Answers are grounded in this paper and cite their sources.
Demonstrates improved RUL prediction in rolling bearings, suggesting enhanced maintenance strategies.
Remaining useful life (RUL) prediction of rolling bearings is essential for condition‐based maintenance of complex rotating machinery. This study proposes a data‐driven prognostic framework that integrates robust signal preprocessing, adaptive feature fusion (AFF) and temporal prediction for dealing with this task. First, a systematic preprocessing pipeline is developed to improve vibration‐signal quality, including chronological concatenation, adaptive wavelet‐threshold denoising, sliding‐window segmentation, and an improved median absolute deviation (MAD) method with dynamic thresholding. By adaptively adjusting the threshold and refining the treatment of non‐numeric (NaN) segments, the improved MAD method effectively suppresses outliers, enhances the signal‐to‐noise ratio, and preserves temporal continuity. Second, a strategy combining AFF and an attention mechanism is introduced to construct a health indicator (HI) that is more sensitive to early weak degradation and more consistent with the underlying degradation trajectory. K‐means clustering is further employed to partition the health states and identify degradation stages, and piecewise linear mapping with normalization is used to generate standardized life labels. Finally, the constructed HI sequence is fed into different temporal predictors for RUL estimation. Experiments on an accelerated life‐test dataset demonstrate that the proposed AFF‐Attention‐HI‐based framework achieves higher prediction accuracy and better long‐horizon stability than conventional AFF‐HI and PCA‐HI schemes. The results further indicate that the proposed HI representation can be effectively utilized by different sequence models, providing reliable support for bearing health assessment and RUL prognostics.
No takes yet. Share an insight, caveat, or question.
Huang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: