Topology optimisation engine: from scene definition to live 3D optimisation and STL export.

What is it?

Phoenix is a Python topology optimisation framework. You define a scene (a design domain with supports, loads, and keep-out regions), pick a physics solver, and watch the optimiser model material in a live 3-D viewer, leaving only the stiffest, lightest, or most efficient shape for the job.

Here is the pipeline:

scene definition  ->  voxel grid  ->  FE solve  ->  density update  ->  live render  ->  STL export

Features

• Scene builder: define domains with boxes, cylinders, unions, and subtractions in JSON or Python
• 8 physics solvers: structural, thermal, convection, vibration, gravity, stress-constrained, max-length, multi-physics (or implement your own)
• Live 3D viewer: watch density evolve in real time
• STL export: marching cubes and smoothing
• JSON settings: tune every solver from a single settings file
• Continuation: coarse -> fine grid refinement for high-res results

Quick start

# Clone and install
git clone https://github.com/dapsvi/phoenix.git
cd phoenix
uv sync

# Run your first optimisation
python cli.py optimize presets/json/bridge.json

The live viewer opens automatically. When it finishes you will find:

• results/bridge_latest.npz: full density field and history
• exports/bridge.stl: smoothed mesh ready for printing

Scene definition

Scenes are built from primitives (boxes, cylinders) with union, subtract, and intersect operations.

JSON format

{
  "name": "bridge",
  "nx": 50, "ny": 24, "nz": 10,
  "suggested_settings": "structural",
  "objects": [
    { "type": "box", "bounds": [0, 49, 0, 23, 0, 9], "kind": "solid" },
    { "type": "box", "bounds": [0, 0, 0, 0, 0, 9], "kind": "solid", "bc": "support", "constraint": "fix" },
    { "type": "box", "bounds": [0, 49, 23, 23, 3, 6], "kind": "solid", "bc": "load", "direction": [0, -1, 0] }
  ]
}

Python API

from scene import Scene, Box

scene = Scene("bridge", 150, 50, 10)
scene.add(Box(0, 149, 0, 49, 0, 9, kind="solid"))
scene.add(Box(0, 0, 0, 0, 0, 9, kind="fixed_solid", bc="support", constraint="fix"))
scene.add(Box(0, 149, 23, 23, 3, 6, kind="fixed_solid", bc="load", direction=(0, -1, 0)))

Included solvers

Example settings file (settings/structural.json):

{
  "type": "structural",
  "material": "aluminum",
  "volfrac": 0.05,
  "penal": 3.0,
  "rmin": 2.0,
  "move": 0.2,
  "max_iter": 100,
  "tol": 0.001
}

CLI

python cli.py optimize <preset>   Run optimisation with live viewer
python cli.py export <result>     Convert .npz result to STL
python cli.py view <result>       Open result in the 3D viewer

CLI optimize flags

Presets

Ready-to-run examples in presets/:

Output

After a run you get:

• .npz result: full 3-D density field, iteration history, and solver snapshots
• .stl mesh: smoothed isosurface, ready for 3-D printing or CFD meshing
• Verification report: FE re-solve confirming the design meets constraints

Dependencies

• numpy, scipy: core numerics
• pyamg: algebraic multigrid preconditioner
• pypardiso: sparse direct solver (Intel MKL Pardiso)
• pyvista: 3-D visualisation and mesh processing
• vpython: interactive viewer support
• Python >= 3.13

License

GNU GPLv3