|
It isn't recommended to explicitly model bridge piers in TUFLOW, or any other 2D or 3D software (the possible exception being CFD software), but the reasons why are a little complicated.<br>
Small scale obstructions to the flow, such as trees, poles, piers, etc. cause additional head losses along a flow path due to their drag characteristics. Historically, dragform loss (or form lossdrag) coefficients for various profile shapes have been determined as a function of Reynold’s number through experimental testing. More recently, computational fluid dynamics (CFD) has been used to attempt to reproduce the velocity field in the wake of such objects,. althoughAlthough providing better results than 2D modelling, the results have not always agreed well with physical tests. In particular, the drag of a given profile depends on the exact location of flow separation points, which in turn depends on the ability of the CFD code to predict the laminar to turbulent transition in the boundary layer, which is many times smaller than the profile shape itself. In general, the form loss results from CFD models show significant sensitivity to mesh size, mesh design, and choice of turbulence model, therefore,. considerableConsiderable caution needs to be exercised even for CFD modelling.<br>
[[File:Flow round a cylinder.png]]
''The point of flow separation around an object has a major bearing on the drag coefficient and is not reliably reproduced by 2D or 3D software.''
Modelling 2D flow around profiles with the 2D or layered 3D form of the shallow water equations (SWE) as used by TUFLOW and other free-surface water flow solvers, is no different in this regard. While mesh-resolved wakes behind the piers using a fine mesh can be seen in the results, the predictions for head losses show the same sensitivities (mesh size, mesh design, choice of turbulence model) as seen in 3D CFD.<br>
The safest and strongly recommended approach with regard to establishing head losses and thereforeconsequently flood levels, is to not resolve the obstructions in the mesh but instead model the effects of such obstructions with form loss (drag) loss coefficients (applied to selected mesh cells) that have been derived from physical testing. This approach has been shown to provide the most consistent results across various mesh resolutions. It also has the added benefit that, by avoiding small cells in the mesh, it will provide much more efficient run times for flow solvers.<br>
== What are the limitations of explicitly modelling bridge piers in TUFLOW? ==
|