Predictive Modeling of Fastening Pressure Based on Geometric Morphological Features of Fasteners

Railway fasteners play a critical role in the railway system, as they are essential for damping dynamic loads and ensuring track stability and alignment. The detection of fastener service status is crucial for improving the overall safety of the railway system. Although visual methods for detecting missing or damaged fasteners have become well-established, assessing the tightness of fasteners remains a significant challenge. Most existing studies primarily rely on a single morphological feature of the fastener to make coarse judgments about its tightness. However, due to the low manufacturing precision of the fastener, these approaches fail to accurately predict the pressure applied by the fastener. In this study, a finite element analysis model of a rail fastener is established, and their deformation behavior during torque application is systematically investigated. To validate the simulation results, an experimental platform equipped with 3D line laser sensors and pressure sensors is developed to quantitatively analyze the correlation between morphological features and fastening pressure. Based on simulation results and point cloud features, a set of fastener morphological extraction algorithms is developed, which extracts 31 key points within the fastener structure. Subsequently, a region-based geometric fusion strategy integrated with Max-Relevance and Min-Redundancy feature selection is proposed, while Gaussian Process Regression is employed for fastening pressure prediction. The proposed method achieves a reduction in the root mean square error of pressure prediction to 0.427 kN, representing a 23.6% reduction compared to traditional fastening pressure prediction methods based on the clip gap.