Haibarakiku/abaqus-ml-skills

# Abaqus LHS Batch Dataset Generator

End-to-end workflow for producing FEA training datasets from a parameterized Abaqus model. Designed for **surrogate model training** (Ridge, MLP, Gaussian Process, etc.) where you need hundreds to thousands of FEA samples covering a design space.

## When to Use This Skill

Activate when the user wants to:
- Build a training set for a surrogate / ML model from Abaqus FEA
- Sweep a multi-dimensional design vector (e.g. force amplitudes, geometry parameters, material properties)
- Run N independent Abaqus jobs in batch with progress tracking and crash recovery
- Generate inputs for inverse-design / topology-optimization pipelines

Do NOT use this skill for:
- Single-case Abaqus runs (use `abaqus-job` instead)

# Abaqus ODB → ML-Ready Grid CSV

The post-processing half of a surrogate-model pipeline. Takes a folder of completed Abaqus FEA cases (typically produced by the **abaqus-lhs-batch-dataset** skill) and produces two flat CSVs ready for `numpy.loadtxt` / `pandas.read_csv` / scikit-learn:

- **`X_amplitude.csv`** — `sample_id, amplitude_0000, amplitude_0001, ..., amplitude_(D-1)` — the input design vectors
- **`Y_grid_uz.csv`** — `sample_id, uz_0000, uz_0001, ..., uz_(N²-1)` — the resulting deformation field flattened from N×N grid (row-major)

These are the two matrices a surrogate model trains on: shape `(N_samples, D)` and `(N_samples, N²)`.

## When to Use This Skill

Activate when the user wants to:
- Convert a directory of `sample_*/Membrane2D1.odb` (or `node_displacement.csv`) cases into ML-ready training matrices
- Resample a fine FEA mesh (e.g. 81×81) onto a coarser learning grid (e.g. 41×41 or 21×21)

# Abaqus Surrogate ↔ FEA Validation Loop

Closes the verification loop on a surrogate-driven inverse design. Surrogates are fast but optimistic — they extrapolate, hallucinate, and reward saturated solutions that the real FEA cannot reproduce. This skill **forces a side-by-side comparison** between what the surrogate thinks it found and what Abaqus actually delivers, on the same target shape.

## When to Use This Skill

Activate when the user wants to:
- Sanity-check a surrogate that was just trained ("is the model trustworthy?")
- Quantify the **surrogate-FEA gap** on a fixed set of validation targets before publishing claims
- Compare optimizer choices (PGD vs. L-BFGS-B vs. multi-start) under equal-FEA-budget conditions
- Generate a reproducible benchmark table for a paper / report
- Debug why an optimizer's surrogate solution looks great but FEA verification fails

Do NOT use this skill for: