Fracture Roughness Characterization from 360 Unrolled Core Images in a Sandstone Reservoir, Case Study, Algeria, Hassi Messaoud

ABSTRACT Various existing indirect measurement methods and techniques in the petroleum industry are used to estimate some of the fracture parameters, such as sonic and electrical image logs. However, none of them can estimate the fracture roughness and the degree of mismatch between the two opposite fracture sides due to the sampling resolution. These two are important parameters to consider when dealing with geomechanics and fluid flow in tight naturally fractured reservoirs because they are important in quantifying shear strength and permeability. A quantification and detailed characterization of the joint roughness coefficient (JRC) and the degree of mismatch has been done on fractures gathered in a total of 23 wells in the Algerian Cambro -Ordovician oil-bearing formation Hassi Messaoud which consists of sandstone, clay, and some volcanic intrusions. These data have been collected from the digitalization of high-resolution unrolled 360-core photographs at the reservoir level. The results have been correlated with rock type, grain size, and fracture dip angle. Fracture roughness is ranging between 5 (low) to 19 (high) on the JRC scale and can be predicted from analysis of the general reservoir rock types (sandstone, eruptive, and clay). The results are applicable to other, especially siliciclastic reservoir lithologies, and will therefore contribute to defining a more accurate aperture and permeability estimation in building reliable Discrete Fracture Network (DFN) models. INTRODUCTION Fracture roughness is a key parameter in characterizing the permeability and fluid flow in petroleum (Camac et al., 2006) and geothermal systems (Porlles & Jabbari, 2022). As the majority of currently used large-scale models still significantly rely on simplified smooth parallel-plate models and its related models for naturally fractured reservoirs with rough walls, it is crucial to quantify the roughness in order to enhance the performance of these models (Bodin et al., 2007).