基于Elman神经网络的深基坑变形智能监测

Intelligent Monitoring on Deformation of Deep Foundation Excavation Based on Elman Neural Networks

Elman神经网络是一种典型的动态递归神经网络,它既可以学习空域模式,又可以学习时域模式,能使训练好的网络具有非线性和动态特性。文章采用11个输入单元和1个输出单元的Elman神经网络。利用Matlab提供的神经网络工具箱编写了Elman神经网络程序,可以通过几种监测数据来推测另一种监测数据。最后以一幢超高层建筑的深基坑工程为例,说明了Elman神经网络方法用于深基坑变形的预测具有较好的可靠性,通过不断加入新的数据,Elman神经网络程序所能判断的数据类型和精度均不断提高。

Abstract：The Elman neural networks are typically dynamic recurrent neural networks. The Elman neural networks are able to learn temporal patterns as well as spatial patterns. Therefore, the Elman neural networks may have the nonlinear and dynamic characteristics when they are trained. In this paper, the Elman neural networks which have eleven input units and one output unit are used. The program of the Elman neural networks is compiled by the use of Matlab toolbox, which can forecast a sort of datum through several sorts of monitoring data. At the final part of this paper, an example of the deep foundation excavation of a high-rise is illustrated to show that it is effective to use the Elman neural networks on the intelligent monitoring on defor- mation of the deep foundation excavation. The type and precision of the data that the Elman neural networks can distinguish are both improved through inputting new data.