Abstract: The rapid development of concrete additive manufacturing provides a new approach for the intelligent construction of hydraulic engineering models. To clarify the variation in the bed surface roughness of hydraulic models produced using different nozzle sizes, in this study, bed surfaces fabricated with four nozzle diameters of 6, 8, 10, and 15 mm were investigated. Uniform open-channel flow experiments were conducted, and three-dimensional laser scanning and planar particle image velocimetry (PIV) were employed to analyze the bed surface geometry and near-wall flow structures. The results reveal that ① the geometric roughness of the bed surface, V, first increases but then decreases with increasing roughness element width b. Moreover, both Manning’s roughness coefficient n and the equivalent sand roughness height k_s first decrease but then increase with increasing b, reaching their minimum values at 10 mm. ② Compared with the bed surfaces fabricated using 6, 8, and 15 mm nozzles, the 10 mm bed surface exhibits relatively large geometric undulations, yet its near-wall high-vorticity activity is restricted in its development toward the outer flow region, resulting in the lowest resistance level. This finding indicates that there is no simple linear correspondence between the geometric roughness V and hydraulic roughness k_s. ③ A roughness coefficient prediction model incorporating the combined effects of k_s and discharge Q was developed, which effectively quantifies the resistance variation characteristics of bed surfaces produced by different additive manufacturing parameters. These findings provide a theoretical basis for optimizing the nozzle size and fabrication parameters in the construction of hydraulic engineering models under target roughness conditions.