基于RBF神经网络的软基短期沉降预测研究

Short-Term Prediction of Soft Ground Settlement Based on RBF Neural Network

RBF(radial basis function)神经网络是一类比较优越的前向式多层神经网络,比传统的BP网络有较快的收敛速度。以深圳湾西部通道填海软基沉降的预测分析为例,探讨采用RBF神经网络解决这一问题的方法。采用插值方法构建时间间隔统一的时间序列数据并进行归一化处理,在此基础上建立了沉降变形时间序列的RBF神经网络模型,通过训练网络模型来预测沉降量。计算实例表明,模型具有运算速度快、预测精度高的特点,是一种具有应用前景的软基预测新方法。

RBF artificial neual network belongs to the kind of forwardtype and multilayer neural network. Compared with the traditional BP network, the RBF neural network is faster in convergence and has a higher application value. The paper takes the prediction of the soft ground settlement of Xibu tunnel filling sea project in Shenzhen as example,and discusses the way to solve the problem by adopting RBF neural network. On the basis of constructing the time series data with the same time interval by interpolation method,and treating the data with normalization method,a time series RBF neural network model about softbase sedimentation is established. The sedimentation is predicted by training the network model. The result shows that the model has quite accurate prediction and fast operation capacity,and hence is a new method to predict soft-ground settlement.