Research

year 2003 
author Kichang Lee 
Keyword manufacturing system design, process model, Petri net, performance model and optimization, simulation 
Abstract In rapidly changing enterprise environments, frequent design activities of a manufacturing system are required. Manufacturing system design decisions have a long term effect on system performance. Moreover, they are so complex problems that coordination and cooperation among departments are necessary to evaluate and improve design alternatives to support the design process, systemic process modeling function and performance analysis/optimization function should be provided in an integrated manner.

In this research, we suggest a new process modeling technique, called PFN(Product Flow Network), as a substitute of previous modeling techniques such as IDEF0, IDEF3, and the ones used in enterprise modeling frameworks. PFN is a directed graph defined for a part to represent the activities and the relationships among them which are necessary to product the part. Connections among nodes can have AND OR XOR types by which various relationships may be captured.

We adopted lead time and manufacturing costs as performance measures. To predict these performance measures for a given design alternatives, Petri net approach is used. We classified a manufacturing system into closed system and open system. In a closed system, the maximum number of parts that may exist in a system is pre-determined. On the other hand, an open system has no such constraints. Petri net models have been constructed for closed and open systems respectively. In addition, an algorithm has been provided to automatically transform a PFM into a Petri net. Before this transformation, XOR elimination algorithm has been suggested, which makes the transformation procedure easy.

In a Petri net for a closed system, as the size of subjective system increases, the computational capability decreased dramatically. To circumvent this shortage, a consolidation approach, which simplifies a Petri net, has been devised. Experimental results are included which show the feasibility of the consolidation. An open system cannot be modeled as a Petri net completely, because the analysis used Markov chain. In this research only service area of a system modeled by Petri net, and arrival area is considered in performance analysis. This approach is new and proved to be effective by simulation experiments. In addition it provides mathematical functions for average lead time and manufacturing costs. Using this functions, we constructed mathematical formulation to minimize average lead time satisfying given manufacturing costs and to minimize total manufacturing costs given average lead time.

In the formulation, lead time function is nonlinear so that the problems cannot be solved directly by commercial optimizer. We adopted linearization procedure in separable programming. By this procedure, nonlinear function can be linearized, which can be solved optimization package such as ILOG CPLEX. We performed simulation experiments for lead time optimization model and manufacturing cost optimization model to validate the formulation. In the result, the mathematical model proved to generate optimized design alternative effectively.

The methodology suggested in this research would make possible to generate various alternatives and to compute optimize performances in the early stage of manufacturing system design process. A computerized environment for this methodology would be more beneficial in supporting the design activities. 
c PhD 

Download :