Assessment of Secondary Collision Hazards Induced by Tsunami Bores Using a Coupled NWT–LS-DYNA Framework
Article information
Abstract
Although tsunami-driven drifting debris poses a critical secondary hazard to coastal infrastructure, quantifying the resulting impact forces is challenging because of the limited post-event observations and a lack of full-scale experimental data. This paper presents a coupled numerical framework integrating a numerical wave tank (NWT) with LS-DYNA to estimate the debris-induced impact forces under tsunami-like inundation conditions. The NWT was used to simulate wave generation, propagation, overtopping, and the resulting nearshore hydrodynamics. In contrast, the arbitrary Lagrangian–Eulerian (ALE) formulation in LS-DYNA was used to resolve debris transport and collision dynamics. Validation against existing hydraulic experiments showed that the proposed framework captures the overall sequence of debris drift and collision reasonably well. Nevertheless, some discrepancies remain in the drift initiation and collision velocity because of the simplified representations of friction and contact. The results further suggest that local contact conditions during collision and the momentum characteristics of the inundation flow strongly influence the maximum impact force. The proposed framework provides conservative estimates of debris-induced impact forces because the numerical model tends to produce surface-to-surface collisions owing to suppressed rotational motion. Overall, the coupled NWT–LS-DYNA framework provides a rational and conservative basis for assessing debris-collision hazards and supporting the structural design of tsunami-prone coastal infrastructure.
1. Introduction
Tsunamis endanger coastal infrastructure through primary hydrodynamic forces and cascading damage from large, drifting debris. During major tsunamis, unanchored objects such as shipping containers, vessels, and vehicles can be dislodged and transported inland, substantially expanding the extent of damage. Post-event surveys of the 2011 Tohoku tsunami documented severe damage to ports and coastal facilities, with numerous containers and other debris transported inland and striking buildings and infrastructure (Chua et al., 2021). Similar observations were reported after the 2015 Illapel tsunami in Coquimbo, Chile, where debris collisions and cascading damage to transportation and lifeline infrastructure were identified as key contributors to component- and network-level failure (Paulik et al., 2021). In addition to causing direct structural damage, drifting debris and debris damming can obstruct roads and utility corridors, hindering emergency response activities and prolonging functional disruption in coastal communities (Kaida et al., 2024; Koh et al., 2024). These events show that the impact forces generated by tsunami-driven debris collisions can be comparable in magnitude to primary wave forces, necessitating their explicit consideration in tsunami risk assessments and disaster mitigation planning (Nistor et al., 2017; Kaida et al., 2024).
The design codes and analytical models for tsunami-induced loads have traditionally emphasized hydrodynamic pressures and flow-induced forces, whereas debris-collision hazards have received less attention. Early design approaches typically relied on simplified assumptions for single-object collisions, and often failed to address repeated-collision scenarios or debris-field effects (Nistor et al., 2017; Kaida et al., 2024). In many earlier studies, the debris effects were approximated indirectly through correlations with flow depth or velocity, without explicitly resolving debris motion or collision processes (Charvet et al., 2015). Subsequent post-disaster investigations and vulnerability analyses have shown that debris-induced collisions can significantly amplify structural damage beyond that caused by the tsunami loading alone (Ma et al., 2021; Jin et al., 2023). Consequently, recent reviews and emerging guidelines emphasize the need to incorporate debris transport, collision probability, and associated impact forces into tsunami hazard assessments and structural design frameworks (Kaida et al., 2024).
Despite growing recognition of the importance of debris collisions, substantial challenges remain in quantifying the impact forces they generate. Post-event field surveys frequently suffer from incomplete or missing information because debris is often displaced or removed during emergency response and recovery operations (Nistor et al., 2017; Kaida et al., 2024). Consequently, controlled laboratory experiments have played a central role in advancing the understanding of debris motion and collision behavior under tsunami-like conditions. Large-scale hydraulic experiments and flume studies have shown that debris mass, buoyancy, stiffness, geometry, and initial placement strongly influence the debris trajectories and the resulting impact forces during collisions (Ko et al., 2014; Nistor et al., 2017; Kim et al., 2023; Hwang et al., 2024; Lee et al., 2024; Deschamps et al., 2025). On the other hand, physical experiments are often constrained by scaling laws and facility dimensions, limiting their ability to reproduce full-scale tsunami–debris interactions, particularly for long inland drift paths and complex debris–structure contact sequences.
