论文部分内容阅读
边坡稳定性分析中,模糊点估计法能同时考虑模糊不确定性和随机不确定性因素。针对传统模糊点估计法计算工作量大的缺点,提出一种神经网络改进模糊点估计法。利用拉丁超立方抽样法和径向基函数神经网络(RBF)建立边坡安全系数的预测模型;对黏聚力和内摩擦角等模糊随机变量取λ截集,并在各截集水平对参数进行组合;利用建立的预测模型对各参数组合的安全系数进行预测;最后由统计矩点估计法计算边坡的可靠度指标。实例分析表明:改进模糊点估计法使用方便、结果可靠,且能通过增加λ截集水平的数目来提高计算精度。对于含有2~4个模糊随机变量的边坡,采用改进模糊点估计法计算可靠度时λ截集水平的数目可近似取25。
In the slope stability analysis, the fuzzy point estimation method can consider both the fuzzy uncertainty and the random uncertainty factors. Aiming at the shortcomings of traditional fuzzy points estimation, this paper proposes an improved fuzzy point estimation method based on neural network. The prediction model of slope safety coefficient was established by using Latin hypercube sampling method and radial basis function neural network (RBF). A λ cut-off was taken for fuzzy random variables such as cohesion and internal friction angle, The safety coefficient of each parameter combination is predicted by using the established prediction model. Finally, the reliability index of the slope is calculated by the statistical moment estimation method. The case study shows that the improved fuzzy point estimation method is easy to use and reliable, and can improve the calculation accuracy by increasing the number of λ cutoff levels. For slopes with 2 ~ 4 fuzzy random variables, the number of λ cut-off levels can be approximated by 25 when using the improved fuzzy-point estimation method to calculate the reliability.