Title: Predicting aluminium using full-scale data of a conventional water treatment plant on Orontes River by ANN, GEP, and DT

Authors: Ruba Dahham Alsaeed; Bassam Alaji; Mazen Ibrahim

Addresses: Faculty of Engineering, Al-Wataniya Private University, Hama, Syria ' Faculty of Civil Engineering, Department of Sanitary and Environmental Engineering, Damascus University, Damascus, Syria ' Faculty of Civil Engineering, Department of Engineering Management and Construction, Damascus University, Damascus, Syria

Abstract: Aluminium sulphate is one of the most common chemicals used to coagulate water. Some studies indicate that it can increase the risk of Alzheimer's disease. This study focused on the relationship between residual aluminium and many parameters. The actual data of Al-Qusayr purification plant in Homs city was used. Three different models were studied, artificial neural networks (ANN), genetic expression technology (GEP) and Decision Tree (DT), to determine the residual aluminium. The models' results were compared. ANN was the best in modelling data when initial turbidity was between 6.5 and 30 NTU, decision tree was better in the range 25 to 60 NTU. In general the best model was ANN, while the most easily generalised one was GEP. The ANN model was found to be the most suitable model to predict residual aluminium with a coefficient of determination R² = 0.88 and RMSE = 0.019 mg/L.

Keywords: aluminium residual; artificial neural networks; gene expression; decision tree; turbidity.

DOI: 10.1504/IJW.2023.133991

International Journal of Water, 2023 Vol.15 No.3, pp.190 - 206

Received: 12 Apr 2022
Accepted: 04 Apr 2023
Published online: 09 Oct 2023 *