Prediction of stress-strain curves of remolded soils in undrained shear test

Through the consolidated undrained triaxial shear testing of remolded soils with different water contents, the normalized features of stress-strain curves were studied, with consideration of two standard normalized factors: the confining pressure and the ultimate value of principle stress difference. Prediction models of stress-strain curves were established using the hyperbolic equation and the BP neural network, respectively. Comparison of the prediction results of the two models shows that the prediction model based on the BP neural network can take into account the influences of the initial water content and confining pressure and has better fitting and prediction accuracy than that based on the hyperbolic equation.