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

Batch manufacturing of discrete parts involves dealing with high uncertainty and variability, the need to coordinate parts and the use of common resources for multiple part types. Since setup times are usually significant, the production lot sizes are important. These influence flowtimes, inventory levels and other performance measures. Recent research has applied queuing relationships to optimize lot sizes. In particular, GI/G/1 queuing relationships have been used. However, these relationships assume independent lot interarrival times; something that is clearly not true in batch production.; This research investigates the effects of auto-correlation using a single stage, two-product scenario. Results show flowtime prediction based on GI/G/1 relationships is poor. A dynamic feedback approach is then developed that adjusts queuing heuristics to compensate for auto-correlation. This approach is evaluated against optimal lot sizes obtained using response surface models based on simulation results. The new approach is found to perform well.