Batch production smoothing with variable setup and processing times

Many companies use mixed-model production systems, running under the just-in-time (JIT) philosophy, in order to efficiently meet customer demands for a variety of products. Such systems require demand be stable and production sequence be leveled. The production smoothing problem (PSP) aims to find level schedules at the final level of a multi-level manufacturing system. The products in a level schedule are dispersed over the horizon as uniformly as possible. In this area, most research has focused on sequencing JIT mixed-model assembly lines where setup and changeover times are assumed negligible. However, in many flow lines in the real life, a significant amount of time needs to be dedicated to setup/changeover among different products. Therefore, for such systems the existing literature falls short of helping to smooth production.; We consider two alternative manufacturing environments, a single machine or a flow-shop, at each level of the manufacturing system; and study both single-level and multi-level versions of the PSP. We allow the products to have arbitrary non-zero processing and setup time requirements on the machines, where the total productive time is limited. Here, one must decide on batch sizes and number of batches for each product, before sequencing the batches. We develop a two-phase solution approach that is applicable on all four models. The first phase finds appropriate batch sizes for the products and the second phase finds a level sequence of the batches of products. We relate the second phase problem to the existing solution methods available in the literature, and focus on the first phase problem.; We build an optimization model for the first phase problem; show that it is NP-complete; devise heuristic methods for its solution; implement meta-heuristic techniques; and develop exact solution procedures based on dynamic programming (DP) and branch-and-bound (B&B) methods. Through computational experiments, we compare the performance of our solution methods. The results show that our exact methods are efficient in solving medium-sized instances of the problem. Also, our meta-heuristic implementations yield near-optimal solutions in almost real-time.