Your battery plant runs automotive quality. The failure mode is aerospace.

A cell passes every end-of-line test. Every parameter within specification. Control plan signs it off, ships it. Eighteen months later it is a thermal event in a customer's driveway. Your PFMEA never modelled that path. It was built for defects you catch at the line, not defects that incubate inside electrochemistry for a year and a half before they try to burn someone's house down.

Two decades in automotive and aerospace quality. IATF 16949 in plants where the defect surfaces at end-of-line or at the customer's dock within a week. AS9100 in environments where the failure might not appear for years, in conditions no test rig can reproduce. Battery manufacturing has chosen the first playbook while the physics of its product demand the second. That gap will produce field events, and the industry will not see them coming because its quality architecture was never designed to look that far ahead.

Why automotive PFMEA logic breaks for batteries

Automotive PFMEA rests on a premise that is sound for metal stamping and injection moulding: defects are detectable, and detection is a function of where you place the control. Map failure modes, rate severity, assess occurrence, evaluate detection, multiply into a risk priority number, build your control plan around the score. The whole logic assumes the defect is visible within the production cycle or shortly after delivery.

Battery defects do not cooperate. Lithium plating builds over charge cycles. Dendritic growth is a function of time and electrochemistry, not a parameter you can probe at end-of-line. A microscopic contaminant in the separator passes every in-line test and then becomes a short circuit after 2,000 cycles in a Phoenix parking lot. The failure is real. It is manufacturing-induced. Your control plan has no horizon for it because IATF was never written to model time-delayed defect propagation.

When I built the greenfield QA/QC operation for 900+ employees at SNOP, the first principle was straightforward: the quality system has to be architected for the specific failure profile of the product, not copied from a template. We cut defect costs by 70% because the control plan matched the physics. Battery plants are doing the opposite right now. They are taking IATF templates designed for stamped metal and applying them to a product that degrades chemically over years, then acting surprised when the playbook misses.

What aerospace got right about latent defects

Aerospace learned about latent defects the hard way. Decades of pressurisation cycles, fatigue crack growth, corrosion inside sealed cavities—failure modes that live on a timeline orders of magnitude longer than the production cycle. AS9100 was built around that reality.

A defect you cannot see at the gate is still a defect when it walks through it. The question is whether your quality system was built to catch what it cannot inspect.

Four elements of that thinking are absent from battery plants. Latent defect modelling: the PFMEA has to include time-delayed failure modes with degradation curves, not just immediate defect signatures. Long-horizon qualification through accelerated life testing that compresses years of degradation into weeks of validation. Lot-level traceability that survives the product's entire service life—when a field event occurs, you reconstruct the production genealogy of that specific cell down to the material lot, the shift, the humidity in the cleanroom. And stress testing beyond specification, because aerospace tests to destruction to learn where zero margin lives. Battery plants test to spec and declare victory.

I have operated inside both systems. The difference is not academic.

The equipment trap

The battery testing equipment market is booming. Formation cyclers, abuse test rigs, thermal propagation chambers, end-of-line diagnostics—billions spent globally. New plants breaking ground across the United States, Europe, and Asia. Every one of those facilities will buy testing rigs, and most will confuse having purchased the hardware with having a quality system.

This is the trap. Testing equipment collects data. It does not predict what it was not architected to measure. A formation cycler tells you the cell's electrical performance at hour zero. Nothing about lithium plating morphology at cycle 1,500. An abuse test rig tells you how the cell responds to nail penetration under controlled laboratory conditions. Nothing about the slow thermal runaway initiated by a micro-contamination your incoming inspection never characterised because the particle sat below the detection threshold of the gauge you bought.

Plants buy capability and call it quality. At SNOP the system worked not because we bought the right gauges—though we did—but because the architecture underneath those gauges was designed for the failure modes the product would actually exhibit. The equipment was the last step in the design, not the first line of the purchase order.

Thermal incidents, field campaigns, recalls—each one will trace back to a control plan built for a world where defects show up on the line, not in a driveway eighteen months later. The quality leaders who see this now, who rebuild the battery PFMEA with aerospace-grade latent defect logic while plants are still scaling, are the ones who will not be explaining themselves to regulators. The workshop to fix this costs €25K. A single thermal field event costs €50M before the lawyers arrive.