Turbidity is critically important in the drinking water treatment process and optimal performance of a sand filter can reduce it significantly. In this research, the optimization of a rapid sand filter was investigated in COMSOL Multiphysics® to decline the output turbidity. The conventional filter changed to a triple-bed filter. For validation of the large-scale model, an experimental model was used. According to 35 experimental data for surface load rate of 120 and 240 m3/m2.day, the average absolute deviation of outlet turbidity was calculated by nearly 7%. Moreover, 243 experiments were considered based on the different height of beds and dimeter of particles. The optimization results showed that the proposed model could remove the turbidity of 10 NTU in the inlet stream to 0.341 NTU in outlet one. Also, when the bed's porosity reaches about 0.2, the bed undergoes a pressure drop, and modeling calculations get slow, which indicates that the bed becomes filled. Therefore, it could be implemented as an index to estimate the filter's valuable working hours or backwash time.