Research

year 1997 
author Sangbok Woo 
Keyword Integrated System, Loading, Tool allocation, Scheduling, Flexible Manufacturing System, Flexibility 
Abstract A Flexible Manufacturing System( FMS ) is an automated manufacturing system which pursues both the productivity of a transfer-line and the flexibility of a job shop simultaneously. In particular, the flexibility of FMS has been recognized as a major element for responding to dynamics of market environment and uncertainties of shop floor. In spite of such an importance, there were few researches which dealt with the effective use of flexibility throughout the production planning and scheduling of FMS. In order to efficiently utilize the flexibility of FMS, it is important to make a pertinent plan for the alternative machines in a given planning horizon. But, this is very complex because we should consider loading and tool-allocation constraints in production planning as well as precedence constraints in scheduling.

There were some researches which divided entire problems into two subproblems - loading and scheduling - and solved them hierarchically. Even a subproblem, however, it is too difficult to find an optimal solution in real time and to establish a relevant goal of the loading problem corresponding with the scheduling performance.

In this thesis, we attempt to solve the loading and scheduling problems of FMS in an integrated manner. So we propose an integrated system and its solution methodologies which can use the flexibility of FMS effectively, and make a decision about tool allocation, machine selection, and sequencing altogether.

The proposed system consists of two main modules, ´schedule generating module´ which makes partial schedules and ´tool-allocation checking module´ which investigates the feasibility of tool-allocation for unscheduled tasks. With interacting two modules, we settle the loading and scheduling problems. In the schedule generating module, we assume FMS scheduling problems as having multiple identical parts and alternative machines. According to characteristics of alternative machines, we classify the schedule generating module into three submodules - identical machines submodule, exclusive machines submodule, and non-identical machines submodule. In the identical machines submodule, we propose an optimal algorithm, a modified nondelay schedule generator which eliminates many duplicating partial schedules not affecting an optimal makespan solution, and show the validity and efficiency of the algorithm by theorems and experiments. In the exclusive machines submodule, we propose some dispatching rules which consider precedence relationships among multiple identical parts, and show the effectiveness of rules by experiments. In the non-identical machines submodule, we propose an schedule generating algorithm which is based on estimated workloads, and some dispatching rules which is devised to efficiently use the non-identical alternative machines, and show the superiority of proposed algorithms and rules by experiments. The schedule generating module can be used independently of the tool-allocation checking module in cases that constraints related to the tool-allocation become meaningless due to the introduction of an automated tool delivery system or a sufficient tool magazine, etc.

In the tool-allocation checking module, we propose a heuristic and a lagrangian relaxation method as well as optimization models. So, it is possible to solve real problems in allowed time limit. The tool-allocation checking module can be used as not only a submodule which solves the loading and scheduling problem, but also a means which indicates the production feasibility of parts selected initially or added in a given tool-allocation status. Experimental results show that in most cases the proposed integrated approach outperforms existing hierarchical approaches in the scheduling performance and the computational time required. In addition to that, the difference between two approaches tends to increase when the number of part types and the number of alternative machines increase and the tool constraints become tight. So, the results show that the proposed approach is very useful for practical problems. 
c PhD 

Download :

번호 c year 제목 author

등록된 글이 없습니다.