Production lot sizing under setup and worker learning

Previous lot-sizing models incorporating learning effects focus exclusively on worker learning. The authors extend these models to include the presence of setup learning, which occurs when setup costs exhibit a learning curve effect as a function of the number of lots produced. The joint worker/setup learning problem can be solved to optimality by dynamic programming. Computational experience indicates, however, that solution times are sensitive to certain problem parameters, such as the planning horizon and/or the presence of a lower bound on worker learning. The authors define a two-phase EOQ-based heuristic for the problem when total transmission of worker learning occurs. Numerical results show that the heuristic consistently generates solutions well within 1% of optimality.