Production Scheduling for a Fabrication-Assembly System.

This dissertation investigates the application of a multi-echelon lot-sizing model to the scheduling of the day-by-day operations of a U.S. automotive stamping plant. The plant consists of a fabrication stage that makes components, and an assembly stage which utilizes these components in producing end items. A hierarchical approach is taken to develop a minimum cost production schedule that meets all dem and s within the available production capacity. The schedule is developed in three steps. First, a production plan is developed using a unique dynamic lot-sizing algorithm for single-facility capacitated systems with sequence dependent set-up time and set-up cost. Second, a similar algorithm is developed to generate the K-minimum cost schedules for each facility over the immediate period. Third, based upon the output produced and resources required by each minimum cost schedule, a detailed daily production schedule is generated for both stages. The problem is modeled as a mixed integer program, which is decomposed by Benders' decomposition, and solved by a branch and bound method. The value of the approach is demonstrated by a case study involving the scheduling of 48 products produced on 8 first-stage facilities and 13 second-stage facilities over a six week period. The results compare favorably with the actual production history. In summary, by relaxing simplifying assumptions and adding resource capacity constraints, this research bridges the gap between theory and practice. Algorithms developed to solve large scale problems are found efficient and robust in practical situations.PhDIndustrial engineeringUniversity of Michiganhttp://deepblue.lib.umich.edu/bitstream/2027.42/160053/1/8412266.pd