Mechanical air entrapment, void compression and void dissolution are the main mechanisms behind void formation in liquid composite molding. Mechanical air entrapment is induced by the highly non-uniform geometry at the meso- and micro-scale and the heterogeneous properties of the reinforcing material. This results in differences in the velocity profile through the porous- and free flow domains, which can lead to air being entrapped in slow-flowing domains, by resin flowing through the faster flowing domains. Depending on the competition between the viscous flow through the free flow domains and the capillary flow through the porous domains, voids occur either in the intra-bundle domain or in the inter-bundle domain, where the competition between the two flow types can be quantified by the capillary number. Once the voids are entrapped, they can change in size due to void compression and -dissolution, all which occur on considerably different timescales. An experimental set-up was developed which uses fast radiation curing to almost instantly cure the resin during injection and thus detach the different stages of void formation from each other. In this research the relationship between the mesoscale structure of a woven fabric and mesoscale void formation was investigated. The mesoscale structure is strongly related to void formation, as viscous flow is related to the inter-bundle domain size in the through-thickness direction of the preform and capillary flow is related to the mesoscale bundle porosity. As the bundle porosity is related to both the bundle permeability, and the capillary pressure, its influences on void formation were considered to be large, and thus bundle porosity became the main research topic. A 2D semi-analytical model based on mechanical air entrapment was constructed which was capable of determining the filling times of the intra- and inter-bundle domains on the mesoscale. A parameter called the competitive number was introduced which was related to these filling times. Once the competitive number was larger than 1, spherical inter-bundle voids should be formed, once it was equal to 1 no voids should be formed and once it was lower than 1 ellipsoidal intra-bundle voids should be formed. This model was validated experimentally. Flow behavior predicted by the model was compared to video footage of the flow front propagation at the surface of the preform at the macro- and mesoscale during multiple injections. Void volumes, types and locations obtained from the model were compared to voids observed in micro-CT scanned samples and were in agreement with each other. The validated model was eventually used to investigate the effects of bundle porosity on void formation. It was found that increased bundle porosity at a constant global preform porosity, leads to increased intra-bundle flow and decreased viscous flow, thus considerably influencing void formation by mechanical air entrapment. In short injection cycles with high macroscopic flow velocities, which can lead to intra-bundle voids, it could be beneficial to switch a fabric with a higher bundle porosity, as that would effectively reduce the intra-bundle void size. For injections with low macroscopic flow velocities, the opposite would apply, thus usage of fabrics with lower bundle porosities could help to reduce the inter-bundle void volumes.