基于动态递归神经网络的地连墙安全预测

对于深基坑开挖时地连墙的安全状态做到及时掌控和预测,是基坑工程的重要内容。通过对深基坑测斜数据的地连墙弯矩反分析,得出地连墙的内力发展状况;基于动态递归神经网络基本原理,预测地连墙水平位移和内力的下一步发展趋势。据此,提出了一套掌握深基坑地连墙变形和内力安全状况,并预测以后安全走势的实用方法。研究发现,该预测方法能够在施工时,从变形和内力两个方面及时了解深基坑地连墙的安全状态,在一定程度上节省基坑监测开支。

Safety Prediction of Diaphragm Wall Based on Dynamic Recurrent Neural Network

To grasp and predict the safety status about diaphragm wall in excavation of deep foundation pit in time is very important for foundation pit engineering. The inner force development condition of diaphragm wall was obtained through the inverse analysis of bending moment in diaphragm wall based on in situ deflection data. The horizontal displacement and further development trend of diaphragm wall were predicted based on the basic principle of dynamic recurrent neural network. Consequently, a practical method was put forward to grasp the deformation and inner force safety status of deep foundation pit diaphragm wall and predict the future safety trend. The results showed that the prediction method can help learning the safety status of deep foundation pit diaphragm wall from two aspects: deformation and inner force; which can save monitoring expenditure of foundation pit to a certain extent.