Warp Simulator

A browser-based thermal warp-likelihood simulator for sliced 3MF projects. Drop a .3mf (or .stl) on the page and it estimates how likely each filament is to warp or lift off the plate for that specific part, paints a corner-lift risk heatmap on the model, and suggests mitigations.

Everything runs client-side — no files leave the browser.

Built and maintained by TalonFab in Sydney, Australia — we design field-ready wildlife recording gear, including parabolic microphones and camera supports, and use this tool to keep our 3D-printed parts flat on the plate.

• Sliced .3mf support — reads the root model, follows production-extension (p:path) object references, and applies build-plate transforms. Plain 3MF and STL (binary + ASCII) also work.
• Multiple build plates — plate layout is read from Metadata/model_settings.config, a plate picker switches between plates, and every object on a plate is simulated separately (each part shrinks toward its own centroid — two small parts far apart are not one huge part). The plate's score is its worst object.
• Your sliced settings as defaults — Metadata/project_settings.config supplies the filaments (pre-selected in the comparison), the selected plate type and its bed temperatures, chamber temperatures, and brim configuration; the printer class (enclosed vs actively-heated-chamber vs open frame) fills in a realistic chamber temperature when the profile doesn't control one. Everything remains overridable in the UI.
• Filament library — PLA, PLA-CF, PETG, PETG-CF, ABS, ASA, ASA-CF (10 % and 20 % fibre grades), PC, PA6-CF, PAHT-CF, TPU, each with lock-in temperature, shrink strain, road stiffness, bed adhesion, and chamber requirements.
• Comparison table — warp score (0–100) and risk bucket per filament under the same conditions, plus per-filament recommendations (brim, enclosure, bed temperature, adhesives, part reorientation).
• 3D heatmap — per-vertex corner-lift risk painted on the part with a sequential colour scale; click a table row to switch filament.
• Per-filament condition overrides — bed temperature, chamber temperature, and brim can be set individually for every filament in the comparison (they're rarely the same); placeholders show each filament's resolved default, clearing an input restores it, and the header brim checkbox sets the whole column. The build-plate surface (textured PEI, smooth PEI, cool, high-tack) is selectable too and scales first-layer grip per polymer family — a cool plate holding ABS at half strength is flagged in the advice.

This is a fast physics-informed heuristic, not FEA. The full write-up — model, calibration evidence, and how to interpret the risk buckets — is in docs/how-it-works.md. In brief, it captures the standard qualitative mechanics of FFF warping:

1. Voxelisation. The part is voxelised (column-parity fill) into up to ~96×96×160 cells to get per-layer cross-sections, the footprint, and each bottom cell's distance from the footprint centroid.
2. Thermal field. While printing, material at height z equilibrates to T_env(z) = chamber + (bed − chamber)·e^(−z/7mm) — the bed dominates near the plate and fades with height.
3. Stress lock-in with a relaxation band. A layer only accumulates differential shrink strain once it cools a whole relaxation band (~20 °C) below the material's lock-in temperature (≈ Tg for amorphous polymers like ABS/ASA/PC, ≈ crystallisation temperature for nylons like PAHT-CF) — material held just under lock-in still creeps its stress away over the minutes-long print. Layers the bed keeps warm shrink together with the part during final cooldown instead, which warps far less. This is why PLA on a 60 °C bed and PETG on a 70 °C bed print flat, why hot beds and chambers work, and why high-Tg ABS/PC and nylons — which sit far below lock-in even on a 100 °C bed — are the materials that warp.
4. Corner lift with a saturating lever. Peel stress at a bottom cell scales with road stiffness × locked-in strain × (distance from centroid)^1.5, where the effective distance saturates at ~90 mm — past a bending wavelength the base flexes instead of prying, so a 250 mm sheet is not radically worse than a 120 mm one. Peel saturates with part height and is resisted by the filament's first-layer adhesion (plus a brim bonus). The score is the 95th-percentile peel-to-adhesion ratio squashed to 0–100. Each object on the plate gets its own field; scores are calibrated against a table of known-outcome anchor prints (tests/anchors.test.ts).

Carbon-filled grades (ASA-CF, PAHT-CF, PETG-CF, PLA-CF) score lower than their base polymers because fibre fill cuts shrink strain by ~3×, which outweighs their higher stiffness. TPU scores low despite high shrink because a soft bead stretches instead of peeling the corner up.

Interpretation: scores are calibrated for relative comparison between filaments and geometries; treat the buckets (low / moderate / high / severe) as guidance, not a guarantee.

cd apps/warp-sim
npm install
npm run dev        # local dev server
npm test           # vitest unit tests (parser, voxeliser, simulation ranking)
npm run build      # type-check + production build to dist/

The app is a standalone Vite + React + TypeScript project with its own dependencies; it is intentionally not part of the root npm workspace.

MIT © TalonFab Pty Ltd. Use it, fork it, ship it — a link back is appreciated.

TalonFab makes field-ready tools for capturing wildlife: parabolic microphone systems for wildlife sound recording, camera supports (ground pods, tripod accessories, hide mounts), and nocturnal lighting mounts — designed, 3D-printed, and tested in Sydney, Australia. This simulator started life as our in-house tool for predicting which printed parts would warp before we burned a spool finding out.