神经网络技术在路堤沉降预测中应用探讨

Study on Prediction of Post-construction Settlement of Expressway on Soft Ground with Neural Network Model

摘要:

针对传统的BP网络模型的不足，应用了改进的BP神经网络模型，把它应用到软基高速公路的路堤沉降预测中，提出了两种构造神经网络训练样本的思路，并分别进行了计算和对比，指出了各自的优、缺点。结果表明改进的BP网络模型比较稳健、收敛快，而且根据时间与对应的沉降量形成的样本训练的网络预测出的工后沉降误差小、精度高。

Abstract:

Aimed at the insufficiency of traditional BP network, the improved BP neural network model is adopted to predict the post-construction settlement of expressway on soft ground. Two methods of forming the training samples are put forward, and their virtues and shortcomings are pointed out by calculations and analyses. It shows that the improved BP neural network model is steady and fast convergence. Furthermore, the network model trained by the samples from time and according settlement has least error and more accurate.