TY - JOUR A1 - Harmening, Jan Hauke A1 - Peitzmann, Franz-Josef A1 - el Moctar, Bettar Ould T1 - Effect of Network Architecture on Physics-Informed Deep Learning of the Reynolds-Averaged Turbulent Flow Field around Cylinders without Training Data N2 - Unsupervised physics-informed deep learning can be used to solve computational physics problems by training neural networks to satisfy the underlying equations and boundary conditions without labeled data. Parameters such as network architecture and training method determine the training success. However, the best choice is unknown a priori as it is case specific. Here, we investigated network shapes, sizes, and types for unsupervised physics-informed deep learning of the two-dimensional Reynolds averaged flow around cylinders. We trained mixed-variable networks and compared them to traditional models. Several network architectures with different shape factors and sizes were evaluated. The models were trained to solve the Reynolds averaged Navier-Stokes equations incorporating Prandtl’s mixing length turbulence model. No training data were deployed to train the models. The superiority of the mixed-variable approach was confirmed for the investigated high Reynolds number flow. The mixed-variable models were sensitive to the network shape. For the two cylinders, differently deep networks showed superior performance. The best fitting models were able to capture important flow phenomena such as stagnation regions, boundary layers, flow separation, and recirculation. We also encountered difficulties when predicting high Reynolds number flows without training data. T3 - Bocholter Hochschulschriften - 1 KW - Physics-informed deep learning KW - unsupervised learning KW - Reynolds-averaged Navier-Stokesequations KW - high Reynolds number flow KW - turbulence modeling Y1 - 2024 ER - TY - CHAP A1 - Harmening, Jan Hauke A1 - Peitzmann, Franz-Josef T1 - Unsupervised physics-informed deep learning of the flow around an airfoil using a mixed-variable network KW - Physics-Informed Deep Learning KW - CFD Simulation KW - High Reynold Numer Y1 - 2024 ER - TY - INPR A1 - Harmening, Jan Hauke A1 - Peitzmann, Franz-Josef A1 - el Moctar, Bettar Ould T1 - Effect of Network Architecture on Physics-Informed Deep Learning of the Reynolds-Averaged Turbulent Flow Field around Cylinders without Training Data Y1 - 2023 U6 - https://doi.org/10.20944/preprints202312.2274.v1 ET - Version 1 ER - TY - INPR A1 - Harmening, Jan Hauke A1 - Pioch, Fabian A1 - Fuhrig, Lennart A1 - Peitzmann, Franz-Josef A1 - Schramm, Dieter A1 - El Moctar, Bettar Ould T1 - Data-Assisted Training of a Physics-Informed Neural Network to Predict the Reynolds-Averaged Turbulent Flow Field around a Stalled Airfoil under Variable Angles of Attack Y1 - 2023 U6 - https://doi.org/10.20944/preprints202304.1244.v1 ER - TY - JOUR A1 - Pioch, Fabian A1 - Harmening, Jan Hauke A1 - Müller, Andreas Maximilian A1 - Peitzmann, Franz-Josef A1 - Schramm, Dieter A1 - El Moctar, Bettar Ould T1 - Turbulence Modeling for Physics-Informed Neural Networks: Comparison of Different RANS Models for the Backward-Facing Step Flow JF - Fluids Y1 - 2023 U6 - https://doi.org/10.3390/fluids8020043 VL - 8 IS - 2 SP - 43 ER - TY - JOUR A1 - Harmening, Jan Hauke A1 - Devananthan, Harish A1 - Peitzmann, Franz-Josef A1 - el Moctar, Bettar Ould T1 - Aerodynamic Effects of Knitted Wire Meshes - CFD Simulations of the Flow Field and Influence on the Flow Separation of a Backward-Facing Ramp JF - Fluids Y1 - 2022 U6 - https://doi.org/10.3390/fluids7120370 VL - 12 IS - 7 SP - 370 ER -