Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
139 tokens/sec
GPT-4o
47 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Physics-informed neural networks for gravity currents reconstruction from limited data (2211.09715v2)

Published 3 Nov 2022 in physics.flu-dyn and cs.LG

Abstract: The present work investigates the use of physics-informed neural networks (PINNs) for the 3D reconstruction of unsteady gravity currents from limited data. In the PINN context, the flow fields are reconstructed by training a neural network whose objective function penalizes the mismatch between the network predictions and the observed data and embeds the underlying equations using automatic differentiation. This study relies on a high-fidelity numerical experiment of the canonical lock-exchange configuration. This allows us to benchmark quantitatively the PINNs reconstruction capabilities on several training databases that mimic state-of-the-art experimental measurement techniques for density and velocity. Notably, spatially averaged density measurements by light attenuation technique (LAT) are employed for the training procedure. An optimal experimental setup for flow reconstruction by PINNs is proposed according to two criteria : the implementation complexity and the accuracy of the inferred fields.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (35)
  1. J. L. Callaham, K. Maeda,  and S. L. Brunton, “Robust flow reconstruction from limited measurements via sparse representation,” Physical Review Fluids 4, 103907 (2019).
  2. V. Mons, J.-C. Chassaing, T. Gomez,  and P. Sagaut, “Reconstruction of unsteady viscous flows using data assimilation schemes,” Journal of Computational Physics 316, 255–280 (2016).
  3. P. Chandramouli, E. Mémin,  and D. Heitz, “4d large scale variational data assimilation of a turbulent flow with a dynamics error model,” Journal of Computational Physics 412, 109446 (2020).
  4. M. Raissi, P. Perdikaris,  and G. E. Karniadakis, “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,” Journal of Computational physics 378, 686–707 (2019).
  5. S. Cai, Z. Mao, Z. Wang, M. Yin,  and G. E. Karniadakis, “Physics-informed neural networks (pinns) for fluid mechanics: A review,” Acta Mechanica Sinica , 1–12 (2022).
  6. E. Haghighat, M. Raissi, A. Moure, H. Gomez,  and R. Juanes, “A physics-informed deep learning framework for inversion and surrogate modeling in solid mechanics,” Computer Methods in Applied Mechanics and Engineering 379, 113741 (2021).
  7. S. Cai, Z. Wang, S. Wang, P. Perdikaris,  and G. E. Karniadakis, “Physics-informed neural networks for heat transfer problems,” Journal of Heat Transfer 143 (2021a).
  8. S. Cuomo, V. S. Di Cola, F. Giampaolo, G. Rozza, M. Raissi,  and F. Piccialli, “Scientific machine learning through physics-informed neural networks: Where we are and what’s next,” Journal of Scientific Computing 92 (2022).
  9. K. Hornik, M. Stinchcombe,  and H. White, “Multilayer feedforward networks are universal approximators,” Neural networks 2, 359–366 (1989).
  10. K. G. Raissi M, Yazdani A, “Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations.” Science 367, 1026–1030 (2020).
  11. S. Cai, Z. Wang, F. Fuest, Y. J. Jeon, C. Gray,  and G. E. Karniadakis, “Flow over an espresso cup: inferring 3-d velocity and pressure fields from tomographic background oriented schlieren via physics-informed neural networks,” Journal of Fluid Mechanics 915 (2021b).
  12. M. Ungarish, An introduction to gravity currents and intrusions (CRC press, 2009).
  13. J. W. L. Paul F. Fischer and S. G. Kerkemeier, “nek5000 Web page,”  (2008), http://nek5000.mcs.anl.gov.
  14. C. Härtel, E. Meiburg,  and F. Necker, “Analysis and direct numerical simulation of the flow at a gravity-current head. part 1. flow topology and front speed for slip and no-slip boundaries,” Journal of Fluid Mechanics 418, 189–212 (2000).
  15. T. M. Özgökmen, P. F. Fischer,  and W. E. Johns, “Product water mass formation by turbulent density currents from a high-order nonhydrostatic spectral element model,” Ocean Modelling 12, 237–267 (2006).
