Pergamon Computers ind. Engng Vol. 26, No. 4, pp. 673~688, 1994 Copyright © 1994 Elsevier Science Ltd 0360-8352(94)E0013-C Printed in Great Britain. All rights reserved 0360-8352/94 $7.00 + 0.00 PERFORMANCE EVALUATION OF MANUFACTURING SYSTEMS: A SPREADSHEET MODEL PYUNG-HOI Koo, COLIN L. MOODIE and JOSEPH J. TALAVAGE School of Industrial Engineering, Purdue University, West Lafayette, IN 47907, U.S.A. (Received for publication 7 March 1994) Abstract--In this paper, we describe the spreadsheet modeling of manufacturing systems which are represented by multi-server and multi-product open networks. The spreadsheet software is characterized by easy data manipulation, on-screen numerical and visual feedback, fast calculation (what-if analysis), and its availability. So far, most spreadsheet models of this type have been used to solve static and deterministic problems in manufacturing systems, ignoring most of the uncertainties. We propose here that a spreadsheet can be implemented for capturing not only static features but also stochastic behavior of manufacturing systems. The experiments show that the results from the spreadsheet model are reasonably close to those from the other existing approaches. The proposed spreadsheet model could be applied by an engineer to the modeling of manufacturing systems, especially for a first-cut design stage, utilizing his/her own spreadsheet software. 1. INTRODUCTION Introduction of advanced manufacturing technologies in the manufacturing environment is creating manufacturing systems which are more complex and more capital-intensive. Modeling of modern manufacturing systems, therefore, has become more important than ever. Manufacturing systems modeling enables one to predict the performance of manufacturing systems and identify the effects of key design parameters on system performance. Three approaches exist for the analysis of manufacturing systems [5]: manufacturing simulation, analytical queueing models, and spreadsheet models. Figure 1 shows the general application area for each modeling approach according to random- ness (deterministic and stochastic) and time dependency (static and dynamic) of the systems. The spreadsheet models have been used for the analysis of manufacturing systems under the static and deterministic environment. Most uses of spreadsheet models for manufacturing systems have been based on the assumption that system characteristics, such as service times and job arrivals, can be modeled using static and deterministic values. System dynamics and uncertainties are usually ignored in this type of model. Static manufacturing problems, such as capacity and utilization analysis [1, 10], material requirements planning [3] and work-load allocation problems [9] are among those application areas of a spreadsheet model. The analytical model, which is mainly based on the queueing network theory, is another modeling approach. Analytical queueing models can describe the systems using mathemetical relationships between workstations in the system and estimate stochastic performance measures as well as static measures. Although they approximate the system performance under some assumptions which may be unrealistic from a manufacturing view point, analytical models are effective, especially for preliminary design, providing a rapid evaluation of the steady-state performance of many system alternatives. A number of computer based, analytical tools for the design and planning of manufacturing systems have been introduced. CAN-Q [14] may have been the first analytical modeling tool. Its theoretical foundation is provided by Jackson [6]. In CAN-Q, the system is modeled as a closed queueing network in which a fixed number of work pieces circulate randomly through the network with a set of routing probabilities. QNA [15] is another queueing network modeling tool. It is designed to approximate performance measures for a system that can be modeled as an open network of queues. A decomposition approximation is implemented in QNA by treating each queue as a GI/G/m queue with multi-servers (machines), unlimited waiting space, and FCFS processing ¢A~E 26/*--E 673