A reliable bearing remaining useful life prediction method based on multi-hierarchy dynamic evaluation and uncertainty amelioration

Due to the synergistic effect of internal and external factors, the degradation process of bearings exhibits strong nonlinearity and high uncertainty, which poses significant challenges for condition monitoring and remaining useful life (RUL) prediction of bearings. Therefore, a reliable RUL prediction method based on multi-hierarchy dynamic evaluation and uncertainty amelioration is proposed in this paper. First, the degradation pattern of the bearing is adaptively determined according to the real-time monitoring data, thereby reducing the reliance on domain-specific prior knowledge of bearing degradation. Subsequently, the health status is iteratively updated with a multi-hierarchy dynamic evaluation mechanism, while a dual-source feedback fine-tuning strategy is designed to collaboratively enhance the model’s predictive performance in real time. Finally, a lifetime uncertainty amelioration technique is developed to integrate lifetime information encoded in uncertainty distributions across multiple hierarchical levels, thereby enhancing the reliability of prediction results. To validate the performance of the proposed method, a comparison with several peer methods is conducted in two experimental bearing datasets, and the outcomes indicate that the proposed method exhibits high accuracy and great reliability.