Abstract: The cokes with Xinjiang coking coals present high activity and poor quality， which can not be used in blast furnace. In order to decrease the coke reactivity index (CRI) and increase the coke strength after reaction (CSR)， a series of coke-making experiments using modified Xinjiang coals were carried out. By the thermo-gravimetric analysis on the blending coals， it is found that the modifying additive is primarily acted on the process of the formation of colloids and the polycondensation of semi-coke. The results show that the oxygenic groups of QM are mainly hydroxyl and alkylether determined by FTIR， while the carboxyl is rarely found. Based on the multiple linear regression analysis by SPSS17.0 software， a prediction model of the cold strength of coke was established， which used the parameters such as the volatile matter (V_daf) and caking index (G) as independent variable and used the crushing strength of coke (M25) and abrasive resistance (M10) as dependent variable. Hence， thermal properties and cold strength of coke can be controlled flexibly by this model.

Key words: coal, modification, coke, cold strength, prediction model