Remaining Useful Life Prediction of Aircraft Engines Based on Multi-Scale Temporal Convolution and Attention Mechanisms

The Remaining Useful Life (RUL) prediction of aircraft engines is a key component of intelligent maintenance strategies, as it can effectively prevent sudden failures, optimize maintenance schedules, and reduce maintenance costs. However, RUL prediction is challenged by noise, outliers in sensor data, and the complexity of modeling diverse operating conditions. To address these issues, this paper proposes an RUL prediction framework for the C-MAPSS multivariate time series dataset. The framework first enhances input data quality by designing an efficient data cleaning method based on Kalman filtering and outlier detection. Next, a multi-scale temporal convolution module is introduced to capture both short-term fluctuations and long-term trend features, and sequence and time attention mechanisms are incorporated to improve the model’s ability to capture key features. Finally, the framework’s effectiveness is validated through model comparisons and ablation experiments.