KC-BiGRUAtt: A Clustering-Enhanced Deep Framework for Machinery Remaining Useful Life Prediction

This paper proposes KC-BiGRUAtt, a generic deep learning framework for machinery remaining useful life (RUL) prediction that decouples domain-specific feature engineering from universal degradation modeling. The framework integrates three components: 1) Configurable feature engineering - Domain-specific temporal/spectral descriptors (e.g., 13 pre-defined vibration features for bearing health monitoring) are configured based on sensor characteristics; 2) Cluster-driven degradation staging detects transition points via K-Means to enable data-driven RUL calibration; 3) Bidirectional GRUs with multi-scale attention model temporal degradation. Validated on PHM-2012 bearing data (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.989 \mathrm{R}^{2}, 0.025$</tex> RMSE), demonstrating effective degradation staging and RUL prediction without prior failure mode assumptions.