Monolithic and partitioned approaches to determine static deformation of membrane structures due to ponding

• Published in: Computers & structures Vol. 244; p. 106419
• Main Authors: Narayanan, N.K., Wüchner, R., Degroote, J.
• Format: Journal Article
• Language: English
• Published: New York Elsevier Ltd 01-02-2021; Elsevier BV
• Subjects: Computational fluid dynamics; Conservation; Convergence; Coupling; Fluid flow; Free surfaces; Hydrostatic load; Incompressible flow; Incompressible fluids; Iterative methods; Membrane structures; Membranes; Methods; Monolithic method; Partitioned method; Ponding; Solvers; Static deformation; Volume-conserving solver; Volume-conserving solver; Partitioned method; Hydrostatic load; Ponding; Monolithic method
• ISSN: 0045-7949; 1879-2243
• DOI: 10.1016/j.compstruc.2020.106419

•A novel monolithic method for ponding analysis on membrane structures is proposed.•Partitioned method can be used for ponding analysis with black box structural solver.•The proposed monolithic method was faster and more robust than the other methods.
This paper proposes monolithic and partitioned methods to calculate the static deformation of a membrane structure due to a given volume of ponding water. The partitioned methods involve coupling of a structural solver for membranes and a volume-conserving solver, modeling static incompressible fluid. Two methods of this type are proposed, either using coupling iterations with convergence accelerator between structural solver and volume-conserving solver or adding the linearized fluid behavior in the structural solver in addition to the external coupling iterations. The monolithic methods solve the system of structural equations under hydrostatic load with the volume conservation behavior of the fluid included in the Newton-Raphson (N-R) iterations of the structural solver. One such method was already discussed in the literature and updates the free surface plane to conserve volume exactly after every N-R iteration. In the second, new monolithic method, the volume conservation constraint is added as an additional equation and solved together with the structural equations. It was found that the partitioned method used with a quasi-Newton convergence accelerator was very robust but slower than the monolithic methods. On the other hand, the new monolithic method proposed in this paper was found to be both computationally efficient and robust.