Optimization of capacitated kanban production systems

The success of Just-in-Time (JIT)/Pull disciplines of manufacturing and distribution operations has led to a vast literature on JIT systems. Kanban systems are studied extensively in the literature as a paradigm for JIT production systems. The number of kanban cards circulating in such a system is of key interest, as this number controls the work-in-process inventory system-wide. While the general objective is to lower this number, too small a number of kanbans may result in shortages or starve the system. Some researchers have approached the allocation of the optimal number of kanbans through trial and error or with simulation models. In this dissertation, we address the optimal allocation of kanbans using a deterministic optimization model and extend the literature by removing the unrealistic assumptions of infinite capacity and zero initial inventory. We provide a formulation for the capacitated problem and investigate the structural properties of the resulting model.;We propose three algorithms to solve the optimization problem under different assumptions on the relevant system parameters (cost and capacity). We perform computational testing to investigate the computational burden of the algorithms, as well as their sensitivity to characteristics of the problem data. We outline the contributions of the dissertation and fist directions for future research on kanban allocation in capacitated pull systems.