BMW uses AI cameras and microphones to check cars during production

A useful factory inspection asks a specific question: is the required item present, is the right variant fitted, or does a newly assembled car make an unexpected sound? BMW’s AIQX platform applies AI to those checks. In June 2026, BMW confirmed visual and acoustic inspection at Spartanburg and said AIQX had become an established company standard.[1]

The setting makes specificity important. BMW’s July 2022 production account describes the new 7 Series sharing a Dingolfing assembly line with several other model families, while accommodating different powertrains. The same account recorded 40 AIQX applications in production of that 7 Series. These were individual checks within a varied manufacturing process.[4]

Capture the right detail first

BMW’s IT team explains that vehicle localisation triggers cameras along the line. Images go to its cloud platform, where deep-learning models trained on examples analyse them. A camera can inspect a badge or a door lock while the system keeps that observation connected to the vehicle’s position in production.[2]

Camera supplier Axis describes the physical requirements behind that process. A rating plate must remain readable against both a light and a dark vehicle. Its 2024 customer account identifies exposure handling, lighting performance and image detail as part of delivering a usable picture to AIQX. The ability to recognise an error begins with an image that actually shows it.[3]

Axis also describes sending analysed results directly to workers’ smart devices. That places the finding where someone can act on it before the vehicle progresses further. A clear record of the observed part and its location is more useful in this setting than a general statement that the car looks unusual.[3]

Listening on the test route

At Dingolfing, BMW’s Acoustic Analytics application records driving sounds with microphones fixed to seats. A trained machine-learning model analyses and classifies the recording, including unwanted noises. BMW describes the application as using the expandable AIQX architecture, alongside automated driving inside the plant.[2]

This extends the evidence available to inspection. A part can be visually present while its movement produces a sound worth investigating. Conversely, an image can reveal a fitting issue without requiring the vehicle to be driven. Combining checks makes sense because the observations answer different questions; a single image score cannot represent them all.

From one inspection to a production system

The 2022 Dingolfing account describes feedback going to employees during assembly, allowing checks of variants, completeness and fitting anomalies. It also separates AIQX from the IPS-i platform that brings together location data. This distinction matters: locating a vehicle or tracing a component supplies context, while an AI inspection interprets the relevant sensor evidence.[4]

The June 2026 update confirms continued use at Spartanburg and says BMW was assessing whether to make AIQX available to suppliers. It does not provide a controlled before-and-after defect rate or a factory-wide cost saving. The documented achievement is a working inspection process and an established platform; broader financial or warranty effects need their own measurements.[1]

For manufacturing, the transferable idea is a chain of useful observations: capture the detail, apply the appropriate check, associate the result with the item being built and deliver it at the point of correction.

Report an error or suggest a correction