To complement experimental investigations, numerical simulations have increasingly been adopted to analyze tsunami-driven debris transport and collision processes. Recent numerical studies using LS-DYNA for fluid–structure interaction (FSI), as well as coupled smoothed particle hydrodynamics–finite element method (SPH–FEM) and computational fluid dynamics–discrete element method (CFD–DEM) frameworks, have reported enhanced capability in resolving debris motion, debris–structure interaction, and the evolution of impact forces during debris-structure collisions under tsunami-like inundation conditions (Yeom et al., 2017; Hasanpour et al., 2021; Xiong et al., 2022; Jin et al., 2023). Furthermore, these numerical investigations suggest that container deformation and time-varying impact forces can significantly influence impact force estimations, underscoring the limitations of simplified impulse-based formulas commonly used in engineering practice. Although these advances are significant, challenges remain in achieving robust coupling strategies, managing computational costs, and ensuring adequate validation against experimental data.
The present study addressed these limitations by implementing a coupled simulation framework integrating a numerical wave tank (NWT) with LS-DYNA to systematically investigate the collision behavior of large waterborne debris driven by tsunami-like inundation. This methodology extends the numerical approach proposed by Kim et al. (2026), which focused on debris-drift characteristics, by incorporating explicit structural collision analysis into the simulation domain. Wave generation, propagation, and overtopping were simulated using the NWT, whereas inland inundation, debris transport, and collision behavior were analyzed in LS-DYNA via FSI. The proposed coupled simulation framework was validated by systematically benchmarking the numerical results against existing experimental data. Subsequently, the drift and collision behavior of a container placed inland behind a vertical revetment are simulated under tsunami-like wave conditions. Finally, the impact forces resulting from collisions between the container and a fixed structure, namely a port crane leg, were analyzed as functions of wave height, container mass, and revetment type. Through this approach, the study contributes to a more comprehensive assessment of secondary hazards associated with tsunami-driven debris collisions.
2. Coupled Numerical Framework
The present study used a coupled numerical framework integrating NWT and LS-DYNA to simulate the collision behavior of large waterborne debris induced by tsunami-like waves. The validity and applicability of the NWT for tsunami-like wave generation, propagation, overtopping, and inland inundation have been verified systematically through comparisons with hydraulic experiments (Lee et al., 2022a, 2022b). In addition, Hwang et al. (2022) evaluated the applicability of debris-collision simulations based on the arbitrary Lagrangian–Eulerian (ALE) method implemented in LS-DYNA. Building on these prior validations, the present study coupled the NWT and LS-DYNA to analyze the drift and collision behavior of large waterborne debris under tsunami-like inundation conditions.
2.1 Numerical Model Configuration
The coupled numerical framework consists of an offshore domain simulated using the NWT and an inland domain simulated using the ALE formulation in LS-DYNA, as shown schematically in Fig. 1. The offshore NWT domain calculates tsunami-like wave generation, propagation, overtopping, and the resulting flow fields up to the shoreline interface. In contrast, the inland ALE domain resolves debris drift and collision processes driven by inundation flows.
Computational domains of the coupled NWT–LS-DYNA framework for simulating waterborne debris collisions
The two domains were coupled through a shared shoreline interface using a one-way coupling approach. In this approach, the time-series data for free-surface elevation and flow velocity calculated in the NWT were transferred to the LS-DYNA domain and prescribed as inlet boundary conditions in the ALE domain. These offshore-to-onshore boundary inputs were used to drive inland inundation and subsequent debris motion. This coupling procedure followed the approach adopted by Kim et al. (2026).
