DiffRUL-CMAPSS

Diffusion-based data augmentation for aero-engine remaining-useful-life estimation

Sharif Center for Information Systems and Data Science — Jan 2025 – Jun 2025 Supervised by Dr. Babak Khalaj and Dr. Mohammad Hossein Rohban

Built a diffusion-based data augmentation model in PyTorch for aero-engine remaining-useful-life (RUL) estimation on the NASA C-MAPSS dataset, based on a 2024 Reliability Engineering & System Safety paper.

Highlights

  • Denoising Diffusion Probabilistic Model (DDPM) for realistic sensor-data augmentation
  • Combined with LSTM and Transformer architectures for downstream RUL prediction
  • Applied to predictive maintenance of batteries and bearings