基于交替迭代算法神经网络评价岩石边坡稳定性

目前边坡工程中常用的稳定性分析方法主要分为极限平衡法和数值分析法2大类，文章对它们各自的主要愿理、特点及其优缺点等进行了阐述。首先，根据经典边坡稳定分析方法存在的局限性，提出有必要建立基于人工神经网络的边坡稳定性预报方法。其次，针对经典算法BP网络存在的某些缺陷，提出了一种交替迭代算法神经网络，以提高其非线性映射能力和泛化能力。交替迭代神经网络算法通过解2个阶数比较低的线性代数方程组，逐步求得连接权值的。以此提高收敛速度，且有利于寻求最优解。作者用FORTRAN语言编制了程序。分析了建立边坡岩体稳定性预测网络模型的建立中应该注意的几个方面。最后，基于已有的40个岩石边坡工程实例进行所建立的神经网络的训练和边坡稳定的预报，结果表明文中所建立的边坡稳定性预报方法具有较高的预报准确度。

Analyzing method of rock slope stability with artificial neural network on alternative and iterative algorithm

The ordinary assessment methods of slope stability are limit equilibrium analysis and numerical simulation.This paper discusses the main principle,character,merit and shortage of each assessment method.Based on the limitation of classical analysis method of slope stability,the authors think it is necessary that the artificial neural network method is introduced in the prediction of slope stability.In view of the limitations of the classic BP Neural Network,an optimal Neural Network algorithm based on alternative and iterative algorithm is put forward in this paper.It can optimize the nonlinear mapping ability and generative ability of the Neural Network algorithm,improve the convergence speed and get the optimization solution through solving two groups of linear equations of lower ladder.After discussing some problems of establishing slope stability prediction models, 40 cases of rock-bed slope are used to test program of FORTRAN.The prediction result illustrates that the newly established method based on the artificial neural network can provide more accurate veracity.