Validation of computational fluid-structure interaction models by comparison with collapsible tube experiments.

摘要

The objective of this thesis was to assess the validity of the fluid-structure interactionud(FSI) facilities in LS-DYNA for the analysis of highly deformable structuresudinteracting with flowing viscous fluids. The collapsible tube experiment was chosenudas a validation tool for FSI since its three-dimensional computational modellingudwould have been impossible if the viscous internal fluid flow were not considered.udAn explicit three-dimensional finite element model of a collapsible-tube wasudconstructed and solved using LS-DYNA. The fully coupled model included internaludfluid flow; external, inlet and outlet pressures; tube wall tension; pre-stressing; andudcontact. The finite element boundary conditions were taken as the recorded values ofudflow rate and pressure from a standard collapsible-tube experiment for both steadyudand unsteady flows.udThe predicted tube geometry in the steady LS-DYNA model showed good agreementudwith the experiment for operating points in the highly compliant region of theudpressure-flow characteristic curve. The comparative position of the pinch at theudoutlet end differed by only 5.6% of the outlet diameter in the worst case.udThis analysis represents an advance on other published work in that previously noudcomparison with experiments have been drawn for FSI models involving highudReynolds number flowing viscous fluids interacting with highly deformable three dimensionaludstructures. This analysis successfully made that comparison and theudexperimental and computational results have combined to form a more detailedudpicture of the collapsible-tube phenomenon by including detailed stress results of theudtube walls and views of the internal fluid flow.udThe collapsible tube model exhibited uncertainty errors due to the use of a coarserudthan desirable mesh and a reduced fluid speed of sound. Although both theseudapproximations caused significant error in the model both were necessary in order toudachieve acceptable solution times. Because of these errors a thorough quantitativeudvalidation could not be achieved although LS-DYNA has been proven to beudqualitatively accurate. Increases in computing speed are required before thoroughudquantitative validation of FSI can be achieved by comparison with the collapsible tubeudexperiments.