In the framework of EGS, using hydraulic stimulation increases pore pressure into the existing fracture networks and can be responsible for slip reactivations. In order to predict the seismic or aseismic character of these slips, it is important to determine the influence of the gouge characteristics on the slip behaviour. Lab or in-situ testing can be very helpful, but experience has shown that it is complicated to install local sensors inside a mechanical contact and that some data remains impossible to obtain. In this study, we are focusing on the physics of the contact inside a granular gouge to understand the mechanisms of slip triggering by using numerical modelling. We implement a 2D granular fault gouge with Discrete Element Modelling (DEM). The model involves two rough surfaces representing the rock walls separated by a granular gouge with realistic angular particles created by the wear of previous slips. A displacement-driven model with dry contact is primarily studied to observe the peak of static friction (shape, slope or duration). In order to spotlight the aseismic and seismic patterns, a second model is carried out to add the stiffness of the rocks (or loading apparatus). Dedicated post processing tools are also used in order to analyse the spatial distribution of the solid fraction and the force chains patterns bringing a new understanding of physics from a granular point of view. Representing realistic morphology of particles (angular shapes) brings an additional level of control with respect to most of the simulations reported in the literature using circular grains. Angular grains lead to higher friction coefficients with different global behaviours. The initial solid fraction of the gouge appears to be a good indicator of the friction peak we have at a slip triggering (Fig.). The more compacted the gouge is, the more effort is needed to disturb the initial grains assembly and to reach a steady state regime. With fractal size distribution, we have also studied the mechanical contribution of small particles in the media. Friction coefficient results let us conclude that the presence or not of very small particles is not as influent as the initial solid fraction of the gouge. Finally, adding a stiffness on the upper rock wall provides information on the transition from stable sliding to slow-slip/stick-slip behaviour.