Abstract: With the fast development of cities, many kinds of pipelines for communication, heat, oil & gas and water transportation are constructed and put into operation. Due to planning, management and partial historical reasons, various pipeline networks cross with each other underground and a huge "spider web" is formed. However, there is no effective unified management organization. Therefore, the difficulty of safety and emergency management in the case of oil and gas pipeline leakage is how to identify and trace accident causes quickly. To solve this problem, a Dynamic Bayesian Network (DBN) based quick identification and in-depth tracing analysis method for oil and gas pipeline accidents was proposed. This method makes full use of the convenience and effectiveness of FT modeling, the reverse inference capacity and the accurate expression of posterior probability of DBN. Then, the fault tree was correspondingly established by analyzing the potential reasons for the leakage of urban oil and gas pipelines. And after it was transformed into DBN model, historical statistical data and multi-source observation information were integrated for the quick identification of accident scene and the tracing of accident cause. The accident analysis results are in line with the later accident investigation conclusions. It is indicated that this method can be used as the guide for accident emergency rescue and accident investigation because of its better accuracy and practicability.