考虑地形与降雨因素的植被混凝土边坡侵蚀量研究  

Research on Slope Erosion of Vegetation Concrete Considering Terrain and Rainfall

作　　者：李灿 周海清 赵尚毅 王庆鑫 彭岳 LI Can;ZHOU Hai-qing;ZHAO Shang-yi;WANG Qing-xin;PENG Yue(Department of Military Facility,Army Logistics University of PLA, Chongqing 401331, China;Chongqing Key Laboratory of Geomechanics & Geoenvironment Protection,Army Logistics University of PLA, Chongqing 401331, China;Chongqing University of Science and Technology,Chongqing 401331, China;China Construction Tunnel Corp. LTD., Chongqing 401147, China)

机构地区：[1]陆军勤务学院军事设施系,重庆401331 [2]陆军勤务学院岩土力学与地质环境保护重庆市重点实验室,重庆401331 [3]重庆科技学院,重庆401331 [4]中建隧道建设有限公司,重庆401147

出　　处：《水电能源科学》2019年第8期96-99,共4页Water Resources and Power

基　　金：重庆市基础研究与前沿探索项目（CSTC2018JCYJAX0373）;中建股份科技研发项目（CSCEC-2017-Z-24）;重庆市自然科学基金项目（cstc2017shmsA0581）;重庆市博士后科研项目（Xm2017006）;重庆市国土资源和房屋管理局科技计划项目（KJ-2017019）

摘　　要：利用室内模型植被混凝土边坡冲刷试验数据54组,建立BP神经网络模型,输入层采用3节点,对应3个受制因子(降雨强度、坡率、降雨历时),隐含层节点数4,输出层为1节点,输出侵蚀模数。样本数据随机分成44组训练样本与10组测试样本,通过建立的神经网络训练学习,对测试样本的计算值与真实值比较,验证计算模型的合理性;同时根据权值矩阵分析受制因子的影响显著性。研究结果表明,大部分样本数据真实值与计算值相对误差小于10%,决定系数R2=0.960,说明BP神经网络能很好应用于植被混凝土边坡泥沙侵蚀模型的计算预估中;受制因子显著性排序为降雨强度>降雨历时>坡率,其中降雨强度显著性为2,远大于降雨历时与坡率。该结果在一定程度上能够对植被混凝土边坡施工与设计提供参考。The 54 groups slope scouring indoor test data of vegetation concrete were used to establish a BP neural network model. The input layer is set 3 nodes, corresponding to 3 factors (rainfall intensity, slope rate, rainfall duration), and the number of hidden layer nodes is set 4. The erosion modulus is taken as output layer with 1 node. The sample data is randomly divided into 44 groups of training set and 10 groups of test set. Through the established neural network for training and learning, the calculated values of the test samples are compared with the real values to verify the rationality of the calculation model. At the same time, the significance of the influencing factor is analyzed according to the weight matrix. The results show that the relative error between the actual value and the calculated value of the most sample data is less than 10%, and the coefficient of determination R2 is 0.960, which indicates that the BP neural network can be well applied to estimation of the sediment erosion model of vegetation concrete slope. The significant order of the influencing factors is: rainfall intensity > rainfall duration > slope rate, where the rainfall intensity is 2, and is far greater than the rainfall duration and slope rate. The result can provide reference for construction and design of vegetation concrete slope to a certain extent.

关 键 词：植被混凝土 坡面侵蚀 侵蚀模数 神经网络 

分 类 号：TU42[建筑科学—岩土工程]