GγSANDnet: A neural network tool for prediction of shear stiffness (G) shear strain (γ) relationship for sands

Abstract

The relationship between shear stiffness modulus and shear strain is a key geotechnical parameter required for prediction of in-service performance of built infrastructure. This software utilises and rapidly trains an Artificial Neural Network (ANN) to generate a shear stiffness degradation curve from basic soil classification data (such as particle size and grain density) for any number and combination of input parameters. This enables geotechnical engineering practitioners or researchers to produce a stiffness degradation curve suitable for design or further modelling on sites with missing or low-quality site investigation data. The software was implemented in MATLAB using the Deep Learning toolbox and is available on Zenodo and GitHub. Also included is the required training dataset; alternatively, users can add their own data through the software's GUI.