Production-ready FTIR spectroscopy analysis tool for grease condition monitoring. Uses a deviation-focused approach that reports factual spectral differences (ΔX, ΔY) rather than quality scores.

✅ Pure Deviation Analysis - Reports WHERE spectra differ (ΔX, ΔY metrics)
✅ Multi-Metric Categorization - Combines correlation + ΔX + ΔY + ratio
✅ Fast Performance - <1 second per sample
✅ Baseline Compatibility - Detects formulation mismatches
✅ AI Translation - Optional LLM for user-friendly language (NOT quality judgments)
✅ PyQt6 GUI - Interactive desktop application
✅ Batch Analysis - Process multiple samples efficiently

CSV Files → Deviation Analyzer → Multi-Metric Categorization → Optional LLM Translation
            (ΔX, ΔY, correlation)   (GOOD/CRITICAL/MISMATCH)    (User-friendly summary)

1. FTIRDeviationAnalyzer (PRIMARY)

   • Pure deviation metrics (ΔX, ΔY)
   • Multi-metric categorization
   • No quality judgments
   • Always reliable (<1s)

2. FTIRPeakAnalyzer (LEGACY)

   • Peak detection
   • Quality recommendations
   • Kept for backward compatibility

# Clone repository
git clone https://github.com/JinLeeGG/YCP_Grease_Analyzer.git
cd YCP_Grease_Analyzer

# Install dependencies
pip install -r requirements.txt

python app.py

1. Load Baseline - Click "Select Baseline CSV"
2. Load Samples - Click "Select Sample CSV(s)" (multiple files OK)
3. View Results - Analysis runs automatically
4. Export - Save graphs and reports

ΔY (Vertical Deviation)

• Intensity difference at same wavenumber
• Thresholds: <0.03A (OK), 0.05-0.10A (major), >0.10A (critical)

ΔX (Horizontal Shift)

• Peak position change
• Thresholds: <5cm⁻¹ (OK), 5-10cm⁻¹ (minor), >10cm⁻¹ (major)

ΔX:ΔY Ratio

• 100: Shift-dominant (chemical change)

• <20: Intensity-dominant (degradation/concentration)
• 20-100: Both types of deviation

Correlation (r)

• ≥0.97: Excellent similarity
• 0.85-0.97: Good to moderate
• <0.85: Poor (possible mismatch)

YCP_Grease_Analyzer/
├── app.py                          # Main PyQt6 GUI application
├── requirements.txt                # Python dependencies
├── modules/
│   ├── ftir_deviation_analyzer.py  # PRIMARY: Deviation analysis
│   ├── ftir_peak_analyzer.py       # LEGACY: Peak detection
│   ├── llm_analyzer.py             # LLM integration (optional)
│   ├── csv_processor.py            # CSV data handling
│   ├── graph_generator.py          # Matplotlib plotting
│   └── peak_detector.py            # Legacy peak detection
├── utils/
│   ├── config.py                   # Configuration settings
│   └── validators.py               # Data validation
├── GUI/
│   ├── Analyzer_main.ui            # Main window UI
│   └── path.ui                     # Path selection dialog
├── docs/
│   ├── FINAL_SYSTEM_SUMMARY.md     # Complete system documentation
│   └── (other documentation)
└── tests/
    ├── test_csv_loader.py
    └── test_llm.py

# In modules/ftir_deviation_analyzer.py
config = DeviationConfig()

# Vertical deviation (ΔY)
config.delta_y_critical = 0.10  # A (absorbance units)
config.delta_y_major = 0.05
config.delta_y_minor = 0.03

# Horizontal shift (ΔX)
config.delta_x_major = 10.0     # cm⁻¹
config.delta_x_minor = 5.0

# Correlation thresholds
config.correlation_excellent = 0.97
config.correlation_good = 0.95

# In utils/config.py
LLM_CONFIG = {
    "enabled": True,
    "model": "llava:7b-v1.6",  # Ollama model
    "timeout": 30
}

To use LLM features:

# Install Ollama
winget install Ollama.Ollama

# Pull model
ollama pull llava:7b-v1.6

Problem: Too many false positives

• Solution: Increase deviation thresholds in DeviationConfig

Problem: BASELINE_MISMATCH errors

• Solution: Verify you're comparing the same grease type

Problem: LLM not working

• Solution: LLM is optional - system works without it
• Check Ollama installation: ollama --version

Problem: Slow performance

• Solution: Disable LLM for batch processing
• Check data file size (should be <10K points)

• FINAL_SYSTEM_SUMMARY.md - Complete system overview
• OPTIMIZATION_CHANGELOG.md - Implementation history
• QUICK_START_NEW_SYSTEM.md - Quick reference guide

"The deviation analyzer does the math; AI makes it human-friendly"

• System reports FACTS (ΔX, ΔY, correlation)
• AI translates to plain language (optional)
• You make decisions based on facts + your expertise

• ✅ Smart BASELINE_MISMATCH detection - Distinguishes different formulations from contamination
• ✅ Degradation pattern recognition - Identifies oxidation + water signature
• ✅ Systematic vs localized deviation analysis - Better context-aware decisions
• ✅ Actionable warnings - "Use baseline from same grease family"
• 📚 See IMPROVED_CATEGORIZATION_SUMMARY.md for details

• ✅ Pure deviation metrics (ΔX, ΔY)
• ✅ Multi-metric categorization
• ✅ ΔX now triggers alerts
• ✅ ΔX:ΔY ratio diagnostic
• ✅ Baseline mismatch detection

• Peak-based analysis
• Quality recommendations
• Still available for compatibility

See LICENSE file for details.

• Development: YCP Lab
• System Design: Based on FTIR spectroscopy best practices
• AI Integration: Claude Sonnet 4.5 consultation

• FTIR Spectroscopy Principles
• Grease Analysis Standards
• PyQt6 Documentation

Last Updated: November 9, 2025
Version: 2.1 - Improved Categorization Logic