Static Elastic Properties of Rocks Obtained by X-ray Microtomography, Inverse Modeling and Surrogate Model

Digital rock analysis uses imaging techniques to obtain information concerning the internal structure of the rock. In the context of obtaining mechanical properties from digital images, much attention is given to determining the elastic parameters of rocks based on their mineralogical composition. While many applications for this type of simulation are available in the literature, the quantification of mineral property variations resulting from external influences has largely been overlooked. This work adopts an inverse modeling approach to estimate such elastic property variations in rocks. This methodology requires a predictive tool and an optimization algorithm to iteratively update the model’s parameters while reducing the discrepancies between modeled and measured data. In this study, a representative sandstone sample from the Botucatu Formation in Brazil was used to evaluate the methodology. Two case studies were considered: a synthetic example was designed for validation of the methodology while the second case applies real-world laboratory test data to estimate the properties that are most supported by the experimental evidence. A MATLAB code was built to integrate a finite element program and a genetic algorithm in a single framework. Additionally, an artificial neural network was used as a surrogate model to reduce the computational time of the numerical forward run. Overall, the results support the robustness of the approach and present a new alternative to obtain the mechanical properties of rock constituents at the pore scale.