[Univ of Cambridge] [Dept of Engineering]

Computational Fluid-Structure Interaction Applied to Long-Span Bridge Design

J B Frandsen


This thesis describes an investigation into the use of numerical methods to predict the aeroelastic behaviour of long-span bridges, and provides an account of measurements of the dynamic behaviour of a suspension bridge obtained on-site in Denmark.

That there is still a problem with the conventional experimental methods in predicting the full-scale behaviour of long-span bridges was confirmed when, as a part of this research, large oscillations were measured on the Great Belt East suspension bridge.  The bridge was instrumented with accelerometers and wind-pressure sensors.  Lock-in samples are presented and evaluated.  For the first time pressures at deck surface and accelerations of the bridge deck are measured simultaneously.  Correlations at and off lock-in are estimated.  The experimental data are compared with the results of the numerical analyses.

The main numerical investigations are based around the finite element method (FEM) using the commercial code Spectrum which employs an arbitrary Lagrangian-Eulerian formulation to stimulate the fluid-structure interaction, using both stationary and moving two-dimensional grids.  The FE-analyses are restricted to those that can be solved on a fast work-station.  In comparative studies, the Discrete Vortex Method (DVM) is applied using the COWIconsult code DVMflow which has been developed for bridge applications.  The aim is to explore whether the FE approach can provide a supplementary tool for designers and can aid the understanding of the physics of flow processes, potentially reducing the large number of expensive physical model tests that are currently required.  The numerical models show the ability to self-excite into vortex-induced oscillations and flutter, and the detailed flow visualisations obtained could contribute to further understanding of the fluid-structure interaction behaviour.  These flow details would be difficult to obtain in a wind tunnel.  In the flow regime associated with vortex-induced vibrations, the numerical model exhibited mesh-dependency and difficulties were found in simulations of the shedding frequency.  Modelling of flutter in the higher wind speed flow regime showed more promising results, even though they were obtained on relatively coarse meshes.

The intended end-use of the FE-method by bridge aerodynamicists forms the motivation behind these studies.  However, this research suggests that successful modelling would require the use of parallel processing, and therefore the discrete vortex method appears to be the more promising approach for application to bridge design.

[Cambridge University | CUED | Structures Group | Geotechnical Group]

This page is maintained by rcb@eng.cam.ac.uk (last update 10 November 2003)