Initial commit

A题/AAA常用/最终内容/fig_generation_p3.py

@@ -0,0 +1,697 @@
+# -*- coding: utf-8 -*-
+"""
+Problem 3: Sensitivity Analysis and Assumption Testing Figures
+All data sourced from 整合输出.md (frozen model specification)
+"""
+
+import matplotlib.pyplot as plt
+import numpy as np
+import seaborn as sns
+from math import pi
+import os
+
+# ==========================================
+# Style Configuration
+# ==========================================
+plt.rcParams['font.family'] = 'Arial'
+plt.rcParams['font.size'] = 11
+plt.rcParams['axes.labelsize'] = 12
+plt.rcParams['axes.titlesize'] = 13
+plt.rcParams['figure.dpi'] = 150
+
+colors = ['#2ecc71', '#3498db', '#9b59b6', '#e74c3c', '#f39c12', '#1abc9c', '#34495e']
+
+os.makedirs('figures', exist_ok=True)
+
+def save_fig(name):
+    plt.tight_layout()
+    plt.savefig(f'figures/{name}.png', dpi=300, bbox_inches='tight')
+    plt.close()
+    print(f"Saved {name}.png")
+
+
+# ==========================================
+# Data from 整合输出.md (FROZEN)
+# ==========================================
+
+# SOBOL_TABLE_v1
+SOBOL_TABLE = {
+    'k_L': {'S_i': 0.412, 'ST_i': 0.445},
+    'k_C': {'S_i': 0.285, 'ST_i': 0.312},
+    'kappa': {'S_i': 0.164, 'ST_i': 0.198},
+    'k_N': {'S_i': 0.042, 'ST_i': 0.065},
+    'R_ref': {'S_i': 0.021, 'ST_i': 0.048},
+    'alpha_Q': {'S_i': 0.011, 'ST_i': 0.032},
+}
+
+# COMPUTE_LOG_v1
+SOBOL_COMPUTE = {
+    'N_base': 512,
+    'D': 6,
+    'N_evals_total': 4096,
+    'seed': 20260201,
+    'sampling_scheme': 'Saltelli',
+}
+
+# SCENARIO_TTE_TABLE_v1
+SCENARIO_TABLE = {
+    'S0': {'desc': 'Baseline', 'TTE': 4.60, 'delta': 0.00, 'reason': 'SOC_ZERO'},
+    'S1': {'desc': 'Brightness Reduced (0.5x)', 'TTE': 5.82, 'delta': 1.22, 'reason': 'SOC_ZERO'},
+    'S2': {'desc': 'CPU Reduced (0.5x)', 'TTE': 5.45, 'delta': 0.85, 'reason': 'SOC_ZERO'},
+    'S3': {'desc': 'Network Reduced (0.5x)', 'TTE': 4.92, 'delta': 0.32, 'reason': 'SOC_ZERO'},
+    'S4': {'desc': 'Poor Signal (Psi=0.2)', 'TTE': 2.78, 'delta': -1.82, 'reason': 'SOC_ZERO'},
+    'S5': {'desc': 'Cold Ambient (0C)', 'TTE': 3.15, 'delta': -1.45, 'reason': 'V_CUTOFF'},
+    'S6': {'desc': 'Hot Ambient (40C)', 'TTE': 4.98, 'delta': 0.38, 'reason': 'SOC_ZERO'},
+    'S7': {'desc': 'Background Cut (0.5x)', 'TTE': 4.74, 'delta': 0.14, 'reason': 'SOC_ZERO'},
+}
+
+# MECH_SIGNATURES_v1
+MECH_SIGNATURES = {
+    'S0': {'avg_P': 3.22, 'max_I': 1.54, 'min_Delta': 8.15, 'avg_R0': 0.108, 'avg_Qeff': 4.00},
+    'S4': {'avg_P': 5.32, 'max_I': 2.45, 'min_Delta': 3.82, 'avg_R0': 0.112, 'avg_Qeff': 4.00},
+    'S5': {'avg_P': 3.28, 'max_I': 1.92, 'min_Delta': 0.85, 'avg_R0': 0.235, 'avg_Qeff': 3.52},
+}
+
+# UQ_SUMMARY_v1
+UQ_SUMMARY = {
+    'mean': 4.6021,
+    'std': 0.0542,
+    'p10': 4.5314,
+    'p50': 4.6018,
+    'p90': 4.6725,
+    'CI95_low': 4.5959,
+    'CI95_high': 4.6083,
+}
+
+# REPRODUCIBILITY_v1
+REPRODUCIBILITY = {
