Neural Network Prediction Of Ground Deformation Caused By Urban Underground Engineering

The ground deformation induced by urban underground engineering is being paid more and more attention in engineering field. The dominating reason is that underground engineering construction by any method inevitably results in stratum movement, especially ground deformation, which will menace the security of underground pipelines and adjacent buildings if the deformation is very large.This paper is based on shield tunnel and excavation. By mechanics analysis of ground deformation, this paper makes a systematic research on the affecting factors of ground deformation of shield tunnel and excavation and establishes a predicted model of ground deformation based on neural network. This model is applied to predict actual ground deformation caused by shield tunnel and excavation in Shanghai, Guangzhou, Nanjing etc. and the predicted result is satisfactory. Finally the hierarchical analysis method based on neural network is applied to analyze the sensitivity of the influential factors. The main content of the thesis is as follows:1. Owing to the various influential factors of ground deformation and the synthesis effect of these influence factors, to consider the ground deformation influenced by many kinds of factors, a predicted model of ground deformation based on neural network is adopted. From point of view of engineering application, this paper discusses design of ANN, processing of data, training of ANN and evaluation of ANN performance etc. Based on the basic arithmetic of BP and Visual C++6.0 procedure, a program is compiled.2. By the analysis of the mechanics of soil disturbance and the ground deformation and the influence factors of the ground deformation of shield tunnel and excavation, main factors of ground deformation is confirmed. These factors consider the ground condition, physical parameter and construction technology.3. A predicted model based on neural network is developed to predict ground deformation caused by shield tunnel. The inputs of this model include 9 influencing factors, such as shield thrust, overburden thickness and soil properties etc and the outputs include maximum settlement and inflection point. The monitored data of three sites are adopted to test the model and the predicted result is satisfactory.4. A predicted model is established to predict ground deformation caused by excavation based on neural network. There are 10 input factors in this model including pile diameter (wall thickness), excavation deep and soil intensity etc. and theoutputs are maximum settlement behind wall and influencing range. The result of actual analysis in many sites indicates that it is feasible to predict the ground deformation behind wall of excavation using the model developed in this paper.5. The hierarchical analysis theory is introduced in brief, and a program is compiled based on Visual C++6.0 procedure. The hierarchical analysis theory is used to analyze the sensitivity of the input parameters of the neural network model of ground deformation of shield tunnel and excavation.