On the basis of analysis of the principle of delay restoration in a disturbed schedule, a heuristic algorithm for rescheduling trains is developed by restoring the total delay of the disturbed schedule. A discrete event topologic model is derived from the original undisturbed train diagram and a back propagation analysis method is used to label the maximum buffer time of each point in the model. In order to analyze the principle of delay restoration, the concept of critical delay is developed from the labeled maximum buffer time. The critical delay is the critical point of successful delay restoration. All the disturbed trains are classified into the strong-delayed trains and the weak-delayed trains by the criterion of the critical delay. Only the latter, in which actual delay is less than its critical delay, can be adjusted to a normal running state during time horizon considered. The heuristic algorithm is used to restore all the disturbed trains according to their critical details. The cores of the algorithm are the iterative repair technique and two repair methods for the two kinds of trains. The algorithm searches iteratively the space of possible conflicts caused by disturbed trains using an earliest-delay-first heuristics and always attempts to repair the earliest constraint violation. The algorithm adjusts the weak-delayed trains directly back to the normal running state using the buffer time of the original train diagram. For the strong-delayed trains,the algorithm uses an utility function with some weighted attributes to determine the dynamic priority of the trains,and resolves the conflict according to the calculated dynamic priority. In the end, the experimental results show that the algorithm produces “good enough” schedules effectively and efficiently in disturbed situations. On the basis of analysis of the principle of delay restoration in a disturbed schedule, a heuristic algorithm for rescheduling trains is developed by restoring the total delay of the disturbed schedule. A discrete event topologic model is derived from the original undisturbed train diagram and a back propagation analysis method is used to label the maximum buffer time of each point in the model. In order to analyze the principle of delay restoration, the concept of critical delay is developed from the labeled maximum buffer time. The critical delay is the critical point of All the disturbed trains are classified into the strong-delayed trains and the weak-delayed trains by the criterion of the critical delay. Only the latter, where the actual delay is less than its critical delay, can be adjusted to a normal running state during time horizon considered. The heuristic algorithm is used to restore all the disturbed trains according to their critical details. The cores of the algorithm are the iterative repair technique and two repair methods for the two kinds of trains. The algorithm searches iteratively the space of possible conflicts caused by disturbed trains using an earliest-delay-first heuristics and always attempts to repair the earliest constraint violation. The algorithm adjusts the weak-delayed trains directly back to the normal running state using the buffer time of the original train diagram. For the strong-delayed trains, the algorithm uses an utility function with some weighted attributes to determine the dynamic priority of the trains, and resolves the conflict according to the calculated dynamic priority. In the end, the experimental results show that the algorithm produces “good enough ” schedules effectively and efficiently in disturbed situations.