【摘要】 传统的有砟道床横向阻力离线检测方法费时耗力且对轨道具有一定破坏性，研究道床横向阻力智能感知方法对提高铁路运维作业效率具有重要意义。该文提出一种融合多体动力学机理与机器学习建模的道床横向阻力智能感知方法。该方法精确表征动力稳定车-稳定装置-有砟轨道十一自由度横向耦合作用，采用基于RBF代理模型优化的参数辨识理论，结合现场试验数据，构建动力学参数与道床横向阻力端到端的深度全连接神经网络（DNN）机器学习模型。通过机理模型与机器学习算法相结合实现横向阻力实时感知。研究结果表明：所提出的道床横向阻力智能感知方法具有较好精度，感知结果与实测数据误差绝对值为2.13%～18.38%。更多还原

【Abstract】 The traditional offline detection method for lateral resistance of ballasted bed is time-consuming,labor-intensive, and has certain destructive effects on the track. The development of intelligent perception method for ballast bed lateral resistance holds significant implications for enhancing railway maintenance and operation efficiency. This study proposes an intelligent perception algorithm for ballast bed lateral resistance that integrates mechanistic modeling of stabilization operations with machine learning algorithm. The proposed method accurately characterizes the 11-DOF lateral coupling interactions among the dynamic track stabilizer,stabilization devices, and ballasted track. An RBF surrogate model-optimized parameter identification approach is developed, integrating field test data to establish an end-to-end deep neural network(DNN)machine learning model correlating dynamic parameters with ballast bed lateral resistance. This hybrid mechanistic-data-driven framework enables real-time monitoring of lateral resistance. The research results indicate that the proposed real-time perception method for lateral resistance of the ballasted bed has good accuracy, with an absolute error of 2.13%-18.38% between the perception results and the measured data.更多还原