A Data-Driven Hybrid Machine Learning and Deep Learning Approach for Remaining Useful Life Prediction of Aircraft Turbofan Engines in Condition-Based Predictive Maintenance Systems

Abstract

Predictive maintenance is a critical approach in the aerospace industry, aimed at reducing downtime, lowering maintenance costs, and improving flight safety. Conventional maintenance schedules often result in unnecessary part replacements or unexpected failures. To address this, this work develops a data-driven hybrid machine learning and deep learning model to predict the Remaining Useful Life (RUL) of aircraft turbofan engines using NASA’s publicly available C-MAPSS dataset. The dataset includes multiple sensor measurements, such as temperature, pressure, and vibration, recorded across different flight cycles under varying operating conditions. Each engine is simulated until failure, providing reliable labels for training, while test data is validated using provided ground truth values. The proposed system involves preprocessing sensor data, extracting relevant features, and applying Random Forest, XGBoost, and Long Short-Term Memory (LSTM) networks to predict RUL. For real-time monitoring, sensor data is processed cycle by cycle, with predictions displayed on a dashboard and alerts issued as engines approach critical stages. This study demonstrates the practical application of predictive analytics in aerospace, highlighting how data-driven methods can enhance reliability, ensure safety, and optimize operational costs.