Multiscale numerical approach to assess impact forces on cylindrical barriers by debris flows

Debris flow is a flow-like landslide involving a mixture of water and earth materials that rapidly moves downslope. Conducting real-scale experiments is costly and resource-intensive, making numerical analysis an efficient alternative for evaluating the performance of countermeasures. This study investigates impact forces on barriers caused by debris flows using a multiscale numerical approach. First, a 3D depth-averaged model simulates debris flow characteristics at a channel-length scale. Then, a 3D Smoothed Particle Hydrodynamics (SPH) model examines debris-barrier interactions at a channel-width scale. Two real-scale debris flow experiments conducted in Pyeongchang, Korea, were used for validation: one without barriers for flow calibration and another with cylindrical baffle-type countermeasures for impact force validation. The depth-averaged model was calibrated to replicate observed runout distances and velocities, while the SPH model simulated interactions with cylindrical baffles. Due to the curved terrain near the first baffle array, debris primarily impacts outer baffles, resulting in higher dynamic impact coefficients. The parametric study reveals that the initial velocity has a negative correlation with the dynamic impact coefficient but a positive correlation with the Froude number. Increased density reduces both the dynamic impact coefficient and pressure gradient. In addition, previously proposed semi-empirical models effectively reproduce the summed impact forces across all baffles upon calibration against the numerical results, but capturing individual baffle’s responses and estimating localized impact forces in open-type barriers still remain challenging. These findings highlight the critical influence of local terrain and flow dynamics on impact force distribution. demonstrating the values of combining numerical simulations and real-scale experiments in geohazard risk assessment and mitigation strategy development.