Material Point Method in Rust

MPM algorithm implemented in Rust, empowered by specs.

Use as a Library

use mpm_rs::*;
use mpm_ply_dump::*;

fn main() {

  // Create a world
  let mut world = WorldBuilder::new(Vector3f::new(1.0, 1.0, 1.0), 0.02)
    .with_system(PlyDumpSystem::new("result", 3)) // Dump to "result"
    .build(); // Build the world

  // Create a boundary to the world with a thickness of 0.06
  world.put_boundary(0.06);

  // Create a ball in the world
  let center = Vector3f::new(0.5, 0.4, 0.5);
  let radius = 0.1;
  let mass = 10.0;
  let num_particles = 10000;
  let youngs_modulus = 10000.0;
  let nu = 0.2;
  world
    .put_ball(center, radius, mass, num_particles)
    .with(ParticleDeformation::new(youngs_modulus, nu));

  // Run 500 steps
  for _ in 0..500 {
    world.step(); // Step once
  }
}

Compile and Run Examples

To compile and run examples, do

$ cargo build --release
$ cargo run --release --example mickey_mouse

Here we prefer release because it's so much faster than debug. mickey_mouse example will output .ply files into the directory result/mickey_mouse. You can visualize the result file using Houdini. You can check out more examples here.

As of an example simulation in a window visualized by kiss3d, you can run

$ cargo run --release --example multi_balls_viewer

Folder structure

A pure MPM simulation framework is implemented in core/.

A .msh loader (tetrahedron mesh loader) is implemented here in lib/msh-rs.

A .ply file exporter is implemented here in lib/ply-dump.

A viewer is implemented here in lib/viewer.

Other examples are located here: examples/examples.

Behind the Hood

mpm-rs is empowered by specs, an Entity-Component-System framework which suits best for doing all kinds of simulations. We structured the code base such that all the Lagrangian particles are individual entities in ECS, and the Eularian grid for doing integral and differentiation as a single resource Grid. Having not a single structure particle but "Arrays-of-particle-properties" empowers us to have various kinds of behaviors of different particles to perform flawlessly across the system. Some particles might have deformation gradient, some might have evolving $J_p$. The Specs ECS can automatically work on them simultaneously while providing lightning speed.