Aiming at the low reliability of blade rolling mill transmission system，comprehensively considered the randomness of the system physical parameters，the geometric dimensions parameters，and rolling force，nonlinear dynamic model of blade rolling mill transmission system is established. The system dynamic responses were obtained by the variable stepsize Runge-Kutta method. The dynamic reliability model of blade rolling mill transmission system was set up by using the theory of probability fatigue damage cumulative. The effect of random variation of parameters on the dynamic reliability of blade rolling mill transmission system and gear parts were calculated and compared with the results of Monte-Carlo method. The feasibility of this method is verified, and the research results provide theoretical basis for reliability optimization design of blade rolling mill gear transmission system.