基于剩余推力法与BP神经网络的玄武岩残坡积土公路边坡稳定性预测

发布时间：2019-03-07 23:56

【摘要】：采用剩余推力法与BP神经网络,以贵州省毕节地区宋阴公路K5+170~K5+220段玄武岩残坡积土边坡作为工程研究对象,对该边坡稳定性展开了计算和预测。选取现场实测剖面作为计算剖面,设置4个计算工况,由剩余推力法得到边坡天然状态(工况1)稳定性系数为1.085 1,当边坡处于16m地下水位+暴雨(工况2)、16m→8m地下变化水位(工况3)和16m→8m地下变化水位+暴雨(工况4)时,边坡稳定性系数均小于1。边坡稳定性敏感因素分析显示,滑带土黏聚力敏感系数平均值为15.9%,内摩擦角为48.3%,地下水位为34.0%,表明滑带土内摩擦角对边坡稳定影响最大,其次是地下水位。选择同一路段其他玄武岩残坡积土滑坡作为训练样本,通过Matlab神经网络ANN工具箱分步骤设计了BP网络,选择加动量学习速率自适应traingdx函数作为训练函数,采用多次预测求均值的方法获取预测结果。BP神经网络预测结果表明,边坡工况1的稳定性系数平均值为1.095~1.139,工况3为0.988~1.021,考虑到暴雨对边坡坡稳定性的影响,工况4时边坡可能发生滑动破坏。神经网络各次预测结果之间误差较大,最大达到45.87%,但求均值后的BP神经网络预测结果与剩余推力计算结果的相对误差大大降低,仅为0.4%~5.2%。将BP网络的输入参数减少为5个后,预测精度反而较高,表明黏聚力、内摩擦角、坡高、坡角、湿重度等因素对边坡稳定性有着实质性的影响,其他因素影响权重则较低。
[Abstract]:In this paper, the residual thrust method and BP neural network are used to calculate and predict the stability of the slope with K _ 5-170~K5-220 section of Songyin Highway in Guizhou Province as an engineering research object. The field measured section is selected as the calculation section, and four calculation conditions are set. The stability coefficient of the natural state of the slope (case 1) is 1.085, and when the slope is in 16m ground water level (case 2), the stability coefficient of the slope is obtained by the residual thrust method. The slope stability coefficient is less than 1. 1 when the changing water level of 16m ~ 8m (condition 3) and the groundwater level of 16m ~ 8m (case 4) are under heavy rain (condition 4). The analysis of sensitive factors of slope stability shows that the average sensitive coefficient of cohesive force of sliding zone soil is 15.9%, the internal friction angle is 48.3%, and the groundwater level is 34.0%, which indicates that the internal friction angle of sliding zone soil has the greatest influence on slope stability. The second is the groundwater level. Other basalt residual slope landslides of the same section are selected as training samples, BP network is designed step by means of ANN toolbox of Matlab neural network, and adaptive traingdx function of momentum learning rate is selected as training function. The result of BP neural network prediction shows that the average stability coefficient of slope condition 1 is 1.095 脳 1.139, and that of condition 3 is 0.988 / 1.021, the prediction results of BP neural network show that the average stability coefficient of slope condition 1 is 1.095 脳 1.139 and 0.988 / 1.021, respectively. Considering the influence of torrential rain on slope stability, sliding failure of slope may occur when working condition is 4. The error between the prediction results of neural network and the residual thrust calculation results is large, the maximum error is 45.87%, but the relative error between the predicted results of BP neural network and the residual thrust calculation results is greatly reduced, only 0.4% to 5.2%. When the input parameters of BP network are reduced to 5, the prediction accuracy is higher, indicating that the factors such as cohesion, internal friction angle, slope height, slope angle and wet degree have a substantial effect on slope stability, while other factors have lower influence weight on slope stability.
【作者单位】：张家口职业技术学院;北京工业大学交通工程北京市重点实验室;中国地质科学院地质力学研究所;
【基金】：河北省科技计划自筹经费项目,项目编号152176267 国家“十二五”科技支撑计划项目,项目编号2012BAK10B02
【分类号】：U416.14

【相似文献】