agentleFS
BaratiLabBaratiLab/mechanical-engineering-skills/skills/meltpool-geometry/SKILL.md

meltpool-geometry

Skill16 starsChanged 16 days agoMITView on GitHub

Predict metal additive manufacturing melt-pool depth, width and length and melt-pool mode (keyhole, lack of fusion, balling, desirable) from laser/electron-beam power, speed, beam size and alloy properties, with Rosenthal estimates, dimensionally consistent power-law fits, and honest cross-study validation. Based on MeltpoolNet (Barati Farimani Lab). Use for melt-pool size estimates, single-track planning, absorptivity and Rosenthal calculations, or adapting a melt-pool model to your own machine with a few measured tracks.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    All of it runs through scripts/meltpool.py (numpy, pandas, scipy, scikit-learn, matplotlib).

    line 15

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Melt-pool geometry (MeltpoolNet)

This skill estimates the melt pool for a metal AM parameter set with three tools that go from most interpretable to most data-driven:

  1. the Rosenthal point-source solution;
  2. a dimensionally consistent power law fitted to experiments;
  3. ML models trained on the MeltpoolNet benchmark (Akbari et al., Additive Manufacturing 55 (2022) 102817).

All of it runs through scripts/meltpool.py (numpy, pandas, scipy, scikit-learn, matplotlib). The data downloads on first use from github.com/BaratiLab/MeltpoolNet into ~/.cache/mechanical-engineering-skills/meltpoolnet, the same cache lpbf-defect-prediction uses.

Units: P in W, V in mm/s, beam/hatch/layer/melt-pool sizes in µm, ρ in kg/m³, Cp in J/kg·K, k in W/m·K, melting T in K.

Pick the tool

QuestionCommand
Quick physics estimate, any alloyrosenthal --material SS316L --power 200 --velocity 800 (or --props ρ,Cp,k,Tm --absorptivity η)
An explicit, explainable formula from dataidentify --target depth|width|length
Best-guess numbers for a known alloytrain once per target, then predict --models depth.joblib,width.joblib ...
How far can I trust it?benchmark --target depth --split paper [--calibrate 3] --plot parity.png
Melt-pool modeclassify --features baseline,onehot --split paper
Adapt to my machinetrain ... --extra my_tracks.csv (your measured single tracks, MeltpoolNet column names)

Features (--features, comma list):

  • baseline: P, V, ρ, Cp, k, Tm
  • energy: line energy and normalized absorptivity
  • abs1: paper Eq. 2; needs beam diameter and minimum absorptivity
  • abs2: paper Eq. 3. It uses the measured width, so it leaks the label; avoid it.
  • hatch, layer
  • onehot: material
  • composition: wt%
  • elemental: composition-weighted element properties

How to report a prediction

  1. Give the ML number and the Rosenthal number side by side, plus W/D. W/D < 1.5 suggests keyhole mode (see lpbf-defect-prediction).
  2. Quote the leave-paper-out validation stored in the model (validation_paper_split), not the random-split score.
    • Without any of the user's own data, a new machine is close to an unseen paper: depth R² ≈ 0 on this data.
    • Say so, and recommend measuring 3–10 single tracks. Retrain with --extra; in benchmark terms that brings R² to about 0.5–0.7.
  3. Flag inputs outside the training range (outside_training_range), unknown alloys (passed --props), and EBM vs. laser (the data is mostly SLM).

Facts to keep straight

  • The Rosenthal depth here uses Q = η_m·P (minimum absorptivity from the data), which is right for conduction mode. It badly under-predicts keyhole depths. The paper's Eq. 8 uses Q with A = 2.5 for AlSi10Mg and 2 otherwise.
  • The paper's random-split R² values (0.96–0.996) and its published equations do not reproduce on the current public CSV. See references/paper-notes.md for the numbers this skill measures. Report those.

Cite: Akbari et al., MeltpoolNet, Addit. Manuf. 55 (2022) 102817.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.