Abstract: Back-propagation (BP) neural network was used to establish a model to predict the removal rate of polycyclic aromatic hydrocarbons (PAHs) in aromatic oil during hydrogenation process. The effects of reaction time, temperature, and pressure on removing PAHs were investigated respectively with single-factor experimental method. Then, the obtained experimental results were used to build a training network by matlab software. The trained BP neural network was adopted to predict the isolated points existed among levels of single-factor experiments. Finally, the optimum process conditions of aromatic oil hydrogenation identified by the experimental verification were determined as follows: a reaction time of 7 h, a reaction temperature of 279 ℃ and reaction pressure of 9 MPa. Under such conditions, the PAHs removal rate reached 47.89%. Moreover, the influences of catalyst to feed ratio and solvent involved were studied based on the optimized conditions. Results showed that the PAHs removal rate could be 51.02% with catalyst to feed ratio of 0.3, and the PAHs removal rate further increased to 54.29% when adding toluene as solvent.