Happy to review before it turns into a launch delay.
We were a few weeks into initial validation on a mid-range ISR drone. One subsystem — a flight computer enclosure — had started kicking off QA flags. Nothing catastrophic, just a slow drip of mounting failures during final torque: a warped PCB here, a misaligned connector there.
Eventually, we traced it back to a stamped stainless bracket that secured the board stack. Dimensions were within spec, the material passed cert, plating was uniform. But the bracket didn’t fit right once it was installed.
The culprit? The stamped slots were aligned to a legacy reference edge — reused from a previous platform. On paper, they looked fine. But the new enclosure’s mounting features were clocked just differently enough to pull everything out of alignment once torqued.
No one had caught it in FAI. And by the time it showed up in validation, the tooling was locked.
The bracket supplier had done their job. They built what was on the print. But no one had stepped back to ask: does this slot orientation still match the real-world assembly?
It was a tooling-driven layout — makes sense for punch sequencing. But it didn’t align with the control unit’s mechanical stack-up. As a result, the entire system was fighting its own geometry under torque.
And here’s the kicker: the drawing didn’t clearly define the slot datum, only the tolerances. No GD&T callouts, no fit interface. The legacy CAD model had skipped over the nuance.
This kind of miss doesn’t get flagged in APQP. It gets flagged when a technician can’t close the lid.
After we froze the part for rework, we asked the supplier to send over the die progression — just to confirm what had gone wrong. Sure enough, slot features were indexed to a front edge that made sense in die but not in use.
We reworked the tooling — shifted the punch reference, reverified slot position to the actual mounting datum. It wasn’t a heroic save. Just methodical. Three weeks lost, but better than letting it roll into field units.
That also pushed us to standardize new part validations: not just dimensional checks, but physical mockups. Brackets now get checked in the assembly jig before we sign off tooling. And we write tighter notes on slot tolerances and plating masks for fit zones.
Looking back, here’s what I now advise procurement to do on UAV subcomponents:
1. Ask for a functional FAI. Not just a CMM report. Include jig-fit or pilot install — especially for brackets, terminals, and EMI covers.
2. Require die progression reviews. Even a basic 2D die strip can reveal punch direction, feature order, and tolerance risks.
3. Flag reused prints for revalidation. If the part comes from an older platform, don’t assume it still fits. Geometry shifts between iterations are common.
Happy to review before it turns into a launch delay.