横向振动条件下高铁梯形螺栓松动 规律与结构优化设计

Trapezoidal bolted joints serve as the primary fastening method for the skirt and bottom plate locks of high-speed train equipment compartments. During service, these joints are subjected to severe vibrational loads, which can lead to preload attenuation and even fatigue fracture, threatening safety. This study aimed to elucidate the loosening mechanism and identify critical conditions under transverse vibration, and conduct structural optimization for enhanced anti-loosening performance and reliability. A high-fidelity finite element model was developed using structured hexahedral meshing for the thread teeth, with mesh sensitivity analysis conducted to ensure convergence. Transverse vibrational loads of varying amplitudes were applied to investigate the loosening behavior and identify critical thresholds, revealing the evolution of contact state and contact stress on the threaded surfaces. A deep neural network (DNN) surrogate model was employed to capture the nonlinear relationship between the bearing surface inclination, clamped component thickness, and preload attenuation rate. Bayesian optimization was subsequently applied to minimize preload attenuation through structural parameter tuning. The DNN model demonstrated superior predictive accuracy compared to surrogate models based on support vector regression, random forest regression, and Gaussian process regression. Results indicate the existence of a critical transverse vibration amplitude. Below this value, the preload attenuation is caused by stress redistribution. Above it, the rotational loosening between internal and external threads can result in progressive preload loss. The optimized joint configuration can reduce preload attenuation by 14.73% and thread surface slip by 78% under 0.2 mm amplitude vibration, significantly improving anti-loosening performance. The integration of the deep neural network surrogate model with Bayesian optimization enables efficient structural parameter identification, substantially enhancing the joint ’ s vibration resistance without additional anti-loosening components. This approach aligns with the lightweight and high-reliability requirements of high-speed rail applications. The results can provide valuable insights for the anti-loosening design and assembly of critical bolted connections such as train skirt locks.