New results in factory physics - Insights from the underlying structures of manufacturing systems.

摘要

The objective of this dissertation is to enhance the overall understanding of practical manufacturing systems by using rigorous analytical approaches, primarily queueing theory, for estimating cycle times. The scope spans from a single manufacturing process to a manufacturing system.;Performance of single machine systems. Queueing theory is commonly used to evaluate the performance of manufacturing systems. However, actual applications of queueing models to single machine systems encounter some practical issues. A real machine is subject to different kinds of interruptions, such as breakdowns, setups and routine maintenance. By systematically classifying different kinds of interruptions, suitable queueing models based on the structure of interruptions are proposed, and a unified model is derived to integrate all types of interruption events. New insights are derived on the application of queueing theory to a single machine system through detailed analysis of different types of interruptions.;From a practical prospective, better models are needed for parallel batch processing in manufacturing systems. The existing G/G/1 based approximate model decomposes the cycle time into three parts: wait-to-batch times, waiting times and service times. However, through detailed examination of the model, a more accurate approximate model is proposed by examining the independence assumption between wait-to-batch times and waiting times. The new model gives considerable improvement over existing approaches.;Performance of manufacturing systems. The behavior of manufacturing systems is explored by first investigating the underlying structure of tandem queues. Due to the non-renewal departure process, this challenging problem is usually tackled by the parametric-decomposition method or by diffusion approximations. To search for a better model, the behavioral structure of tandem queues has been investigated. The identified structures, termed the intrinsic gap and intrinsic ratio, come with very nice properties, which we call the heavy-traffic and nearly-linear relationship properties. Based on those properties, new approximate models are developed for two single-server stations in series. Then the model is extended to approximate the queueing time of many single-server stations in series. Based on extensive simulation results, the new models considerably outperform earlier methods based on the parametric-decomposition and diffusion approximations. By using the identified structure, a way to analyze the intricate dependence among tandem queues has been proposed.;By extending the results from tandem queues, an approximate model for an entire manufacturing system is derived. The model is used to quantify the performance of manufacturing systems and has been evaluated using complicated simulation models from an industrial partner. The results demonstrate that the new method has strong capability to describe the performance of practical manufacturing systems precisely.