code/visualize_t4_structural_dynamics_v1.py

"""Create the publication preview for the T4 v1 structural-dynamics dataset."""

from __future__ import annotations

import json
from pathlib import Path

import h5py
import matplotlib.pyplot as plt
import numpy as np
from matplotlib.tri import Triangulation

try:
    from .t4_structural_dynamics_v1 import CASE_DIR, DATA_DIR, configuration_design, load_config
except ImportError:
    from t4_structural_dynamics_v1 import CASE_DIR, DATA_DIR, configuration_design, load_config


def main() -> None:
    rows = configuration_design(load_config())
    colors = {
        "train": "#4C78A8",
        "validation": "#59A14F",
        "test_id": "#F28E2B",
        "excitation_ood": "#E15759",
        "parameter_ood": "#B07AA1",
    }
    fig, axes = plt.subplots(2, 2, figsize=(13.5, 8.2), constrained_layout=True)
    ax = axes[0, 0]
    for row in rows:
        role = row["protocol_role"]
        key = row["split"] if row["split"] != "test" else ("test_id" if role == "id" else role)
        p = row["parameters"]
        ax.scatter(p["length_m"], p["height_m"] * 1e3, s=28, c=colors[key], alpha=0.88)
    labels = (
        ("train", "Train · 96 configurations"),
        ("validation", "Validation · 16"),
        ("test_id", "Test ID · 8"),
        ("excitation_ood", "Test excitation OOD · 4"),
        ("parameter_ood", "Test parameter OOD · 4"),
    )
    for key, label in labels:
        ax.scatter([], [], s=34, c=colors[key], label=label)
    ax.set(xlabel="Beam length (m)", ylabel="Beam height (mm)", title="Leakage-safe physical configurations")
    ax.legend(frameon=False, fontsize=8, ncol=2)

    frequencies = []
    frequency_colors = []
    for row in rows:
        with h5py.File(CASE_DIR / f"{row['case_id']}.h5", "r") as h5:
            frequencies.append(float(h5["modal/frequencies_hz"][0]))
        role = row["protocol_role"]
        key = row["split"] if row["split"] != "test" else ("test_id" if role == "id" else role)
        frequency_colors.append(colors[key])
    ax = axes[0, 1]
    ax.scatter(range(128), frequencies, c=frequency_colors, s=24)
    ax.axvline(95.5, color="#777777", lw=1, ls="--")
    ax.axvline(111.5, color="#777777", lw=1, ls="--")
    ax.set(xlabel="Configuration ID", ylabel="First natural frequency (Hz)", title="Dynamic diversity and frozen split")

    case_path = CASE_DIR / "config_0119.h5"
    with h5py.File(case_path, "r") as h5:
        group = h5["trajectories/near_resonant_sine"]
        time = np.asarray(group["time_s"])
        sensors = np.asarray(group["sensor_displacement_m"])
        field_time = np.asarray(group["fields/time_s"])
        displacement = np.asarray(group["fields/displacement_m"])
        geometry = np.asarray(h5["common/reference_geometry_m"])
        topology = np.asarray(h5["common/topology"])

    ax = axes[1, 0]
    for index, fraction in enumerate((0.2, 0.4, 0.6, 0.8, 1.0)):
        ax.plot(time, sensors[:, index] * 1e3, lw=1.25, label=f"x/L={fraction:.1f}")
    ax.set(xlabel="Time (s)", ylabel="Vertical displacement (mm)", title="Five virtual sensors · near-resonant loading")
    ax.legend(frameon=False, fontsize=8, ncol=3)

    sampled_tip = np.interp(field_time, time, sensors[:, -1])
    frame = int(np.argmax(np.abs(sampled_tip)))
    deformed = geometry[:, :2] + 12.0 * displacement[frame, :, :2]
    values = displacement[frame, :, 1] * 1e3
    triangulation = Triangulation(deformed[:, 0], deformed[:, 1])
    ax = axes[1, 1]
    collection = ax.tripcolor(triangulation, values, shading="gouraud", cmap="coolwarm")
    length = float(np.max(geometry[:, 0]))
    height = float(np.max(geometry[:, 1]))
    ax.plot([0, length, length, 0, 0], [0, 0, height, height, 0], color="#555555", lw=0.7, ls="--")
    ax.set_aspect("equal")
    ax.set(xlabel="x (m)", ylabel="y (m)", title=f"Full-field response · t={field_time[frame]:.3f} s · deformation ×12")
    colorbar = fig.colorbar(collection, ax=ax, shrink=0.85)
    colorbar.set_label("Vertical displacement (mm)")

    fig.suptitle(
        "AgentFEM Structural Dynamics · 128 configurations · 512 complete trajectories",
        fontsize=17,
        fontweight="bold",
    )
    DATA_DIR.mkdir(parents=True, exist_ok=True)
    output = DATA_DIR / "preview.png"
    fig.savefig(output, dpi=180, bbox_inches="tight")
    plt.close(fig)
    print(json.dumps({"preview": str(output), "bytes": output.stat().st_size}))


if __name__ == "__main__":
    main()