Answers are grounded in this paper and cite their sources.
Purpose Plurigaussian simulation (PGS) is a powerful approach in generating discrete stochastic structures of complex multiphase systems across a range of scales. Originating as an extension of truncated Gaussian simulation, PGS allows for the detailed representation of phase relationships and connectivity. A key element of PGS is the identification of truncation rules to dictate the geometric and facies relationships and traditionally involves a trial-and-error approach based on the physical characteristics of the modelled structures. The purpose of this study is to present an approach to learning such rules. Design/methodology/approach This study introduces the use of generative adversarial networks, specifically using the Wasserstein generative adversarial network with gradient penalty architecture, to automate and enhance the selection of truncation rules through a neural truncation rule. Findings By learning directly from the characteristics of the training data, this approach aims to bypass the limitations of manual truncation rule selection, offering a more efficient and precise method to replicate the desired structural features. Originality/value This study highlights the efficacy of generative machine learning in identifying suitable truncation rules for PGS, significantly reducing the uncertainty in truncation rule choice and enabling the accurate realisation of complex structural characteristics without the ill-informed trial-and-error process traditionally undertaken.
No takes yet. Share an insight, caveat, or question.
Ricketts et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: