Abstract: Considering the transport efficiency, all the spaces including solids and void space need to be described for Monte Carlo transport codes. Since the void space needs to be described accurately, the void space modeling is more complicated than the solids modeling. For system with complex geometry, the void filling may generate too complicated Monte Carlo description for efficient modeling and calculation. Based on the multi-physics coupling analysis modeling program SuperMC/MCAM, an improved void filling algorithm based on convex solids was developed.This algorithm would test all the convex solids generated by decomposing the CAD solids with sub-spaces and generate the void description by complementary describing convex volumes intersected with the sub-spaces. This algorithm was tested with ITER reference neutronics model. The result shows that this algorithm simplifies the description of the void to 1/60, accelerates the void space modeling by 25 times and speeds up calculation by 1.4 times comparing to benchmark algorithm.