Following the experimental configurations reported by Kim et al. (2023) and Hwang et al. (2024), two types of revetment structures were considered. A vertical revetment (VR) with a freeboard Fb of 5 cm was adopted as the baseline case. The wave-absorbing revetment (WAR) configuration was formed by installing a porous wave-absorbing structure in front of the VR, while maintaining the crest elevation of the VR.
The wave-absorbing structure consisted of a rubble mound with a 1:2 slope and a porous armor layer installed over it. The rubble layer had a porosity of 0.46 and an average unit weight of 2.2 g. Two layers of tetrapods were placed on top of the rubble mound, each with a porosity of 0.5 and an average unit weight of 368 g.
2.2 Numerical Wave Tank
2.2.1 Governing equations
The NWT incorporates incompressible viscous flow and is governed by the continuity equation for mass conservation and the Navier–Stokes equations for momentum conservation. The volume of fluid (VOF) method was used to simulate complex free-surface motions, including wave breaking (Hirt & Nichols, 1981). The VOF function F represents the volume fraction of fluid within a computational cell, where F = 1 denotes a fully fluid-filled cell, F = 0 denotes an air cell, and 0 < F < 1 corresponds to a free-surface cell.
The governing equations are given by
where xi represents the Cartesian coordinate system (x, y, and z), with the x-axis aligned with the wave-propagation direction and the z-axis pointing vertically upward. t denotes time. vi (u, v, and w) are the velocity components, and γv and γi represent volume and surface porosities, respectively. ρf represents the fluid density. p is the pressure. νT = ν+νt is the total viscosity (where ν is the kinematic viscosity coefficient, and νt is the eddy viscosity coefficient from the turbulence model), and Dij is the strain-rate tensor.
The additional terms are defined as follows:
Si: surface tension force based on the continuum surface force model (Brackbill et al., 1992)
Ri: fluid resistance term in porous media
Gi: gravitational acceleration term (= γvg, where g is the is the gravitational acceleration)
Oi: wave energy damping term, (= −βw, where β is the damping factor), active in sponge layers
Qi: wave source term (
The wave source q* is represented by:
where q is the flux density, and Δxs is the grid spacing in the x-direction at the wave source location xs.
2.2.2 Tsunami-like wave generation
Tsunami-like waves are generated using a source-function method (Lee et al., 2022c) based on Brorsen and Larsen (1987), which produces waves propagating upstream and downstream of the source. The source-strength formulation proposed by Ohyama and Nadaoka (1991) was used to account for this bidirectional propagation and to suppress spurious reflections as follows:
where U0 is the horizontal velocity, ηs and η0 are the free-surface elevations at the source location and the target wave, respectively, and h is the still-water depth.
The prescribed free-surface elevation and horizontal velocity at the source were determined according to Dean and Dalrymple (1991):
where A0 is the incident wave amplitude (i.e., the maximum free-surface elevation ηmax). ε = A0/h is the relative wave height, and C is the wave celerity.
2.3 LS-DYNA Model
2.3.1 Arbitrary Lagrangian–Eulerian method
The ALE method combines Eulerian and Lagrangian descriptions by allowing partial mesh motion relative to material motion. Although Eulerian formulations are suitable for fluids and Lagrangian formulations are effective for solids, the ALE approach alleviates mesh distortion during large deformations (Souli et al., 2000; Belytschko et al., 2014). In the present study, turbulence effects were neglected to focus on inertia-dominated debris–wave interactions under long-wave tsunami forcing. Therefore, the ALE formulation was adopted to capture debris drift and collision behavior during tsunami inundation.
2.3.2 Governing equations
In the ALE framework, mesh motion is decoupled from material motion, and the material derivative is expressed as
where Xi and xi denote the Lagrangian and Eulerian coordinates, respectively, and wi = vi − ui is the relative mesh velocity.
The conservation equations are formulated as follows:
where ρ is the material density; σij is the Cauchy stress tensor; bi is the body force, and E is the internal energy per unit mass.
