Lot streaming and scheduling heuristics for m-machine no-wait flowshops

The objective of this paper is to minimize makespan in m-machine no-wait flowshops with multiple products requiring lot streaming. A ‘product’ here implies many identical items. ‘Lot streaming’ creates sublots to move the completed portion of a production lot to downstream machines so that machine operations can be overlapped. For the single product case with fixed number of sublots we obtain optimal continuous-sized sublots and then use a heuristic to find integer-sized sublots. For the multi-product continuous-sized sublots case we show that the optimal sequencing of products may be attained by solving a traveling salesman problem. We then construct another heuristic to yield integer-sized sublots. Finally, we evaluate the use of genetic algorithmic meta-heuristics for the interacting decision phases in simultaneous lot streaming and sequencing. We conclude that while GA may deliver makespans comparable in quality to those given by heuristic methods that cleverly exploit problem features particular to lot streaming, GA loses out in computational efficiency. On the other hand, GA can optimize the number of sublots for each product — a task for which neither an analytical nor a heuristic method presently exists.