+    'seed': 20260201,
+    'M': 300,
+    'theta': 1/600,
+    'sigma': 0.02,
+    'dt': 1.0,
+}
+
+# Assumption Robustness Data (from document analysis)
+ASSUMPTION_ROBUSTNESS = {
+    'Baseline': {'TTE': 4.60, 'delta': 0.0, 'delta_pct': 0.0},
+    'OCV Linear→Poly': {'TTE': 4.68, 'delta': 0.08, 'delta_pct': 1.7},
+    'CPL→CC': {'TTE': 5.12, 'delta': 0.52, 'delta_pct': 11.3},
+    'Lumped→FEM': {'TTE': 4.48, 'delta': -0.12, 'delta_pct': -2.6},
+    'Signal Exp→Lin': {'TTE': 5.49, 'delta': 0.89, 'delta_pct': 19.3},
+    'OU θ Change': {'TTE': 4.63, 'delta': 0.03, 'delta_pct': 0.7},
+}
+
+# Physical Decoupling Data
+DECOUPLING_DATA = {
+    'Full Model (All Coupling)': {'TTE': 4.60, 'delta_pct': 0.0},
+    'No Thermal (dT/dt=0)': {'TTE': 4.85, 'delta_pct': 5.4},
+    'No CPL (I=const)': {'TTE': 5.12, 'delta_pct': 11.3},
+    'No Signal (Psi=0.9)': {'TTE': 6.42, 'delta_pct': 39.6},
+    'No R(T) (R0=const)': {'TTE': 4.73, 'delta_pct': 2.8},
+    'No Coupling (Linear)': {'TTE': 7.21, 'delta_pct': 56.7},
+}
+
+# Extreme Scenario Data
+EXTREME_SCENARIOS = {
+    'E0: Baseline': {'TTE': 4.60, 'delta_pct': 0, 'confidence': 5},
+    'E1: Arctic (-10°C)': {'TTE': 1.68, 'delta_pct': -63.5, 'confidence': 3},
+    'E2: Basement (Weak Signal)': {'TTE': 1.85, 'delta_pct': -59.8, 'confidence': 4},
+    'E3: Desert (50°C)': {'TTE': 3.78, 'delta_pct': -17.8, 'confidence': 3},
+    'E4: Perfect Storm': {'TTE': 0.92, 'delta_pct': -80.0, 'confidence': 2},
+    'E5: Aged Battery (SOH=70%)': {'TTE': 3.22, 'delta_pct': -30.0, 'confidence': 4},
+    'E6: Gaming Max': {'TTE': 1.54, 'delta_pct': -66.5, 'confidence': 3},
+}
+
+# Usage Fluctuation Data (from OU process analysis)
+FLUCTUATION_DATA = {
+    'Low (σ=0.01)': {'mean': 4.60, 'std': 0.027, 'p10': 4.57, 'p90': 4.63, 'cv': 0.59},
+    'Baseline (σ=0.02)': {'mean': 4.60, 'std': 0.054, 'p10': 4.53, 'p90': 4.67, 'cv': 1.18},
+    'High (σ=0.04)': {'mean': 4.60, 'std': 0.108, 'p10': 4.46, 'p90': 4.74, 'cv': 2.35},
+    'Extreme (σ=0.08)': {'mean': 4.60, 'std': 0.215, 'p10': 4.32, 'p90': 4.88, 'cv': 4.67},
+}
+
+# Load Model Comparison (CPL vs CC vs CR)
+LOAD_MODEL_COMPARISON = {
+    'CPL (Ours)': {'TTE': 4.60, 'end_current': 1.01, 'current_increase': 46},
+    'CC (Constant Current)': {'TTE': 5.12, 'end_current': 0.69, 'current_increase': 0},
+    'CR (Constant Resistance)': {'TTE': 5.38, 'end_current': 0.59, 'current_increase': -14},
+}
+
+# Signal Mapping Validation Data
+SIGNAL_MAPPING = {
+    0.9: {'measured': 0.78, 'exp_model': 0.80, 'lin_model': 0.82},
+    0.5: {'measured': 2.15, 'exp_model': 2.18, 'lin_model': 1.52},
+    0.2: {'measured': 5.32, 'exp_model': 5.28, 'lin_model': 2.22},
+    0.1: {'measured': 8.15, 'exp_model': 8.21, 'lin_model': 2.92},
+}
+
+# Amplification Factor by SOC
+AMPLIFICATION_BY_SOC = {
+    '1.0-0.7 (Early)': 1.8,
+    '0.7-0.4 (Mid)': 2.7,
+    '0.4-0.2 (Late)': 3.5,
+    '0.2-0 (Critical)': 4.2,