2.4 Simulation Conditions
A 1/40-scale, ISO-standard 20-ft container, identical to those used in previous experiments, was adopted for the simulations. Table 1 lists the specifications of the two container models, which have identical external dimensions but different loading conditions: Model-A (partially loaded) and Model-B (fully loaded). A single leg of a J100 ship-to-shore (STS) crane (Kosbab, 2010) was placed 100 cm inland from the shoreline, with its top connected to a load cell to measure the impact forces generated during debris collisions. In the model, the original crane leg dimensions (1,340–2,830 mm in height × 1,730 mm in width) were simplified to a square cross-section of 1,600 mm × 1,600 mm and scaled to 40 mm × 40 mm. The crane leg height was limited to 560 mm in the scaled configuration.
Specifications of the 1:40-scale ISO 20-ft container models
The material properties were configured to match the acrylic used in the experiments, specifically a density of 1,180 kg/m3 and a Young’s modulus of 2,690 MPa. The fluid in both computational domains was freshwater at 20°C, with a density of 998.21 kg/m3, a dynamic viscosity of 1.002 × 10−3 Pa·s, and a kinematic viscosity of 1.004 × 10−6 m2/s.
For the bed-contact condition in LS-DYNA, RIGIDWALL_PLANAR was used to represent the flat bed surface. The friction coefficients for contact between the acrylic container and the bed were set to 0.1 and 0.2 for Model-A and Model-B, respectively, according to the same conditions adopted in Kim et al. (2026).
Table 2 lists the tsunami-like wave conditions, excluding cases without overtopping or debris drift. Incident wave amplitude A0 corresponds to the maximum wave elevation ηmax at the wave source. The relative wave height (ε) A0/h, relative freeboard A0/Fb, and wave steepness A0/Le are also given.
Incident wave conditions and the corresponding relative wave parameters
Containers were initially placed inland at detached distances of 10 cm, 50 cm, and 80 cm from the shoreline, resulting in separation distances G between the container and crane leg of 83.9 cm, 43.9 cm, and 13.9 cm, respectively. These correspond to nondimensional separation distances G/h of 1.79, 0.93, and 0.30. Thirty-six simulation cases were analyzed by combining two revetment types, three wave conditions, three detached/separation distances, and two container models.
2.5 Validation of Coupled Framework
The numerical results were compared with the computer vision–based experimental data of Kim et al. (2023) to validate the proposed coupled simulation method for modeling debris drift and collision behavior induced by tsunami-like bores and inland inundation. Fig. 2 presents a direct comparison of the time-sequenced image frames captured by an overhead camera in a wave-flume experiment with the corresponding numerical snapshots obtained from the coupled NWT–LS-DYNA simulations.
Time-sequenced comparison of the plan-view drift behavior of waterborne debris (Model-A) induced by a tsunami-like bore under the VR condition (left: coupled simulation; right: hydraulic experiment)
The selected tsunami condition corresponds to ε = 0.13, with Model-A placed inland at a detached distance of 10 cm behind VR. Fig. 2 shows the sequence of overtopping, bore-driven inundation, and the consequent drift behavior of the container. The numerical results accurately reproduced the dominant translational drift characteristics observed experimentally, including the overall drifting direction and the magnitude of the drifting speed.
A discrepancy was observed in the rotational motion of the container. In the experiment, noticeable rotation occurred during drifting, primarily because of spatially nonuniform bottom friction and asymmetric contact between the container and the bed. By contrast, the numerical simulations assumed uniform bottom friction and symmetric hydrodynamic loading, suppressing the development of rotational motion.
Despite this limitation, the present comparison confirmed that the coupled simulation framework captures the primary translation-dominated drift behavior of waterborne debris during bore-driven tsunami inundation. Consequently, the coupled approach is considered validated with respect to the debris drift kinematics. Nevertheless, its limitations in reproducing rotational motion must be taken into account when interpreting the results.
