Determination of buffer size and location in scheduling systems

Competition among companies is becoming more and more fierce. In addition to increased quality conformance, companies with short manufacturing (cycle) times can gain competitive advantages. Short cycle times are achieved by having viable scheduling systems that require minimal inventories. Many scheduling systems such as Kanban, Drum-Buffer-Rope (DBR), and CONWIP, as well as a new system referred to as Dynamic Flow Control (DFC), have been developed that seek to achieve these competitive advantages (objectives). But, most of these systems fall short of achieving these objectives because they do not clearly identify where and how much inventory should be employed. Obviously, zero inventory is inefficient and some inventory is needed in a manufacturing line. However, the exact determination of total inventory quantities and the determination of the location and size of work-in-process (WIP) inventories is difficult. It is difficult, since it is impossible to forecast the impact of all statistical fluctuations, which is needed to determine work-in-process inventory structures. Many real-world manufacturing lines side-step the problem by employing high finished inventory levels. However, there is a tradeoff in setting inventory levels and inventories should not necessarily be set at high levels. On the one hand, too low inventory levels starve work centers which reduces line performance. On the other hand high WIP inventory levels result in long manufacturing cycle time, which operates counterproductive to the competitive advantage of a firm. This dissertation deals with problems of WIP inventory management and provides a simulation-search procedure which can be used to determine inventory size and location requirements of a firm. It shows that the size and location of WIP inventory is a combinatorial problem which is a function of buffer sizes at each operation in a manufacturing line. This study identifies specific inventory locations and minimal sizes, referred to as "inventory profiles," which support manufacturing scheduling systems. Because of the complexity of the profiling process, simulation, combined with an artificial intelligence based search heuristic, is employed to address the problem. Simulation is used to model the manufacturing facility under study, while a Tabu Search metaheuristic is used to analyze the simulated results, to direct the search process in the combinatorial environment, and to identify optimal or near optimal buffer profiles.