A discontinuous future for numerical modelling in geomechanics?

Abstract

Finite-element or finite-difference methods are commonly used in geomechanics to model the response of soil or rock at an engineering site. These methods assume that the material is a continuum (although known discontinuities can be included explicitly). There are two drawbacks with continuum methods. Firstly, an appropriate stress-strain law for the material may not exist, or the law may be excessively complicated with many obscure parameters. Secondly, the natural development of cracks and rupture surfaces is not well-handled by continuum approaches. It is suggested that the future trend for numerical modelling in soil and rock may consist of the replacement of continuum methods by particle methods. Assemblies of discrete particles (bonded together to represent rock, and unbonded to represent soil) capture the complicated behaviour of actual material with simple assumptions and few parameters at the micro level. Complex overall behaviour arises as an emergent property of the assembly. The application of particle methods to large-scale problems is currently difficult or impossible because of high computational demands. It is shown that such applications should be feasible within ten years, and certainly within 20 years. Examples are given of the simulation of granular material and rock by particle methods.