Abstract: [Objectives] Aiming at the nonlinearity of trimaran rolling, a system identification modeling method combining CFD and support vector regression (SVR) is proposed. [Methods] Based on the STAR-CCM+ platform, numerical simulations of prescribed rolling motion of the trimarans with different lateral layouts are performed. Then the SVR method is used to identify the parameters of the hydrodynamic moment history curve and calculate the additional mass and damping coefficient of the hull at different frequencies. [Results] The results show that the damping coefficient of the trimaran shows strong frequency dependence; the nonlinear characteristics of the damping are profound at low frequencies, and the damping component of the bilge keel accounts for a large proportion. [Conclusions] Compared with potential theory, the SVR method presented in this paper can accurately capture the nonlinear characteristics of the trimaran's rolling motion, and take the interaction of the flow among hulls into account.