The work aims to reveal the effects of three-dimensional roughness parameters of sandblasted pretreatment surface on bond strength of thermal spray coating. The effect analysis was carried out on the surfaces of 45# steel specimens, which were pretreated with sandblasting at different distances and speeds as the processing parameters. The three-dimensional topography and roughness parameters of the sandblasted surfaces were obtained by the three-dimensional optical microscope. The bond strength of thermal spraying coatings was tested with the testing methods of "bonding-stretching". Using the regression analysis method, the mathematical model between the bond strength and the three-dimensional roughness parameters of coating was established, and the effect mechanism of the three-dimensional roughness parameters was further analyzed. The results indicated that many irregular peaks and pits were randomly distributed on the sandblasted surface without fixed orientation. The average surface roughness and bonding strength of the coating were 4.84 μm and 32.8 MPa, respectively. The relationship between bond strength and 3D roughness parameters is linked by nonlinear correlation. The effect of three-dimensional roughness parameters Sa, Sdr, Sdq, Sq on the bond strength of the coating is significant. The contact area of coating with the substrate increases with the increasing of Sa and Sdr. The larger the Sdq is, the sharper the pits and peaks of the blasted surface are , which is beneficial to provide more anchor points for mechanical bonding of the coating. The surface topography of the combined interface with larger Sq mainly consists of deep pits, which leads to the formation of residual porosity defects in the depth of the pit and reduces the wetting effect and bonding performance between coating and substrate. Sa, Sdr, Sdq, Sq are the main influencing factors for bond strength of thermal spraying coating, and the regression relationship is nonlinear. The effect mechanism and trend of each three-dimensional roughness parameter on the bond strength of the coating are consistent with the regression mathematical model.