AMPathWeaver

The lightest solver in the stack computationally — and the most critical to get right, because every downstream solver reads its output. PathWeaver holds no physics: no PDEs, no thermal field, no material state. Pure computational geometry and combinatorics.

Seven sequential stages take geometry to trajectory. Stage 4 — zone assignment — activates only for multi-laser machines, where each laser owns a build-area zone and the output carries a laser_id per segment.

FIG.01 · pathweaver_internal_pipeline geometry in · time-stamped trajectory out · zero physics state

Six features that define what a trajectory file actually carries downstream.

Configurable scan strategies

Choose how each layer is filled — applied uniformly or per-region:

• Stripe · long parallel hatch bands
• Island · checkerboard tiles to break up thermal accumulation
• Chessboard · alternating rotated patches for stress homogenisation
• Configurable hatch-rotation increment per layer (e.g. 67° golden-angle)

Each region of a layer can carry its own fill strategy, power, and hatch spacing.

Multi-laser machine support

Same solver, single- or multi-laser machines:

• Optional zone-assignment stage partitions the build area per laser
• Every output segment is tagged with a laser_id
• Per-laser parameters (power, jump speed) can vary across the same trajectory
• Concurrent scanning is preserved end-to-end through the pipeline

laser_id = 1 laser_id = 2 laser_id = 3 laser_id = 4

Build-area zones distributed across all available lasers — concurrent scanning preserved in the output trajectory.

Per-region process parameters

Different physical regions of the part can carry independent process parameters:

• Volume infill — high-throughput hatching
• Contour pass — finer power and spacing for surface finish
• Up-skin / down-skin — tuned for surface quality
• Supports — fast, low-power scanning isolated from the part

Volume

P = 280 W
h = 0.10 mm

Contour

P = 180 W
h = 0.05 mm

Up-skin

P = 220 W
h = 0.08 mm

Supports

P = 120 W
h = 0.15 mm

Pre-slice validation

PathWeaver checks the oriented part against the machine envelope before slicing — oversized jobs fail early, not after an hour of compute:

• Build envelope & orientation-aware checks
• STL mesh repair and watertightness verification
• Face-tagging (up-skin / down-skin / wall / support) sanity checks
• Strategy compatibility — e.g. island size vs. minimum feature size

✓ STL watertight ........ 12,488 faces
✓ Build envelope ......... fits 250×250×300
✓ Face tagging ........... 4 surface classes
✓ Strategy compatible .... island ≥ 2.0 mm
⟶ ready to slice (247 layers)

Single trajectory output contract

One trajectory file carries everything downstream solvers need:

• Ordered segment list with t_start / t_end timestamps
• laser_id tag for multi-laser machines
• Per-segment power, scan speed, jump moves
• Region metadata (volume / contour / skin / support)
• Compact trajectory.json + an index file for random-access reads

{
  "layer": 247,
  "z_mm": 9.880,
  "segments": [
    { "id": 14201, "type": "contour",
      "p0": [10.2, 4.8], "p1": [10.4, 4.9],
      "P_W": 180, "v_mm_s": 950,
      "t_start": 12.4801, "t_end": 12.4824,
      "laser_id": 2 },
    ...
  ]
}

Lightweight computational stack

No FEM mesh, no voxeliser, no PDE solver. PathWeaver runs on a pure geometry stack:

• trimesh — STL/STEP ingest, repair, face tagging
• Shapely — 2D polygon decomposition, hatch generation, contour offsets
• scipy — ordering, nearest-neighbour jump-move optimisation

Runs in seconds for typical parts — even larger builds finish before the lab kettle boils.

trimesh Shapely scipy

Three input groups go in; one trajectory comes out. Because PathWeaver carries no material state, FusionCore and StressForge each consume the same contract without needing to know anything about the original geometry.

FIG.02 · pathweaver_io_contract geometry · machine · strategy → trajectory.json

PathWeaver opens the AM pipeline. Its trajectory drives the melt-pool solve, which feeds microstructure and residual-stress prediction, converging in part qualification.

No installation. No licensing servers. No workstation needed. Trajectories generate on SolidNetics cloud infrastructure — your local machine stays free for design work.

Three steps from upload to a machine-ready trajectory file.

Built for LPBF process engineers, AM machine builders, and research teams developing new scan strategies — from production-grade metal parts to thermal-stress sensitive geometries and qualification builds.