Forecasting System In Excavation Ditch Based On Neural Network And FLAC3D Analysis

Excavation deformation prediction of foundation excavation in advance is very important, It enables the design and construction staff to know the health of foundation in advance and provide the basis for support design and construction changes. This paper is to explore the process of building a real-time forecasting system in Excavation ditch. This system can repeated amend the model of Excavation to go to a best model conforms to reality using the kernel FLAC3D.The best model can be used to forecast the future excavation. The ultimate aim is to guidance the excavation construction to prevent foundation pit cutting of the accident. The mainly research of this article include:(1)Establish different finite-difference numeric model based on different soil physical parameters, calculate the different model and get the three types of displacement response of the foundation pit cutting. Take the orthogonal experimental design using different numeric model. Obtain different degree of influence of the soil physical parameters through the direct analysis of the 16 group orthogonal test.(2)To get the soil physical parameters by inversion research using back propagation neural network, Study samples are designed by the 16 group test data values to get the better soil parameters conforms to reality.(3)Get the value of the displacement after next step excavation. Train the network again based on numeric data sample and get the more accurate soil parameters after inversion research.(4)Get a best model conforms to reality after twice inversion research, Forecast the next step excavation using the numeric model and calculate the displacement and the supporting structure internal force of the foundation pit cutting to guidance the excavation construction.