We implemented the object-based method using marked point processes to generate a natural fracture network honoring an assumed fracture characteristics’ distribution where the fractures are two-dimensional zero-thickness circular disks. Fractures are divided into two groups based on their alignment which is acquired by Monte Carlo sampling from two Gaussian distributions with 90-degree shift in the mean value; this hypothesis is validated considering the frequently observed checker-board fracture patterns in the outcrops. The growth of the fractures in the second group or the secondary (daughter) fractures can be terminated by a criterion derived from the distribution of the fractures in the first group or the primary (parent) fractures. For data assimilation purposes, a smooth seismic distribution for fracture density is mimicked by simple krigging which inherently possesses a smoothing nature. Then, the generated seismic data is honored by revising the fracture distribution such that in areas with less fracture density we have fewer fractures. This work provides a novel, yet easy and fast workflow to stochastically model a natural fracture network following the attributes offered by seismic data and concludes an orthogonal or bidirectional fracture pattern. This pattern can be easily extended for multi-directional fracture patterns using the proposed framework.
One of the big concerns in fractured reservoirs characterization is the representation of the subsurface fractures due to large uncertainty and extremely limited direct measurements pertaining to exact spatial distribution of fractures. Stochastic approaches allow us to realize fractures discretely . This approach embraces three different methods, object based simulation, hierarchical fracture modeling, and multiple point statistics based algorithms . The current work concentrates on the first method.