+}
+
+
+# ==========================================
+# Fig 14: Sobol Sensitivity Indices
+# ==========================================
+def plot_fig14():
+    """
+    Sobol sensitivity indices from SOBOL_TABLE_v1
+    Shows first-order (S_i) and total-order (ST_i) indices
+    """
+    params = list(SOBOL_TABLE.keys())
+    param_labels = [r'$k_L$ (Screen)', r'$k_C$ (CPU)', r'$\kappa$ (Signal)', 
+                    r'$k_N$ (Network)', r'$R_{ref}$ (Resistance)', r'$\alpha_Q$ (Capacity)']
+    S1 = [SOBOL_TABLE[p]['S_i'] for p in params]
+    ST = [SOBOL_TABLE[p]['ST_i'] for p in params]
+    interaction = [st - s for st, s in zip(ST, S1)]
+    
+    fig, ax = plt.subplots(figsize=(11, 6))
+    
+    x = np.arange(len(params))
+    width = 0.35
+    
+    bars1 = ax.bar(x - width/2, S1, width, label=r'First-order $S_i$', color=colors[1], alpha=0.8, edgecolor='black')
+    bars2 = ax.bar(x + width/2, ST, width, label=r'Total-order $ST_i$', color=colors[4], alpha=0.8, edgecolor='black')
+    
+    # Add interaction values as text
+    for i, (xi, sti, inter) in enumerate(zip(x, ST, interaction)):
+        ax.annotate(f'Int={inter:.3f}', xy=(xi + width/2, sti + 0.01), 
+                    ha='center', fontsize=9, color='gray')
+    
+    ax.set_ylabel('Sobol Index')
+    ax.set_title(f'Global Sensitivity Analysis: Sobol Indices (N={SOBOL_COMPUTE["N_evals_total"]})')
+    ax.set_xticks(x)
+    ax.set_xticklabels(param_labels, rotation=15, ha='right')
+    ax.legend(loc='upper right')
+    ax.set_ylim(0, 0.55)
+    
+    # Add cumulative line on secondary axis
+    ax2 = ax.twinx()
+    cumsum = np.cumsum(ST) / np.sum(ST) * 100
+    ax2.plot(x, cumsum, 'o--', color='red', alpha=0.7, linewidth=2, markersize=8, label='Cumulative %')
+    ax2.set_ylabel('Cumulative Contribution (%)', color='red')
+    ax2.tick_params(axis='y', labelcolor='red')
+    ax2.set_ylim(0, 110)
+    
+    # Mark 75% threshold
+    ax2.axhline(75, color='red', linestyle=':', alpha=0.5)
+    ax2.text(2.5, 77, '75% (Top 3 params)', fontsize=9, color='red')
+    
+    # Stats box
+    stats_text = (f"SOBOL_TABLE_v1:\n"
+                  f"Top contributor: k_L (ST={SOBOL_TABLE['k_L']['ST_i']:.3f})\n"
+                  f"Top 2 total: {SOBOL_TABLE['k_L']['ST_i'] + SOBOL_TABLE['k_C']['ST_i']:.1%}\n"
+                  f"Max interaction: kappa ({max(interaction):.3f})")
+    ax.text(0.02, 0.98, stats_text, transform=ax.transAxes, fontsize=9,
+            verticalalignment='top', bbox=dict(boxstyle='round', facecolor='wheat', alpha=0.9))
+    
+    save_fig('fig14_sobol_indices')
+
+
+# ==========================================
+# Fig 15: Assumption Robustness Waterfall
+# ==========================================
+def plot_fig15():
+    """
+    Bar chart showing impact of changing modeling assumptions on TTE.
+    MCM O-Award style: clean, professional, data-driven visualization.
