PhyNet: Physics Guided Neural Networks for Particle Drag Force Prediction in Assembly

Paradigm Shift Through the Integration of Physical Methodology and Data Science

“…Mulalindhar et al [67] demonstrated that in designing a convolutional neural network to predict fluid properties, a network architecture in which the intermediate layers output the velocity and pressure fields of the fluid (physics guided neural networks) improve the interpretability of the computational process of the model, and demonstrated that it can achieve higher prediction performance than the physical model even when there is an insufficient amount of training data. Furthermore, Greydanus et al [68] demonstrated that when predicting the behavior of a dynamical system, the total energy of the system is conserved and the dynamic properties of the model output does not break down by using Hamiltonian neural networks that learn the Hamiltonian instead of learning the behavior itself, which improves the properties and performance of the model by devising the intermediate and output quantities of the data-driven model.…”

“…The integration methods mentioned in this study have made significant contributions to prediction. The methods proposed by Read et al [45], Kubo et al [49], Mulalindhar et al [67], and Rasp et al [71] directly improved the prediction accuracy of the physical quantities of interest, whereas the methods proposed by Karpatne et al [37] and Jia et al [44] allowed us to leverage on the flexible prediction capabilities of data-driven models even when sufficient labeled data were not available. Additionally, the computational speedup achieved by Zhang et al [41], Ichimura et al [42], and Hijazi et al [66] extended the scope of analysis and prediction.…”

Physics-Integrated Variational Autoencoders for Robust and Interpretable Generative Modeling

“…Physics+ML in model design Integration of physics models with machine learning models has been a subject of study for a long time and is becoming a very active area, with deep neural networks being employed in various scientific applications. Examples range from classical ones (such as Psichogios and Ungar, 1992;Rico-Martínez et al, 1994;Thompson and Kramer, 1994) to recent attempts (Young et al, 2017;Raissi, 2018;Long and She, 2018;Wan et al, 2018;Nutkiewicz et al, 2018;Ajay et al, 2018;de Bézenac et al, 2019;Zeng et al, 2019;Wang et al, 2019;Roehrl et al, 2020;Le Guen and Thome, 2020;Muralidhar et al, 2020;Belbute-Peres et al, 2020;Sengupta et al, 2020;Rackauckas et al, 2020;Li et al, 2020). Most of them focus on prediction, and the generative modeling perspective has been less investigated.…”

“…• physics-based output y is used as input features of machine learning model (Karpatne et al, 2017b;Pawar et al, 2020;Muralidhar et al, 2020;Zhang et al, 2021;Belbute-Peres et al, 2020);…”

Surrogate modeling of fluid dynamics with a multigrid inspired neural network architecture

“…More interestingly, it has also found great utility in the modeling of engineering systems such as in fluid dynamics. Neural networks have been shown to complement numerical models in areas such as turbulence modeling, force prediction and flow reconstruction Holland, Baeder, & Duraisamy, 2019;Ling, Kurzawski, & Templeton, 2016;Muralidhar et al, 2020;Ooi et al, 2020;Parish & Duraisamy, 2016;Umetani & Bickel, 2018;Ye et al, 2020). In particular, the convolutional neural network (CNN) has been shown to be effective for prediction across different flow scenarios while providing flexibility in the treatment of variable geometries.…”