Thumbnail Image


umi-umd-1759.pdf (1.5 MB)
No. of downloads: 1394

Publication or External Link






This dissertation addresses two research questions relating to the role of setups in discrete parts manufacturing. The first research topic uses a carefully designed simulation study to investigate the role of setup economies in the factory-wide conversion of functional layouts (job shops) to cellular manufacturing. The model-based literature shows a wide dispersion in the relative performance of cellular manufacturing systems as compared to the original job-shop configurations, even when the key performance measure is flow time and the assessment tool used is simulation. Using a standardized framework for comparison, we show how this dispersion can be reduced and consistent results can be obtained as to when the conversion of the job shop is advantageous.

The proposed framework standardizes the parameters and operational rules to permit meaningful comparison across different manufacturing environments, while retaining differences in part mix and demand characteristics. We apply this framework to a test bed of six problems extracted from the literature and use the results to assess the effect of two key factors: setup reduction and the overall shop load (demand placed on the available capacity). We also show that the use of transfer batches constitutes an independent improvement lever for reducing flow time across all data sets. Finally, we utilize the same simulation study framework to investigate the benefits of partial transformation, where only a portion of the job shop is converted to cells to work alongside a remainder shop.

The second research question examines the role of dispatching rules in the reduction of setups. We use queueing models to investigate the extent of setup reduction analytically. We single out the Alternating Priority (AP) rule since it is designed to minimize the incidence of setups for a two-class system. We investigate the extent of setup reductions by comparing AP with the First-Come-First-Served (FCFS) rule. New results are obtained analytically for the case of zero setup times and extended to the case of non-zero setup time through computational studies.