✈ Aircraft Engine RUL Prediction

Predict Remaining Useful Life (RUL) of aircraft turbofan engines using Deep Learning.

Built with PyTorch LSTM + Temporal Attention on NASA's CMAPSS Turbofan Engine Degradation Simulation dataset.

---

🎯 Project Overview

Feature	Details
Dataset	NASA CMAPSS FD001 (100 train engines, 100 test engines)
Model	2-Layer Stacked LSTM with Temporal Attention
Framework	PyTorch
Dashboard	Streamlit (interactive web UI)
Metrics	RMSE, MAE, NASA Scoring Function

🚀 Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Train the Model

python main.py --mode all

This will:

• Load & preprocess the CMAPSS FD001 dataset
• Train the LSTM model with early stopping
• Evaluate on the test set
• Generate prediction plots

3. Launch the Dashboard

streamlit run dashboard/app.py

📊 Dashboard Pages

1. 🏠 Overview — Key metrics, dataset info, engine lifecycle visualization
2. 📊 Sensor Explorer — Interactive sensor degradation charts, correlation heatmap
3. 🔮 RUL Predictions — Actual vs Predicted analysis, scatter plots, error distribution
4. 🔬 Model Insights — Training curves, attention weight heatmaps
5. ⚙️ Live Inference — Upload sensor data for real-time RUL prediction

🏗 Architecture

Input (batch, 30, 14 features)
  → LSTM Layer 1 (128 hidden) + Dropout
  → LSTM Layer 2 (64 hidden) + Dropout
  → Temporal Attention
  → FC(64 → 32) → ReLU → FC(32 → 1)
Output: Predicted RUL (cycles)

📁 Project Structure

engine-anomaly/
├── data/CMAPSSData/        # Dataset files
├── src/
│   ├── data_loader.py      # Data loading & preprocessing
│   ├── dataset.py           # PyTorch Dataset (sliding window)
│   ├── model.py             # LSTM + Attention model
│   ├── train.py             # Training loop
│   ├── evaluate.py          # Evaluation metrics
│   ├── predict.py           # Inference
│   └── utils.py             # Plotting & helpers
├── dashboard/app.py         # Streamlit dashboard
├── models/                  # Saved checkpoints
├── results/                 # Predictions & plots
├── config.py                # Hyperparameters
├── main.py                  # CLI entry point
└── requirements.txt

📈 Metrics

• RMSE: Root Mean Squared Error (target: ≤ 15 cycles)
• MAE: Mean Absolute Error
• NASA Score: Asymmetric penalty — late predictions penalized more heavily (safety-critical)

📚 References

• NASA CMAPSS Dataset
• Saxena, A., et al. "Damage Propagation Modeling for Aircraft Engine Run-to-Failure Simulation." (2008)