以边坡高度、边坡角度、岩土重度、粘聚力、内摩擦角等作为输入模式变量，建立BP人工神经网络训练样本集以之用作滑坡稳定性评价。通过对网络学习参数的优化，如学习速率为0.9，学习步长为0.7，在迭代12 589次网络训练后样本收敛。以此为基础，建立BP神经网络各隐含层的连接权重和阈值，进行模式识别，完成了鱼洞河边坡状态和稳定系数的计算。计算结果表明，鱼洞河边坡处于破坏（不稳定）状态，稳定系数为1.100 5

A BP artifical neural networks model was establishedusing slope height and angle, rock gravity-density, cohe
sion and internal friction angle as input variables. The model is used for stability evaluation of Yudonghe Landslide. By means of the optimization of trainningparameters, e.g., the learing ratio being 0.9, the learing step being 0.7
,after 12 589 times trainning the samples approached to convergence. The linking weights and threshod values were determined to recognize mode shape, and thenthe calculation of the slop state and stability coefficients was completed. The calculated results show that the Yudonghe landslide is in danger state and its stability coefficient only 1.100 5