PREDICTIVE MODEL FOR NORMALIZED SHEAR MODULUS OF COHESIVE SOILS

摘要

Evaluating dynamic properties of geomaterials is an essential step for solving geotechnical earthquake engineering problems. The shear stiffness-reduction curves of soils are commonly presented in normalized form and have many applications in equivalent-linear and nonlinear dynamic analyses. In this study, a radial basis function (RBF) neural network model was developed to predict normalized shear modulus of cohesive soils. The most important factors that affect this parameter include effective confining pressure, shear strain, and plasticity index. The comprehensive database used for the development of the model was obtained from previously published experimental results. Validation of the model performance was carried out by using centrifuge tests results. A parametric analysis was then performed to evaluate sensitivity of the proposed model to variations of the influencing parameters. The results indicate that the neural network model could provide predictions more accurate than those obtained by the previous models.