Data-Driven Modeling in Geomechanics

Abstract

The framework of data-driven computational mechanics offers a novel avenue to solve problems in geomechanics, including challenging ones that involve failure and localized deformation. Free from the uncertainty of the classical constitutive modeling approach and the caveats of machine learning models, the data-driven formulation offers an alternative paradigm for computation. This chapter reviews the framework for the case of simple and non-simple (polar), elastic and inelastic media, which represent common descriptions for geomaterials. It discusses data mining from experiments and high-fidelity lower scale simulations, while highlighting remedies for data scarcity (adaptive data sampling). The chapter presents the representative examples of a flat punch indentation and a rupture through a soil layer. It also provides a link to open-source Python code.