+    Data from ASSUMPTION_ROBUSTNESS
+    """
+    # Extract data (skip baseline for comparison)
+    assumptions_full = list(ASSUMPTION_ROBUSTNESS.keys())
+    assumptions = assumptions_full[1:]  # Exclude 'Baseline'
+    tte_vals = [ASSUMPTION_ROBUSTNESS[a]['TTE'] for a in assumptions]
+    delta_pcts = [ASSUMPTION_ROBUSTNESS[a]['delta_pct'] for a in assumptions]
+    
+    baseline = 4.60
+    
+    fig, ax = plt.subplots(figsize=(10, 6))
+    
+    x = np.arange(len(assumptions))
+    
+    # Vibrant color coding: bright green for robust, orange for moderate, bright red for critical
+    bar_colors = []
+    for dp in delta_pcts:
+        if abs(dp) > 10:
+            bar_colors.append('#e74c3c')  # Red - Critical
+        elif abs(dp) > 5:
+            bar_colors.append('#f39c12')  # Orange - Moderate
+        else:
+            bar_colors.append('#27ae60')  # Green - Robust
+    
+    # Create bars
+    bars = ax.bar(x, tte_vals, width=0.6, color=bar_colors, alpha=0.9, 
+                  edgecolor='#2C3E50', linewidth=1.5)
+    
+    # Baseline reference line
+    ax.axhline(baseline, color='#2c3e50', linestyle='--', linewidth=2.5, 
+               label=f'Baseline CPL Model ({baseline:.2f}h)', zorder=1)
+    
+    # Add value labels ABOVE bars (not overlapping)
+    for i, (bar, tte, dp) in enumerate(zip(bars, tte_vals, delta_pcts)):
+        height = bar.get_height()
+        # Position label above the bar
+        y_pos = height + 0.12
+        sign = '+' if dp > 0 else ''
+        label = f'{tte:.2f}h\n({sign}{dp:.1f}%)'
+        ax.text(bar.get_x() + bar.get_width()/2, y_pos, label,
+                ha='center', va='bottom', fontsize=10, fontweight='bold')
+    
+    # Styling
+    ax.set_xticks(x)
+    ax.set_xticklabels(assumptions, rotation=20, ha='right', fontsize=10)
+    ax.set_ylabel('Time to Exhaustion (hours)', fontsize=11, fontweight='bold')
+    ax.set_xlabel('Modeling Assumption Change', fontsize=11, fontweight='bold')
+    ax.set_title('Assumption Robustness Analysis: Impact on TTE Prediction', 
+                 fontsize=13, fontweight='bold', pad=15)
+    ax.set_ylim(0, 6.8)
+    
+    # Add legend with color explanation - positioned at upper right, outside plot area
+    from matplotlib.patches import Patch
+    legend_elements = [
+        plt.Line2D([0], [0], color='#2c3e50', linestyle='--', linewidth=2.5, label=f'Baseline ({baseline:.2f}h)'),
+        Patch(facecolor='#27ae60', edgecolor='#2C3E50', label='Robust (|Δ| < 5%)'),
+        Patch(facecolor='#f39c12', edgecolor='#2C3E50', label='Moderate (5-10%)'),
+        Patch(facecolor='#e74c3c', edgecolor='#2C3E50', label='Critical (|Δ| > 10%)')
+    ]
+    ax.legend(handles=legend_elements, loc='upper right', fontsize=9, 
+              framealpha=0.95, edgecolor='gray', bbox_to_anchor=(0.99, 0.99))
+    
+    # Grid for readability
+    ax.yaxis.grid(True, linestyle=':', alpha=0.6)
+    ax.set_axisbelow(True)
+    
+    plt.tight_layout()
+    save_fig('fig15_assumption_robustness')
+
+
+# ==========================================
+# Fig 16: Physical Coupling Decoupling Analysis
+# ==========================================
+def plot_fig16():
+    """
+    Shows impact of disabling each physical coupling mechanism.
+    Data from DECOUPLING_DATA
+    """
+    experiments = list(DECOUPLING_DATA.keys())
+    tte_vals = [DECOUPLING_DATA[e]['TTE'] for e in experiments]
+    delta_pcts = [DECOUPLING_DATA[e]['delta_pct'] for e in experiments]
+    
+    fig, ax = plt.subplots(figsize=(10, 6))
+    
+    # Color by magnitude
+    bar_colors = [colors[0] if d == 0 else (colors[3] if d > 10 else colors[1]) for d in delta_pcts]
+    
+    bars = ax.barh(experiments, tte_vals, color=bar_colors, alpha=0.8, edgecolor='black', height=0.6)
+    
+    ax.axvline(4.60, color='red', linestyle='--', linewidth=2, label='Full Model (4.60h)')
+    
+    # Add percentage labels
+    for bar, pct in zip(bars, delta_pcts):
+        width = bar.get_width()
+        label = f'+{pct:.1f}%' if pct > 0 else ('Baseline' if pct == 0 else f'{pct:.1f}%')
+        ax.text(width + 0.15, bar.get_y() + bar.get_height()/2, label,
+                va='center', fontsize=10, fontweight='bold')
+    
+    ax.set_xlabel('TTE (hours)')
+    ax.set_title('Physical Coupling Decoupling Analysis: TTE Overestimation When Ignoring Feedback')
+    ax.set_xlim(0, 8)
+    ax.legend(loc='lower right')
+    ax.grid(axis='x', alpha=0.3)
+    
+    # Key insight box
+    insight_text = ("Key Insight:\n"
+                    "Signal-Power coupling: +39.6%\n"
+                    "CPL feedback: +11.3%\n"
+                    "All coupling off: +56.7%")
+    ax.text(0.98, 0.02, insight_text, transform=ax.transAxes, fontsize=9,
+            ha='right', va='bottom', bbox=dict(boxstyle='round', facecolor='lightyellow', alpha=0.9))
+    
+    save_fig('fig16_decoupling')
+
+
+# ==========================================
+# Fig 17: Extreme Scenario Stress Testing
+# ==========================================
+def plot_fig17():
+    """
+    Bar chart showing TTE under extreme conditions.
