Abstract
Batch processing machine plays an important role in many manufacturing environments. For example, the workstation in the liquid crystal injecton(LC injection) process in LCD manufacturing consists of identical parallel batch processoing machines. Each machine can process a certain number of panels simultaneously. The objective of this scheduling problem is to balance “the expected delay of lots in queue” and “the capacity utilization of batch machines” under the constraints of sequence dependent setup times. Machines require different setups. That the less times the machines spend on setups and the larger batch sizes are processed can improve machine utilization. Therefore, to balance the trade-off, we set the objective of the problem is to minimize the total costs. By utilizing “the future arrivals information” as proposed by Weng and Leachman【1993】, we use the arrival <a href="http://www.ntsearch.com/search.php?q=time&v=56">time</a> prediction and make decisions whether we should start a batch or wait. The quality of production decision can be measured by a cost function that is the sum of inventory costs, backorder costs, and setup costs. Two heuristic <a href="http://www.ntsearch.com/search.php?q=search&v=56">search</a> algorithms, stochastic tabu <a href="http://www.ntsearch.com/search.php?q=search&v=56">search</a> (STS) and simulated annealing (SA), are used to solove the schedule problem. Finally, to validate the quality and to evaluate the speed of the two algorithms, extensive experiments were performed. The result shows that STS is better than SA in terms of the quality of final solutions, but STS is worse than SA at the early state of computation.