Baselining a Three-Chemistry Car Deck

Commissioning a detection layer on a deck that carries one battery chemistry is a solved problem. This exercise covered a deck that carries three — NMC, LFP and sodium-ion — with nothing on the manifest to say which vehicle is which, and a cargo mix that changes every voyage. The finding is structural rather than numerical: on a mixed deck you cannot baseline per chemistry at all.

What a baseline can and cannot learn

It characterises background, not signal — and confusing the two is the trap. On a loaded commercial vessel you cannot stage a thermal runaway to see what one looks like, so nothing observed during a commissioning window is an event. What the window actually produces is a picture of normal: the thermal and gas conditions a deck presents when nothing is wrong, across ventilation states, loading operations, weather and cargo mix. That is genuinely valuable, because an alarm threshold is meaningless without it. But it is only half the problem. Whether the resulting envelope would catch a real event is a question the bench rig and its staged-event catalogue answer, not the ship. The two are complementary and neither substitutes for the other.

You cannot baseline per chemistry

Because the ship is never told which chemistry it is carrying. Under IMDG Special Provision 961, vehicles carried in flag-approved ro-ro vehicle spaces sit outside most of the Code where its conditions are met, so no chemistry, state of charge or condition is declared. The manifest is a vehicle count. That has a hard consequence for commissioning: there is no way to tag a deck position as NMC, LFP or sodium-ion and tune it accordingly, because the tag does not exist and would change on the next voyage anyway. So the baseline is established per position — what this location on this deck looks like under normal conditions — and the alarm envelope is then set not to the cargo average but to the weakest signal that could plausibly be sitting there. On a three-chemistry deck, that is a materially different number from the two-chemistry case.

The floor moved when sodium-ion entered the mix

Adding a third chemistry did not widen the envelope evenly — it tightened it on two axes at once. Under a two-chemistry assumption the weakest case was LFP: cooler than NMC at around 446°C peak vent-gas temperature against 1050°C, but with a hydrogen-rich vent stream at roughly 41% that gave the gas channel plenty to work with. Sodium-ion is weaker on both. Peak vent-gas temperature around 265°C erodes the thermal channel considerably further, and a hydrogen fraction near 15% removes the compensation LFP provided. An envelope set for LFP is therefore not automatically adequate once sodium-ion is in the cargo mix, and the commissioning exercise had to be re-run against the lower floor rather than inherited from the earlier two-chemistry configuration.

What the commissioning regime became

• Baseline per deck position across multiple voyages rather than a single window, so the picture of normal spans several cargo mixes instead of one.
• Set the envelope against the weakest chemistry that could be present on the trade, and record which chemistry that assumption names — an envelope whose basis is undocumented cannot be reviewed when the mix changes.
• Treat a change of trade as a trigger to re-examine the baseline, not merely a change of route. Different rotations carry different mixes and therefore different backgrounds.
• Keep the staged-event catalogue and the shipboard baseline as separate artefacts. The ship tells you what normal looks like; only staged testing tells you whether the envelope catches an event.
• Extend the bench catalogue to cover the chemistry actually being carried. A catalogue built when the fleet was NMC-and-LFP does not by itself demonstrate coverage of a sodium-ion event, and that gap should be closed deliberately rather than assumed away.

Sources

• RoRoSAFE mixed-chemistry commissioning record (operator under NDA) — Asia–Europe finished-vehicle rotation, two enclosed vehicle decks, baseline established per deck position across multiple voyages with varying cargo mix; alarm envelope set to the weakest chemistry that could be present rather than to the cargo average.
• Chemistry comparison figures — peak vent-gas temperature approximately 265°C for NVPF sodium-ion, 446°C for LFP and 1050°C for NMC; hydrogen fraction of the vent stream approximately 15% for sodium-ion, 41% for LFP and 19% for NMC; per-Ah vent-gas volume approximately 0.05, 0.02 and 0.07 mol/Ah respectively — Batteries (MDPI), 2025, 11(9):323.
• IMDG Code Special Provision 961: vehicles carried on ro-ro ships with flag-approved vehicle spaces fall outside most of the Code where its conditions are met — the reason no chemistry, state of charge or condition reaches the ship, and therefore the reason per-chemistry baselining is not available — imo.org.
• IMDG Code Amendment 42-24: sodium-ion powered vehicles carried under their own UN entry, bringing a third chemistry formally into the cargo mix — imo.org.
• Companion RoRoSAFE analysis — 'Do Sodium-Ion Cells Need New Thresholds?' (the physics that moved the floor), 'Inside the Bench-Rig Validation Program' (the staged-event catalogue this exercise deliberately does not substitute for), and 'LFP vs NMC: Tuning Detection Thresholds' (the two-chemistry envelope this replaces).