Modeling and analysis of non-explosive underwater shock loading using a PD-SPH coupling method

The evaluation method of ship’s explosion shock resistance is challenged by some key mechanical problems, such as strong nonlinear fluid-structure coupling, large-deformation and failure evolution of solid structure. By coupling the respective advantages of peridynamics (PD) and smoothed particle hydrodynamics (SPH), an efficient PD-SPH numerical method suitable for underwater explosion shock simulations was developed. The SPH method was employed to simulate underwater shock wave propagation and fluid-structure interaction, while the PD method accurately characterized the complete mechanical behavior of solid structures from elastic deformation to progressive damage failure. A PD-SPH numerical model was established for non-explosive underwater shock loading devices. In the non-ordinary state-based peridynamics (NOSB-PD) framework, the Johnson-Cook damage model was introduced. To suppress the occurrence of numerical instability, the artificial stiffness form was introduced by increasing the internal constraints between particles. To improve the computational efficiency in large-scale simulations, a multi-GPU (graphics processing unit) parallel computing framework based on domain decomposition and data-communication mechanisms was established. The domain decomposition was carried out through the Eulerian format. When particles move from one domain to another, the physical quantities of the particles were exchanged for information. Model validation and parallel efficiency tests demonstrate that the proposed method can accurately predict shock wave wall pressure and target dynamic deformation, successfully reproduce typical crack propagation patterns in thin-plate structures and simulate the entire damage process of complex grid sandwich structure. In complex fluid-structure coupling scenarios with more than 5 million particles, the 8*RTX4090 achieved an acceleration ratio of 4.13 compared to a single RTX4090, with a parallel efficiency of 51.6%. The actual computation time can be reduced to nearly 1 hour. Meanwhile, compared with traditional CPU (central processing unit) parallelism, the multi-GPU parallelism can achieve an acceleration ratio of more than 9 times. The research outcomes provide a high-precision and efficient numerical analysis tool for the design of explosion-resistant naval structures, offering significant reference value for engineering applications of fluid-structure interaction in underwater explosion problems.