Degradation Twin Modeling of Bearing With Local Fluctuations at Spatial–Temporal Scale

Degradation modeling of machinery accurately is a crucial foundation for achieving predictive maintenance in industrial systems. In real-world applications, addressing local fluctuations within the degradation process poses a significant challenge to modeling accuracy, particularly when these fluctuations occur near the failure threshold. Traditional modeling approaches usually address these fluctuations through denoising techniques during the modeling phase. However, these uncertainty fluctuations contain essential degradation-related information before sudden failure for bearing health assessment. In this article, the data-model combined concept and method of degradation twin modeling with local fluctuations are proposed. By establishing correlations between degradation features at various temporal scales and physical models of fault evolution at various spatial scales, the spatial-temporal scale model for assessing and predicting bearing performance degradation is developed. The results not only provide an evolutionary simulation of multistage failure events at the spatial scale, offering a reinterpretation of nonlinear degradation trends, but also introduce a new five-stage degradation assessment method that leverages degradation uncertainties across various temporal scales. This highlights the significance of adopting spatial-temporal scale perspective when assessing the reliability of performance degradation of physical systems. The effectiveness of the method is verified through experimental cases, affirming its practical utility and robustness.