The fibrous filter with random structure has been established based on the Matlab program and the Journal file in preprocessing software, Gambit. The gas-solid two phases flow characteristics in four fibrous filters with random structure are simulated by computational fluid dynamics (CFD) technology, and the predicted results are compared with the values of the empirical models. The results indicate that utilization of the modeling method proposed can obtain a filter geometry that is similar to the actual filter. The predicted values of the dropped pressure are in excellent agreement with the experimental correlation, and the error are less than 2%. Meanwhile, the predicted values of the collection efficiencies show a similar tendency as theoretical model, and the capture mechanism changes with different particle sizes. When the particle size is smaller than 0.5 μm, Brownian diffusion plays a significant role in the filtration process, while the inertial impaction becomes an important capture mechanism when the particle size is greater than 1.5 μm. When the particle size is in the range from 0.5 to 1.5 μm, the Brownian diffusion and the inertial impaction are both relatively weak. Additionally, the fiber diameter and the solid volume fraction (SVF) are important geometry parameters in the filter. The collection efficiencies of geometry 1(in which the fiber diameter and SVF are decreased along the gas flow) and geometry 3 (in which the fiber diameter and SVF are increased along the gas flow) are higher than those of geometry 2 (in which the fiber diameter is decreased and SVF is increased along the gas flow) and geometry 4 (in which the fiber diameter is increased and SVF is decreased along the gas flow), and for the pressure drop, the results are just the contrary. Meanwhile, the four geometries have different collection efficiencies for different particle size. For large particles, the collection efficiency of geometry 1 is higher than that of geometry 3. While for submicron particles, the result is the opposite. As for all types of particles, the collection efficiency of geometry 4 is higher than that of geometry 2.