The REFFpro project, launched on 21 September 2020, aimed to make small and medium‑sized German foundries more environmentally sustainable by harnessing data science and artificial intelligence. The grant recipient was Kemptener Eisengießerei Adam Höonig AG, with the research team led by Prof. Dr. Dierk Hartmann and supported by Florian Huber, M.Sc. The consortium worked over a three‑year period, concluding in 2023, and was financed through a German research grant.

At the core of the project was the creation of a unified data architecture that collected process‑relevant information from all stages of the casting chain. An architecture diagram (Figure 2) shows how sensor data, machine logs, and manual entries were streamed into a central repository. From this pool, the team derived energy profiles for individual cast parts. By applying a custom energy model (Figure 4), they could quantify the electricity and heat consumption of each casting operation. The resulting part‑specific energy profiles were visualised in a mobile application (Figures 5 and 6), allowing operators to see real‑time energy usage during forming and pouring.

The most significant scientific contribution was the development of AI‑based decision support tools. First, a classification model was trained to detect the wear state of crucibles from electrical signals and visual inspections (Figure 10). The model achieved a classification accuracy of 92 %, enabling early intervention before catastrophic failure. Second, a predictive model forecasted the wear trajectory of crucibles over time (Figure 11), allowing maintenance to be scheduled proactively. Third, a scheduling optimisation algorithm was applied to the mold machine (Formautomat). By analysing historical utilisation data and energy consumption, the algorithm generated a production plan that reduced idle time and balanced energy demand across shifts (Figures 13–15). The optimisation led to a measurable drop in energy consumption for the mold machine, as shown in the energy evaluation charts (Figure 14).

The combined effect of these tools was a substantial environmental impact. Kemptener Eisengießerei reported a reduction of 1 138 200 kg of CO₂ emissions compared with 2019 levels, a figure that directly reflects the energy savings realised through the AI‑driven recommendations. The project also produced a set of best‑practice guidelines and a mobile app that is now in routine use within the foundry, ensuring that the benefits persist beyond the project’s formal end.

Collaboration was a key element of REFFpro’s success. The consortium brought together the foundry’s production engineers, data scientists from the university, and software developers from a local tech partner. Regular workshops and sprint meetings facilitated knowledge transfer and rapid prototyping. The project’s outcomes were disseminated through conference presentations, industry journals, and media coverage, including a feature in the “Bayern Innovativ” magazine and a dedicated article on Foundry Planet. The consortium’s evaluation of the project highlighted its significant and sustainable impact, and the foundry plans to continue building on the insights gained, extending the AI tools to other production lines and exploring further optimisation opportunities.