An Advanced Multimodal Reliability Assessment for Predicting Remaining Useful Life of Bearings With Degradation Analysis

ABSTRACT

Rolling bearings are critical components of many machines, like high‐speed trains, wind turbines, aerospace systems, and various industrial applications, whose safety and reliability highly rely on their states. Accurately predicting components' remaining useful life is essential to avoid uninterrupted operation and reduce potential problems. Traditional single‐modality techniques to predict remaining useful life often involve complex feature extraction, demanding substantial manual effort, and falling short in effectively processing machinery sensor data because of the immense nature of temporal dynamics of component degradation. The proposed multimodal multi‐task learning framework enables simultaneous bearing degradation assessment and remaining useful life prediction by leveraging complementary information from multiple sensing modalities to address the mentioned limitations. The framework integrates heterogeneous sensor information through a hybrid deep‐learning pipeline that combines time‐frequency representation learning with joint prognostic modeling. Specifically, vibration and temperature signals are transformed into informative representations to capture degradation‐sensitive patterns, while a shared feature‐learning backbone extracts spatial and temporal dependencies relevant to both health‐state assessment and RUL estimation. Instead of claiming novelty in the individual network blocks, the main contribution lies in the coordinated integration of multimodal fusion, joint learning, and maintenance‐oriented risk interpretation within a unified prognostic framework. The proposed model provides an effective and reliable tool for rolling bearing degradation analysis and supports timely maintenance planning to improve the operational safety, reliability, and efficiency of complex machinery.