Soft Computing Method for Settling Velocity Prediction of Fine Sediment in Retention Structure

Most of the retention structures were polluted by various pollutants, particularly fine sediment carried by rainwater due to erosion. Fine sediment is the main cause of siltation which may cause numerous health and environmental problems. To date, the study on fine sediment is limited due to the technology constraint. Therefore, there is a need to formulate a mathematical model which is able to provide an acceptable level of accuracy for the settling velocity prediction of fine sediment. In this study, Radial Basis Function Network (RBFN) and Gradient Boosted Trees (GBT) models were developed and trained by using the experimental data from Particle Image Velocimetry (PIV) tests. Flow rate, particle sizes, vertical displacement and maximum depth were considered as the input while the settling velocity of fine sediment was kept as the output. The developed models were evaluated using a series of statistical analyses. For RBFN and GBT, model VI and model V respectively, has achieved the best performance in terms of coefficient of determination, mean absolute error, and root mean squared error. The findings show that GBT is more suitable than RBFN for the settling velocity prediction of fine sediment with a R2 value of 0.9591, MAE value of 0.000177 and RMSE value of 1.03E-05.