In our group, we develop machine learning (ML) methods and algorithms for process engineering applications. A particular focus is combining flexible, data-driven ML algorithms with explicit physical knowledge or established physical models to develop powerful hybrid methods. In addition to theoretical and methodological work, we carry out laboratory experiments, including measuring thermophysical properties, NMR analysis of complex mixtures, and the operation of chemical plants.

Our group currently has three main research areas:

• Development of data-driven and hybrid ML methods for the prediction of thermophysical properties as well as their interpretation and targeted improvement using design-of-experiment strategies
• Thermodynamic modeling of complex mixtures of unknown composition based on NMR spectroscopic fingerprints and ML algorithms
• Machine learning on chemical process data, in particular for automatic error detection in process engineering processes and the development of self-learning plants

• DFG Emmy-Noether Junior Research Group "Hybrid Thermodynamic Models"
  Integration of explicit physical knowledge into ML algorithms for the prediction of thermophysical properties of mixtures and design of experiments using active learning strategies
• DFG Research Unit KI-FOR 5359"Deep Learning on Sparse Chemical Process Data"
  Automated fault detection in chemical processes, self-learning plants, generation of chemical process data with physical, data-driven, and hybrid approaches
• DFG Research Training Group GRK 2908 "Recyclable Wastewater"
  Thermodynamic modeling of phosphorus recovery from wastewater, prediction of the thermophysical properties of electrolyte solutions, fingerprinting of unknown mixtures
• DFG Priority Program SPP 2331 "Machine Learning in Chemical Engineering" (associated)
  Interpretability and explainability of ML methods in thermodynamics, gaining knowledge from data and models
• DFG Priority Program SPP 2363 "Molecular Machine Learning"
  Interpretability and explainability of ML methods in thermodynamics, gaining knowledge from data and models
• CZS Perspectives project "Process Engineering 4.0"
  Merging physical modeling with ML methods, text and image recognition in a process engineering context, apparatus and process modeling
• FEI project "Benchtop-NMR in Winemaking"
  Development of model-based analytics for the quality assessment of grapes for winemaking using benchtop NMR