Built on RadialBasisFunctions.jl operators with native AD rules for Enzyme.jl and Mooncake.jl ā the numerical core of every simulation is differentiable.
That is the whole workflow. Getting Started walks through each step in detail, and Custom PDEs shows how to swap SolidEnergy for an equation of your own.
Steady-state temperature on a unit square, and displacement magnitude in a cantilever beam under end shear. See the Examples page for the code behind both.
Meshless methods operate on scattered point clouds with no connectivity requirements:
Simple geometry handling ā drop points on the boundary and fill the interior; no element quality concerns
Easy refinement ā add more points where you need accuracy; no remeshing required
Natural for moving boundaries ā points move freely without topological constraints
Macchiato.jl uses radial basis function (RBF) collocation, where differential operators are approximated at each point using its local neighborhood of nearest neighbors.
API Reference ā full type and function documentation
Hero: steady heat conduction solved on a point-cloud moka pot ā geometry from 3D Printable Moka Pot by CNC Kitchen (CC BY), rendered by examples/moka_pot.jl.