Subsidence Displacement Prediction of Underground Engineering on the Basis of FLAC3D and ANN

摘要

Because the problem of estimating the settlement of underground engineering is very complex and not yet entirely understood, traditional methods of settlement prediction of underground engineering are far from accurate and consistent. In this paper, combining FLAC3D based on finite difference method with manual neural network adopted by Visual C++6.0, it constructed the analysis method by connecting direct computation with back analysis. At the same time, in the direct computation, soil parameters are generated by the random method to avoid the disturbance of artificial factors, which reflected the uncertainty of soil inherent property and the prediction capacity of ANN on uncertainty. Comparing the prediction by improved MBP neural network with observed values, the maximum and minimum error is 27.7% and 0.26% respectively, which proved that the scientificity and accuracy of associated application of ANN with FLAC3D on predicting subsidence displacement by underground engineering, and it is supplied that scientific evidence to reduce damages of ground existing structures induced by underground construction.