Dr. Andreas Lintermann

Andreas Lintermann is a postdoctoral researcher and group leader of the SDL FSE in the High-Performance Computing in Applied Science and Engineering Division at the Jülich Supercomputing Centre, Forschungszentrum Jülich. He is coordinating the European Center of Excellence in Exascale Computing ‘‘Research on AI- and Simulation-Based Engineering at Exascale’’ (CoE RAISE), leads the EU-funded activities in the EuroCC/EuroCC2 , interTwin , and SPECTRUM projects and co-leads the EU-project HANAMI, the BMBF project StroemunsRaum, and BMWK project nxtAIM from Jülich’s side, and is involved in the Industry Relations Team of the institute.

Amongst others, the research in his group concentrates on encoder-decoder Convolutional Neural Networks (CNNs) for the prediction of aeroacoustic fields, convolutional autoencoders and Recurrent Neural Networks (RNNs) for the compression and reconstruction of actuated turbulent boundary layer flows, and for predicting the associated net power savings and air resistance reduction. In addition, the SDL developed a convolutional defiltering model to reconstruct turbulent wall-shear stresses, to be used to resolve near-wall regions in low-fidelity simulations, and a super-resolution network to increase the resolution of under-resolved computer tomography image data. The SDL trained a proximal policy optimization and a multi-agent Deep-Q Learning (DQL) algorithm for shape optimization of simple and complex geometries using various physics-based target functions. Ongoing work concentrates on training Physics-Informed Neural Networks (PINNs), Deep Neural Networks (DNNs), and Graph Neural Networks (GNNs) to predict flow fields for flow-field initialization in large-scale simulations.