The formation of water columns inside the gas diffusion layer (GDL) of the proton exchange membrane fuel cell (PEMFC), which is harmful phenomenon, can be controlled by the GDL's microstructure and material. Using computational fluid dynamics (CFD), a three-dimensional model is developed to monitor the impacts of the GDL's porosity and permeability on the maximum GDL liquid removal. In this regard, twenty-four different cases are simulated at the GDL contact angle of 110 degrees. Results indicate that higher permeabilities and porosities improve the GDL liquid removal and the performance of the system. Obtaining the simulation data, an artificial neural network (ANN) model is trained at the current density of 0.41 A/ cm(2) and the voltage of 0.6 V to predict the maximum GDL liquid removal in 300000 points and to perform the optimization. The ANN model is trained with four neurons with the respective mean squared error values 6.32422e-6, 1.00637e-5, and 4.12086e-6 for the training, validation, and testing, which approves the accuracy of the model. Using a fitted curve and the ANN model, the optimum values of the porosity and the permeability are computed to be 0.9 and 1.481e-11 (m(2)), respectively, to reach the maximum GDL liquid removal of 0.373 (kg/m(3)s). (c) 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license