【作者】 李小龙；

【导师】 宋旭明； 程丽娟；

【作者基本信息】 中南大学 ， 土木水利（专业学位）， 2025， 硕士

【摘要】 随着我国交通运输需求增长,工程用地紧张问题日益突出,新建工程在既有高铁桥墩周边设置弃土场的状况比较常见。为保证高速铁路安全运营,本文以福厦高铁西溪特大桥沿线弃土场为工程背景,对影响高铁桥墩附加位移的因素进行敏感性分析,建立支持向量机(Support Vector Machine,SVM)响应面模型对高铁桥墩附加位移进行预测,计算高铁桥墩附加位移的超限概率。完成的主要工作如下:(1)选取莫尔-库伦土体本构模型及桩界面单元模拟桩土相互作用关系,建立土体-桥梁空间有限元模型,采用扰动分析法及灰色关联度法分析堆载等级、堆载距离及多个土层参数对高铁桥墩附加位移的敏感性。结果表明:对墩顶横向附加位移影响最大的是上层软弱土层的泊松比,对墩顶竖向附加位移影响最大的是持力层的弹性模量。结合两种分析方法,确定堆载等级、堆载距离、土层弹性模量、泊松比等10个参数为主要影响参数。(2)根据Box-Behnken Design(BBD)法设计161组试验样本,构建SVM模型拟合10个主要影响参数与高铁桥墩附加位移之间的隐式关系,并研究SVM模型特性。结果表明:采用RBF核函数的拟合效果更好;通过参数寻优后,SVM模型预测值的最大相对误差不超过10%;高斯核的宽度g取值在0<g≤0.1范围内时SVM模型预测精度较高。(3)当训练样本减少至101组时,SVM模型预测值最大相对误差仍不超过10%且变化幅度较小;当训练样本减少至91组时,由于训练样本与预测样本在空间中的分布出现偏移,SVM模型预测值最大相对误差超过15%。(4)对比分析SVM模型与多项式响应面模型的预测结果。结果表明:对于墩顶横向附加位移,SVM模型的预测精度稍高;对于竖向附加位移,两者的预测精度相差不大,SVM模型在预测小数值数据时波动性较大。(5)分别以SVM模型和多项式响应面模型构建结构功能函数,结合蒙特卡罗法计算高铁桥墩附加位移可靠概率。结果表明:两种计算方法所得高铁桥墩附加位移可靠概率均大于99%。在保证高速铁路安全运营条件下,背景工程铁路沿线弃土场堆载距离分别为5m、10m、15m时,对应的堆载等级应分别小于57k Pa、72k Pa、87k Pa。当堆载距离为20m～25m,堆载等级在90k Pa内时,铁路沿线堆载对高速铁路的运营安全影响很小。图51幅,表20个,参考文献89篇更多还原

【Abstract】 With the increasing demand for transportation in China,the issue of route intersections and overlaps has become more prominent,and it is common for new construction projects to set up spoil sites around existing high-speed railway piers.To ensure the safe operation of high-speed railways,this study takes the spoil sites along the Xixi Grand Bridge of the Fuzhou-Xiamen High-Speed Railway as the research background.A sensitivity analysis is conducted on the factors influencing the additional displacement of high-speed railway piers.A Support Vector Machine(SVM)response surface model is established to predict the additional displacement of high-speed railway piers,and the probability of displacement exceeding the limit is analyzed.The main work completed is as follows:(1)A three-dimensional finite element model of the soil-bridge system was established using the Mohr-Coulomb constitutive model for the soil and interface elements to simulate the pile-soil interaction.The disturbance analysis method and the grey correlation method were used to analyze the sensitivity of additional displacement in high-speed railway bridge piers to factors such as load level,load distance,and multiple soil layer parameters.The results indicate that the Poisson’s ratio of the upper weak soil layer has the greatest impact on the lateral additional displacement at the pier top,while the elastic modulus of the bearing layer has the greatest impact on the vertical additional displacement at the pier top.Based on the two analysis methods,ten key influencing parameters were identified,including load level,load distance,soil layer elastic modulus,and Poisson’s ratio.(2)According to the Box-Behnken Design(BBD)method,161groups of test samples were designed,and the implicit relationship between the 10 main influencing parameters and the additional displacement of the high-speed rail pier was constructed to construct the SVM model,and the characteristics of the SVM model were studied.The results show that the radial basis function(RBF)kernel provides a better fitting effect.After parameter optimization,the maximum relative error of the predicted values by the SVM response surface model does not exceed10%.When the Gaussian kernel width is in the range of(0<g≤0.1),the prediction accuracy of the SVM response surface model is relatively high.(3)When the number of training samples is reduced to 101,the maximum relative error of the predicted values by the SVM response surface model remains within 10%with minor fluctuations.However,when the training samples are further reduced to 91,a spatial distribution deviation between the training and prediction samples occurs,leading to a significant increase in the maximum relative error of the predicted values.(4)The prediction results of the SVM model and the polynomial response surface model were compared and analyzed.The results show that for the lateral additional displacement at the pier top,the SVM response surface model exhibits slightly higher prediction accuracy.For vertical additional displacement,the prediction accuracy of both models is similar,though the SVM response surface model shows greater fluctuations when predicting small values.(5)Structural performance functions were constructed using both the SVM response surface model and polynomial response surface models.The Monte Carlo method was then applied to calculate the reliability probability of additional displacement in high-speed railway bridge piers.The results show that the reliability probabilities of the additional displacements of high-speed railway piers,obtained by both calculation methods,are greater than 99%.Under the condition of ensuring the safe operation of high-speed railways,the allowable load levels for spoil grounds along the railway are as follows:when the load distance is 5 m,10 m,and 15 m,the corresponding load levels should be less than 57 k Pa,72 k Pa,and 87 k Pa,respectively.When the loading distance is between20 m and 25 m and the loading level is within 90 k Pa,the impact of embankment loading along the railway on the operational safety of high-speed railways is minimal.With 51 graphs,20 sheets and 89 references更多还原