3. Numerical Results and Experimental Comparison
3.1 Drift and Collision Behavior
Fig. 3 shows the time-sequenced drift and collision behavior of waterborne debris (Model-A) under a representative test condition, namely the VR configuration with ε = 0.21. This condition corresponds to the lowest wave intensity among the tested cases that still resulted in overtopping and subsequent debris motion. Hence, it is well-suited for examining the onset and early-stage evolution of debris drift.
Time-sequenced comparison of the side-view drift and collision behavior of waterborne debris (Model-A) induced by a tsunami-like bore under the VR condition (left: coupled simulation; right: hydraulic experiment)
The coupled NWT–LS-DYNA simulations and the hydraulic experiments indicated that, once overtopping occurred, the debris motion was predominantly translation-dominated, with translational displacement governing the motion while rotational components remained limited. The debris then progressed toward a collision with the fixed structure. The overall sequence, comprising bore arrival, initiation of debris motion, drift, and eventual collision, was consistently reproduced by the numerical model, demonstrating its capability to capture the fundamental kinematics of tsunami-like bore-induced debris transport.
Although the qualitative agreement was strong, a discrepancy was observed in the timing of debris motion. Under the present condition, the numerical simulation predicted an earlier onset of drift and an earlier arrival at the collision location than those observed experimentally. This phase lead suggests that the numerical model initiates sliding and transport under a lower effective hydrodynamic threshold. Such behavior was attributable to the simplified representation of bed–debris interactions, in which bottom friction is assumed to be spatially uniform and symmetric.
In contrast, the experimental results showed delayed drift initiation and more complex near-bed interactions, reflecting the influence of spatially nonuniform bottom friction and asymmetric contact conditions inherent to the physical setup. These differences affected the timing of mobilization but did not alter the overall drift–collision sequence. Accordingly, Fig. 3 confirms that the coupled simulation framework reproduces the essential drift and collision behaviors under low-to-moderate tsunami-like forcing, while highlighting the limitations in accurately capturing the timing of debris motion initiation.
3.2 Collision Velocity Characteristics
Fig. 4 presents a quantitative comparison of the measured and simulated collision velocities for waterborne debris colliding with the fixed structure across different container models and revetment types. Overall, the coupled NWT–LS-DYNA simulations showed strong agreement with the experimental measurements, exhibiting clear linear correlations across all tested conditions. The correlation coefficients were approximately 0.94 and 0.95 for Model-A and Model-B, respectively, indicating that the numerical framework reliably captured the relative variations of collision velocity.
Comparison of the measured and simulated collision velocities for waterborne debris impacting the fixed STS crane leg. The colors indicate the revetment type (VR: blue; WAR: red); the symbol shapes represent the separation distance, and the filled and open symbols denote debris mass (Model-A and Model-B, respectively).
The influence of the container mass was consistently observed in the experiments and simulations. Model-B consistently exhibited higher collision velocities than Model-A across all tested conditions. Moreover, this mass-dependent separation was clearly reproduced by the coupled simulations, showing their ability to represent the relative differences in collision velocity associated with debris loading conditions.
Regarding the revetment type, the VR and WAR cases showed good agreement between measured and simulated collision velocities. No distinct clustering or systematic deviation associated with revetment type was identified (Fig. 4), suggesting that the coupled model captures the effects of the structural type on the collision velocity without introducing significant distortion.
The collision velocity exhibited a discernible dependence on the G/h ratio. Higher collision velocities correlate with increased G/h ratios for both container models, suggesting that debris initially located farther from the target structure underwent a more extended acceleration path within the inundation flow prior to collision. Under tsunami-like bore conditions, debris motion exhibits an acceleration-dominated phase before reaching a quasi-steady drift state, during which the drift velocity increases (Kim et al., 2023). Therefore, collisions occurring during this acceleration phase reflect the degree of acceleration attained before impact, which increases with the separation distance.
