In this paper, FMS scheduling problems are modeled by timed Petri nets. The timed Petri nets can model the important characteristics of FMSs such as routing flexibility, shared resources, precedence relationships among operations, concurrency, synchronization and lot sizes. So the Petri net model of an FMS scheduling problem which reflects the given constraints would limit the search within the feasible schedule space. By making use of a heuristic scheduling algorithm which satisfies the lower bound condition of the algorithm A*, the search space can be further reduced and an optimal path can be found. This new scheduling approach which combines Petri net modeling with heuristic search can effectively handle with the structural complexity and computational complexity of FMS scheduling problems.