Queuing models for performance evaluation and optimal design of fabrication/assembly systems

Manufactured products are typically assembled from multiple components. Each of these components is often manufactured in fabrication lines involving a series of operations. After completing these operations, the components wait at a kitting station until all the components required for an assembly are available. Although the release of the components into these fabrication lines is synchronized, the stochastic nature of manufacturing operations often results in components waiting at the kitting station resulting in high inventory holding costs. The kitting delays also result in the costly shut down of assembly operations and customer dissatisfaction due to delay in order completion. Therefore, the design of efficient fabrication/assembly systems that minimize such delays is an important problem in supply chains.; In this dissertation, efficient analytical models are developed for performance evaluation of fabrication/assembly systems, in which end products are assembled from several components. The components are supplied by fabrication lines, that are composed of one or more manufacturing stations. The total population of each component present in the system at anytime is limited by using a kanban control policy. The system is modeled as a closed queuing network having fork/join stations and is analyzed using parametric decomposition method.; The first part of this dissertation analyzes the dynamics of kitting operations by modeling it as a fork/join station operating in isolation. The insights from the analysis are useful in understanding the kitting operation and also form useful building blocks in analyzing larger systems. The second part of this dissertation analyzes fabrication/assembly systems with multiple end products wherein components are first manufactured in production lines prior to being kit. Here, parametric decomposition techniques are developed for performance analysis. Subsequently, the approach is extended to analyze multi-product systems with reduced computational effort. The final portion of the dissertation demonstrates, the use of queuing models developed in the prior chapters in identifying optimal designs of fabrication/assembly systems. Game theoretic models of simple assembly systems are formulated to understand design tradeoffs in central.