Prediction analysis of strata deformation by subway engineering based on artificial intelligence theory

Because the complication of subway engineering and uncertainty of influencing factors, traditional prediction methods of strata deformation by subway engineering are far from accuracy. In this paper, Visual C++6.0 is adopted to combine FLAC3D based on finite difference method with artificial neural network, to construct the artificial analysis method connecting normal analysis with back analysis. In normal analysis, soil parameters are generated by the random method considering the uncertainty of soil inherent properties, which also validates the prediction capacity of ANN on uncertainty. Comparing the predicting vertical and horizontal displacement by improved back analysis neural network (MBP) with the corresponding observed values, this paper concluded that the maximum and minimum error of vertical displacement is 9.75% and 0.18% respectively, and those of horizontal displacement is 8.92% and 0.08% respectively, which proved the scientificity and accuracy of the artificial intelligence theory applied on predicting strata deformation by construction of subway engineering.