Against this background, the ambition arose to rethink simulation – not as an isolated analysis tool, but as an integral component of operational optimisation. The objective was to create a digital twin that precisely reflects real production conditions, delivers reliable results in a short time and actively contributes to process improvement.
At Krones, the digital twin emerged from a very concrete industrial requirement. Simulation has long been an established tool in the planning and optimisation of filling and packaging lines. However, as system complexity increased, traditional simulation approaches increasingly reached their limits. In particular, complex flow and fluid processes in the beverage industry often resulted in simulation runs taking several hours. This was insufficient for fast engineering decisions and flexible production control.
Agentic Digital Twins: simulation and AI working together
Together with Ansys (part of Synopsys), CADFEM, Microsoft, NVIDIA and SoftServe, Krones has developed a new generation of agent‑based digital twins. These so‑called Agentic Digital Twins combine highly accurate, real‑time physical simulations with AI‑supported decision‑making and optimisation functions. The digital twin runs simulations independently, analyses and evaluates the results autonomously.
Instead of performing isolated simulation runs, the digital twin operates continuously and automatically. It responds to changes in the process and dynamically adapts production parameters in real time. This significantly increases the level of automation and opens up possibilities that previously required a substantial amount of manual effort.
Technological progress through cloud computing and accelerated processing
A key advancement lies in the drastic reduction of computing time. Simulations that previously took several hours can now be completed within a matter of minutes. This is made possible by the use of Microsoft Azure Cloud combined with NVIDIA’s accelerated computing platform. Highly parallelised GPU computing processes and continuous real‑time model synchronisation ensure that optimisations are not only calculated faster, but can also be implemented almost immediately.
This speed fundamentally changes the practical use of simulation. Optimisations can be calculated and applied directly during live operation, without long lead times or complex manual interventions. As a result, the digital twin evolves from a planning tool into an active component of production.
Value creation for the beverage industry
The technology is particularly relevant for the beverage industry, where complex flow and fluid processes require precise real‑time control. At the same time, it enables measurable savings in energy and resources, more stable production processes, shorter maintenance cycles and reduced downtime.
In parallel, the digital twin supports the development of new machines and lines. Scenarios can be evaluated more quickly, variants compared more efficiently and decisions made on a more informed basis. This approach therefore contributes to improved efficiency, quality and sustainability alike.
International recognition through the MIMA
For this project, Krones AG was honoured as Overall Winner at this year’s Microsoft Intelligent Manufacturing Award (MIMA). The award is presented annually by Microsoft and the management consultancy Roland Berger and is regarded as one of the most important recognitions in the field of industrial digitalisation in the EMEA region. The jury particularly acknowledged the combination of artificial intelligence, highly precise simulation and real industrial applicability.
“Together with our partners, we are setting new benchmarks in the beverage industry,” emphasises Markus Tischer, Member of the Executive Board of Krones AG. “The Agentic Digital Twins demonstrate that AI‑based optimisation is no longer a vision of the future, but a practical and economically effective component of modern production.”
A step towards autonomous production
At Krones, the digital twin is not an isolated innovation project, but part of a long‑term technological development. The new generation of agent‑based twins opens up perspectives for shorter commissioning times, data‑driven performance delivery and more sustainable production processes. At the same time, it marks an important step towards increasingly autonomous manufacturing systems, in which analysis, decision‑making and execution are closely interlinked.
