二维随机场下边坡稳定性的径向基函数神经网络分析法

发布时间：2019-01-12 08:37

【摘要】：岩土参数的随机性会直接影响边坡稳定性评价结果的精度。首先,依据边坡参数的常用分布特征,利用拉丁超立方抽样法生成若干组边坡土性参数和几何参数的随机样本,用有限元强度折减法求解各组样本对应的边坡安全系数。再考虑土性参数的空间变异性,在二维随机场模型下将蒙特卡罗模拟和有限元强度折减法相结合求解各组样本对应的边坡失效概率。然后,利用样本数据及其安全系数和失效概率对径向基函数(RBF)神经网络进行训练和测试,从而建立边坡安全系数和失效概率的预测模型。算例表明,二维随机场模型能相对精确地考虑参数的空间变异性;在此基础上建立的神经网络模型对边坡的安全系数和失效概率具有较高的预测精度,且能极大地节省边坡稳定性分析的时间。
[Abstract]:The randomness of geotechnical parameters will directly affect the accuracy of slope stability evaluation results. Firstly, according to the common distribution characteristics of slope parameters, the Latin hypercube sampling method is used to generate some random samples of slope soil parameters and geometric parameters, and the slope safety factor corresponding to each sample is solved by using the finite element strength reduction method. Considering the spatial variability of soil parameters, the Monte Carlo simulation and the finite element strength reduction method are combined to solve the slope failure probability of each group of samples under the two-dimensional random field model. Then, the radial basis function (RBF) (RBF) neural network is trained and tested by using the sample data, its safety factor and failure probability, and the prediction model of slope safety factor and failure probability is established. The numerical examples show that the two-dimensional random field model can consider the spatial variability of the parameters relatively accurately. The neural network model established on this basis has high prediction accuracy for slope safety factor and failure probability, and can greatly save the time of slope stability analysis.
【作者单位】：华中科技大学土木工程与力学学院;
【基金】：国家自然科学基金项目(No.51278217)
【分类号】：TU43

【相似文献】