Fracture networks exert a strong influence on fluid flow in the subsurface. We present a 2-D linear elastic finite element model that generates fracture patterns in incremental iterative steps. A subcritical failure criterion is applied to simulate quasi-static multiple crack propagation. We study their impact on fluid flow as a function of fracture density. Fractures are represented by closed polygons. Geomechanical apertures are a by-product of growth and depend on the current stress state. Fracture arrest, closure, and coalescence are handled by a geometric kernel. Traction and cohesion along fracture walls are not taken into account. All patterns are generated assuming plane strain and applying displacement tensile boundary conditions. We assume randomly distributed flaw positions with a uniform probability distribution and Gaussian-distributed flaw lengths. A piecewise fracture permeability is derived from the parallel plate law. We measure the effective permeability, k_eff, and fracture-matrix flux ratio, q_f/q_m across the percolation threshold. Before the percolation threshold we observe an increase in k_eff of up to two orders of magnitude. Models with fixed apertures overpredict k_eff by up to six orders of magnitude, as they disregard variations in the aperture distribution due to fracture interaction. After percolation our model predicts steady linear increase in effective permeability. The q_f/q_m ratio better captures the initial increase in hydraulic conductivity of the system as opposed to the k_eff measurements. Results corroborate that fracture percolation and stress-dependent aperture distribution due to mechanical interactions control the evolution of the k_eff of the system. Results depend on the number of initial flaws used for fracture set growth.