+    Data from EXTREME_SCENARIOS
+    """
+    scenarios = list(EXTREME_SCENARIOS.keys())
+    tte_vals = [EXTREME_SCENARIOS[s]['TTE'] for s in scenarios]
+    delta_pcts = [EXTREME_SCENARIOS[s]['delta_pct'] for s in scenarios]
+    confidences = [EXTREME_SCENARIOS[s]['confidence'] for s in scenarios]
+    
+    fig, ax = plt.subplots(figsize=(11, 6))
+    
+    # Color by severity (green to red)
+    norm_delta = [(d - min(delta_pcts)) / (max(delta_pcts) - min(delta_pcts) + 1e-6) for d in delta_pcts]
+    bar_colors = plt.cm.RdYlGn_r(norm_delta)
+    
+    bars = ax.bar(scenarios, tte_vals, color=bar_colors, edgecolor='black', alpha=0.85)
+    
+    ax.axhline(4.60, color='gray', linestyle='--', alpha=0.7, linewidth=2, label='Baseline (4.60h)')
+    
+    # Add labels with confidence stars
+    for bar, d, c in zip(bars, delta_pcts, confidences):
+        height = bar.get_height()
+        stars = '*' * c
+        label = f'{d:+.0f}%\n[{stars}]' if d != 0 else f'Baseline\n[{stars}]'
+        ax.text(bar.get_x() + bar.get_width()/2, height + 0.15, label,
+                ha='center', va='bottom', fontsize=9)
+    
+    ax.set_ylabel('TTE (hours)')
+    ax.set_title('Extreme Scenario Stress Testing (Confidence: * to *****)')
+    ax.set_ylim(0, 6)
+    plt.xticks(rotation=30, ha='right')
+    ax.legend(loc='upper right')
+    
+    # Highlight perfect storm
+    ax.annotate('Perfect Storm\n(DANGER)', xy=(4, 0.92), xytext=(5, 2),
+                arrowprops=dict(arrowstyle='->', color='red', lw=2),
+                fontsize=10, color='red', fontweight='bold')
+    
+    save_fig('fig17_extreme_scenarios')
+
+
+# ==========================================
+# Fig 18: Usage Pattern Fluctuation Impact
+# ==========================================
+def plot_fig18():
+    """
+    Shows how usage pattern volatility affects TTE uncertainty.
+    Data from FLUCTUATION_DATA
+    """
+    labels = list(FLUCTUATION_DATA.keys())
+    means = [FLUCTUATION_DATA[l]['mean'] for l in labels]
+    stds = [FLUCTUATION_DATA[l]['std'] for l in labels]
+    p10s = [FLUCTUATION_DATA[l]['p10'] for l in labels]
+    p90s = [FLUCTUATION_DATA[l]['p90'] for l in labels]
+    cvs = [FLUCTUATION_DATA[l]['cv'] for l in labels]
+    
+    fig, ax = plt.subplots(figsize=(9, 6))
+    
+    x = np.arange(len(labels))
+    bar_colors = [colors[0], colors[1], colors[4], colors[3]]
+    
+    # Plot 90% confidence intervals
+    for i, (m, low, high, c) in enumerate(zip(means, p10s, p90s, bar_colors)):
+        ax.plot([i, i], [low, high], color='black', linewidth=3, alpha=0.4)
+        ax.plot(i, m, 'o', markersize=14, color=c, markeredgecolor='black', markeredgewidth=1.5)
+        ax.plot([i-0.15, i+0.15], [low, low], 'k-', linewidth=2)
+        ax.plot([i-0.15, i+0.15], [high, high], 'k-', linewidth=2)
+    
+    # Add width and CV annotations
+    for i, (low, high, cv) in enumerate(zip(p10s, p90s, cvs)):
+        width = high - low
+        ax.annotate(f'Width: {width:.2f}h\nCV: {cv:.1f}%',
+                    xy=(i, high + 0.03), ha='center', fontsize=9,
+                    bbox=dict(boxstyle='round', facecolor='white', alpha=0.8))
+    
+    ax.set_xticks(x)
+    ax.set_xticklabels(labels)
+    ax.set_ylabel('TTE (hours)')
+    ax.set_title('Usage Pattern Fluctuation: Impact on TTE Uncertainty (90% CI)')
+    ax.set_ylim(4.1, 5.1)
+    ax.axhline(4.60, color='gray', linestyle='--', alpha=0.5, label='Mean TTE')
+    ax.legend(loc='upper left')
+    ax.grid(axis='y', alpha=0.3)
+    
+    # Robustness conclusion
+    ax.text(0.98, 0.02, 'Robust: CV < 2.5% for reasonable volatility',
+            transform=ax.transAxes, ha='right', va='bottom', fontsize=10, fontweight='bold',
+            bbox=dict(boxstyle='round', facecolor='lightgreen', alpha=0.8))
+    
+    save_fig('fig18_fluctuation')
+
+
+# ==========================================
+# Fig 19: CPL vs CC vs CR Load Model Comparison
+# ==========================================
+def plot_fig19():
+    """
+    Compares three load models: CPL, CC, CR.
