Abstract: Coal seam fracture roughness is a key factor affecting gas migration characteristics. Accu⁃ rate characterization of rough fracture structures is of great engineering significance for the optimization of coal mine gas drainage and disaster prevention and control. To clarify the action mechanism of roughness on gas migration，the PerlinN-GF-RA algorithm was adopted to construct three-dimensional stochastic fracture models with controllable roughness，and a mathematical model of roughness is established com⁃ bined with the response surface regression method. The determination coefficient of the model reaches 98.92%，showing favorable fitting accuracy. Numerical simulations are carried out by using COMSOL， and four groups of roughness gradients and three inlet flow velocity working conditions are set to analyze the influence law of fracture roughness on gas flow parameters. The results show that with the increase of roughness，the uniformity of pressure distribution in fractures decreases，the extreme pressure rises from 5.59×1010 Pa to 7.07×1010 Pa，and the fluctuation of pressure head intensifies. The increase of roughness enhances fluid disturbance and deteriorates the stability of gas flow velocity. Under the same flow velocity condition，gas emission capacity is negatively correlated with roughness. When the roughness Ra is 8 and the flow velocity is 4.5 m/s，the peak gas emission capacity is 0.28 m3 /s. The promoting effect of initial flow velocity on gas emission gradually weakens with the increase of roughness. This study clarifies the internal mechanism of fracture roughness affecting gas migration，and can provide theoretical references for mine gas prediction，outburst risk assessment and gas drainage scheme optimization.

Key words: random roughness, gas migration, PerlinN-GF-RA algorithm, fracture structure, numer? ical simulation