Despite these strong correlations, the simulated collision velocities were generally lower than the measured values. The mean negative bias was approximately 0.17–0.20 m/s. This underestimation remained consistent across container models and revetment types. Therefore, the persistence of this bias indicated a systematic tendency of the numerical framework rather than case-specific discrepancies. Nevertheless, the high correlation and consistent ranking across conditions showed that the coupled simulation approach provides a robust basis for a comparative assessment of the collision velocity characteristics, even though the absolute values were modestly underestimated.
3.3 Time Histories of Impact Forces
Fig. 5 shows the time histories of the horizontal impact force acting on the STS crane leg during collisions involving Model-A under the same experimental and numerical conditions considered in Fig. 4. For a direct comparison, the measured and simulated force signals were temporally aligned with respect to the instant of the maximum impact force.
Comparison of the measured and simulated time histories of the horizontal impact force acting on the STS crane leg during collision with waterborne debris (Model-A) under the VR condition
The experimental measurements and the coupled NWT–LS-DYNA simulations exhibited a distinct impulsive force peak at the moment of collision, followed by a gradually decaying oscillatory response. The initial peak corresponds to the primary contact between the container and the crane leg and represents the dominant component of the impact loading. The numerical results reproduced the timing and magnitude of this primary impulse with reasonable agreement.
Differences were observed in the force evolution immediately before and after the primary collision event. The numerical simulation showed that the debris drift initiated earlier than in the experiment, leading to discrepancies in the hydrodynamic loading conditions before collision. These differences manifest as variations in pre-impact force fluctuations and in the post-impact oscillatory response. In particular, the simulated force histories showed attenuated secondary peaks and a smoother decay compared with the experimental records.
The observed discrepancies were attributed to differences in the representation of FSI during the post-collision phase. In the hydraulic experiments, residual flow motion, local free-surface disturbances, and spatially nonuniform bed friction contributed to complex transient loading on the structure following impact. In contrast, the numerical model adopted idealized contact conditions and spatially uniform friction, which tended to suppress high-frequency force components and secondary impact features.
Despite these differences in the detailed force evolution, the dominant impulsive component associated with the primary collision was consistently captured by the coupled simulation framework. Therefore, the time-history comparison presented in Fig. 5 provides a basis for examining the characteristics of the maximum impact force, which were addressed separately in Section 3.4.
3.4 Maximum Impact Force Characteristics
Fig. 6 compares the maximum debris-induced impact forces acting on the fixed structure obtained from the hydraulic experiments and the coupled numerical simulations. Overall, the numerically predicted maximum impact forces were on average 49.42% larger than the experimental measurements. When classified using the container model, the simulations overestimate the experimental values by 42.83% and 56.01% for Model-A and Model-B, respectively. Compared with the collision velocity results shown in Fig. 4, the maximum impact force showed substantially greater variability in the experimental data, indicating greater sensitivity to the collision conditions and localized hydrodynamic effects.
Comparison of the measured and simulated maximum impact forces acting on the fixed STS crane leg during collisions with waterborne debris
3.4.1 Effects of revetment type
The influence of the revetment type on the maximum impact force was examined by comparing the WAR and VR cases under identical wave and container conditions. In the experiments, the WAR/VR ratio of the maximum impact force spans a finite range because of the mixed debris–structure contact configurations during collisions. For Model-A, the experimental WAR/VR ratio ranged from 0.83 to 1.49, with a mean value of 1.14, while for Model-B, it ranged from 0.77 to 1.47, with a mean of 1.15. Regarding the revetment type, the experimental results revealed a tendency toward higher average maximum impact forces under WAR than under VR.
In the coupled simulations, the maximum impact force exhibited a consistent response across test conditions. The resulting WAR/VR ratios showed systematic amplification under WAR. The simulated mean WAR/VR ratio was 1.48 and 1.45 for Model-A and Model-B, respectively, indicating that the simulations successfully reproduced the overall tendency toward larger maximum impact forces under WAR.
