According to the configurationally characteristics, a bearing-gear coupling non-linear dynamics model of two-stage split herringbone gear trains used on special equipment will be presented, and this model includes bearing stiffness, mesh stiffness，mesh errors and gear backlash. Then the Runge-Kutta step-by-step integration method will be used to solve the non-linear differential equations of motion of a two-stage split herringbone gear train, and the time travel curve and frequency spectrum of the dynamic meshing force and dynamic bearing force can be obtained involving random mesh stiffness and mesh error excitations, and further, the statistical characteristic of them can also be obtained by Monte Carlo simulation to analyze the influences of the discrete degree of random mesh stiffness and mesh errors on dynamic mesh force and dynamic bearing force. The research results lay a foundation for the dynamic optimization and dynamic reliability optimization of two-stage split herringbone gear trains.