Accurate prediction of the roughness coefficient of sediment-containing drainage pipes can help engineers optimize urban drainage systems. In this paper, the variation of the roughness coefficient of circular drainage pipes containing different thicknesses of sediments under different flows and slopes was studied by experimental measurements. Back Propagation Neural Network (BPNN) and Genetic Algorithm-Back Propagation Neural Network (GA-BPNN) were used to predict the roughness coefficient. To explore the potential of artificial neural networks to predict the roughness coefficient, a formula based on drag segmentation was established to calculate the roughness coefficient. The results show that the variation trend of the roughness coefficient with flow, hydraulic radius, and Reynolds number is consistent. With the increase of the three parameters, the roughness coefficient decreases overall. Compared to the traditional empirical formula, the BPNN model and the GA-BPNN model increased the determination factors in the testing stage by 3.47 and 3.99%, respectively, and reduced the mean absolute errors by 41.18 and 47.06%, respectively. The study provides an intelligent method for accurate prediction of sediment-containing drainage pipes roughness coefficient.