Furthermore, this tendency can be explained by the momentum-flux characteristics of inundation. Previous studies have shown that, although WAR effectively dissipates the incident wave energy, relatively strong onshore momentum flux can persist during the inundation phase compared with VR (Lee et al., 2022b). Under WAR conditions, enhanced horizontal flow components act on drifting debris before collision, leading to larger maximum impact force responses.
3.4.2 Effects of container mass
The effects of container mass on the maximum impact force were evaluated by comparing Model-B and Model-A under identical wave and revetment conditions. In the experiments, the Model-B/Model-A ratio of the maximum impact force spans a finite range, reflecting the variability of the debris–structure contact modes, including surface-to-surface, line-to-surface, and point-to-line impacts (Hwang et al., 2024). Under VR, the experimental Model-B/Model-A ratio ranged from 0.64 to 1.41, with a mean of 1.10, while under WAR, it ranged from 0.57 to 1.53, with a mean of 1.12. These results suggest that Model-B generally produces larger maximum impact forces than Model-A, while exhibiting substantial experimental scatter.
In the coupled simulations, larger maximum impact forces were consistently obtained for Model-B than for Model-A. The simulated mean Model-B/Model-A ratio was 1.24 and 1.15 under VR and WAR, respectively, showing that the numerical results capture the mass-dependent tendency observed in the experiments. Nevertheless, the absolute magnitude of the maximum impact force remains sensitive to the contact idealization.
3.4.3 Effects of separation distance
The effects of separation distance on the maximum impact force were examined using G/h under identical incident-wave and revetment conditions. Regarding the separation distance, the maximum impact force did not show a clear, monotonic dependence on the separation distance for either container model. Considerable scatter was observed across different G/h values, suggesting that the separation distance alone does not determine the maximum impact force response.
Although a larger G/h generally leads to a higher collision velocity, an increase in collision velocity does not necessarily translate into a proportional increase in the maximum impact force. This observation reflects the nonlinear nature of debris–structure collision, in which the maximum impact force is influenced by the contact conditions and transient force transmission during impact rather than by velocity alone. Under tsunami-like bore conditions, debris motion undergoes an acceleration-dominated phase before reaching a quasi-steady drift state, during which the drift velocity increases (Kim et al., 2023). Therefore, collisions during this transient phase occur under highly unsteady inundation-flow conditions, leading to substantial variability in the peak impact force.
Additional variability was observed in the hydraulic experiments because of mixed modes of debris–structure contact. Even in numerical simulations, where surface-to-surface collisions occurred owing to suppressed rotational motion of the debris, the maximum impact force remained sensitive to the transient contact duration and local hydrodynamic conditions. Consequently, while the separation distance affects collision velocity, its influence on the maximum impact force was outweighed by the momentum of the inundation flow and the collision contact characteristics.
4. Discussion
The present study examined debris–structure collisions induced by tsunami-like inundation as a secondary hazard characterized by localized, transient impact loading arising from interactions between the inundation flow and mobilized debris. Unlike primary tsunami hazards, which are governed by large-scale hydrodynamic forces, debris-induced collisions involve short-duration, contact-dominated loading, posing inherent challenges for quantitative evaluation within conventional design frameworks. Furthermore, the results suggest that the severity of debris-induced impacts is closely linked to the momentum characteristics of the inundation flow. As the tsunami-like wave intensity increases, the depth of inland inundation and onshore-directed momentum are enhanced, accelerating drifting debris and increasing collision velocity. These results suggest that the dynamics of the inundation bore strongly govern debris transport and collision kinematics.
