The Containment Step That Saves You When a Battery Pack Fails Incoming Inspection

Quarantine the pack before anyone opens it, attach a numbered nonconformance tag, photograph the as-received condition, and preserve the exact test record that failed it. All four, inside the first shift. That sequence is the containment step, and it is the difference between a supplier defect and an internal escape, meaning a bad pack that gets absorbed into your build, ships bolted inside a robot, and fails in front of someone who paid for the machine.

I have stood at that bench with the crate lid half off and a line lead behind me asking how long. The pack is on the ESD mat, the tester is still showing red, the foam is on the floor where it fell, and the pull to grab a screwdriver is enormous, because opening the thing feels like progress.

It is not progress. The moment the housing comes off, you have destroyed the only evidence that was free.

A tag is a fence, not paperwork

You walk the pack to the quarantine area. Not the corner by the wall, not the top of a cabinet, an actual marked bay with tape on the floor, a sign, and a log sheet, so that a second-shift material handler cannot mistake it for WIP and roll it to the line. Then the tag, zip-tied to the lift handle and not taped to a surface you intend to photograph.

What goes on it: NCR number, date and time received, PO and line item, supplier, part number and revision, date code and serials, quantity received, quantity failed, the test ID that failed and the actual reading, your initials, and the word HOLD. That is not bureaucracy. That is the object that survives the week when three people have all touched the pack and nobody remembers who did what.

While you are at it, hold the rest of the receipt. The single most common containment mistake I have watched is one unit getting tagged while the other eleven on the pallet get pulled to the line because the build schedule was already tight. You have not contained anything until you have stopped the shipment, not the unit.

And put it somewhere that a thermal event would be survivable. A suspect pack on a shelf above cardboard is not quarantined, it is staged.

Photograph the pack you received, not the one you expected

You need the outer carton on all six sides, labels, any crush or puncture, the shock and tilt indicators, the humidity indicator card, and the dunnage exactly as it lay when the lid came off. Then the inner packaging, then the pack from four sides, the terminal block, a macro shot of the connector pins, the date code stamps, the BMS board revision sticker, the sealant bead, the mounting bosses. Anything dimensional gets calipers in the frame with the reading visible.

I do this before the pack goes to the test bench, not after, because the second unit out of the same carton is the one worth filming as you open it, and you only get one first opening.

The photos go into the NCR record the same day. A camera roll on somebody's phone is not a record, it is an anecdote with a timestamp.

Four ways a pack fails at the dock

Shipping damage. Crushed dunnage on one corner, a tripped shock indicator, a hairline crack at a mounting boss, a terminal lug bent two degrees, a connector shell with a chip out of the keyway. Look at the pallet and the stretch wrap, not only the pack. The pack tells you what happened; the packaging tells you how hard.

Cell voltage outliers. You rest it and measure again before you call anything. A pack reading low after three weeks in a container is telling you about self-discharge, not about your meter. If the spread between cell groups widens across a rest period, that points at cells. If it snaps back to nominal the moment you re-seat the probe, it points at your measurement path, and you should be glad you found that out before you wrote the SCAR.

BMS communication faults. Before the pack is condemned, you fully seat the connector, swap the harness, put a known-good pack on the same channel, and confirm the tester itself is not the problem. A meaningful share of the communication faults I have chased back were a bent pin in our own cable or a wake sequence the fixture never asserted. That check takes ten minutes and has saved me two embarrassing escalations.

Dimensional nonconformance. Hole pattern, overall height against the bay, connector clocking, weight. Measured against the drawing revision you actually released, with that revision number written on the NCR, because arguing about a tolerance against the wrong revision is a special kind of wasted month.

The question that decides whether it is one pack or a hundred

Every escalation I send opens with the same question: is this unit-specific or lot-specific?

It is the question suppliers least want to answer, and it is the one that determines whether you have a bad afternoon or a fleet problem. While they work on it, you do your half without waiting: pull every pack from that receipt, trace the serials already installed from previous receipts in the same date code range, and decide whether those units need inspection in the field. That costs real money and it is still cheaper than finding out from a customer.

Escalation with a clock on it

A phone call is not an escalation. You send a written SCAR or 8D request, dated, with the failure data and photos attached, and you ask for three separate things: their containment, meaning what they are doing to their own stock and their other shipments from that build window; a root cause with the method named; and a corrective action with an owner and a date. Give them a response due date and a date for closing the 8D.

Then log the date you notified them. RMA and DOA windows are frequently counted from receipt of goods, and a supplier who is being difficult will absolutely count them that way.

When the line cannot wait, write the concession down

Sometimes you have one failed unit out of two hundred, a single-source supplier, and a line that stops costing more per hour than the whole pack. Releasing the remainder of the lot under a documented engineering concession can be the right call. The concession gets signed, the serials get listed, incoming inspection gets tightened, and the failed unit stays on the shelf no matter who asks for it.

The concession is not the failure. The undocumented concession is.

Which is why the DOA and RMA language matters before you need it: who pays freight, how long the supplier has to evaluate, whether you get a replacement or a credit, and whether the failed pack goes back at all or is scrapped in place. Shipping a failed lithium pack back across the country has its own hazards and paperwork, and negotiating that while a unit is sitting in quarantine is negotiating from the worst possible position.

Two weeks from now, when the supplier says they have never seen this before, you will walk to that shelf. The tag is on the handle, the photos are in the record, the shock indicator is in a bag with a date on it, and the SCAR has a due date on it. That is the whole trick. You are not building a case, you are just refusing to throw away the evidence while it is still free.