Abstract
There are three important objectives for the production management of an apparel factory. The overall throughput is affected by the productivity of the bottleneck operation. The first objective of an apparel factory is to maximize its bottleneck, the sewing process, whose productivity is determined by sewing schedule. The second objective of the factory is to maximize fabric utilization that is determined by cut order plan. Given the decisions by the optimization of the two objectives, the third objective is to minimize the cut pieces WIP inventory through an effective cutting scheduling. The focus of this study is developing an effective and efficient cutting scheduling method that minimizes cut piece WIP area under the constraint of no shortage for sewing operations. This study compares six methods, mixed integer programming model proposed by Hung and Chang (2011), an enumerative method, and the four meta-heuristic search methods - simulated annealing, tabu search, genetic algorithm, and ant colony optimization - for solving the cutting scheduling problem. To improve the efficiency of the meta-heuristic search methods, an initial solution provided by the heuristic algorithm proposed by Hung and Chang (2011) is used. A dominant sequence is defined as a sequence that a fabric lay is cut only when one of the produced cut piece type is required. It is proved that an optimal cutting schedule must be a dominant sequence. Thus, we do not need to consider non-dominant sequence when searching for the solution. Also, a proposed algorithm is used to modify a non-dominant sequence into a dominant one. An algorithm that determines the cutting time of each fabric lay from a given fabric lay sequence and the cutting times determined as such are proved to be optimal. The experiment results show that an enumerative method is not capable of solving the problem that contains more than 20 fabric lays within a reasonable time. Furthermore, the heuristic algorithm proposed by Hung and Chang (2011) is more effective than the common practice of arbitrary selection of fabric lay. When normalized CPU time before 0.4, ACO performs better than TS. Whereas, when normalized CPU time is greater than 0.4, TS surpass ACO.