In this paper, an optimal artificial neural network determined by a self-adaptive differential evolution approach is applied to model and optimize the removal of copper from wastewater by an ion-exchange process. The Purolite S930 1 resin with iminodiacetic group was used in batch mode for Cu(II) removal from synthetic aqueous solutions in different working conditions (initial solution pH, stirring rate, initial concentration of copper, temperature, contact time and resin amount). The obtained results indicated that the used methodology was able to provide good models for the studied process, the mean squared error in the testing phase obtained by the best network being 0.0034. In addition, the optimal combination of parameters leading to the maximization of removal efficiency determined with the proposed approach was experimentally validated, the prediction being in correlation with the observed data.