Ice accretion on an aircraft occurs when supercooled droplets in a cloud impinge on the body and threatens navigation safety. Especially, icing on an airfoil reduces the aerodynamic performance by changing its shape. Although the precise icing prediction is necessary, it is difficult to establish the numerical simulation model since the phenomenon is based on a multi-physics nature. In addition, in the glaze ice condition, where the ambient temperature is about −10 to −3 ℃, the impinged droplets do not freeze instantly but flow downstream along the airfoil surface (so-called runback). Furthermore, in supercooled large droplet (SLD) conditions, where the diameter of impinged droplets is over 40 μm, the splashing and the rebounding of droplets generate secondary droplets. These phenomena complicate the ice shape, making prediction challenging. For the simulation of ice accretion, Toba et al. suggested the hybrid method of the grid- and particle-based methods. In this scheme, the flow field and the droplet trajectories are computed using the grid-based method, and the behavior of impinging droplets and the icing process is simulated using the explicit-moving particle simulation (E-MPS) method. This method enables simulations to consider unique phenomena of glaze and SLD icings without empirical icing models based on experimental results. However, there is an operational issue: the computational cost is too high for real-scale conditions. In the present study, we developed an acceleration method of the grid-particle coupling simulation for SLD icing. This scheme was introduced as the subsequent step of the particle-based method in multi-shot simulation. The simulation time of the icing process was shrunk to one-fifteenth, and the ice shape was estimated by extrapolation based on the increasing ice mass ratio. The inner computational particles in the ice layer were deleted for the next computational cycle. The developed method was applied to an SLD icing simulation on the NACA0012 airfoil under a real-scale condition. As a result, the computational time taking for whole icing process was reduced by about 97 % compared to the previous method. The predicted ice shape on the leading edge agreed well with the experimental results. The present method will provide a more practical icing simulation in the actual design process of an aircraft.