Estimation of Particle Froude Number in Deposited Bed Condition Using Hybrid Machine Learning Models

In hydrology, maintaining self-cleaning capabilities in drainage systems is crucial to prevent sediment deposition at the bottom of channels. This deposition can disrupt the hydraulic capacity of the channels. This study accurately estimates the particle Froude number in non-deposition with deposited beds condition. A dataset was collected from various studies and analyzed using hybrid machine learning (ML) models. Specifically, Additive Regression (AR) and Random Committee (RC) models were employed. These ML models utilized Random Forest (RF) and Random Tree (RT) as base regressors, enhancing the estimation performance. The results were evaluated using multiple performance metrics. Through comparisons, it was determined that the proposed AR-RF (MAE = 0.456, NSE = 0.930, RMSE = 0.603, and \(R^2\) = 0.931) ML model outperformed the other models. The AR-RF accurately estimates the particle Froude number, which is important for self-cleaning sewer systems.