Nevertheless, the relationship between the collision velocity and maximum impact force is inherently nonlinear because different controlling factors govern the two quantities. Although the collision velocity is influenced primarily by inundation-flow momentum, debris mass, and the acceleration history associated with the initial separation distance, the resulting maximum impact force depends critically on the collision contact conditions and transient force transmission during impact. The experimental observations revealed substantial variability in impact force because debris–structure collisions may involve diverse contact modes, including surface-to-surface, line-to-surface, and point-to-line interactions. In contrast, the numerical simulations tended to produce surface-to-surface collisions because the spatially uniform bed-friction condition suppressed rotational motion of the debris during drift. This tendency increased the effective contact area and promoted more direct force transmission during impact, resulting in consistently higher maximum impact forces. Accordingly, collision velocity alone is insufficient to determine the peak impact force uniquely. Instead, the effective contact duration and contact configuration play a dominant role in governing force transmission during debris–structure interaction (Mizutani et al., 2005; Haehnel & Daly, 2004). From an engineering perspective, this modeling characteristic provides a conservative representation of debris-induced secondary hazards by emphasizing the plausible upper-bound loading scenarios.
By contrast, the revetment type affects the debris impact severity, primarily through its influence on inundation momentum rather than collision kinematics alone. For example, WAR sustains stronger onshore-directed momentum during inundation, leading to larger impact forces than VR. The debris mass also plays a consistent role, with heavier debris producing larger impact forces because of increased inertia, a tendency reproduced by experiments and simulations despite differences in the absolute force magnitude. Moreover, the influence of the initial separation between the debris and the structure further underscores the localized and transient nature of debris impacts. Although greater separation allows debris to experience a longer acceleration phase before collision and can lead to higher collision velocity, the maximum impact force does not exhibit a clear or systematic dependence on the separation distance. Once the inundation flow entrains debris, the collision response is governed predominantly by instantaneous flow momentum and contact behavior at the moment of impact, resulting in considerable variability in the maximum impact force.
Overall, debris-induced collisions during tsunami-like inundation constitute a distinct secondary hazard dominated by localized, transient, and nonlinear loading mechanisms. Although the contact representation remains simplified, the coupled NWT–LS-DYNA framework provides a robust, conservative foundation for estimating debris-induced impact forces. Consequently, this approach effectively complements conventional assessments focused on the primary tsunami loading and facilitates the quantification of secondary hazard demands acting on coastal structures.
5. Conclusions
This study evaluated a coupled NWT–LS-DYNA numerical framework to assess large waterborne debris collisions induced by tsunami-like inundation, with a particular focus on debris-induced impact forces as a secondary hazard. By coupling offshore wave generation and inundation processes with inland debris drift and collision dynamics, the applicability and limitations of the framework were examined by a systematic comparison with hydraulic experiments. The main conclusions are as follows:
(1) The coupled NWT–LS-DYNA framework successfully reproduced the fundamental sequence of debris transport, acceleration, and collision under tsunami-like inundation conditions. Although minor discrepancies were observed in the timing of drift initiation and collision velocity because of simplified representations of bed friction and contact conditions, the overall drift and collision behavior were consistently captured.
(2) Experimental evidence suggests that fully floating debris most frequently undergoes line-to-surface or point-to-line collisions. In contrast, surface-to-surface collisions occur less frequently but produce the most significant impact forces. Because the coupled numerical framework inherently emphasizes surface-to-surface contact, the resulting impact force predictions can be interpreted as conservative upper-bound estimates of physically plausible worst-case collision scenarios.
(3) Debris–structure contact conditions can vary substantially, as observed in the experiments. Hence, the relatively conservative impact forces associated with the surface-to-surface collision tendency in the coupled NWT–LS-DYNA framework may be useful for assessing the structural safety of coastal structures against secondary collision hazards under tsunami-like inundation conditions.
Future work should focus on improving the representation of debris–bed interactions, floating-state dynamics, and contact-mode variability. In addition, the parameter range should be extended to conditions that allow debris to undergo a longer acceleration phase before collision. These efforts are expected to clarify the role of the separation distance in debris impact assessments.
Notes
Woo-Dong Lee is an editorial board member of the Journal of Ocean Engineering and Technology. However, he was not involved in the decision-making process for publishing this article. No potential conflicts of interest related to this article have been reported.
This work was supported by the National Research Foundation of Korea (NRF) grants funded by the Korean government (MSIT) (RS-2024-00356327).