TFCA-GRU: A temporal and feature cooperative attention GRU for remaining useful life prediction of bearings across conditions

Accurate prediction of the remaining useful life (RUL) of rolling bearings under different operating conditions is essential to enable proactive maintenance of industrial systems. To address the limitations of traditional recursive models in capturing complex time dependencies and distinguishing sensor contributions, we propose a temporal and feature cooperative attention-based gated recursive unit (TFCA-GRU). By integrating temporal and feature channel attention mechanisms, the model adaptively focuses on degradation-related time steps and key sensor features to build more robust RUL models from multidimensional time-frequency fusion data. Therefore, the TFCA-GRU enables sensitive extraction of critical features from long-term dependencies and local feature information. The validity of the proposed TFCA-GRU is verified using the bearing dataset from the XJTU-SY and the IEEE PHM 2012 Prognostic Challenge. This paper also provides the model’s intrinsic interpretability through structured non-uniform attentional learning across time and feature dimensions. This reveals TFCA-GRU's selective attention to information time intervals and feature dimensions, which improves the transparency of the model. Experimental results show that the proposed TFCA-GRU performs better than some existing prediction methods in bearing RUL prediction across operating conditions.