Remaining Useful Life Prediction
Aircraft Engine Health Monitoring
Machine Learning Research Project
Accurate prediction of Remaining Useful Life (RUL) in aircraft engines is critical for aviation maintenance — enabling early detection of potential failures, timely component replacement, and improved operational safety.
Overview
This project focuses on preprocessing sensor data from the C-MAPSS (Commercial Modular Aero-Propulsion System Simulation) dataset and building machine learning models to predict RUL. The goal: uncover underlying degradation patterns from raw sensor streams and build reliable predictive models for a mission-critical aviation challenge.
Technical Approach
- Data preprocessing: RUL calculation, normalization, feature engineering from multi-sensor time series
- Models explored: LSTM-based sequence models, gradient-boosted trees, and classical regression baselines
- Evaluation: Root mean square error (RMSE) on held-out engine run-to-failure trajectories
Key Results
- Achieved competitive RMSE on C-MAPSS benchmarks
- Identified dominant sensor channels contributing to degradation signal
- Demonstrated data-driven approach is viable for fleet-scale predictive maintenance