Application of Artificial Neural Networks for the prediction of undrained shear modulus in cohesive soils

The paper presents a study carried out in a Hollow Cylinder Apparatus (HCA) to determine the shear modulus Gu in undrained conditions. Selected results of tests performed for undisturbed cohesive soils are presented. Values of undrained shear modulus Gu have been determined at shear strain of 0.1% and 0.5%. Laboratory tests were performed on slightly overconsolidated clay (Cl) and sandy silty clay (sasiCl) with an overconsolidation ratio OCR of about 3.5 and 2.7, and a plasticity index Ip of 77.6% and 34.7%, respectively. HCA tests were performed with anisotropic consolidation and shearing in undrained conditions. Results of laboratory tests have allowed to assess the influence of the principal stress rotation on the values of undrained shear modulus Gu using Artificial Neural Networks (ANNs).