地连墙变形的神经网络多步预测研究

结合润扬长江公路大桥南汊北锚碇深基坑工程 ,提出并应用神经网络多步预测方法来研究地连墙施工变形的预测问题。系统介绍了基于时间窗口的神经网络多步滚动预测技术 ,并详细讨论了输入输出层的设计、隐层神经元数以及预测时间步长等一些基本预测技术问题。该预测方法应用于润扬长江公路大桥南汊北锚碇深基坑围护工程 ,取得了较好的工程效果。

DISPLACEMENT PREDICTION OF UNDERGROUND CONTINUOUS WALL BY THE MULTI STEP NEURAL NETWORK METHOD

Across the Yangtze River, the Run-Yang highway bridge which is under construction will link up Yangzhou city with Zhengjiang city. Separated by the island in the river center, the bridge is designed as two parts. The north section is diagonal tension bridge and the south section is suspended cable bridge. The north pit foundation of the suspended cable bridge is a key engineering for the whole bridge. How to predict the deformation of the support construction-underground continuous wall in the process of pit construction is very important to the safety of the engineering. In order to solve this issue, a prediction method of multi-step neural network is put forward. In this paper, the artificial neural network prediction method of multi-step tumble skill based on time window is introduced systematically. A series of key technique problems, such as input and output layer design, the proper number of hidden layer nodes and time-step of prediction are discussed in the paper. Using this method, the deformation of underground continuous wall in the north pit foundation of the Run-Yang highway bridge is predicted in the process of construction. This has good effects on the engineering.