城市轨道交通运营异常具备状态连续演化、跨专业协同处置和恢复时长差异显著等特征。从连续事件过程中识别影响恢复及时性的关键条件，有助于深化对异常演化与恢复关系的认识，为异常初判及应急响应资源前置提供依据。鉴于此，以上海地铁2016—2025年官方运营公告为数据源，从52 291条公告中重构806起具有明确异常起点和恢复端点的事件，识别运营影响扩大、行车组织调整、客运组织调整和应急公交接驳等过程状态。采用log-logistic AFT模型分析恢复时长，并以60 min为阈值，采用Firth惩罚似然Logistic回归模型识别长时运营影响。结果表明，恢复时长中位数为16.87 min，87起事件持续60 min及以上，占比10.79%；供电、线路和信号设备故障的恢复时长分别为车辆故障的2.512、1.836和1.778倍，运营启动时段和多线路影响事件的恢复时长分别为日间平峰和单线路事件的1.679和3.871倍。系统设备故障、外部及其他事件、运营启动时段和多线路影响与长时运营影响显著相关，其中多线路影响事件长时影响率为62.50%、恢复时长中位数为98.24 min。总体上，故障发生频率与恢复难度并不同步，系统设备属性、运营启动时段和多线路影响是恢复效能降低的主要关联特征，多线路影响呈现低频高后果特征。

Urban rail transit operational disruptions are characterized by successive state evolution, cross-functional coordinated response, and substantial variation in recovery duration. Identifying factors associated with recovery timeliness from continuous event processes can deepen the understanding of the relationship between disruption evolution and recovery and support preliminary disruption assessment and pre-positioning of response resources. Given this, official operational bulletins issued by Shanghai Metro from 2016 to 2025 were used as the data source. A total of 806 events with clearly defined disruption onset and recovery endpoints were reconstructed from 52,291 bulletins,and process states including expansion of operational impacts, train operation adjustment, passenger flow management adjustment, and emergency bus bridging were identified. A log-logistic AFT model was used to analyze recovery duration, and a Firth penalized-likelihood logistic regression model with a 60-min threshold was employed to identify long-duration operational impacts. The results showed that the median recovery duration was 16.87 min, and 87 events lasted 60 min or longer, accounting for 10.79% of the sample. The recovery durations of power supply, track equipment, and signaling equipment failures were 2.512, 1.836, and 1.778 times those of vehicle failures, respectively. Events occurring during the service start-up period and events affecting multiple lines had recovery durations 1.679 and 3.871 times those of daytime off-peak events and single-line events, respectively. System equipment failures, external and other events, the service start-up period, and multi-line impacts were significantly associated with long-duration operational impacts; multi-line events had a long-duration impact rate of 62.50% and a median recovery duration of 98.24 min. Overall, disruption frequency didn′t correspond to recovery difficulty; system-equipment characteristics, the service start-up period, and multi-line impacts were the main features associated with lower recovery performance, while multi-line impacts showed a low-frequency, high-consequence pattern.