The performance of three anion-exchange membranes (AEMs) in the fluoride ions reduction by electrodialysis (ED) is performed on real and synthetic water. The electric potential method measures the potential difference (PD) between two synthetic anion solutions separated by ACS, AFN and AXE membranes. The selectivity of these three AEMs coupled with the membrane CMX, is a cation-exchange membrane (CEM) towards different ions. The removal rate is influenced by the thickness of the polarization layer (PL) which reduces the material transfer and provides an additional barrier. The greater the thickness δ of the PL, the longer the passage time and consequently the removal rate of anions is small. Using the unstirred layer model, δ for each ion will be determined. According to the potential measurement method, none of the tested AEMs are selective to fluoride ions and the order of selectivity is as follows: AFN> AXE> ACS. Best membrane couple selected for fluoride ion removal is ACS/CMX and ion selectivity follows the order: Cl−> NO−3>F−> HCO−3> SO42−. For ACS membrane, both the demineralization rate (DR) and δ of fluoride ions are influenced by the initial concentration of the co-ion according to the following order: NO−3> Cl−> HCO−3> SO2−4.

In this work, we investigated the feasibility of ammonium removal by electrodialysis (ED), a well‐known electro‐membrane process, based on the selective migration of anions and cations through anion exchange membranes (AEMs) and cation exchange membranes (CEMs). ED experiments are performed using a laboratory pilot. The ion exchange membranes (IEMs) pair used is AXE/CMX, AXE as an AEMs and CMX as a CEMs. The first tests are performed with real groundwater solutions from Kenitra city (Morocco), spiked with an initial concentration of 3 mg/L NH₄Cl. The results gave a specific demineralization (SD) to NH₄⁺ ions of 84.60 %; for a demineralization rate (DR) of 80 % and a time of 60 min. The multivariate principal component analysis (PCA) presents a total inertia of 99.23 %, the majority of the variables of which are positively correlated on the C1 axis with a variance of 95.5 % than that of C2 of 3.78 %. The quality of the diluted water determined by the Legrand‐Poirier method showed that the water was aggressive and that the addition of 4.54 mg/L Ca²⁺ was necessary to balance the water and make it fit for human consumption.