. Dimensions that were on spec before the HT cycle are off by 0.005-0.015 inch after. Critical cavity features end up undersize on one axis and oversize on another. The root cause is rough machining left residual stress in the block, the stress relieved during the austenitizing soak, and the part warped to a new equilibrium shape. The stress was always there. The HT cycle made it visible. The inspection is a dimensional check after rough machining and a stress relief sign-off on the traveler before the part moves to finish machining. Stress relief is 650-700°C for 2-4 hours, slow cool, documented with a cycle trace. The check is whether the stress relief actually ran, not whether someone wrote it on the traveler. On asymmetric or thin-section parts, a second stress relief between semi-finish and HT is cheap insurance.

Both screwups share a pattern: the build looks fine on paper but the geometry the part actually has differs from the geometry on the print, either because the as-built feature is wrong or because the residual stress state warps the feature to a new position during HT.