We present a simulation-based approach to capture the interactions between train operations and passenger behavior during disruptions in urban rail transit systems. The simulation models the full disruption and recovery cycle. It is based on a discrete-event simulation framework to model the network vehicles movement. It is paired with an agent-based model to replicate passenger route choices and decisions during both the undisrupted and disrupted state of the system. We demonstrate that optimizing and exibly changing the train dispatch schedules on specific routes reduces the impact of disruptions. Moreover, we show that demand uncertainty considerably changes the measures of performance during the disruption. However, the optimized schedule still outperforms the non-optimized schedule even under demand uncertainty.