Development and Evaluation of Machine Learning Models for Vortex Tube Sediment Ejector

Abstract

The vortex tube silt ejector (VTSE) is an important hydraulic device for reducing sediment deposits in canals, providing a very effective alternative to manual sediment removal. The complex nature of the silt removal process occurs from the spatially varied flow (SVF) in the channel and the rotational flow within the tube. Conventional models often struggle to accurately predict silt removal efficiency due to these complexities. However, recent advancements in machine learning (ML) present robust alternatives for understanding and modelling such complex hydraulic processes. In this study, we explore the application of various ML models, i.e. Support vector machine (SVM), Random forest (RF), and Random tree (RT) to quantify the efficiency of vortex tube silt ejector using laboratory datasets. Comparative analysis are conducted with conventional models to predict the efficacy of ML based models. The findings of the study reveal that the RT model, exhibiting a Root Mean Square Error ( RMSE) of 2.165 and Nash-Sutcliffe Efficiency (NSE) of 0.98 outperforms the other applied ML models, demonstrating suppier accuracy with fewer errors. Sensitivity analysis focuses on the extraction ratio as a critical parameter in computing vortex tube silt ejector removal efficiency. The outcomes derived from the ML- based moldes presented in this study hold effective implications for hydraulic engineers and researchers involved in assessing the sediment removal efficiency of vortex tube silt ejectors. Nevertheless, to formulate a more universally application models for the comprehensive research, both within the same field and in related areas, may be imperative.