Abstract: The lightweight bogie is a core technology for achieving “weight reduction, energy conservation, quality improvement, and efficiency enhancement” in high-speed trains. However, the lightweight design alters the dynamic environment of the axlebox bearings, resulting in more complex and concealed fault characteristics. The scarcity of on-site data further restricts the development of intelligent maintenance for axlebox bearings. To address this, this paper proposes a few-shot cross-domain fault diagnosis method assisted by lightweight bogie bearing mechanism simulation. Firstly, a rigid-flexible coupling dynamic model of a lightweight bogie with carbon fiber reinforced plastics (CFRP) side beams is established. Considering the dynamic loads on the bearings from various components, a joint simulation analysis of the bogie-axlebox bearing is conducted to construct a bearing simulation dataset containing typical faults. Secondly, a collaborative model integrating VMamba and cross deformable transformer (VMCDT) is constructed, combined with a teacher-student network knowledge distillation framework, to achieve in-depth mining and smooth transfer of complex cross-domain features. Finally, by introducing a label smoothing strategy, few-shot fault diagnosis of bearings in cross-domain scenarios is realized. Experimental results from six different testing tasks demonstrate that the proposed method achieves fault diagnosis accuracies exceeding 89% and 93% under 4-way 1-shot and 5-shot conditions, respectively, outperforming conventional few-shot learning approaches. This approach overcomes the dual challenges of significant distribution differences between simulation and measured data and the high difficulty of few-shot cross-domain diagnosis, providing a new paradigm for fault diagnosis of axlebox bearings in lightweight bogies of high-speed trains.