+    Shows TTE prediction differences and end-of-life current behavior.
+    MCM O-Award style: clear comparison with proper annotations.
+    """
+    models = list(LOAD_MODEL_COMPARISON.keys())
+    tte_vals = [LOAD_MODEL_COMPARISON[m]['TTE'] for m in models]
+    currents = [LOAD_MODEL_COMPARISON[m]['end_current'] for m in models]
+    
+    # Baseline values
+    baseline_tte = 4.60  # CPL model TTE
+    nominal_current = 0.69  # Nominal operating current
+    
+    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 5.5))
+    
+    # Distinct colors for each model
+    bar_colors = ['#27ae60', '#3498db', '#9b59b6']  # Green, Blue, Purple
+    
+    # ===== Left Panel: TTE Comparison =====
+    x1 = np.arange(len(models))
+    bars1 = ax1.bar(x1, tte_vals, width=0.6, color=bar_colors, alpha=0.85, 
+                    edgecolor='#2c3e50', linewidth=1.5)
+    
+    # Baseline reference line (CPL model)
+    ax1.axhline(baseline_tte, color='#e74c3c', linestyle='--', linewidth=2, 
+                label=f'CPL Baseline ({baseline_tte:.2f}h)')
+    
+    # Add value labels ABOVE bars
+    for i, (bar, t) in enumerate(zip(bars1, tte_vals)):
+        height = bar.get_height()
+        delta_pct = (t - baseline_tte) / baseline_tte * 100
+        if abs(delta_pct) < 0.1:
+            label = f'{t:.2f}h\n(Baseline)'
+        else:
+            label = f'{t:.2f}h\n({delta_pct:+.1f}%)'
+        ax1.text(bar.get_x() + bar.get_width()/2, height + 0.12, label,
+                ha='center', va='bottom', fontsize=10, fontweight='bold')
+    
+    ax1.set_xticks(x1)
+    ax1.set_xticklabels(models, fontsize=10)
+    ax1.set_ylabel('Time to Exhaustion (hours)', fontsize=11, fontweight='bold')
+    ax1.set_title('(a) TTE Prediction by Load Model', fontsize=12, fontweight='bold')
+    ax1.set_ylim(0, 6.5)
+    ax1.legend(loc='upper left', fontsize=9)
+    ax1.yaxis.grid(True, linestyle=':', alpha=0.5)
+    ax1.set_axisbelow(True)
+    
+    # ===== Right Panel: End-of-Life Current =====
+    x2 = np.arange(len(models))
+    bars2 = ax2.bar(x2, currents, width=0.6, color=bar_colors, alpha=0.85,
+                    edgecolor='#2c3e50', linewidth=1.5)
+    
+    # Nominal current reference line
+    ax2.axhline(nominal_current, color='#7f8c8d', linestyle='--', linewidth=2, 
+                label=f'Nominal Current ({nominal_current:.2f}A)')
+    
+    # Measured range shading (28-45% increase from literature)
+    measured_low = nominal_current * 1.28
+    measured_high = nominal_current * 1.45
+    ax2.axhspan(measured_low, measured_high, alpha=0.25, color='#27ae60', 
+                label=f'Measured Range ({measured_low:.2f}-{measured_high:.2f}A)')
+    
+    # Add value labels ABOVE bars
+    for i, (bar, c) in enumerate(zip(bars2, currents)):
+        height = bar.get_height()
+        delta_pct = (c - nominal_current) / nominal_current * 100
+        label = f'{c:.2f}A\n({delta_pct:+.0f}%)'
+        ax2.text(bar.get_x() + bar.get_width()/2, height + 0.03, label,
+                ha='center', va='bottom', fontsize=10, fontweight='bold')
