Comparative hydrodynamic study of rigid-lid and level-set methods for LES of open-channel flow

Computational modeling of open-channel flows at low Froude (F) numbers is often carried out using the rigid-lid assumption to prescribe the water surface location. However, the presence of wall-mounted hydraulic structures could cause F to locally exceed unity, inducing a transition from subcritical to supercritical flow, locally accelerate the flow generating high-energy turbulent eddies at a constriction region, and lead to the formation of backwater at upstream regions. Under such circumstances, the rigid-lid assumption may not be adequate to obtain reliable flow field simulations. Previous investigation to examine the effect of a rigid-lid assumption on large-eddy simulation (LES) of a low-Reynolds open-channel flow showed that the second-order turbulent statistics are altered depending on the free-surface resolving method. In this work, we seek to study the rigid-lid effect on high-Reynolds LES applicable to natural river flows. We employed an in-house Virtual Flow Simulator (VFS-Geophysics) model in its LES mode to simulate the flow field in a similar test case that is scaled up to obtain a high Reynolds of 7.9 × 10⁴. We performed LES evaluating the two free-surface resolving approaches: rigid-lid and level-set. The flow field and free-surface computations were validated using a series of measured data that we obtained by carrying out laboratory experiments. Our LES results indicate that at near-bed regions of the flume and around the bridge abutment, both the first- and second-order turbulent statistics are influenced by the method selected to prescribe the water surface.