Gas sweetening is a fundamental step in gas treatment processes. Acid gas loading in alkanolamine solutions is one of the most important and commonly used parameters for monitoring the performance of gas treating units, and therefore should be closely monitored to prevent operational problems, such as excessive energy consumption and corrosion in units. In this article, a new method based an artificial neural network for prediction of rich amine concentration is presented. H2S, H2O, and CO2 mole fractions in sour gas and H2S, H2O, and CO2 diethanolamine mole fractions in lean amine by their flow rates have been input variables of the network and have been set as network output. To check the artificial neural network model, the samples have been divided into three groups. Among the 130 data set, 92 data have been implemented to find the best artificial neural network structure as train data (group 1). Nineteen data have been used to check generalization capability of the trained artificial neural network named validation data (group 2) and 19 data have been used to test optimized network as test data (group 3). The results of this study include the calculation of R value and mean squared error between the experimental data and artificial neural network predictions that show good accuracy of this type of modeling.