A real-time dynamic lot-sizing heuristic for a manufacturing system subject to random setup times

A heuristic is derived for real-time dynamic lot sizing at the bottleneck of a high volume manufacturer of many part types. This bottleneck resource is subject to significant, random setup times. The problem is motivated by the scheduling requirements of a medical device manufacturing division of Johnson & Johnson, where meeting a common due date for all the current week’s production is critical. The lot-sizing heuristic is one component of a hierarchical dispatching policy in which lot sequencing dominates lot sizing. A sample problem is developed and modelled as a dynamic program in order to demonstrate the form of an optimal solution. Numerical tests are provided which compare the dynamic program policy, a derived heuristic and other controllers. The results indicate the possibility of closed feedback control as an efficient method for dispatching parts. A prototype controller which was developed based on the heuristic control policy is briefly described.