1. Introduction
Tsunami events such as the one occurred in Japan in 2011 in Tohoku [
1] and the ones in Indonesia in 2004 and 2018 [
2] resulted in catastrophic consequences on coastal communities and structures causing many casualties. Studies in the literature often focused on the impact of tsunamis on the coastal region [
3,
4,
5,
6,
7]. However, recent research investigated the effects that debris transported by such tsunamis can have on existing structures, especially after the damage documented during the 2011 Tohoku tsunami. Using the evidence found after this event a debris hazard classification was developed by Naito et al. [
8] by analysing the damages caused by different debris types. Among the types with most energetic estimated impact forces on structures were shipping containers, especially loaded ones. Forces between 1500 and 6400 kN were estimated by Naito et al. [
8] on structures by applying the FEMA P646 [
9] guidelines for the design of structures. These recommend a formula for the calculation of the maximum force exerted by floating debris based on the Haehnel and Daly [
10] relationship. In the wake of the 2011 Tohoku tsunami, further relationships were incorporated in the American Society for Civil Engineers (ASCE) [
11] design codes and provide a simplified model for large floating debris impact forces for structural design, accounting also for impact orientation based on the mean impact orientation of Haehnel and Daly [
10]. However, important factors such as debris material are not yet taken into account in any design guideline to date.
In order to have more detailed models for large floating debris dynamics, a large number of studies focused on investigating their behaviour when entrained in a flow, e.g., trajectories, spreading, orientation etc. [
1,
12,
13,
14,
15] or on analysing their impact forces during a collision with a structure [
16,
17,
18,
19]. More recently Derschum et al. [
20] and Stolle et al. [
21,
22] studied both processes in conjunction by using a single “smart debris” [
23] shaped as a 1:40 geometrically scaled 20 ft standard container to analyse both the trajectories and the impact forces on a structure with the aim of improving the reliability of design guidelines for structures that may be exposed to such hazards and to serve as benchmark tests for future numerical simulations.
The numerical modelling of these processes is rare in the literature [
24] with significant limitations in the numerical methods used. Initial studies on solid-fluid interactions for debris were conducted by Wu et al. [
25]. They used Navier–Stokes equations coupled with Volume Of Fluid (VOF) to track the free surface, an approach very often used in modelling fluid-structure impacts (see, e.g., De Finis et al. [
26]), and a large eddy simulation turbulence model. The motion of the floating body was modelled with a Discrete Element Method (DEM) model validating it with laboratory experiments. However, implementations of floating body mechanisms on mesh-based models are usually very challenging, requiring ad hoc formulations for specific cases and long processing times [
27]. Due to these difficulties, Lagrangian particle-based methods, such as the Moving Particle Semi-implicit method (MPS) [
28] or the Smoothed Particle Hydrodynamics method (SPH) [
29] are often used as good alternatives to mesh-based methods [
30,
31]. SPH, due to its formulation, is well suited to investigate high energy phenomena since it is capable of modelling violent flows as well as flow particle detachments in the context of tsunamis [
32,
33]. However, even if these methods produced accurate results for solid–fluid interactions [
31,
34] they do not inherently account for solid body contact laws, i.e., this contact requires a specific sub- or coupled model. Canelas et al. [
35] and successively [
36] expanded an SPH solver via coupling with a DEM model which was validated with PVC cubes subject to a flow, but showed limitations in correctly simulating the friction between moving and boundary elements of the simulations. This was further shown in Goseberg et al. [
37] which modelled experiments from Nistor et al. [
15]. The combined use of SPH and DEM clearly resulted in negligible variations of debris dynamics when changing the kinematic friction coefficient
fc of the materials. Additionally, DEM implementation was also prone to numerical instabilities, also resulting in high computation times.
In order to address the aforementioned issues, the newly implemented DualSPHysics-CHRONO coupling in DualSPHysics v5.0 [
38,
39] is used to model the container trajectories measured in Stolle et al. [
21,
22]. This is, additionally, a first necessary step to study the feasibility of such approach and successively extending their use for flow-debris-structure interactions. DualSPHysics is chosen due to its advantage over other SPH numerical models of using the computational power of the Graphics Processing Unit (GPU) of a PC to speed up simulation times up to 146 times [
38]. CHRONO [
40,
41] is a multi-physics engine that allows to solve rigid or deformable contacts and impacts between solids. In addition, it includes the possibility of integrating different types of constraints as recently done by Tagliaferro et al. [
42] to simulate a planing hull. In contrast to the DEM implementation, CHRONO allows for a realistic frictional behaviour due to the inclusion of a full Coulomb sliding/sticking/rolling model. Additionally, the application of different types of constraints such as joints, hinges and springs allows to include complex structures allowing to realistically model the ones found in coastal regions or in harbours. Another advantage is the increased numerical stability gained when compared with DEM coupling.
The main objective of this work is to validate the coupling of DualSPHysics and CHRONO with the experiments presented in Stolle et al. [
22]. First, the modelling of the hydrodynamics of the flow is validated, then a sensitivity analysis of the modelling of the details of the debris geometry and description of the material characteristics on the debris dynamics is carried out. Finally, the results of the calibrated models are presented and discussed.
The remainder of the manuscript is organized as follows. The explanation of the methodology, illustrating the numerical setup along with a brief introduction on the numerical methodology used is presented in
Section 2. In
Section 3 the main results of the initial hydrodynamics validation and subsequent calibration and validation of the debris dynamics are presented. Finally, in
Section 4, the results are analysed, and the feasibility of this numerical approach is discussed in relation to potential applications for debris-structure interactions and conclusions of the work are given.
4. Discussion and Conclusions
This section discusses the findings of this study with the focus on analysing the future applicability of this numerical methodology in a broader context, as introduced in
Section 1. First, the accuracy of the simulation results was compared with the accuracy reached by the container experimental measurement system as additional benchmark. Stolle et al. [
22], as measurement system, employed an optical tracking algorithm introduced in Stolle et al. [
57] where it is investigated its accuracy using the parameters introduced in
Section 2.5. The mean values of the parameters calculated by Stolle et al. [
57] for the experimental validation tests with a single container are summarised in
Table 6.
and
values for the simulations of the present study are additionally calculated for both OR1 and OR2. This helps to highlight the very similar accuracy of the present numerical model and the experimental tracking system in the description of the container trajectory giving further confidence on the employed numerical method.
In conclusion, the coupled models DualSPHysics and CHRONO can accurately simulate the behaviour of a debris transported by a dam break flow. Note that the modelling accuracy is referred to the mean trajectory among the ones resulting from repetitions of the experiment. This is because the numerical model was calibrated using the mean flow and container kinematic conditions, which was necessary given the highly variable characteristics of this phenomenon. Additionally, the calibration process of the model has shown that the details of the container internal structure and, above all, mass distribution is essential to achieve accuracy due to the importance of the container rotation in its kinematics. This rotation is, in fact, responsible for the container drift from the flume axis as in its absence the container would otherwise follow an almost one-dimensional trajectory. The difference in drift from the flume axis between simulations and experiments during the initial stage of the container motion, in turn affecting the velocity, suggests a lower accuracy in the forces if an impact would occur during this stage. However, as the trajectory stabilises the accuracy significantly improves especially for OR1. This has important consequences in modelling the impact of these debris on structures, which is the natural next step of this investigation. It is expected that impact forces are more accurately predicted if the impact occurs at later stages of the transport.