Mixed uncertainties quantification of rolling bearings based on active Kriging model

Rolling bearings are critical components widely utilized in industries such as high-speed railways, aviation, and maritime sectors, where they significantly impact the stability and safety of equipment. However, the complex failure dynamics of rolling bearings pose challenges in analytically representing the limit state function, often necessitating finite element simulations for accurate establishment. Methods based on numerical simulations are computationally intensive and insufficient when describing the operational state of rolling bearings solely through static random uncertainty. Incorporating interval and time-dependent correlations is essential. Consequently, these methods require extensive sampling, further increasing computational demands. To enhance the efficiency of bearing reliability analysis, this study introduces a method for assessing time-dependent failure probability and predicting Mean Time to Failure (MTTF) using active learning Kriging models. The proposed method, suitable for mixed uncertainties, is initially validated through a numerical example. It is then applied to quantify the mixed uncertainties of a double-row cylindrical roller bearing.