Implicit SPH in LS-DYNA for Automotive Water Wading Simulations

The explicit SPH solver implemented in LS-DYNA is well fitted for numerical simulations involving hypervelocity impacts, explosions and other transient events, but is unsuitable for slower fluid-flow simulations such as water wading. In this work, we introduce an implicit SPH formulation specifically developed for handling large-scale incompressible fluid simulations. The method is based on a traditional projection scheme: Intermediate velocities are first predicted based on external and viscosity forces contribution, and a Poisson equation is then solved to obtain pressure forces such that incompressibility is maintained up to a given tolerance. All the surfaces composing the structure are automatically sampled with SPH particles by LS-DYNA using a user-supplied maximum interparticle distance, and the fluid-structure interaction is embedded in the SPH solver directly. This aspect of the simulation does not require any contact card to be setup. As the simulation evolves, the initial domain decomposition performed by LS-DYNA can become inefficient, triggering increasing communications across processors and poor load balancing, resulting in an increasing CPU time per simulation cycle as the SPH fluid particles intermix. A new feature has been developed based on the full-deck restart capability of LS-DYNA. The objective is to re-decompose the domain across processors at regular intervals, based on the updated geometry of the problem. This results in a more constant simulation time and overall improved performance.