Advanced search
1 file | 229.39 KB

Partitioned solution of the unsteady adjoint equations for the one-dimensional flow in a flexible tube

Author
Organization
Abstract
For gradient-based optimization, the gradient of the objective function needs to be calculated repeatedly. If the number of design variables is high, this gradient can be obtained efficiently from adjoint equations. This research focuses on the gradient calculation for an objective function which involves a fluid-structure interaction (FSI) simulation. The interaction can be calculated in a partitioned way by coupling a flow solver with a structural solver. In this work, quasi-Newton coupling iterations with an approximation for the inverse of the Jacobian from a least-squares model (IQN-ILS) are employed for the state equations as well as for their adjoint equations. The problem at hand is the unsteady, one-dimensional flow of an incompressible, inviscid fluid in an elastic tube. Special attention has been given to the interface variables which are exchanged between the adjoint flow and structural solver, to avoid the communication of system matrices between them.
Keywords
Adjoint, partitioned, Fluid-structure interaction, quasi-Newton

Downloads

  • (...).pdf
    • full text
    • |
    • UGent only
    • |
    • PDF
    • |
    • 229.39 KB

Citation

Please use this url to cite or link to this publication:

Chicago
Degroote, Joris, Majid Hojjat, Electra Stavropoulou, Roland Wüchner, and Kai-Uwe Bletzinger. 2012. “Partitioned Solution of the Unsteady Adjoint Equations for the One-dimensional Flow in a Flexible Tube.” In 10th World Congress on Computational Mechanics, Proceedings, 1–18.
APA
Degroote, Joris, Hojjat, M., Stavropoulou, E., Wüchner, R., & Bletzinger, K.-U. (2012). Partitioned solution of the unsteady adjoint equations for the one-dimensional flow in a flexible tube. 10th World Congress on Computational Mechanics, Proceedings (pp. 1–18). Presented at the 10th World Congress on Computational Mechanics (WCCM - 2012).
Vancouver
1.
Degroote J, Hojjat M, Stavropoulou E, Wüchner R, Bletzinger K-U. Partitioned solution of the unsteady adjoint equations for the one-dimensional flow in a flexible tube. 10th World Congress on Computational Mechanics, Proceedings. 2012. p. 1–18.
MLA
Degroote, Joris, Majid Hojjat, Electra Stavropoulou, et al. “Partitioned Solution of the Unsteady Adjoint Equations for the One-dimensional Flow in a Flexible Tube.” 10th World Congress on Computational Mechanics, Proceedings. 2012. 1–18. Print.
@inproceedings{3099993,
  abstract     = {For gradient-based optimization, the gradient of the objective function needs to be calculated repeatedly. If the number of design variables is high, this gradient can be obtained efficiently from adjoint equations. This research focuses on the gradient calculation for an objective function which involves a fluid-structure interaction (FSI) simulation. The interaction can be calculated in a partitioned way by coupling a flow solver with a structural solver. In this work, quasi-Newton coupling iterations with an approximation for the inverse of the Jacobian from a least-squares model (IQN-ILS) are employed for the state equations as well as for their adjoint equations. The problem at hand is the unsteady, one-dimensional flow of an incompressible, inviscid fluid in an elastic tube. Special attention has been given to the interface variables which are exchanged between the adjoint flow and structural solver, to avoid the communication of system matrices between them.},
  author       = {Degroote, Joris and Hojjat, Majid and Stavropoulou, Electra and Wüchner, Roland and Bletzinger, Kai-Uwe},
  booktitle    = {10th World Congress on Computational Mechanics, Proceedings},
  isbn         = {9788586686696},
  keywords     = {Adjoint,partitioned,Fluid-structure interaction,quasi-Newton},
  language     = {eng},
  location     = {São Paulo, Brazil},
  pages        = {1--18},
  title        = {Partitioned solution of the unsteady adjoint equations for the one-dimensional flow in a flexible tube},
  year         = {2012},
}