Chengxing Yang

The increasing demand for lightweight and safe rail transit structures has highlighted negative Poisson's ratio (NPR) honeycombs as promising energy-absorbing architectures due to their superior failure mechanism and crashworthiness. This study introduces a novel graded concave octagonal NPR honeycomb (G-ICOH), evolved from the conventional inner concave hexagonal honeycomb (ICHH), to enhance deformation stability and energy dissipation efficiency. Specimens were fabricated from AlSi10Mg using selective laser melting followed by T6 heat treatment with residual stresses markedly reduced, confirming reliable material properties. Quasi-static compression tests and finite element simulations revealed that G-ICOH develops a distinct "X-shaped" deformation mode accompanied by a strong NPR effect. Compared to ICHH, G-ICOH demonstrates a ~ 50% reduction in initial peak stress and a smoother plateau stress response, leading to more stable energy dissipation and up to a 7.58% improvement in specific energy absorption. A parametric analysis showed that the foundation thickness-to-beam arm ratio significantly affects plateau stress and energy absorption, while the influence of the intermediate beam gradient and concave angle is comparatively minor. These results provide explicit guidelines for parameter optimization. The proposed G-ICOH structure combines NPR effect with enhanced crashworthiness, offering practical potential for buffer devices, impact protection systems, and lightweight energy-absorbing components in rail transit and broader transportation engineering applications.

The locomotive coupler usually adopts semi-permanent non-compact coupler, which has the characteristics of strong yaw attitude compared with the compact coupler of urban railway. The state of the coupler has an extremely strong randomness in influencing the vertical and lateral mechanical behaviors of the vehicle, which directly affects the climbing and derailment risk of the train. In order to study the effects of coupler vertical deflection and lateral deflection on the collision mechanical behavior of locomotive, a modeling method of locomotive collision dynamics under lateral/vertical offset attitude of coupler is presented, and its effectiveness was verified by experiments. Then, the influences of coupler with vertical and lateral deflection on the collision mechanical behavior of locomotive were carried out. Finally, through the construction of surrogate model, the influence law of the lateral/vertical offset attitude of coupler on vehicle body collision was intuitively analyzed, at the same time, the main effect and interaction effect of coupler deflection on the collision response were also analyzed. The results show that the lateral/vertical offset of the couplet seriously affects the posture of vehicle after collision, thereby affecting the energy absorption of the crushing deformation of the energy-absorbing components at the end of the vehicle. The influence of the lateral deflection of the coupler on the energy absorption of the collision is stronger than that of the vertical deflection. The vertical deflection of the coupler has a significant effect on the vehicle climbing and the wheelset lifting, which may lead to vehicle climbing and train derailment in severe cases.

The pursuit of higher rail speeds and increasing traffic density have intensified collision risks, making effective crashworthiness design a critical safeguard for mitigating accident consequences. Modern multi-car trains present a particular challenge, as their safety during collisions is governed by complex inter-vehicle interactions. While finite element (FE) methods offer high accuracy in simulating these scenarios, their prohibitive computational cost has limited their utility in rapid safety design and optimization. To address this issue, four machine learning-based framework model for rapid prediction of structural dynamic response was proposed. The surrogate model is coupled with multi-body dynamics to enable efficient multi-objective optimization of train crashworthiness. The AE-GRU model demonstrated excellent convergence, with final losses below 0.0009. More importantly, when applied to large-scale optimization involving 10,000 iterations, the computational cost was reduced to under 25% of that required by conventional FE analysis. This approach provides a robust and efficient data-driven foundation for enhancing collision safety in modern train systems.

Additive manufacturing (AM) has emerged as a transformative approach to fabricating complex geometries with tailored architectures, offering significant advantages in terms of design freedom, material efficiency, and on-demand production

Conducting real-world train crash experiments is the most straightforward and effective method to research the train’s crashworthiness and enhance passive safety protection. As a non-contact measurement method, high-speed camera can efficiently capture the evolving motion patterns of trains under the high-speed states. Traditional data extraction methods rely on expert-based manual annotations, which are susceptible to factors such as illumination changes, scale variance, and shock debris. Inspired by the tremendous success of Deep Neural Networks in the computer vision community, we firstly collect 75 real-world train crash scenes and manually annotated them to form the Crash2024 dataset, enriching the community’s data resources. Moreover, we propose a novel Gradient-guided Joint representation loss with Adaptive neck Detection network (GJADet). At the macro level, we embed the adaptive module into the Path Aggregation Feature Pyramid Network, which combines multiple self-attention mechanisms to achieve scale-awareness, spatial-awareness, and task-awareness, improving the detector’s representation ability and alleviating the dense-small characteristics of the ‘point’ class without significant computational overhead. At the micro level, due to the extreme imbalance of ‘point’ class compared to other classes, we propose a gradient-guided joint representation classification loss to mitigate the long-tailed detection issue. Moreover, the classification and regression are joint representation to maintain consistency between training and inference phase. On the Crash2024, GJADet achieves the performance improvement of 3.5AP and significantly alleviate the accuracy loss problem for rare categories. Our code are open source at https://github.com/YanJieWen/GJADet-crash-pytorch.