Individual Multiple-Unit Train Damage Prognosis Method Based on Digital-Twin Concept

Digital twin (DT) framework based on dynamic Bayesian network (DBN) has established a novel paradigm for damage prognosis. This study focuses on the multidimensional uncertainty characterization in damage prognosis of individual multiple-unit trains (IMUTs). The structural feature perception model is developed, integrating a probabilistic load equivalence quantification model based on a generalized durability load spectrum for IMUT, a normalized fatigue crack growth rate model with uncertainty propagation and a probabilistic description model for equivalent initial flaw size. A hybrid uncertainty quantification method is employed to achieve synergistic modeling of deterministic and stochastic parameters. DT experimental platform centered on bogie welded structures is established, implementing a closed-loop validation mechanism between physical tests and virtual models. Experimental results demonstrate tracking errors ≤ 6.8% for damage parameters ( $m,\text{logC}$) and a 90.5% improvement in prediction accuracy compared to conventional methods.