Research Of Application Of General Regression Neural Network To Deformation Monitoring Of Foundation-Pit

In the Excavation monitoring, the future of the monitoring is predicted trends through the past records of the monitoring data processing and analysis, however, because of the complexity of the factors affecting the pit, the functional relationship is very complex, highly nonlinear relationship so that there is not a constitutive model can be a good description of it. And the Generalized Regression Neural Network is a nonlinear function model. It has strong nonlinear approximation capability. It is ideal to apply to the deformation analysis and data forecasting of the deep foundation pit monitoring.On the basis of the research of the theory of the excavation monitoring and comparison of the common data processing methods,we can conclude that the Neural Network is good at forecasting the data of the excavation monitoring, which has a large number of monitoring data.The main tasks of this paper are as follows:(1) Based on the study of the Excavation monitoring, the common data processing methods used in the excavation are summarized.And we conclude that the deep excavation monitoring system is a highly complex non-linear model and there are many impact factors in it so that it is difficult to use a model describe it and it requires a model, which must have a strong nonlinear approximation ability.(2) Based on the study of the basic principles of the Neural Network, we find that the Neural Network has strong nonlinear approximation ability, extremely suitable for application in deep foundation pit monitoring. The model of the BP network in deep foundation pit monitoring is also established and programmed. And verify the feasibility of the model through a simulate example.(3) Study the theory of the Radial Basis Function Neural Network further. Based on the study of the theory of the General Regression Neural Network, a model of the application of the General Regression Neural Network in the deep foundation pit monitoring is established and programmed. And verify the reliability of the model through a simulate example.(4) Establish the model of the General Regression Neural Network in one foundation pit engineering. And compared with the results of the BP Network model, the AR model and the Stepwise Regression model, we can conclude that the accuracy of the General Regression Neural Network is higher than the Stepwise Regression model, which is slightly higher than the BP Network and the AR model. And the convergence rate of the General Regression Neural Network is faster than the BP Network, meanwhile, its overall nonlinear approximation ability is stronger than the AR model. Therefore, the model of the General Regression Neural Network to monitoring the Foundation-Pit is feasible and reliable.