NASA-CMAPSS-RUL-Prediction

RUL prediction using BiLSTM on NASA C-MAPSS FD001 dataset

NASA-CMAPSS-RUL-Prediction

This project predicts the Remaining Useful Life (RUL) of aircraft engines using a Bidirectional LSTM model on the ASA C-MAPSS FD001 dataset.

Features

1. BiLSTM-based deep learning model
2. Sliding window time-series approach
3. Robust training using Huber Loss
4. GroupKFold cross-validation by engine ID
5. Final model achieves MAE < 10 cycles

Final Result

Final MAE : 9.77 cycles
Test Set: FD001 subset of NASA C-MAPSS dataset

Model Architecture

1.Bidirectional LSTM (128 → 64) 2.Dropout & BatchNormalization layers 3.Dense layers with ReLU activation 4.Huber Loss + Adam Optimizer 5.EarlyStopping and ReduceLROnPlateau callbacks

Dataset

Source: NASA C-MAPSS Dataset (FD001) Files Used: 1.train_FD001.txt – Engine sensor data until failure 2.test_FD001.txt – Partial engine run data 3.RUL_FD001.txt – Ground-truth remaining life

Libraries Used

• Python 3.x
• TensorFlow / Keras
• Scikit-learn
• Pandas & NumPy
• Matplotlib

How to Run

1. Clone this repository
2. Open and run NASA-CMAPSS-RUL-Prediction.ipynb using Jupyter or Google Colab
3. Download or provide access to:
   • train_FD001.txt
   • test_FD001.txt
   • RUL_FD001.txt
4. View final MAE and prediction visualizations

Author

Yarlagadda Ramakrishna