<p>It is crucial in geomechanics to understand how fracture surface roughness influences shear failure mechanism, fluid flow, and solute transport through rocks. In this regard, performing identical experiments on fractures poses difficulties due to their inherent heterogeneity and anisotropy. It is challenging to replicate natural rock samples with controlled roughness. This study investigates how artificial rock samples can be generated with controlled roughness (joint roughness coefficient (JRC)), using a combination of three-dimensional (3D) printing and mold-and-cast methods for a specific rough surface as an example. A sample with a certain JRC value using fractal theory was used to create synthetic rough fractures. The fractures were then replicated using 3D printing and concrete molds. Profilometry and photogrammetry were both used to evaluate the accuracy and precision of roughness replication. Both methods resulted in the successful replication of rough fractures. However, photogrammetry demonstrated superior accuracy (absolute mean difference of roughness (<i>Ᵹ</i> = 1.1), compared with profilometry (<i>Ᵹ</i> = 2.8)). Due to the limitations of the JRC parameter, various statistical parameters were used to evaluate the accuracy of the simulation of 3D roughness surfaces with laboratory samples. Amplitude-based, spatial-based, and hybrid methods, along with empirical cumulative distribution function plots, were utilized in the statistical analysis, revealing high similarities between the original and replicated samples. This study shows that 3D printing may be implemented to produce artificial rock samples with identical roughness for studying shear failure, fluid flow, and solute transport in rough fractures in the laboratory. Additionally, to assess the joint surface roughness, photogrammetry is more accurate than profilometry, due to the reduction of human error and higher precision.</p>