Numerical approach to performance evaluation of cylindrical baffles in real-scale debris flow experiment

Abstract

Debris flow is a flow-like landslide in which mixtures of water and earth materials move downward rapidly. Installation of physical countermeasures, such as check dams, mitigates the damage caused by debris flows. Thus, performance evaluation of these countermeasures through real-time and large-scale debris flow physical field experiment is critical. However, field experiments require lengthy preparations, a human workforce, and a high cost-benefit ratio. Thus, numerical analysis of these real-scale experiments becomes an efficient alternative performance evaluation solution. This research collects topographic data from a read-scale debris flow experiment in Pyeongchang, South Korea, through an Unmanned Aerial Vehicle-Light Detection And Ranging (UAV-LiDAR) survey. The Three-Dimensional Dynamic Analysis (DAN3D) numerical model simulates the debris flow test without countermeasures. The runout distance and frontal velocity from the experiment and simulation without countermeasures show good agreement when the turbulence parameter ($\zeta$) is 1100 m/s2, and the friction coefficient (f) is 0.11. In contrast, the Smoothed-Particle Hydrodynamics (SPH)-based DualSPHysics code simulates the experiment with two rows of cylindrical baffles installed along the channel width as countermeasures for debris flows. The simulated frontal velocity along the runout distance of the debris flow, with cylindrical baffles, shows good agreement with the experimental result when the Bingham viscosity ($\mu_B$) is 9.05 Pa·s, and the yield shear stress ($\tau_y$) is 19.05 Pa. Frontal velocity of debris is reduced by topographical characteristics and cylindrical baffles. The more left the baffle in the first array is located, with respect to the upstream direction, the larger impact load is exerted. Furthermore, due to topographical characteristics, the impact load reaches the second peak value earlier as the baffle is located on the left with respect to the upstream direction. However, no significant differences in impact load trend are not observed in the second array. Comparing the local impact loads from the DualSPHysics simulation and the experiment, the limitations of the numerical approach used in this study is discussed. Therefore, this study provides the valuable information to real-scale experiment design and performance evaluation of countermeasures.