Abstract: The traditional Group Contribution method and Semi-empirical Correlation do not consider the interaction of group bonds and lack of calculation methods for the critical properties of heavy distillates in direct coal liquefaction oil(DCLO).So Artificial Neural Network and Group Bond Contribution Coupled Model(ANN-GBC)is constructed to predict critical properties of DCLO.Three layer network structure is adopted,with 45 group bonds and atmospheric boiling point of DCLO,a total of 46 as the number of neurons in the input layer of ANN-GBC Model,the optimum number of neurons in the hidden layer is determined to be 40 by the difference method,and the critical properties is used as the output layer.The correlation between the critical properties of 15 narrow fractions of DCLO and their molecular structure are studied.The critical properties of 20 model compounds are checked and validated by ANN-GBC Model,the relative error between the calculated value and the theoretical deviation is below 2.5%,and the correlation coefficient is 0.999 69.The results show that ANN-GBC Model has good function of simulating and accurately identifying the isomer.The critical temperature and critical volume of DCLO increase with the increase of distillation cutting temperature,and the critical pressure increase firstly and then decreased with the increase of cutting temperature.The results of model prediction and semi empirical correlation are basically the same,and the content difference of different components in each narrow fractions leads to the jump of individual results.The ANN-GBC Model not only reveals the quantitative relationship between the critical properties of DCLO and the molecular structure,but also provides a new and effective method for the prediction of the critical properties of other complex systems.