Unsupervised Autoencoders Combined with Multi-Model Machine Learning Fusion for Improving the Applicability of Aircraft Sensor and Engine Performance Prediction

Predicting aircraft engine performance and sensor data is crucial for ensuring safety and optimizing maintenance schedules, which in turn extends the life of aircraft components. Effective prediction improves operational efficiency and is essential for managing safety. This study focuses on estimating the Remaining Useful Life (RUL) of aircraft engines based on the advanced machine learning models, which is critical for scheduling maintenance proactively. Accurate predictions of RUL allow organizations to plan maintenance based on the engine’s actual condition rather than fixed intervals, helping to minimize unexpected downtimes and optimize resource use. Our approach integrates advanced machine learning techniques, using autoencoders for feature extraction combined with various predictive models, to enhance the accuracy of RUL predictions. The experimental results demonstrated the effectiveness of the method. This method leverages detailed patterns from sensor data to improve maintenance strategies and increase aircraft reliability and availability. Implementing such predictive analytics makes maintenance operations more efficient and cost-effective, significantly benefiting fleet management and safety.