Abstract The paper focuses on spatial modelling of composites with discontinuous reinforcement. The algorithm for creating a representative volume element (RVE) must consider random distribution and size of reinforcing particles (RP), prevention of RP interpenetration, and maintaining the desired volume fraction of the reinforcing phase (Vp) in the composite microstructure. Assuming fixed RVE dimensions and randomly determined RP size, the actual Vp value needs to be continuously determined. If the assumed (desired) Vp is lower than the current value, additional reinforcement is added to the RVE. As the RP location is random, some particles may extend beyond the RVE limits, affecting Vp calculation. The research aims to determine the RP volume within the RVE boundaries when RP extends outside. The RVE was discretized with N points, and the number of Ni points within the area occupied by RP was determined. The sought value was calculated using the ratio Ni /N = Vp /VRVE, where VRVE, is the volume of the RVE. Two discretisation methods, systematised (RI) and random (Monte Carlo (MC)), were employed. The study investigated the effects of discretisation type and number N points on calculation accuracy and microstructure generation time for particle-reinforced composites in sphere, cylinder, and ellipsoid shapes. Systematised discretisation yielded higher accuracy/stability, with number N dependent on RP dimensions. The MC method reduced generation time but introduced instability and significant errors.