Background: Oil pollution is one of the serious environmental problems that can cause irreparable environmental damages. Petrochemicals are almost unavoidable. In this study, the results of phenol removal by activated carbon adsorbent using perceptron neural networks with the variable number of neurons -1 to 20 neurons- were estimated and predicted. In this network, the data of initial concentrations like time, adsorbent weight, pH and temperature being reported in the background section of the study were selected as input variables, while the phenol removal efficiency values were considered as the network`s outputs. Methodology: The laboratory collected 79 data from different sources. In order to develop and validate this extended model, the entire database was randomly divided into three types: 70% (55 data points), 15% (12 data points) and 15% (12 data points) were respectively used as training, validation, and test. In order to improve the performance of the artificial neural network, input values and target values were normalized in the range of -1 to 1. Results: The results of the prediction showed that the developed artificial neural network model provides an accurate prediction of the removal efficiency values with a total correlation coefficient of 0.8986 and a total squared error of 0.0002788. Conclusion: The study of the ability of the artificially trained neural network to predict the effect of input variables in the artificial neural network on the removal efficiency values confirmed the good performance of the developed artificial neural network model.
