Optimal lot-sizing with capacity constraints and auto-correlated interarrival times

There have been recent advances in using queuing relationships to determine lot sizes that minimize mean flowtimes when multiple product types are being produced at capacity-constrained resources. However, these relationships assume lot interarrival times are independent, which is not the case in most manufacturing scenarios. This study examines the performance lot-sizing optimization relationships based on GI/G/1 relationships when lot interarrival times are auto-correlated. Simulation and response surface modeling are used to experimentally determine optimal lot sizes for a sample problem. The flowtimes for "optimal" lot sizes determined analytically are found to compare poorly with the best flowtimes obtained experimentally. An approach is then developed that uses feedback during simulation to adjust parameters within queuing heuristics that support dynamic lot-size optimization. Performance using this approach compares well with the best performance obtained using the much more difficult experimental approach.