We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Modeling of fluoride removal by nanofiltration: coupled film theory model with Nernst–Planck equation and artificial neural network.
- Authors
Addar, F. Z.; Fahid, B.; Belfaquir, M.; Tahaikt, M.; Elmidaoui, A.; Taky, M.
- Abstract
Fluoride ions are necessary and beneficial for human beings, however high levels of fluoride in groundwater can be toxic and harmful to human health and water resources, causing serious problems on teeth and bones. This situation is widespread in many parts of the world, including Morocco. According to World Health Organization (WHO), the maximum allowable concentration of fluoride in drinking water is 1.5 mg/L. The objective of this work is to understand the mechanism of transfer of fluoride ions in nanofiltration (NF) membranes. Two methodologies have been used, one mathematical, Nernst–Planck equation combined with film theory (NP-FT), and the other statistical artificial neural network (ANN). This study has been realized with experimental data obtained from three NF membranes (TR60, NF270 and NF90) that have been used for the removal of fluoride from NaF-doped groundwater with different initial concentrations of 5, 10, 15 and 20 mg/L. In a first part, the NP-FT model that is proposed as a contribution to the modeling of the concentration polarization phenomenon and to study the influence of the initial concentration (Ci) of fluoride on the reflection coefficient (σ), the permeability of the solute (Ps) and the thickness of the boundary layer (δ) has been made. In a second part, the two models ANN and NP-FT are used to predict the fluoride rejection as a function of the permeate flux. In addition, a comparison is made in terms of coefficient of determination R² between the two models. The results obtained showed that the influence of concentration bias is more important for NF270 than for TR60 and NF90, and high reflection coefficients, almost equal to unity, were obtained for NF90, indicating that convective transport is almost completely hindered. Similarly, reflection coefficients of about 0.97 and 0.89 were found for TR60 and NF270, respectively, suggesting that both mechanisms are present, but that diffusion is more pronounced for NF270 than for TR60. Thus, the thickness of the boundary layer and permeability change inversely to one another. It is about 10–5 m for NF270 and 10–6 m for TR60. On the other hand, the value obtained for the NF90 membrane is 10–14 m which is very low. The obtained solute permeability follows the order: R60 > NF270 > NF90. Although the experimental data agree well with the data predicted by ANN and NP-FT, which indicates that both models are appropriate for the prediction and satisfactory R². However, the ANN model reveals a slight superiority over the NP-FT model.
- Subjects
MOROCCO; GROUNDWATER purification; WORLD Health Organization; NERNST-Planck equation; ARTIFICIAL neural networks; FILM theory; NANOFILTRATION; MODEL theory; WATER filtration
- Publication
Desalination & Water Treatment, 2022, Vol 257, p76
- ISSN
1944-3994
- Publication type
Article
- DOI
10.5004/dwt.2022.28398