Reliable Prediction of the Aeration Efficiency of Venturi Flumes Using Intelligent Approaches

Abstract

This research utilizes experimental data derived from various Venturi flume setups to assess the Venturi flume aeration efficiency (VFAE20). This evaluation encompasses dimensional factors (such as discharge per unit width denoted as q, throat width as W, throat length as F, flume length as L, flume height as H, and submergence ratio as S) as well as nondimensional parameters including the ratio of throat length to width denoted as F/W, the ratio between flume length and throat width as L/W, the ratio of flume height to throat width as H/W, the reciprocal of the submergence ratio as 1/S, and functions incorporating the Reynolds number R, Hb gauge, and hydraulic radius R represented as RHb/R, and another function incorporates the Froude number F, Hb gauge, and throat width W represented as [(F23/W)Hb]3. The study compares empirical relations from multiple nonlinear regression (MNLR) and multiple linear regression (MLR) with proposed artificial intelligence (AI) models, including neural networks (NN), deep neural networks (DNN), extreme learning machines (ELM), gradient-boosting machines (GBMs), and neuro-fuzzy systems (NFS). Using performance metrics and graphical evaluators, GBM consistently outperforms all models (dimensional and nondimensional data), followed by NFS_Tri. Despite minor variances, all the suggested machine learning (ML) models exhibit commendable performance. Uncertainty analysis identifies GBM as the top performer, closely followed by NFS_Tri while existing relations perform poorly. Analysis of variance (ANOVA) results indicate insignificant disparities between experimental and predicted VFAE20 values across all proposed AI models but notable differences within existing relations in dimensional and nondimensional data sets. Sensitivity analysis highlights (q) and (RHb/R) as the most influential factors affecting VFAE20 in dimensional and nondimensional data sets, respectively, supported by correlation diagrams and Shapley values. Further, we also investigated how the flumes performed in terms of aeration discharge and water flow characteristics. Venturi flumes find diverse applications in environmental engineering, aquaculture, hydroelectric power generation, and irrigation systems, providing precise measurement and regulation of water flow rate, oxygen levels, and air infusion. Their effective aeration capabilities bolster treatment efficacy in wastewater facilities, foster the growth of aquatic organisms in aquaculture settings, optimize turbine functionality in hydroelectric plants, and mitigate clogging in irrigation setups. This versatility renders Venturi flumes indispensable across multiple domains where precise water flow and oxygen control are essential.