This paper focuses on an innovative modeling approach for the damage growth mechanism for structures made of composite materials exposed to sloshing. This approach is based on finite particles, which provide similar properties as finite elements in classical structure mechanics. It is implemented into a computationally efficient numerical multi-body simulation program, which is flexible in geometry and dynamic load spectrum. It features a particle-based liquid model and provides time accurate hydrodynamic pressure results in three dimensions and 6 Degrees-of-Freedom. Based on the results of dam break tests and sloshing analysis of LNG tank geometries, unsteady impact loads are computed for a model tank made of carbon fiber reinforced plastic (CFRP). The initial damage conditions and damage evolution laws are discussed with respect to the intensity of sloshing. In contrast to the classical fluid-related physical modeling of sloshing based on energy dissipation in the liquid, the present approach emphasizes also the energy transfer into the structure. A special focus is set on the strength degradation and the interaction with the sloshing liquid due to the sloshing impact pressures.
Sloshing in partially filled tanks is a complex physical process. It covers phenomena like wave propagation and liquid-structure interaction; see (Yamamoto and Kataoka and Shioda and Ashitani, 1995). Applying continuum fluid mechanics for a simulation approach, Euler and Navier-Stokes solvers provide a time accurate simulation only if an appropriate free surface determination and propagation is implemented. In consequence, computation time and required memory resources increase with increasing code complexity and accuracy requirements Recent approaches consider statistical methods for the wall pressure determination see (Gervaise and de Sèze and Maillard, 2009). Briefly, there is still a big amount of uncertainty concerning the physical background of liquid sloshing effects. An important driver to improve numerical methods is the fluid-structure interaction (FSI) with the supporting structure. With increasing tank size, the structural loads become an important factor in case of partially filled tanks where the liquid is able to slosh. In this paper, an innovative approach is highlighted that is able to predict the hydrodynamic wall interaction resulting from liquid impact forces on a composite tank. Especially the structural aspects of sloshing, i.e. crack initiation and crack propagation, are not entirely understood for composite tanks consisting of multi-layer brittle materials such as Carbon-Fiber Reinforced Polymers (CFRPs). Hence a closed-loop simulation between the physics of liquid sloshing and structural response is required to predict structural integrity in presence of highly unsteady hydrodynamic loads. The present approach does not follow the classical continuum mechanics methods, but is Lagrangian in nature. A large number of microscopic particles (molecules, ions), both for liquid and tank structure, are supposed to be concentrated in particle-clusters of macroscopic dimensions. These clusters obey interaction forces according to Newton's fundamental law. The liquid particle-clusters are connected to each other by a Lennard-Jones potential originally in use in molecular dynamics. The potential parameters have been transformed into macroscopic domain using fundamental mechanical and chemical relations. Introducing Fourier's law of heat transfer and an energy dissipation scheme that obeys the conservation of momentum fulfills thermodynamic laws. Numerically, the cluster trajectories are computed using a Velocity-Verlet integration scheme. An innovative neighborhood list has been designed that permits a linear dependence of the computation time from the number of particle-clusters. A variable time-step scheme allows a drastic reduction of computation time. The particle-clusters representing the solid structure however obey the principle of the peridynamicss approach established by (Silling, 2000). Here, the interaction forces are governed by the force- elongation ratio and Poisson's ratio for a specific neighborhood of particle-clusters (Silling, 2003).
