sync benchmarking

benchmarking/neuralfoil_point_comparison.py

@@ -0,0 +1,225 @@
+import aerosandbox as asb
+import aerosandbox.numpy as np
+from aerosandbox.tools.string_formatting import eng_string
+from neuralfoil import get_aero_from_airfoil
+from pathlib import Path
+from matplotlib.colors import LinearSegmentedColormap
+from tqdm import tqdm
+
+af = asb.Airfoil(name="HALE_03", coordinates=Path(__file__).parent / "assets" / "hale_03mod.dat")
+alphas_xfoil = np.linspace(-5, 15, 50)
+alphas_nf = np.linspace(-6, 17, 300)
+Re_values_to_test = [1e4, 8e4, 2e5, 1e6, 1e8]
+
+# Obtain data
+aeros = {}
+
+aeros["xfoil"] = [
+    asb.XFoil(
+        airfoil=af,
+        Re=Re,
+        timeout=30,
+        xfoil_repanel=False,
+        max_iter=100,
+    ).alpha(alphas_xfoil)
+    for Re in tqdm(Re_values_to_test, desc="XFoil")
+]
+
+nf_model_sizes = [
+    "medium", "xxxlarge"
+]
+nf_linestyles = ["--", "-"]
+xfoil_linestyle = "k:"
+
+for model_size in nf_model_sizes:
+    aeros[model_size] = [
+        get_aero_from_airfoil(
+            airfoil=af,
+            alpha=alphas_nf,
+            Re=Re,
+            model_size=model_size
+        )
+        for Re in tqdm(Re_values_to_test, desc=f"NeuralFoil {model_size}")
+    ]
+
+import matplotlib.pyplot as plt
+import aerosandbox.tools.pretty_plots as p
+
+fig, ax = plt.subplots(figsize=(8,5))
+plt.xscale('log')
+
+cmap = LinearSegmentedColormap.from_list(
+    "custom_cmap",
+    colors=[
+        p.adjust_lightness(c, 0.8) for c in
+        ["orange", "darkseagreen", "dodgerblue"]
+    ]
+)
+colors = cmap(np.linspace(0, 1, len(Re_values_to_test)))
+transparency = 0.7
+
+for i, (Re, color) in enumerate(zip(Re_values_to_test, colors)):
+
+    nf_line2ds = []
+
+    for model_size, linestyle in zip(nf_model_sizes, nf_linestyles):
+        nf_aero = aeros[model_size][i]
+        nf_line2d, = plt.plot(
+            nf_aero["CD"],
+            nf_aero["CL"],
+            linestyle,
+            zorder=4,
+            color=color,
+            alpha=transparency,
+        )
+        nf_line2ds.append(nf_line2d)
+
+    xfoil_line2d, = plt.plot(
+        aeros["xfoil"][i]["CD"],
+        aeros["xfoil"][i]["CL"],
+        linestyle=":",
+        color=p.adjust_lightness(color, 0.4),
+        alpha=1
+    )
+
+
+    annotate_x = np.max(np.array([
+        aero[i]["CD"][-1]
+        for aero in aeros.values()
+    ]))
+    annotate_y = np.median(np.array([
+        aero[i]["CL"][-1]
+        for aero in aeros.values()
+    ]))
+
+    plt.annotate(
+        f" ${{\\rm Re}} = \\mathrm{{{eng_string(Re)}}}$",
+        xy=(annotate_x, annotate_y),
+        color=p.adjust_lightness(color, 0.8),
+        ha="left", va="center", fontsize=10
+    )
+
+plt.annotate(
+    text="Note the log-scale on $C_D$, which is unconventional - it's\nthe only way to keep it readable given the wide range.",
+    xy=(0.01, 0.01),
+    xycoords="figure fraction",
+    ha="left",
+    va="bottom",
+    fontsize=8,
+    alpha=0.5
+)
+
+from matplotlib.lines import Line2D
+
+legend_handles = [
+    Line2D([], [], color="k", linestyle=xfoil_line2d.get_linestyle(), label="XFoil (ground truth)"),
+    *[
+        Line2D([], [], color=color, linestyle=linestyle, label=f"NF \"{model_size}\"")
+        for model_size, linestyle, color in zip(nf_model_sizes, nf_linestyles, colors)
+    ]
+]
+
+for h in legend_handles:
+    h.set_color("k")
+
+plt.legend(
+    handles=legend_handles,
+    title="Analysis Method",
+    # loc=(0.8, 0.15),
+    loc='lower left',
+    # bbox_to_anchor=(0.5, 0., 0.5, 0.5),
+    fontsize=11,
+    labelspacing=0.3, columnspacing=1.5, handletextpad=0.4,
+    framealpha=0.8,
+)
+
+plt.xlim()
+plt.ylim(bottom=-0.8)
+
+afax = ax.inset_axes([0.76, 0.802, 0.23, 0.23])
+afax.fill(
+    af.x(), af.y(),
+    facecolor=(0, 0, 0, 0.2), linewidth=1, edgecolor=(0, 0, 0, 0.7)
+)
+afax.annotate(
+    text=f"{af.name} Airfoil",
+    xy=(0.5, 0.15),
+    xycoords="data",
+    ha="center",
+    va="bottom",
+    fontsize=10,
+    alpha=0.7
+)
+
+afax.grid(False)
+afax.set_xticks([])
+afax.set_yticks([])
+# afax.axis('off')
+afax.set_facecolor((1, 1, 1, 0.5))
+afax.set_xlim(-0.05, 1.05)
+afax.set_ylim(-0.05, 0.28)
+afax.set_aspect("equal", adjustable='box')
+
+
+plt.suptitle("Comparison of $C_L$-$C_D$ Polar for NeuralFoil vs. XFoil", fontsize=16,y=0.94)
+plt.title(f"On {af.name} Airfoil (out-of-sample)", fontsize=12, alpha=0.7)
+
+plt.xlabel("Drag Coefficient $C_D$")
+plt.ylabel("Lift Coefficient $C_L$")
+
+p.show_plot(
+    None,
+    legend=False,
+    savefig="neuralfoil_point_comparison.svg",
+    savefig_transparent=False,
+    rotate_axis_labels=False
+)
+
+# p.plot_color_by_value(
+#     nf_aero_xl["CD"],
+#     nf_aero_xl["CL"],
+#     c=nf_aero_xl["analysis_confidence"],
+#     clim=(0.8, 1),
+#     colorbar=Re == Re_values_to_test[-1],
+#     colorbar_label="Analysis Confidence" if Re == Re_values_to_test[-1] else None,
+#     cmap="turbo_r",
+#     # "--",
+#     # color=color,
+#     zorder=4,
+# )
+
+# plt.plot(
+#     nf_aero_xl["CD"],
+#     nf_aero_xl["CL"],
+#     "--",
+#     color=color, alpha=transparency
+# )
+
+# nf_aero_m = get_aero_from_airfoil(
+#     airfoil=af,
+#     alpha=alphas_nf,
+#     Re=Re,
+#     model_size="medium"
+# )
+#
+# plt.plot(
+#     nf_aero_m["CD"],
+#     nf_aero_m["CL"],
+#     ":",
+#     color=color, alpha=transparency
+# )
+
+# xfoil_aero = asb.XFoil(
+#     airfoil=af,
+#     Re=Re,
+#     timeout=30,
+#     max_iter=100,
+# ).alpha(alpha=alphas_xfoil, start_at=5)
+#
+# plt.plot(
+#     xfoil_aero["CD"],
+#     xfoil_aero["CL"],
+#     "--",
+#     color="k", alpha=0.4
+# )
+# xfoil_aero = nf_aero