What UL 2271, UN 38.3, and IEC 62133-2 Prove on a Robot Pack

UL 2271, UN 38.3, and IEC 62133-2 each answer a different question, and a robot pack needs all three answers kept separate. UL 2271 addresses the safety of the battery system as a product. UN 38.3 addresses whether that pack can move in transport, and it is documented by a test summary rather than a certificate issued by anyone. IEC 62133-2 addresses cell-and-pack safety against the international standard a customer may name in its own purchase requirement. Put them in the RFQ as three separate lines with three named deliverables, then check each document against the exact model, cell, and configuration on the quote.

The submittal that taught me this arrived as one tab in a shared drive. One PDF, forty-odd pages, filename ending in compliancepackfinalv3, with a cover page carrying a UL mark. Around page nine sat a UN 38.3 test summary. Page thirty was a CB test certificate. The supplier's program manager had written across the top: "Fully certified — UL, UN, IEC."

Pull the quote up beside it. That is the entire exercise, and it takes about four minutes if you know where to look. The UL certificate named a model family, a base designation followed by a capacity range. The quote was for one capacity at the bottom of that range, with a BMS from a different vendor than the one listed in the certificate's critical components table. The UN summary listed a cell part number and a pack designation that did not appear anywhere in the quote. The IEC certificate was three years old and written to the original issue of the standard.

Nothing there was fraudulent. It was worse than fraudulent, in a way — it was sincere and useless. Three documents, three different questions answered, none of them answering the question I was actually asking, which was: are the packs arriving at my dock the same packs that got tested?

What UN 38.3 actually tells you

UN 38.3 is a transport safety test. It exists because a lithium battery in a cargo hold is a fire risk to the aircraft and everyone in it, and the regulations require evidence that the design survives the ordinary abuses of shipping. The document you receive is a test summary — eight test results, sequentially numbered, with the cell and battery identities written into it and a date. Nobody stamps it. It is not a certificate in the sense a UL file is a certificate. It is a manufacturer's summary of testing, and the transport regulations specify what has to be in it.

The window it covers is short. A few days in a freight lane. Altitude, thermal cycling, vibration, shock, short circuit, crush, overcharge, forced discharge. What it does not cover is the next six years of a pack riding in an AMR that does 40 km a shift in a freezer aisle. It says nothing about whether your enclosure survives a tip-over, whether the BMS disconnects at the right threshold, or whether the charger you paired it with is doing something the tested configuration never anticipated.

So when a supplier hands me a UN summary as general proof of safety, I put it down gently. It is proof of shippability. That is not nothing — you cannot move a container without it, and a summary missing the specific pack designation will stop a shipment at the freight forwarder's desk. It is just a narrow proof, and it is the one document of the three that has a hard operational deadline attached to it.

What UL 2271 is buying you

UL 2271 is a product safety certification for the battery system in a light electric vehicle. The unit of evaluation is the pack, not the cell — construction, enclosure, BMS function, protection against overcharge, overdischarge, and overtemperature, mechanical abuse, environmental exposure. Somebody sat with the design and asked what happens when this thing is dropped, crushed, charged wrong, or left on a dock in August.

The part buyers underestimate is the follow-up service. A listing comes with factory surveillance — the production site gets visited, and the sample on the test bench has to keep resembling what ships. That is a fundamentally different kind of assurance from a one-time type test, and it starts to matter around the second purchase order, when the pilot units are gone and you are buying two hundred packs a year from a line you have never stood in front of.

The limitation is in the name. Light electric vehicle. An industrial AMR is not a scooter, and the scope conversation with the certifier is real — duty cycle, vibration profile, operating environment, all of it gets argued. What matters to me on the buyer's side is not the argument, it is the model list. The certificate has to name your part number, or a family whose constructional limits genuinely contain your configuration. "Similar construction" is a sentence that should make you ask for the file and the critical components list.

What IEC 62133-2 is, and who is asking for it

IEC 62133-2 is the international safety standard for secondary cells and batteries, covering requirements at both cell and pack level. If your customer is a European OEM or an Asian integrator, there is a decent chance their own purchase requirement names it, and their quality function will ask for a CB test certificate with national deviations accounted for. That is the practical reality: this is the standard that shows up in someone else's specification, not necessarily in yours.

Which means it cannot be substituted. A UL 2271 file does not satisfy a customer who wrote IEC 62133-2 into their contract, and a CB certificate to IEC 62133-2 does not satisfy a North American customer who wants a listing with follow-up service. Both can be true of the same pack at the same time. Neither proves the other.

The check I run on every document

Same three columns every time. Part number, cell, BMS.

On the part number: does the document name the thing on the quote, exactly, including capacity and connector variant? A family designation with a range is not the same as your part. On the cell: is the cell make and model the one your supplier plans to build with, or the one they built with eighteen months ago before a sourcing change? On the BMS: this is where it goes wrong most often, because BMS boards get revised quietly, firmware gets updated, and a protection IC swap can move you outside the tested configuration without anyone thinking of it as a design change.

Then two more. Standard revision — the current issue of each standard is printed on the certificate, and an old issue is not automatically disqualifying but is a conversation. And scope: the page count of the certificate matters. The model list page is the page you read.

How it goes into the RFQ

Three lines, worded so that a supplier cannot answer two of them with one document.

• UL 2271 certificate of compliance, including the model list pages that cover the proposed part number, plus the file number and issue date.
• UN 38.3 test summary for the pack itself — not the cell alone — with the pack designation matching the quoted part number, current to the design being offered.
• IEC 62133-2 CB test certificate with national deviations identified, if the customer or the destination market requires it.

Then the clause that ties it together: written change notification, with re-testing at the supplier's cost, before any substitution of cell, BMS, protection device, or enclosure material. Because every one of those three documents is a snapshot of one configuration, and the failure mode I actually worry about is not a missing certificate. It is a valid certificate for a pack that is no longer the pack.

One more housekeeping point that has bitten me: hold the documents yourself. Not a link to the supplier's portal, not an email attachment from a program manager who leaves in March. Copies, in your own system, indexed by part number and revision. Auditors and freight forwarders do not accept "I can get it from them."

What none of the three will tell you

Field durability. Cycle life at your actual duty cycle, which is almost never the cycle life on the datasheet. Whether the BMS talks to your fleet controller the way the integration guide claims. Thermal behavior inside your enclosure, in your aisle, at your ambient. Production consistency across lots. Second-source cell qualification.

Those come from the pilot build and the factory audit, and no certificate substitutes for them. The three documents tell you the pack is not obviously dangerous, is legally movable, and meets the standard your customer named. That is a floor, and it is a floor worth having. It is not a performance claim.

So the stack on my desk stays three separate stacks. When I sign the review sheet at the top of a submittal, I write the same line every time: three questions, three answers, same part number on all three. When those line up, we have something to work with. When they do not, I would rather find out in a conference room than in a loading dock.