Simulating complex fluids with smoothed particle hydrodynamics

Abstract

Complex fluid dynamics encompasses a large variety of flows, such as fluids with non-Newtonian rheology, multiphase and multi-fluid flows (suspensions, lather, solid/fluid interaction with floating objects, etc.), violent flows breaking waves, dam-breaks, etc.), fluids with thermal dependencies and phase transition or free-surface flows. Correctly modeling the behavior of such flows can be quite challenging, and has led to significant advances in the field of Computational Fluid Dynamics (CFD). Recently, the Smoothed Particle Hydrodynamics (SPH) method has emerged as a powerful alternative to more classic CFD methods (such as finite volumes or finite elements) in many fields, including oceanography, volcanology, structural engineering, nuclear physics and medicine. With SPH, the fluid is discretized by means of particles and thanks to the meshless, Lagrangian nature of the model, it easily allows the modeling and simulation of both simple and complex fluids, simplifying the treatment of aspects that can be challenging with more traditional methods: dynamic free surfaces, large deformations, phase transition, fluid/solid interaction and complex geometries. In addition, the most common SPH formulations are fully parallelizable, which favors implementation on high-performance parallel computing hardware, such as modern Graphics Processing Units (GPUs). We present here how GPUSPH, an implementation of the SPH method that runs on GPUs, can model a variety of complex fluids, highlighting the computational challenges that arise in its applications to problem of great interest in volcanology.