The Nernst-Planck equation describes ionic movements in saturated porous materials subject to chemical concentration gradients in the pore fluid and applied electric fields. When applied to a macroscopic system the equation supposes that the flux of each ion is independent of every other one. However, owing to the ionic exchange among different or similar species, there are ionic potentials that affect the final flux. These will distort the applied electric field and thus keep the electroneutrality of the sum of all the ionic species involved. The applied potential will fall linearly across the sample but an additional 'membrane' potential will change with position and time. This paper summarises a theoretical and experimental investigation into the application of the non-linear electric membrane potential to the simulation of the migration of chlorides in concrete. A new electrochemical test has been developed and carried out with different samples of concrete blended with pulverised fuel ash (PFA) and ground granulated blastfurnace slag (GGBS). During the tests the transient current and the electrical membrane potential were measured. A numerical model developed previously using the classical equations, but including changes in the concrete membrane potential, was optimised using an artificial neural network (ANN). Based on the experimental results and the simulations, the intrinsic diffusion coefficients of chloride, hydroxide, sodium and potassium were obtained. Also, the initial hydroxide composition of the pore solution, the porosity, and the chloride binding capacity were determined. The results showed good agreement with the theory and can help to explain the complex phenomena that occur during a concrete migration test.