Fraunhofer automates EV battery X-ray analysis — 5 min to evaluate, ~15 min per car, our maths says 32 cars a shift

What is new: the software, not the hardware

The AIR system (Antriebsbatterieinspektion mittels Röntgentechnik, traction battery inspection using X-ray technology) is not new. Fraunhofer presented the hardware in 2024 and had already run it past more than 100 privately owned electric cars. What electrive reported on 22 September is the piece that was missing: a software library that evaluates the 2D projections by itself.

The software locates cylindrical, prismatic and pouch cells inside the pack and compares cell-edge geometry against a reference. Displaced cells, overlapping edges, bulges and swelling then show up as deviations rather than as something a human has to squint at for an hour. The inspection geometry looks like a car wash: an X-ray source above the vehicle, a detector in the floor, and the pack is penetrated in place.

Timing matters here. Fraunhofer states around ten minutes to scan a complete vehicle, and roughly five minutes for the automated evaluation of one pack in its test environment. Actual total time depends on pack size, measurement setup and the number of projections required. Note the ratio: the software consumed five of roughly fifteen minutes. The bottleneck is physics and vehicle handling, not image processing — a detail worth remembering the next time someone sells you "AI diagnostics" as a speed problem.

The crashed ID.3, and the pack that failed the same test

The demonstration case is a VW ID.3 driven at 79 km/h into the rear of a stationary VW Tiguan under laboratory conditions at DEKRA. Despite heavy deformation at the front, the software found the pouch cells in the nine battery modules and compared the curvature of selected cell edges: no critical deviations, no overlaps, no bulges.

But the images did not clear the pack on their own. The final decision combined a hazard evaluation under DEKRA guidelines, Event Data Recorder and BMS data, plus further electrical tests — and only then was the battery classified in the lowest EUCAR hazard level 0 and approved for further use. A second, cylindrical-cell pack from a different crash produced a clearly deformed area with cells shifted against the reference. No swelling was found, yet the pack was still not recommended for reuse because of the mechanical deformation.

That contrast is the whole point of the method. Swelling argues itself; displacement is a geometry argument. State of Health will not tell you either of them.

Our maths: 15 minutes per car does not scale to a fleet

Aviloo recently evaluated more than 500,000 EVs and found SoH varying significantly even between identical models with comparable mileage. Electrical testing does that at minutes per car, with a dongle and no shielded hall.

Now run the X-ray numbers. Ten minutes of scanning plus five minutes of evaluation is 15 minutes per vehicle, ignoring set-up, marshalling, cool-down and the fact that not every car arrives with a clean reference dataset: that is four cars per hour, 32 cars in an eight-hour shift, or about 8,000 packs per year per bay at 250 working days. To radiograph Aviloo's 500,000 vehicles you would need one bay running for 62 years. Ten parallel bays bring that to roughly six years of single-shift work — and you still have not touched the millions of BEVs already on European roads.

So the value sits in the tail, not in the fleet: accident-damaged cars, insurance write-offs, packs being triaged before second life or recycling. We have run this kind of arithmetic on battery and charging claims before — Sunwoda's 15C pack works out to 5.8C once you apply the stated 10–97% window, and Geely's four-minute charge claim assumes a 2.2 MW connection most sites do not have. Radiography has the same shape of problem: brilliant physics, awkward throughput. The same logic applies to reuse economics, as our look at CATL and Octopus's truck battery swapping showed — a pack only earns money in a second application if it is mechanically sound enough to survive repeated handling.

What a used-EV buyer should take from this

An SoH certificate is a capacity statement, not a safety statement. Ageing, vibration or an undeclared crash can shift modules without moving the SoH needle by a single percentage point. If you are buying a used EV in the EU, ask for the crash and insurance history, look for underfloor repair invoices, and treat a suspiciously cheap car with a spotless SoH report as exactly that. Our EV catalog lists pack sizes and WLTP ranges for 67 models, so you can at least check what the percentage on the certificate is a percentage of.

Battery passport, STK and what Europe will actually get

The EU Battery Regulation's digital battery passport, applicable from 18 February 2027 for EV batteries, will carry SoH and other lifecycle data — not radiographs. And anyone hoping for X-ray at their next Czech STK or German HU should hold: Fraunhofer has announced no concrete plans for periodic technical inspection use.

The roadmap points elsewhere: Fraunhofer EMI's high-speed X-ray system for dynamic processes inside individual large-format cells is scheduled for Volkswagen's PowerCo plant in Salzgitter in 2028, and the planned "Giant Eye" CT system in Fürth should allow three-dimensional analysis of complete vehicles. What remains open is the boring part — cost per scan, who pays, and whether insurers and auction houses adopt it. Those numbers, not image resolution, will decide whether AIR becomes routine.

Source: https://www.electrive.com/2026/09/22/fraunhofer-automates-x-ray-analysis-of-ev-batteries/