Buffer sizing in manufacturing production systems with complex routings

A three-phase heuristic approach has been developed to determine the total buffer storage and to allocate the buffers to each stage in complex manufacturing systems. This cost-based approach extends the classical response surface methodology to explore the complex relationship between cost function and buffer storage. In phase one, queuing statistics are used to determine the rough-cut buffer capacity and its allocations to each buffer stage. Based on the confidence interval concept, a modified steepest descent search is applied to identify new buffer sizes for the cost improvement. In the second phase, the theoretical optimum buffer settings are determined by constructing a second-order model of the response surface and the buffer sizes. In phase three, an integral solution for the buffer sizing problem is obtained by the derivative approximation. The computational detail is illustrated by an example. Numerical studies are reported for 20 buffer sizing problems, including serial flow lines, multi-entry multi-part manufacturing systems, and single-entry multi-part manufacturing systems.