In this paper, we develope a genetic algorithm for the erection scheduling in shipbuilding. Erection, the final manufacturing stage of shipbuilding, involves the landing and joining of blocks at drydock. Since several ships are built simultaneously at the same drydock and they compete with the common constrained production resource such as labor, space, and crane, we should consider both ship-specifics and common resource constraints for the desirable erection scheduling. Ship erection should also satisfy the predetermined dock cycle given from higher level production planning. Thus, erection schedule of a ship can be represented as a PERT/CPM project network with its own deadline. Since production resources are shared, the erection scheduling become the so-called multi-project scheduling problem with limited resources, which can not be solved easily due to the large size of project network. We propose a function as a minimization of load index which represented the load deviation over time horizon considering the yard production strategy. For the optical parameter setting, we tried various experiments. We verified that the proposed approach was effective to deal with the erection scheduling problem in shipbuilding.