Aiming at the problem of insufficient and uncertain information in the initial stage of emergency decision-making of sudden accidents in high-speed railway and the limitation of emergency decision-making ability of rescue departments, this paper proposes an emergency decision-making method based on intuitionistic fuzzy theory and Bayesian inference network. Firstly, the inducement, accident classification and emergency decision-making characteristics and process of high-speed railway emergency are analyzed., then the intuitionistic fuzzy number is used to represent the evidence information of Bayesian network nodes, and an intuitionistic fuzzy Bayesian network (IFBN) emergency decision-making model is given. Finally, this paper takes the EMU for emergency decision-making in the case of a sudden fire accident on the viaduct as an example to reason with BN and IFBN respectively. After several groups of reasoning, it is found that when the information hesitation is large, BN reasoning will produce the opposite conclusion to the actual situation, but IFBN overcomes this shortcoming, and the reasoning result is still accurate, which verifies the superiority of using IFBN reasoning model to make emergency decision.