Modern battery systems rely on complex electrochemical processes that vary spatially within the cell. Local differences in current density, temperature, and lithium concentration significantly affect performance, aging, and safety.

We develop reduced-order, physics-based models that retain high accuracy while enabling real-time implementation in battery management systems. These models support advanced state estimation and diagnostics, including reliable predictions of critical operating conditions during fast charging.

By enabling robust estimation of key indicators such as State of Charge (SOC) and State of Health (SOH), digital twins facilitate intelligent charging strategies, improved monitoring, and extended battery lifetime.

Fuel cells and electrolysers are key technologies for future energy systems. Their performance and lifetime are strongly influenced by degradation mechanisms that depend on local operating conditions within the cell.

Since internal measurements are typically not feasible, our approach combines external sensor data with high-resolution models to reconstruct internal states using model-based observers. This allows for accurate assessment of system health and degradation processes.

The resulting digital twins enable adaptive operating strategies that improve energy efficiency and significantly extend system lifetime - even under dynamic conditions.

Alongside injection moulding and calendering, extrusion is one of the most important shaping processes in the elastomer industry. It enables the manufacture of a wide range of products, such as industrial hoses, seals and handrails. Product quality depends largely on the internal conditions within the extruder, particularly on the development of the melt temperature along the screw. However, these internal conditions are generally not directly measurable during extrusion; mathematical models of the process are therefore required to make them accessible.

The strong coupling of the rheological, thermal and fluid dynamic processes means that high-resolution simulations of the extrusion process involve considerable computational effort and are therefore usually not suitable for real-time applications. However, by using suitable model reduction methods, real-time-capable surrogate models can be derived from these complex models.

When combined with process measurement data available online, these reduced models form the basis for a digital twin of the extrusion process. By combining these models with real-time measurement data, deviations in the simulated variables can be continuously corrected by using differences between measured and modelled variables to update states and parameters. The digital twin thus enables not only the reconstruction of internal states that cannot be measured directly, but also predictive analyses, state-based process control and the early detection of deviations or faults.

Extrusion is a key manufacturing process in the elastomer industry, where product quality strongly depends on internal process conditions such as melt temperature distribution. Because these internal states cannot be directly measured, we develop reduced-order models derived from high-fidelity simulations. These models enable real-time capable digital twins that integrate live process data.

This approach allows continuous correction of model predictions, early detection of deviations, and implementation of predictive and condition-based process control—leading to improved product quality and process stability.