Abstract

The evaluation of geological resources in deposits requires 3D models that allow to estimate the law, volume or tonnage. Geostatistics, the main methodology for evaluating reservoirs, uses spatial location and distances between data to deliver estimation and uncertainty measurements of reserves. Due to the presence of folds and faults generated after the mineralization, the spatial continuity of the data is altered and the geostatistics loses the usefulness, so it is proposed to implement a structural restoration previous to the geostatistical analysis. In order to perform these restorations there are several methodologies, which are based mainly on stress calculations, mechanical equations, geometric principles and the geologist can incorporate a series of interpretations that limit the deformations, thus converging to a valid solution. This paper presents the geological validation of a new unfolding methodology based on mechanical learning, combinatorial optimization and data science, which is capable of delivering geological restorations with a minimum of interpretations entered by the expert. This methodology is validated through the restoration of synthetic cases. These cases are based on methods such as simple shear, fault-parallel-flow and trishear, which allow to construct rollover anticlines and fault-propagation folds. The results obtained by this methodology are validated in the Move2017.1 geological restoration software, widely used in academia and industry. Keywords: Unfolding, Modelos Mecánicos, Geoestadística, Optimización.