by Steven Lottes1, Tanju Sofu1, Kornel Kerenyi2, Junke Guo3, Bushra Afzal3, and Bhaskar Tulimilli4
Analysts at Argonne National Laboratory's Transportation Research and Analysis Computing Center (TRACC) and researchers at the Turner-Fairbank Highway Research Center (TFHRC), the University of Nebraska, and Northern Illinois University are collaborating to develop computational fluid dynamics (CFD) methodologies for conservative prediction of bridge scour using validated, commercial CFD software. Scour depth estimation is an essential element of foundation design for bridges and of scour failure risk assessment for bridges during floods. CFD approaches to scour evaluation apply the fundamental laws of conservation of mass, momentum, and energy directly in the analysis except where limits of the theory or computational resources require models to make the analysis feasible. The work pursued in this program is the start of an effort to facilitate the transfer of the fruits of academic research on scour and scour mechanisms into the realm of reliable commercial CFD applications that can be routinely used in the engineering analysis of scour risk at bridges.
The objectives of the initial effort were to predict the depth and contour of scour holes that form in an initially flat sand bed under inundated laboratory scale bridge decks and to compare the model predictions with experimental data from the TFHRC hydraulics laboratory. TFHRC experiments start with an unobstructed open channel flow in a flume with velocity just below the critical value. When the bridge deck is immersed in the flume to imitate the flood conditions, the flow redistribution due to the obstruction of the deck causes erosion of the sediment layer in the vicinity of the deck. The local shear stress under the deck exceeds the scour-critical limit typically by a factor of two to four resulting in accelerated sediment pickup.
Numerous modeling options ranging from single-phase to multi-component and multi-phase flow techniques in major commercial CFD codes are being evaluated to simulate pressure-flow scour and other scour mechanisms. As one of the modeling options, a single-phase flow approach with a moving sediment bed interface is being investigated as a basis for developing and testing models with more detailed physics. Solutions for the flow field, turbulence, and bed shear stress using an effective bed roughness value, are based on the Reynolds Averaged Navier-Stokes (RANS) equations, and a k–ε turbulence closure model. The bed shear stress distribution is used to compute bed displacements where the shear exceeds the critical shear stress. The bed is displaced, and the process is repeated until the bed shear stress is everywhere at or below critical. Two studies were carried out using the STAR-CD and STAR-CCM+ commercial CFD software packages. One used a two dimensional geometry with the bed modeled as a porous medium, and the other used a full three dimensional geometry with the bed modeled as a rough wall.
In the porous media model, the packed sediment layer was represented as a porous medium using distributed resistance coefficients suitable for sand with 1 and 2 mm mean diameters used in TFHRC scour experiments. Although the porous sediment layer creates a high flow resistance, the no-slip condition does not hold at the interface between the free flow region and the stationary bed. In addition, some of the mean flow diverted down by the flooded bridge deck blockage enters the top layers of the bed upstream of the deck and re-emerges farther downstream under the deck. This low velocity flow within the porous bed generates body forces on particles in the bed that can contribute to the initiation of particle motion causing scour. The flow pattern computed with the bed modeled as a porous media is different from that obtained when modeling the bed as a rough wall.
The three dimensional scour modeling work tackled the problem of maintaining the volume computational mesh quality while displacing the bed surface based on the computed bed shear stress. This process was automated to allow a sequence of CFD runs, bed displacement, and computational grid re-meshing iterations to be carried out without human intervention until a scoured bed topology with shear stress at or below the critical value is found. Without the automation a dozen iterations of the process took about two weeks to complete. With the automation 100 iterations have been completed within two days. The three dimensional scoured bed topology is compared with TFHRC experimental data for a series of bridge deck inundation depths for pressure flow scour with 1 and 2 mm sand. The agreement is reasonably good for this type of scour using a RANS turbulence model to determine bed shear stress. The single phase model does not include sediment transport and therefore there is no settling and redeposition of sediment in the downstream. The results under these simplified conditions yield a conservative prediction of the scour hole depth that does not under predict the scour, and the three dimensional effects caused by the presence of the flume walls are present in the results of the simulations.
1 Argonne National Laboratory
2 FHWA's Turner-Fairbank Highway Research Center
3 University of Nebraska-Lincoln
4 Northern Illinois University
Link to complete paper Link to complete paper