Detecting EV Faults in the Yard, Not at Sea
Four shore-side pilots — OEM yard, port terminal, BMS-pull, battery warehouse — show detection catching EV faults before the cargo reaches a deck.
The cheapest place to catch a vehicle-deck fire is before the vehicle is on the deck. Four shore-side pilots put the marine detection architecture on land — an OEM finished-vehicle yard holding roughly 14,000 cars, a port pre-load terminal, a closed-loop BMS-data-pull trial, and an adjacent battery-storage warehouse — and each caught, or was built to catch, a fault while the cargo was still under terminal control and a fault could still simply be driven away.
Why the pre-load yard is the gap
Between the production line and the ship, EVs sit unmonitored. A port terminal's pre-load yard holds vehicles for 3–14 days awaiting a sailing, often with no thermal coverage at all — a fault there would be found by sight or smoke, with the nearest CCTV operator terminals away. An OEM's finished-vehicle yard is the same blind spot at larger scale: thermal coverage on the production line and on cell-stock racks, but nothing on the 14,000 cars in rotation between them. Both pilots closed that gap with the same sensor cells and segment masters used at sea, mounted on relocatable columns instead of deck members, with per-row coverage and longer baseline windows suited to a yard that is far more thermally stable than a rolling deck.
What the yard pilots caught
The OEM yard caught two confirmed early-warnings in 90 days with zero false alarms after a two-week calibration, both handed to the incident-response team and isolated before any visible signature; that pilot is expanding to a second plant in 2026 H2. The port terminal, running twelve sensor columns over about 2,400 vehicles, caught one pre-load EV before any smoke and had it isolated and removed from the yard within eighteen minutes. In a yard, unlike at sea, a caught fault does not need firefighting — the vehicle is simply driven away.
The BMS-pull closed loop
The pre-load yard is also where pre-fire detection can become a closed loop. In a trial with an OEM and a terminal operator, per-vehicle BMS state — state-of-charge, cell-voltage spread, internal temperature — was pulled and fused with the yard's external thermal layer across roughly 2,800 vehicles. On four staged, BMS-instrumented abuse events, three were detected at Stage 1 from the BMS data alone — internal cell anomalies no external sensor can see — and all four at Stage 2 by external cells after the BMS trip. Fusing the internal and external signals tightened the false-positive rate sharply, because the two sensor classes corroborate or contradict each other within seconds. Where a single OEM's BMS is reachable and the cargo is under terminal control, that loop is achievable now; the shipboard equivalent waits on the contractual frameworks to catch up.
How far the architecture generalises
The same rolling-baseline, cross-cell-coherence and segment-master-ring architecture also transferred to a problem with no vehicles at all. A 60-day battery-storage-warehouse pilot of roughly 280 modules adapted the baseline to per-module rather than per-vehicle: the rolling baseline, coherence and segment ring carried across, while the vehicle-grade IR optics did not, because racks sit much closer to the sensor than a car does. It recorded zero events but produced a clean baseline dataset that informed the second-generation cell optics. RoRoSafe is not pursuing stationary storage as a market directly, but the pilot showed the architecture generalises cleanly where the geometry differs.
What it means for owners and OEMs
For terminal operators and OEMs, the shore-side pilots reframe where detection pays off: a fault caught in the yard is a vehicle driven aside, not a casualty fought at sea. For the maritime chain, the pre-load yard is also the natural place to verify pre-shipment state-of-charge and to prove the BMS-plus-external fusion the shipboard closed loop will eventually depend on. Detection that begins at the terminal rather than the loading ramp shortens the risk window on both sides of loading — and catches the used or damaged battery the manifest never declared before it is ever lashed to a deck.
Sources
- RoRoSafe deployment records (clients under NDA) — OEM finished-vehicle yard: ~14,000 vehicles in rotation across six yard blocks, two confirmed early-warnings in 90 days, zero false alarms after two-week calibration; expanding to a second plant 2026 H2.
- RoRoSafe deployment records — European port vehicle terminal: ~2,400 vehicles, twelve sensor columns, one confirmed pre-load early-warning isolated and removed within 18 minutes.
- RoRoSafe deployment records — pre-loading BMS-pull trial (OEM + terminal, Asia-Pacific): ~2,800 vehicles; 3 of 4 staged events detected at Stage 1 from BMS data alone, 4 of 4 at Stage 2 by external sensors; internal+external fusion tightened the false-positive rate.
- RoRoSafe deployment records — battery-storage-warehouse adjacent pilot: ~280 modules, 60-day run, zero events; baseline dataset informed second-generation cell optics.
- IMO SOLAS Chapter II-2 (2026) and IUMI EV-carriage guidance — the pre-loading interface (state-of-charge, damaged-cell screening) as a recognised risk window. [VERIFY: all pilot figures are internal RoRoSafe results, clients under NDA, not independently published.]
Questions, answered
Can EV fires be detected before vehicles are loaded onto a ship?+
Yes — that is the point of shore-side monitoring. Vehicles sit for days in pre-load yards with no thermal coverage. In these pilots the marine detection architecture, mounted on relocatable columns, caught three pre-load early-warnings across an OEM yard and a port terminal, each before any visible signature. A fault caught in the yard is isolated by driving the vehicle away rather than fighting a fire at sea.
How does yard monitoring differ from shipboard detection?+
It uses the same sensor cells and segment masters but mounted on relocatable columns instead of deck members, with per-row rather than per-vehicle coverage and longer baseline windows, because a yard is far more thermally stable than a rolling deck. The detection logic and calibration approach are the same; the mounting, coverage granularity and baseline timing are adapted to the environment.
What does pulling BMS data add to yard detection?+
Internal visibility. Pulling per-vehicle state-of-charge, cell-voltage spread and internal temperature exposes Stage-1 cell anomalies no external sensor can see. In the trial, 3 of 4 staged events were caught at Stage 1 from BMS data alone, and fusing internal with external signals tightened the false-positive rate because the two sensor classes corroborate or contradict each other within seconds.
Does the detection architecture work outside vehicles, like battery storage?+
It generalises. A 60-day battery-storage-warehouse pilot adapted the rolling-baseline approach to per-module rather than per-vehicle; the baseline, cross-cell coherence and segment-master ring transferred, while vehicle-grade IR optics did not, since racks sit closer to the sensor. It caught no events but produced a clean dataset. RoRoSafe is not pursuing stationary storage directly, but the architecture transfers where the geometry differs.
Continue the thread
The BMS Handshake: Pre-Fire Shutdown
In ESS markets, off-gas vendors tie the sensor into the BMS to shut down the affected stack. The marine analogue is harder — but the principle holds.
Used-EV Terminal: Pre-Loading Screening
A trial screening mixed-make used EVs at unknown state-of-health at an export terminal — external thermal and damage checks before they reached the deck.
The Argument Over State-of-Charge at Loading
No public standard sets the state of charge of an EV as it rolls onto a vehicle carrier. The number is argued hard behind closed doors — and it matters.