Disrupting rock engineering concepts: is there such a thing as a rock mass digital twin and are machines capable of learning rock mechanics?

Authors: Elmo, D; Stead, D Download Paper Open access courtesy of:

DOI https://doi.org/10.36487/ACG_repo/2025_34

Cite As:
Elmo, D & Stead, D 2020, 'Disrupting rock engineering concepts: is there such a thing as a rock mass digital twin and are machines capable of learning rock mechanics?', in PM Dight (ed.), Slope Stability 2020: Proceedings of the 2020 International Symposium on Slope Stability in Open Pit Mining and Civil Engineering, Australian Centre for Geomechanics, Perth, pp. 565-576, https://doi.org/10.36487/ACG_repo/2025_34



Abstract:
Introduced by NASA engineers in the early 1960s during the Apollo space program, the term ‘digital twin’ has recently gained more visibility and exposure thanks to the development of the related new concept of the Internet of Things, IoT. A digital twin is not merely a model of a real physical asset but represents the actual connection between the physical world and the virtual (digital) reality. In this context, numerical models of a rock mass are virtual prototypes of that rock mass, which can be used to guide the design process and to estimate possible ground behaviour; they cannot however be considered true rock mass digital twins. Rock engineering differs from other engineering disciplines since often design must be completed prior to developing access to rock exposures, forcing engineers and practitioners to rely on information obtained from severely limited 1D sampling methods. Even if engineers were to have unlimited resources, the natural variability of the rock mass, combined with the limited knowledge of the rock mass in the early phases of a project, would be such that the design outcome would still be influenced by what we do not know rather than by what we effectively know. The authors strongly believe that if initial 1D information to be complemented by 2D sampling during slope excavation, and smart sensors embedded within the rock mass, the opportunity exists to update virtual models and to compare models to the data from the smart sensors on a periodic basis. Note that the smart sensors would have to provide more information than deformation alone, since failure of intact rock bridges in engineered slopes is a progressive damage process and as such, requires location of the source where damage is accumulating. This paper outlines the challenges facing the rock engineering community if we really want to truly transform and improve virtual (digital) geological and geomechanically rock mass models. The authors also provide a critical discussion on the potential use of machine learning algorithms based on empirical methods and the use of qualitative to semi-quantitative scales of measurements that are inappropriate for a full statistical analysis.

Keywords: machine learning, digital twin, rock engineering

References: