Volume 25, Issue 3
Data-Driven, Physics-Based Feature Extraction from Fluid Flow Fields using Convolutional Neural Networks

Carlos Michelén Ströfer, Jin-Long Wu, Heng Xiao & Eric Paterson

Commun. Comput. Phys., 25 (2019), pp. 625-650.

Published online: 2018-11

[An open-access article; the PDF is free to any online user.]

Preview Full PDF 370 2481
Export citation
  • Abstract

Feature identification is an important task in many fluid dynamics applications and diverse methods have been developed for this purpose. These methods are based on a physical understanding of the underlying behavior of the flow in the vicinity of the feature. Particularly, they require the definition of suitable criteria (i.e. point-based or neighborhood-based derived properties) and proper selection of thresholds. However, these methods rely on creative visualization of physical idiosyncrasies of specific features and flow regimes, making them non-universal and requiring significant effort to develop. Here we present a physics-based, data-driven method capable of identifying any flow feature it is trained to. We use convolutional neural networks, a machine learning approach developed for image recognition, and adapt it to the problem of identifying flow features. This provides a general method and removes the large burden placed on identifying new features. The method was tested using mean flow fields from numerical simulations, where the recirculation region and boundary layer were identified in several two-dimensional flows through a convergent-divergent channel, and the horseshoe vortex was identified in threedimensional flow over a wing-body junction.

  • Keywords

Machine learning feature identification fluid dynamics convolutional neural network.

  • AMS Subject Headings

76G25 76M99

  • Copyright

COPYRIGHT: © Global Science Press

  • Email address
  • BibTex
  • RIS
  • TXT
@Article{CiCP-25-625, author = {}, title = {Data-Driven, Physics-Based Feature Extraction from Fluid Flow Fields using Convolutional Neural Networks}, journal = {Communications in Computational Physics}, year = {2018}, volume = {25}, number = {3}, pages = {625--650}, abstract = {

Feature identification is an important task in many fluid dynamics applications and diverse methods have been developed for this purpose. These methods are based on a physical understanding of the underlying behavior of the flow in the vicinity of the feature. Particularly, they require the definition of suitable criteria (i.e. point-based or neighborhood-based derived properties) and proper selection of thresholds. However, these methods rely on creative visualization of physical idiosyncrasies of specific features and flow regimes, making them non-universal and requiring significant effort to develop. Here we present a physics-based, data-driven method capable of identifying any flow feature it is trained to. We use convolutional neural networks, a machine learning approach developed for image recognition, and adapt it to the problem of identifying flow features. This provides a general method and removes the large burden placed on identifying new features. The method was tested using mean flow fields from numerical simulations, where the recirculation region and boundary layer were identified in several two-dimensional flows through a convergent-divergent channel, and the horseshoe vortex was identified in threedimensional flow over a wing-body junction.

}, issn = {1991-7120}, doi = {https://doi.org/10.4208/cicp.OA-2018-0035}, url = {http://global-sci.org/intro/article_detail/cicp/12823.html} }
TY - JOUR T1 - Data-Driven, Physics-Based Feature Extraction from Fluid Flow Fields using Convolutional Neural Networks JO - Communications in Computational Physics VL - 3 SP - 625 EP - 650 PY - 2018 DA - 2018/11 SN - 25 DO - http://dor.org/10.4208/cicp.OA-2018-0035 UR - https://global-sci.org/intro/cicp/12823.html KW - Machine learning KW - feature identification KW - fluid dynamics KW - convolutional neural network. AB -

Feature identification is an important task in many fluid dynamics applications and diverse methods have been developed for this purpose. These methods are based on a physical understanding of the underlying behavior of the flow in the vicinity of the feature. Particularly, they require the definition of suitable criteria (i.e. point-based or neighborhood-based derived properties) and proper selection of thresholds. However, these methods rely on creative visualization of physical idiosyncrasies of specific features and flow regimes, making them non-universal and requiring significant effort to develop. Here we present a physics-based, data-driven method capable of identifying any flow feature it is trained to. We use convolutional neural networks, a machine learning approach developed for image recognition, and adapt it to the problem of identifying flow features. This provides a general method and removes the large burden placed on identifying new features. The method was tested using mean flow fields from numerical simulations, where the recirculation region and boundary layer were identified in several two-dimensional flows through a convergent-divergent channel, and the horseshoe vortex was identified in threedimensional flow over a wing-body junction.

Carlos Michelén Ströfer, Jin-Long Wu, Heng Xiao & Eric Paterson. (2020). Data-Driven, Physics-Based Feature Extraction from Fluid Flow Fields using Convolutional Neural Networks. Communications in Computational Physics. 25 (3). 625-650. doi:10.4208/cicp.OA-2018-0035
Copy to clipboard
The citation has been copied to your clipboard