This paper concentrates on the scheduling problem in seru production system (SPS), where seru is a successful new-type production mode arising from the Japanese labor-intensive electronic assembly industry. Motivated by the practical situations, the sequence-dependent setup time and DeJong’s learning effect are considered in seru scheduling problems, and the objective is to minimize the makespan. The seru scheduling problem is formulated as a mixed-integer programming (MIP), and then reformulated to a set partitioning master problem and some independent subproblems by employing the logic-based Benders decomposition (LBBD) method. Subsequently, the set partitioning master problem is used to assign jobs to serus of SPS, and the subproblems are applied to find the optimal schedules in each seru given the assignment of the master problem. Finally, computational studies are made, and results indicate that the LBBD method is able to return high-quality schedules for solving seru scheduling problems.