Foreign Material in Canned Foods: Metal Detection Failure Analysis
A metal detector that passes its test pieces can still miss real contamination. This reconstruction follows a fictional cannery through a customer complaint, a failed rejection mechanism, and the uncomfortable discovery that “validated” equipment had never actually been challenged.
Background
HarvestLine Foods — a fictional canned-vegetable processor in the Pacific Northwest, running sweet corn and green beans on a single high-speed line at 400 cans per minute — operated a pipeline metal detector positioned after the seamer and before the retort. The unit was a balanced-coil system with automatic reject, tested at the start of each shift and every 2 hours with ferrous, non-ferrous, and stainless-steel test wands at 1.5 mm, 2.0 mm, and 2.5 mm respectively. Logs showed every test passing for the eleven months since installation.
In early October, a retail customer returned a complaint with physical evidence: a 6 mm fragment of stainless-steel wire embedded in a can of green beans, photographed and retained by the customer’s QA lab. A second complaint arrived eleven days later — a different lot, same product, a 4 mm ferrous sliver. Two foreign-material complaints with retained samples in under two weeks triggered the plant’s serious-incident protocol.
Investigation
The first assumption was that the detector had drifted out of calibration. It hadn’t — the unit passed all three test pieces at the start of the complaint investigation, exactly as the logs said it would. That was the first real clue: the detector could find the test wands but had missed real fragments.
QA reconstructed the fragment’s likely journey. The stainless wire matched the diameter and alloy of the mesh on the plant’s blanching conveyor — and maintenance records showed a 30 cm section of that mesh had been replaced in September after a tear. Ferrous slivers pointed downstream: the can seamer’s first-operation roll had been reground in August, and fine metal dust from seamer tooling is a classic canned-foods contaminant. Both sources sat upstream of the detector, so in theory both should have been caught.
The breakthrough came from orientation testing. The plant’s QA manager ran the retained 6 mm wire fragment — and thirty more wire pieces cut to match — through the detector in different orientations, embedded in product cans. Fragments passing through the aperture aligned with the coil axis, the so-called “blind orientation,” were detected only 40% of the time. Test wands, mounted in rigid plastic cards, always crossed the aperture in the most detectable orientation. The equipment had been validated against a geometry that real contaminants never respect. Worse, the reject mechanism itself was degraded: timing trials showed the air-blast reject firing 180 milliseconds late at full line speed, pushing borderline cans past the reject gate before the blast arrived — a fault invisible during the slow, careful pace of test-piece verification.
Root Cause
The root cause was a validation gap, not an equipment failure in the narrow sense. The metal detector’s installation qualification had used only the manufacturer’s standard test cards in a single orientation, and the plant’s ongoing verification copied that method exactly — shift after shift, for eleven months. Nobody had ever challenged the system with realistic contaminants in realistic orientations. The late-firing reject was a maintenance issue (a partially blocked solenoid valve, dropping air pressure to the blast nozzle), but it had gone undetected because verification was always performed at reduced line speed, where the timing error didn’t matter.
Corrective Actions
The fix started upstream. The blanching conveyor mesh was replaced plant-wide with a heavier-gauge specification and put on a monthly integrity check; seamer tooling went on a wear-measurement schedule with replacement limits instead of run-to-failure. On the detection side, the plant commissioned a proper validation study: 200 seeded cans with wire fragments, shavings, and screen pieces in random orientations, run at full line speed. Sensitivity was re-established at 1.2 mm ferrous, 1.5 mm non-ferrous, and 2.0 mm stainless in all orientations — a tighter standard than the original. The reject solenoid was replaced, air pressure to the reject was put on a monitored alarm, and — the change that mattered most — the every-2-hour verification now runs test pieces through in three orientations at full operating speed, with a seeded “challenge can” of unknown-to-the-operator timing once per shift. Product hold: the two complaint lots and all production between the last good orientation challenge and the fix — roughly 1.2 million cans — were 100% re-screened through an offline detector; 14 additional contaminated cans were found and destroyed.
Lessons Learned
Verification is only as good as the challenge it poses. A detector that passes its test pieces in one easy orientation, at walking pace, has proven it can detect the test piece — nothing more. Real validation means realistic contaminants, realistic orientations, and realistic line speed, a principle embedded in the Codex General Principles of Food Hygiene and every credible HACCP text: monitoring must reflect operating conditions. The second lesson is about the reject, not the detector: most metal-detection programs audit the sensing head and ignore the kick-off mechanism, yet a perfect signal with a late reject is the same as no detector at all. Time your reject at speed, or you haven’t verified anything.
Educational reconstruction. All facility names, people, and production figures are fictional.
