Abstract: Visual fluid simulation is an important research direction in computer graphics. Among them, free surface fluid based on the lattice Boltzmann method (LBM) has attracted widespread attention for its accuracy and efficiency. In view of the problems of existing methods in stability, fluid detail capture, time efficiency improvement and boundary processing, this paper first proposes a free surface LBM fluid simulation method based on a multi-relaxation model of adaptive viscosity, which can capture more fluid details than the single relaxation model; secondly, the problem of excessive time step reduction in the existing adaptive time step method is optimized, and a flexible and efficient time step scaling strategy is proposed to improve the computational efficiency; finally, in order to solve the problem of fluid attachment on the surface of moving obstacles, the free slip boundary conditions and free surface normal calculation method are improved to improve the accuracy of fluid-solid interaction. Experimental results show that the proposed multi-relaxation model based on adaptive viscosity can maintain a volume conservation rate of more than 90%; the proposed adaptive time step optimization method improves the efficiency by 7.25%~25.85% compared with the comparison method without changing the accuracy; the improved boundary conditions make the interaction between obstacles and fluids more in line with physical laws. In summary, The free surface fluid simulation method proposed in this paper has advantages in accuracy and computational efficiency compared with the comparative method, which has positive significance for promoting the visualization simulation application of LBM.