宝兰客运专线路堤段影响车致地面振动的土体参数敏感度分析及反演

Sensitivity Analysis and Inversion of Soil Parameters Affecting Train-induced Ground Vibration in Embankment Section of Baoji-Lanzhou Passenger Dedicated Line

采用敏感度分析与数值模拟相结合的方法,对影响高速列车引起地面振动的土体参数敏感度进行分析及参数反演。以宝兰客运专线榆中站附近一段路堤的地面垂向振动现场试验为依托,建立车辆-轨道-路基-地基数值模型与神经网络。以土体参数样本集为输入源,各测点地面垂向振动加速度有效值作为输出,基于敏感度理论计算各测点处的参数敏感度。结果表明:影响地面振动的土体参数敏感度顺序为弹性模量>阻尼系数>泊松比>密度>内摩擦角>黏聚力;阻尼系数敏感度随距中心线距离增大呈上升趋势,其余参数随距离增大变化不大;基于敏感度分析结果,选取相关参数作为待反演参数,将现场实测的各测点垂向振动加速度有效值输入神经网络,反演得出的各参数值与现场实际参数值相对误差均小于5%,在工程可接受的范围之内,证明了神经网络反演方法的可行性。

Based on the combining method of sensitivity analysis and numerical simulation,the sensitivity of soil parameters in ground vibration caused by high speed railway was analyzed and the parameter inversion was carried out.According to the ground vertical vibration field test of one embankment section near Yuzhong station of Baoji-Lanzhou passenger dedicated line,the numerical model and neural network of vehicle-track-subgrade-foundation were established.Taking the soil parameter sample set as the input source and the effective value of ground vertical vibration acceleration of each measuring point as the output,the parameter sensitivity of each measuring point was calculated based on the sensitivity theory.The results show that the order of soil parameters sensitivity affecting ground vibration is elastic modulus> damping coefficient> Poisson ratio> density> internal friction angle> cohesion,the sensitivity of damping coefficient show an increase trend with the distance from the center line increasing,while the other parameters have little change with the distance.Based on the results of sensitivity analysis,the relevant parameters are selected as the parameters to be inverted,and the effective value of vertical vibration acceleration of each measuring point measured in the field is as input into the neural network.The relative error between the inverted parameters and the actual parameters in the field is less than 5%,which is within the acceptable range of the project,proving the feasibility of the neural network inversion method.