Making rock without a sample: petrology-inspired digital twins for reservoir characterization

When core is unavailable for laboratory characterization, rock properties can be constrained using grain- and pore-scale digital twins, provided that the twins faithfully reproduce rock microstructure. This talk presents a workflow for generating three-dimensional microstructure by forward modeling deposition and diagenesis. The resulting digital rocks are used as inputs for physics-based simulations of fluid transport and geomechanical behavior. Across benchmark cases, predictions from simulated twins match the accuracy obtained when physics simulations are based on microCT-derived rock models. Because synthetic twins do not require a physical specimen, the workflow greatly broadens the applicability of digital rock analysis. For example, it makes it possible to reconstruct fluid-transport and mechanical properties through geologic time or to predict rock properties in unsampled regions. Additionally, the approach provides a platform for conducting numerical experiments that quantify sensitivity of rock properties to composition, texture, and diagenetic overprint.

We summarize the petrologic basis for constructing sandstone twins and benchmark results against microstructural observations and measurements from sandstone reservoirs spanning diverse framework compositions and degrees of diagenetic alteration. Beyond sandstones, the workflow generalizes to other clastic materials, including mudrocks and conglomerates, lithologies that have remained challenging for conventional digital rock methods. To illustrate the versatility of the approach, we present a proof-of-concept mudrock twin that undergoes substantial changes in rock properties during progressive burial.