Industrial AI uses machine learning and scientific computing to develop innovative solutions for industry, services and medicine and to drive data-driven discoveries.

Industrial Artificial Intelligence (Industrial AI for short) uses tools from both machine learning and scientific computing to develop new methods for scalable, domain-aware, robust, reliable and interpretable learning and data analysis. Industrial AI (IAI) is critical to driving the next wave of data-driven scientific discovery in the physical sciences and engineering. Industrial Artificial Intelligence is a subfield of Scientific Machine Learning, with a specific application focus on industry, services and medicine.

Domain-aware machine learning: We combine artificial intelligence with proven knowledge. By integrating physical principles, simulations and expert feedback, we create models that are more precise, easier to understand and faster to train. These approaches help us to reduce data requirements and make more reliable predictions.

Modelling, simulation and optimisation with ML: We use machine learning to make physical processes and simulations more efficient and adaptable. Our methods make it possible to solve complex physical equations such as the Navier-Stokes equations or particle approaches (direct element modelling, DEM) with high accuracy - often in fractions of a second. These models are not only suitable for fast simulations, but also for gradient-based optimisations, such as the improvement of components or processes. Thanks to modern neural networks, we can calculate solutions flexibly and scalably without complex grids or discretisations. With generative models (e.g. LLMs), we are able to automatically generate code for parameterised simulation models in order to efficiently evaluate design options.

Reliability and trustworthiness in ML: Robust and trustworthy ML models are essential for use in safety-critical areas such as industry or medicine. We develop methods that not only deliver precise predictions, but also clearly quantify the uncertainties of the results. With these approaches, we create trust in data-driven models and ensure high standards of verification, validation and reproducibility.

Explainability of machine learning models: We make machine learning models more transparent and easier to understand. By integrating existing knowledge, we ensure that the models remain comprehensible and at the same time offer valuable insights. Innovative visualisation and analysis tools help us to better understand complex relationships and make the results tangible for users.

Last updated: 12.03.2025