  16. T. M. Özgökmen, T. Iliescu,  and P. F. Fischer, “Reynolds number dependence of mixing in a lock-exchange system from direct numerical and large eddy simulations,” Ocean Modelling 30, 190–206 (2009).
  17. A. T. Patera, “A spectral element method for fluid dynamics: laminar flow in a channel expansion,” Journal of computational Physics 54, 468–488 (1984).
  18. C. Marshall, R. Dorrell, S. Dutta, G. Keevil, J. Peakall,  and S. Tobias, “The effect of schmidt number on gravity current flows: The formation of large-scale three-dimensional structures,” Physics of Fluids 33, 106601 (2021).
  19. S. Balasubramanian and Q. Zhong, “Entrainment and mixing in lock-exchange gravity currents using simultaneous velocity-density measurements,” Physics of Fluids 30, 056601 (2018).
  20. B. Pérez-Díaz, P. Palomar, S. Castanedo,  and A. Álvarez, “Piv-plif characterization of nonconfined saline density currents under different flow conditions,” Journal of Hydraulic Engineering 144, 04018063 (2018).
  21. J. García-Alba, J. F. Bárcena,  and A. García, “Zonation of positively buoyant jets interacting with the water-free surface quantified by physical and numerical modelling,” Water 12, 1324 (2020).
  22. S. Akbari and S. Taghavi, “Injection of a heavy fluid into a light fluid in a closed-end pipe,” Physics of Fluids 32, 063302 (2020).
  23. J. M. Holford and S. B. Dalziel, “Measurements of layer depth during baroclinic instability in a two-layer flow,” Applied Scientific Research 56, 191–207 (1996).
  24. Y. Dossmann, C. Shakespeare, K. Stewart,  and A. M. Hogg, “Asymmetric internal tide generation in the presence of a steady flow,” Journal of Geophysical Research: Oceans 125, e2020JC016503 (2020).
  25. S. W. Kim, I. Kim, J. Lee,  and S. Lee, “Knowledge integration into deep learning in dynamical systems: An overview and taxonomy,” Journal of Mechanical Science and Technology 35, 1331–1342 (2021).
  26. M. Raissi and G. E. Karniadakis, “Hidden physics models: Machine learning of nonlinear partial differential equations,” Journal of Computational Physics 357, 125–141 (2018).
  27. A. G. Baydin, B. A. Pearlmutter, A. A. Radul,  and J. M. Siskind, “Automatic differentiation in machine learning: a survey,” Journal of Marchine Learning Research 18, 1–43 (2018).
  28. D. E. Rumelhart, G. E. Hinton,  and R. J. Williams, “Learning representations by back-propagating errors,” nature 323, 533–536 (1986).
  29. M. Raissi, A. Yazdani,  and G. E. Karniadakis, “Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations,” Science 367, 1026–1030 (2020).
  30. Z. Mao, A. D. Jagtap,  and G. E. Karniadakis, “Physics-informed neural networks for high-speed flows,” Computer Methods in Applied Mechanics and Engineering 360, 112789 (2020).
  31. P. Ramachandran, B. Zoph,  and Q. V. Le, “Searching for activation functions,” arXiv preprint arXiv:1710.05941  (2017).
  32. D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980  (2014).
  33. X. Glorot and Y. Bengio, “Understanding the difficulty of training deep feedforward neural networks,” in Proceedings of the thirteenth international conference on artificial intelligence and statistics (JMLR Workshop and Conference Proceedings, 2010) pp. 249–256.
  34. P. Wesseling, Principles of computational fluid dynamics, Vol. 29 (Springer Science & Business Media, 2009) Chap. 4.
  35. C. Wu, M. Zhu, Q. Tan, Y. Kartha,  and L. Lu, “A comprehensive study of non-adaptive and residual-based adaptive sampling for physics-informed neural networks,” Computer Methods in Applied Mechanics and Engineering 403, 115671 (2023).
Citations (9)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now