+    
+    ax2.set_xticks(x2)
+    ax2.set_xticklabels(models, fontsize=10)
+    ax2.set_ylabel('Current at SOC=0.1 (A)', fontsize=11, fontweight='bold')
+    ax2.set_title('(b) End-of-Life Current Surge', fontsize=12, fontweight='bold')
+    ax2.set_ylim(0, 1.3)
+    ax2.legend(loc='upper right', fontsize=9)
+    ax2.yaxis.grid(True, linestyle=':', alpha=0.5)
+    ax2.set_axisbelow(True)
+    
+    # Add key insight annotation on right panel
+    ax2.annotate('CPL captures\nreal-world surge', 
+                xy=(0, 1.01), xytext=(0.8, 1.15),
+                fontsize=9, ha='center',
+                arrowprops=dict(arrowstyle='->', color='#27ae60', lw=1.5),
+                bbox=dict(boxstyle='round,pad=0.3', facecolor='#d5f5e3', edgecolor='#27ae60'))
+    
+    plt.tight_layout()
+    save_fig('fig19_cpl_comparison')
+
+
+# ==========================================
+# Fig 20: Signal-Power Mapping Validation
+# ==========================================
+def plot_fig20():
+    """
+    Validates exponential signal-power mapping against literature.
+    Shows linear model failure at low signal quality.
+    """
+    psi_vals = list(SIGNAL_MAPPING.keys())
+    measured = [SIGNAL_MAPPING[p]['measured'] for p in psi_vals]
+    exp_model = [SIGNAL_MAPPING[p]['exp_model'] for p in psi_vals]
+    lin_model = [SIGNAL_MAPPING[p]['lin_model'] for p in psi_vals]
+    
+    fig, ax = plt.subplots(figsize=(9, 6))
+    
+    x = np.arange(len(psi_vals))
+    width = 0.25
+    
+    ax.bar(x - width, measured, width, label='Measured (Literature)', color='gray', alpha=0.7, edgecolor='black')
+    ax.bar(x, exp_model, width, label='Exponential Model (Ours)', color=colors[0], alpha=0.8, edgecolor='black')
+    ax.bar(x + width, lin_model, width, label='Linear Model', color=colors[3], alpha=0.8, edgecolor='black')
+    
+    ax.set_xlabel('Signal Quality Ψ')
+    ax.set_ylabel('Network Power (W)')
+    ax.set_title('Signal-Power Mapping: Model Validation')
+    ax.set_xticks(x)
+    ax.set_xticklabels([f'Ψ={p}' for p in psi_vals])
+    ax.legend(loc='upper left')
+    
+    # Calculate and annotate errors
+    for i, (m, l) in enumerate(zip(measured, lin_model)):
+        err = (l - m) / m * 100
+        if abs(err) > 20:
+            ax.annotate(f'Error: {err:.0f}%',
+                        xy=(i + width, l), xytext=(i + width + 0.3, l + 0.8),
+                        arrowprops=dict(arrowstyle='->', color='red', lw=1.5),
+                        color='red', fontsize=9, fontweight='bold')
+    
+    # Add conclusion box
+    ax.text(0.98, 0.98, 'Linear model fails at Ψ < 0.3\n(Error > 50%)',
+            transform=ax.transAxes, ha='right', va='top', fontsize=10,
+            bbox=dict(boxstyle='round', facecolor='lightyellow', alpha=0.9))
+    
+    save_fig('fig20_signal_validation')
+
+
+# ==========================================
+# Fig 21: 3D Sensitivity Framework Radar
+# ==========================================
+def plot_fig21():
+    """
+    Radar chart summarizing sensitivity across six dimensions.
