Study on deformation property of soft soil based on neural networks

Based on the triaxial test results of soft soils, an error back propagation network predicting model for deformation property of soft soil is built. Improved BP neural network model is trained by additional momentum term, adaptive learning rate and Bayesian regularization performance function. Research shows that improved BP neural network model applied to predict soft soil foundation settlement, has fast computation, high accuracy, strong generalization ability, and good capability of matching the real data and the measured one. According to test data, the creep models can avoid any artificial assumption of complex constitutive equation, and can reflect nonlinear creep properties of soft soil objectively, thus has better fault-tolerance and more convenient than the traditional method.