The process of train operation is affected by many disturbance factors, which may cause train delay and affect passengers. In order to identify the disturbance and estimate the influence, a novel approach based on historical data of urban rail transit is proposed. According to the emergency disturbance and its related multi-source data, such as train operation record and automatic fare collection data, the generation and influence law of the disturbance were analyzed. The main contents include the classification, causes and occurrence law of the disturbance, as well as the duration, spatial range and relative passenger flow rate under the disturbance. The case study of a city shows that: disturbance events were more likely to occur in peak hours, and vehicle failure was the most common cause. More than half of the disturbance events could be eliminated in 30 minutes. The delay caused by disturbance mainly occurred on the follow-up trains and at the stations in front of the occurrence location. The passenger flows decreased in the first half hour, then increased or even exceeded the normal standard ones, which coincided with each other about two hours.