降雨及时效影响下高边坡复杂位移多点神经网络监控模型

HIGH SLOPE COMPLEX DISPLACEMENT MULTI-POINT NEURAL NETWORK MONITORING MODEL CONSIDERING RAIN AND TIME EFFECT

为揭示高边坡位移在降雨和时效影响下的复杂变化规律,采用积分型降雨因子和时效函数构建径向基函数(RBF)神经网络监测模型的输入层,以测斜孔内多个测点位移为输出层向量,根据实测降雨、位移时间序列,以模糊C均值聚类(FCM)算法确定RBF计算中心,建立高边坡位移多测点神经网络监控模型。实例表明:采用合理降雨、时效输入层因子及FCM算法的模型可获得理想效果,能有效捕捉复杂变化的位移发展过程,并具有较高的预测精度。

To analyze the complex rules of high slope displacement under rain and time effect,integral rain factors and time effect functions were used to form the input layer of radial basis function（ RBF） neural network,multi-point displacement was taken as its output layer. Based on rain and displacement time series survey data,slope multi-point displacement neural network monitoring model was established with fuzzy C-means algorithm（ FCM）,which was used to determine RBF centers. Instances showed that the model using suitable rain and time effect input factors together with FCM had good training results. The model could describe the displacement complex changing process well,also had good forecast precision.