Zukunftslabor2030
Using artificial intelligence and digital twins to predict the spoilage of meat products and reduce food loss.
The project
Every year large quantities of meat are thrown away, often out of caution, because there is no reliable way to tell how fresh a product still is. Between discarding too early and leaving it on sale too long, a dependable basis for that decision has been missing.
Zukunftslabor2030 developed a digital, data-driven risk assessment system for this. Using the Internet of Things and artificial intelligence, a digital twin was created along the supply chain for each product, a digital image of its chemical, physical and microbiological properties. The data for it came from new measurement methods such as mass spectrometry, gene sequencing, spectroscopy and micro gas sensors.
Learning models use that to predict when a product is likely to spoil. This makes it easier to judge when a product is still safe, and less produce is lost unnecessarily. The approach was trialled on meat products such as minced pork, turkey escalope and raw sausages.
The contribution from benelog
In the project we were responsible for the traceability platform on which the whole system comes together.
A digital twin is only as good as the data that flows into it. That data arises in many places: measurements and laboratory analyses from the analytics partners, plus product master, transport, storage and life cycle data for the food in question. It sits in different systems and only forms a coherent picture once it comes together in the context of the processing and transport chain.
For that we developed and operated an open, scalable platform on the basis of open GS1 standards for the event-based exchange of supply chain data. Alongside the unbroken chain of custody, meaning the path a product takes through the chain, it also covers the product's actual quality and spoilage. Through it the measurement, analysis and production data were captured, brought together and made retrievable along the chain.
The learning, self-updating digital twins could thus be supplied with new data continuously and keep their spoilage prediction up to date. The platform was laid out openly so that further use cases and participants can be connected. The same foundation of open GS1 standards also carries benelog's own implementation, OpenEPCIS.
More about the project
