To minimize the average tardiness of the jobs processed by the bottleneck station, a semi-online scheduling method was proposed, considering the re-entrant manufacturing essence and the mastery of the production information which could be classified into historical information, predictable information about future arrival jobs and real-time information. Firstly, we analyzed the sources of the jobs processed by the bottleneck station, and provided a method to establish the task set of the bottleneck station in scheduling horizon. Then we described the re-entrant flows on parallel tools in bottleneck area with disjunctive graph and developed an ant colony optimization (ACO) algorithm to find a satisfactory solution in a reasonable computation time. Finally, three analytical experiments were used to demonstrate the effectiveness of the proposed method.