Modeling and Solution of Realistic Facility Layout Problems Involving Random Elements

点击翻译按钮获取中文摘要

In this project, we propose a new framework of facility layout problems which takes the production efficiency into account explicitly. In the classical framework of facility layout problems, layouts are evaluated only by the material handling cost, and any aspects as to how layouts affect the production efficiencies are not considered. In real systems, however, the layout of facilities do affect the production efficiencies such as throughput, lead time and so on. Thus, in general, there should exist the "best" layout which absorbs the variability involved in the system and attains the highest production efficiency. In short, our problem is to find the layout of facilities which results in the maximal production efficiency. We refer this type of layout problems as the "Stochastic Facility Layout Problem (SFLP)". For examples, in case when the physical size of a buffer space can not be ignored compared with the facility itself, allocating buffer spaces to a facility affects the distances … More between facilities. In such a case, the production efficiency depends on both buffer space allocation and facility layout in a complex way. Our problem gives rational solutions for such a situation.In the first model, we choose the manufacturing lead time as the performance measure, formulate an optimal facility layout and buffer space allocation problem for job-shop type production systems in which relatively large size parts are transfered by AGVs and processed via variable routes in FCFS manner. Our approach is based on an approximate Markov analysis using a decomposition technique for evaluating the lead time and the genetic algorithm for searching the optimal combination of buffer space allocation and layout. In the second model, we consider production systems with a feed-forward configuration and variable processing times. Our objective is to efficiently find Pareto optimal solutions for both throughput and material handling cost. We assume that work transfer is performed by belt-conveyors or unlimited vehicles. Since the facility layout does not affect the throughput under these assumptions, we first calculate the throughput for every buffer space allocation using Markov analysis. Then the set of Pareto optimal facility layouts and buffer space allocations is searched using a Genetic Algorithm. Less