+    Higher = more sensitive / less robust (risk indicator)
+    """
+    categories = ['Temperature\nSensitivity', 'Signal\nSensitivity', 'Load\nSensitivity',
+                  'Assumption\nRobustness', 'Fluctuation\nRobustness', 'Extreme\nResilience']
+    
+    # Values: higher = worse (more sensitive or less robust)
+    # Based on analysis: Temp=4.8 (63.5% impact), Signal=4.5 (59.8%), Load=3.2, 
+    # Assumption=4.0 (robust to most), Fluctuation=4.5 (CV<2.5%), Extreme=2.5 (E4 problematic)
+    values = [4.8, 4.5, 3.2, 4.0, 4.5, 2.5]
+    
+    N = len(categories)
+    angles = [n / float(N) * 2 * pi for n in range(N)]
+    values_plot = values + values[:1]
+    angles += angles[:1]
+    
+    fig, ax = plt.subplots(figsize=(8, 8), subplot_kw=dict(polar=True))
+    
+    ax.plot(angles, values_plot, 'o-', linewidth=2, color=colors[2], markersize=8)
+    ax.fill(angles, values_plot, alpha=0.25, color=colors[2])
+    
+    # Risk zones
+    theta = np.linspace(0, 2*pi, 100)
+    ax.fill_between(theta, 4, 5, alpha=0.15, color='red')
+    ax.fill_between(theta, 0, 2, alpha=0.15, color='green')
+    
+    ax.set_xticks(angles[:-1])
+    ax.set_xticklabels(categories, size=10)
+    ax.set_ylim(0, 5)
+    ax.set_yticks([1, 2, 3, 4, 5])
+    ax.set_title('3D Sensitivity Framework Summary\n(Higher = More Sensitive / Less Robust)', pad=20)
+    
+    # Add legend
+    from matplotlib.patches import Patch
+    legend_elements = [Patch(facecolor='red', alpha=0.3, label='High Risk Zone (>4)'),
+                       Patch(facecolor='green', alpha=0.3, label='Safe Zone (<2)')]
+    ax.legend(handles=legend_elements, loc='upper right', bbox_to_anchor=(1.3, 1.0))
+    
+    save_fig('fig21_framework_radar')
+
+
+# ==========================================
+# Fig 22: Fluctuation Amplification by SOC
+# ==========================================
+def plot_fig22():
+    """
+    Shows how fluctuation amplification increases at low SOC.
+    CPL feedback causes "last 20% drops fast" phenomenon.
+    """
+    soc_ranges = list(AMPLIFICATION_BY_SOC.keys())
+    beta_vals = list(AMPLIFICATION_BY_SOC.values())
+    
+    fig, ax = plt.subplots(figsize=(9, 5))
+    
+    # Color gradient (green to red)
+    colors_grad = plt.cm.Reds(np.linspace(0.3, 0.9, len(beta_vals)))
+    
+    bars = ax.bar(soc_ranges, beta_vals, color=colors_grad, alpha=0.85, edgecolor='black')
+    
+    ax.axhline(1.0, color='gray', linestyle='--', alpha=0.6, label='No Amplification (β=1)')
+    ax.set_ylabel('Amplification Factor β')
+    ax.set_xlabel('SOC Range')
+    ax.set_title('Fluctuation Amplification by Battery State: Why "Last 20% Drops Fast"')
+    ax.set_ylim(0, 5)
+    ax.legend(loc='upper left')
+    
+    for bar, b in zip(bars, beta_vals):
+        ax.text(bar.get_x() + bar.get_width()/2, b + 0.15, f'{b:.1f}×',
+                ha='center', fontweight='bold', fontsize=11)
+    
+    # Physical explanation
+    ax.text(0.98, 0.98, 'Physics: I=P/V causes\npositive feedback at low V',
+            transform=ax.transAxes, ha='right', va='top', fontsize=10,
+            bbox=dict(boxstyle='round', facecolor='lightyellow', alpha=0.9))
+    
+    ax.grid(axis='y', alpha=0.3)
+    
+    save_fig('fig22_amplification')
+
+
+# ==========================================
+# Main Execution
+# ==========================================
+if __name__ == "__main__":
+    print("=" * 60)
+    print("Problem 3 Figure Generation (Data from 整合输出.md)")
+    print("=" * 60)
+    
+    print("\n[1/9] Fig 14: Sobol Sensitivity Indices...")
+    plot_fig14()
+    
+    print("[2/9] Fig 15: Assumption Robustness Waterfall...")
+    plot_fig15()
+    
+    print("[3/9] Fig 16: Physical Decoupling Analysis...")
+    plot_fig16()
+    
+    print("[4/9] Fig 17: Extreme Scenario Testing...")
+    plot_fig17()
+    
+    print("[5/9] Fig 18: Usage Pattern Fluctuation...")
+    plot_fig18()
+    
+    print("[6/9] Fig 19: CPL vs CC vs CR Comparison...")
+    plot_fig19()
+    
+    print("[7/9] Fig 20: Signal Mapping Validation...")
+    plot_fig20()
+    
+    print("[8/9] Fig 21: 3D Framework Radar...")
+    plot_fig21()
+    
+    print("[9/9] Fig 22: SOC Amplification...")
+    plot_fig22()
+    
+    print("\n" + "=" * 60)
+    print("All Problem 3 figures generated successfully!")
+    print("Output directory: figures/")
+    print("=" * 60)