scripts/plots/03_batch_clusters.py

#!/usr/bin/env python3
"""Batch-cluster overview — the canonical "all batches at a glance" figure,
replacing the old overlaid composite (which, at 76-80 curves, was unreadable
spaghetti; see git history / README).

Each specimen collapses to a single point at its stress-strain **peak**:
  x = strain at peak, y = ultimate strength (peak stress).
Peak is derived as max(stress) over the saved curve rather than read from the
persistent.h5 PeakStress field — TestWorks often fails to detect the peak for
D638 tensile and leaves that scalar null (see CLAUDE.md), but the full curve is
always saved, so the maximum is robust.

Each SLS batch (plus the FormLabs PA12GF benchtop reference) becomes a cluster
of its specimens' peak points, wrapped in a translucent convex-hull boundary in
the batch color with a legend entry. This carries the headline ultimate-strength
number and shows how tightly each batch clusters (strength-vs-ductility), with
none of the overlaid-curve clutter. To read the actual stress-strain curves for
one batch, see the per-batch detail figures (scripts/plots/04_batch_details.py,
assets/batches/{standard}_{batch}.png).

Excluded, same as the old composite: PLA/PETG filament controls, the FormLabs
Nylon 12 White control, and Batch M's Type IV specimens (different gauge
geometry) — all of which have their own dedicated figures.

Outputs assets/{standard}_batch_clusters.png (dpi=1200) + .pdf.
"""
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.lines import Line2D
from matplotlib.patches import Ellipse
from matplotlib.colors import to_rgba

from _lib import (BATCH_COLORS, FILAMENT_CONTROLS, FORMLABS_COLOR, NYLON_CONTROLS,
                  ORDERED_BATCHES, OUT_DIR, ROOT, TYPE_LINESTYLES, load_standard, save_figure,
                  style_axes)

# Cluster boundary size, in standard deviations along each principal axis. 1.0
# draws the ±1 SD spread of each batch's peak points (matching the ±1 SD
# convention in 02_batch_averages.py) — some points fall outside by design.
N_STD = 1.0

# Group order: SLS batches (print chronology) then the FormLabs reference last.
GROUP_ORDER = ORDERED_BATCHES + ["PA12GF_FL"]
GROUP_LABELS = {**{b: f"Batch {b}" for b in ORDERED_BATCHES}, "PA12GF_FL": "FormLabs PA12GF"}
GROUP_COLORS = {**BATCH_COLORS, "PA12GF_FL": FORMLABS_COLOR}


def group_key(row: dict) -> str | None:
    """Cluster membership: SLS batch label, or the FormLabs PA12GF reference.
    Everything with its own dedicated figure returns None (excluded)."""
    if row["material_class"] == "PA12GF_FL":
        return "PA12GF_FL"
    if row["material_class"] in FILAMENT_CONTROLS or row["material_class"] in NYLON_CONTROLS:
        return None
    if row["astm"].get("type") in TYPE_LINESTYLES:
        return None  # Type IV etc. — different geometry, own figure
    return row["batch_label"] or None


def peak_point(spec: dict) -> tuple[float, float]:
    """(strain_at_peak, ultimate_stress_mpa) from the specimen's saved curve."""
    stress = spec["stress_mpa"]
    peak_i = max(range(len(stress)), key=lambda i: stress[i])
    return spec["strain"][peak_i], stress[peak_i]


def confidence_ellipse(points: list[tuple[float, float]], ax, color, n_std: float = N_STD):
    """Draw a covariance-based n_std confidence ellipse for a cluster of points.
    Requires >=3 points for a non-degenerate covariance; the 1-2 point cases are
    handled by the caller (segment / bare marker)."""
    pts = np.asarray(points, dtype=float)
    cov = np.cov(pts, rowvar=False)
    vals, vecs = np.linalg.eigh(cov)  # ascending eigenvalues, orthonormal vecs
    order = vals.argsort()[::-1]
    vals, vecs = vals[order], vecs[:, order]
    vals = np.clip(vals, 0.0, None)   # guard tiny negative from round-off
    angle = np.degrees(np.arctan2(vecs[1, 0], vecs[0, 0]))
    width, height = 2 * n_std * np.sqrt(vals)  # full axis lengths
    ax.add_patch(Ellipse(
        xy=pts.mean(axis=0), width=width, height=height, angle=angle,
        facecolor=to_rgba(color, 0.15), edgecolor=to_rgba(color, 0.75),
        linewidth=1.4, zorder=2))


def render(standard: str, groups: dict[str, list[tuple[float, float]]],
           ylabel: str, title: str) -> None:
    fig, ax = plt.subplots(figsize=(9, 6))
    handles = []
    for key in GROUP_ORDER:
        pts = groups.get(key)
        if not pts:
            continue
        color = GROUP_COLORS[key]
        xs = [x for x, _ in pts]
        ys = [y for _, y in pts]

        # Cluster boundary: ±N_STD confidence ellipse (>=3 pts), a connecting
        # segment (2 pts), or nothing (1 pt — just the marker).
        if len(pts) >= 3:
            confidence_ellipse(pts, ax, color)
        elif len(pts) == 2:
            ax.plot(xs, ys, color=color, linewidth=1.4, alpha=0.7, zorder=3)

        ax.scatter(xs, ys, s=48, color=color, edgecolor="white", linewidth=0.6,
                   zorder=5, alpha=0.95)
        handles.append(Line2D([0], [0], marker="o", linestyle="none", color=color,
                              markeredgecolor="white", markeredgewidth=0.6, markersize=8,
                              label=f"{GROUP_LABELS[key]} (n={len(pts)})"))

    ax.set_xlabel("Strain at peak (mm/mm)")
    ax.set_ylabel(ylabel)
    ax.set_title(title)
    style_axes(ax)
    ax.legend(handles=handles, loc="upper left", bbox_to_anchor=(1.02, 1.0), borderaxespad=0)

    out_path = save_figure(fig, OUT_DIR / f"{standard}_batch_clusters")
    plt.close(fig)
    print(f"wrote {out_path.relative_to(ROOT)}  ({len(handles)} clusters)")


def plot_standard(standard: str) -> None:
    specs = load_standard(standard)
    groups: dict[str, list[tuple[float, float]]] = {}
    for s in specs:
        key = group_key(s["row"])
        if key is None:
            continue
        groups.setdefault(key, []).append(peak_point(s))

    ylabel = {"D638": "Ultimate tensile strength (MPa)",
              "D790": "Ultimate flexural strength (MPa)"}.get(standard, "Ultimate strength (MPa)")
    title = {"D638": "ASTM D638 — tensile batch clusters (peak point per specimen)",
             "D790": "ASTM D790 — flexural batch clusters (peak point per specimen)"}.get(
                 standard, standard)
    render(standard, groups, ylabel, title)


if __name__ == "__main__":
    plot_standard("D638")
    plot_standard("D790")