Chrono::GPU, Modeling and Simulation of Granular Dynamics using GPU Computing

Modeling granular system of large degree of freedom poses high computation cost. Chrono::GPU is designed to simulate granular material via the framework of Discrete Element Method (DEM) to produce realistic results. The physics-based features include Hertzian contact, Coulomb friction with stick-slip condition, a history-based model for tangential friction force, rolling friction and cohesion. Both first and second order time integrators are available. A wide variety of benchmark tests, such as rotating drum tests, cratering tests, and direct shear tests, have been carried out to ensure the physical soundness of the solver.

General triangle mesh in OBJ format can be interfaced with the granular media to perform co-simulation with Chrono, supporting simulations that involve the interplay between granular material and complex mechanical systems, such as wheeled and tracked vehicles.

Chrono::GPU is built with efficiency in mind. The framework utilizes mixed data types, domain decomposition , shared memory and CUB library. Performance scales linearly with problem up to 130 million frictional particles.

The ongoing work focuses on multi-GPU support for larger problem size and poly-dispersed granular material.

Contributors: Luning Fang, Ruochun Zhang, Radu Serban, Dan Negrut