Ultrasound seems to be viable

> A study on structural quality control of Swiss‐type cheese with ultrasound is presented. We used a longitudinal mode pulse‐echo setup using 1–2MHz ultrasonic frequencies to detect cheese‐eyes and ripening induced cracks. Results show that the ultrasonic method posses good potential to monitor the cheese structure during the ripening process. Preliminary results indicate that maturation stage could be monitored with ultrasonic velocity measurements.

https://pubs.aip.org/aip/acp/article-abstract/894/1/1328/953...

Note that this is already industry-standard for non-destructive testing of metal, concrete and composites. Handheld ultrasonic tomography devices are commercially available and could plausibly be used on cheese with only minor modification.

https://acs-international.com/instruments/ultrasonic-pulse-e...

Industrial CT would probably be pretty effective at giving a density distribution and identifying voids, but that may be overkill - a couple planes of X-ray imaging may be sufficient.

Ultrasound would also be a solid bet, but it depends how many points you need to sample if it would be time efficient.

Microphones and spectral analysis. Or the same equipment used for flaw detection of welds with ultrasound.

Just doing the same thing they do, basically. Have a tiny transducer "hit" the cheese and listen for echo back.

Now, I don't think ultrasound would work (it's a harder cheese than Swiss cheese), also the mechanical interface would be a complication (you can't gel the cheese). CT would work but be expensive

But tiny percussions and analysis of echo/transmission delays would work in principle

A sonogram… it doesn’t even need to be AI

Other commenters have given what's probably the right answer, but in the spirit of exploration, I wonder about electrical impedance tomography. I'd think a salty cheese would be a pretty good medium to pass electricity through, and since the wheels are very uniform in shape, you could make a chuck that held electrodes in contact with the wheel. We don't really need precise information about the shape and location of the voids, we're not going to do surgery on the cheese, so the coarseness of EIT might not be an issue.

(Admittedly without knowing much about it) I'll throw Ground Penetrating Radar onto this armchair metaphorical ideas whiteboard. The industrial CT scanner idea uses x-rays, while GPR is more in the UHF/VHF frequencies which probably means cheaper/easier? The tech seems to have some tunability for the specific application:

> Thus operating frequency is always a trade-off between resolution and penetration. Optimal depth of subsurface penetration is achieved in ice where the depth of penetration can achieve several thousand metres (to bedrock in Greenland) at low GPR frequencies. Dry sandy soils or massive dry materials such as granite, limestone, and concrete tend to be resistive rather than conductive, and the depth of penetration could be up to 15 metres (49 ft). However, in moist or clay-laden soils and materials with high electrical conductivity, penetration may be as little as a few centimetres.

https://en.wikipedia.org/wiki/Ground-penetrating_radar

It wouldn't. Not for now

It's not only the sound, it's the sound, the bounce, the response to different strengths, the smell, the color. Humans are multimodal, machines are not, yet.

The moment we have a Michelin star level robot cook, then we can start thinking about automating this kind of stuff. For now, we have better results with humans!

Italians have absolutely zero problems replacing manual processes with technology. Creating each wheel is more science than art, everything is done in highly sterilized environments with exact temperature control, as an example.

Preparing a Michelin star worthy meal is orders of magnitude harder than checking a cheese for defects.

It requires the same basic modality

It feels simplistic, but if the primary problem is fissures and pockets forming, I'd assume they would be full of air. If you know the volume of the wheel and the density per pound, you could weigh it on a very precise scale