OR/MS Today - August 2003|
Versatile modeling package
By Gian Frontini and Andrew Burns
The Centre for Automotive Materials and Manufacturing (CAMM) at Queen's University is an industry, academia and government collaborative program dedicated to research and development in advanced materials, manufacturing methods and manufactured products in the automotive industry. One of the key programs in this centre focuses its research on the dynamics of cost in manufacturing processes and in the supply chain.
A complex array of advanced manufacturing technologies is routinely employed in manufacturing plants, including advanced quality systems, statistical process control, automation, robotics, lean and agile manufacturing, and just-in-time processing. The impact of these techniques on the quality and reliability of consumer products is impressive; the impact on productivity and profitability of the manufacturers is unproven.
Recent statistics on total factor productivity, which takes into account the impact of the capital employed, suggest that the manufacturing industry is heading toward capital deepening and smaller margins, rather than toward a more prosperous future. Our research confirms that cost dynamics is a critical factor in the long-term viability of a business. Even with the best process control techniques, variation due to common causes renders manufacturing systems profitable sometimes, but not all the time!
One way to study the dynamics of product cost is through simulation and comparison of the simulation with actual performance. CAMM's manufacturing economics research team uses system dynamics (SD), discrete-event (DE) or hybrid software depending on the nature of the problem.
In dynamic modeling of product cost, the end objective defines a "precision spectrum." At one end, we have the "30,000-foot view," with long-term time horizons and less detailed modeling of the process, where systems dynamics applications like iThink are typically employed. At the other end, we have the "plant floor" view, with extremely detailed representations of all aspects of the process in question, where we would use Simul8 and Arena.
The learning curve for a user of GoldSim is relatively steep, and can be traversed with relative ease, thanks to a clear description of the individual model elements and very good pop-up menus. The recent addition of tutorials and on-line help has been especially helpful for new users, and both nicely compliment the excellent printed documentation and example models. Less obvious are the more subtle uses of delays and discrete events to simulate the more complex interactions of resources and information. These features, designed for more advanced users, add major capabilities to this software and are not commonly available in other products that we have seen.
GoldSim does not have capability to identify entities and entity attributes; this necessitates fairly complex logic statements to accurately handle the flow of products through shared resources. In cases where there are a significant number of different products flowing through the system, we generally turn to pure discrete-event software.
Other features of GoldSim make it a good tool for our research on cost dynamics. The capability of easy read-write to and from spreadsheets is a major asset, and it solves one of the problems in introducing dynamic software to industrial managers. Product managers and R&D portfolio managers are familiar with business analysis done in spreadsheets, but consider dynamic software not worth the complexity of dealing with unfriendly interfaces.
The "control panel" common to SD software and easily created in most DE software is not highly developed in GoldSim, but we suggest that energy spent in this direction would not be well rewarded. We have found "control panels" and "flight simulators" singularly unpopular with mangers, who regard sliders and moving graphs with suspicion, because of the lack of transparency in the underlying logic.
As we become more familiar with the software, we have developed a "wish list" of features that we would like to see developed in GoldSim. It is our impression from discussions with the software developers that they are interested in taking user feedback and building new and effective capabilities into an expanding suite of tools.
We have tested GoldSim in teaching a manufacturing business strategy course with fourth-year engineering undergraduates, and the response has been so positive that we have included GoldSim models in the interactive CD tutorials of our textbook, "Manufacturing in Real-Time" .
Clicking on the plus sign of each container reveals the workings of each stage. In Figure 2 we have expanded the high temperature firing stage, which illustrates the handling of a discrete event (kiln firing), triggered by information which prioritizes the use of resources and tracks their utilization. Note that all of the process and economic inputs in this stage and right through the model are statistical distributions and not single numbers.
Expanding the second level container on the logic for loading the kiln (Figure 3) shows the complexity of the logic associated with sharing resources and prioritizing tasks. Developing and modeling this logic is not an easy task in GoldSim, and this particular example demonstrates this shortcoming.
The outputs can be shown with the use of a control panel (Figure 4), which is quickly built from user-friendly menus, and the statistical results are viewable by clicking on a button.
Some of the outputs demonstrate the variation in business conditions and the relationships between "time in system" (indicated by shipping frequency), cost and price. Figures 5a and 5b summarize the financial performance of the business with net operating profit before tax (NOPAT) and economic value added (EVA). These key business indicators will vary with inputs from the control panel sliders and allow a thorough analysis of business conditions. GoldSim does not have a built-in optimizer. However, data is easily exported for post-processing.
GoldSim is an excellent bridge between the worlds of system dynamics and discrete-event simulation, providing a means of including otherwise difficult scenario descriptions within strategic models. Overall, we feel that GoldSim offers a tool that is expanding in functionality; one that addresses the needs of the community of users who apply dynamic simulation to operations research and management science.
Gian Frontini is a senior research fellow and adjunct professor at Queen's University School of Applied Science in Kingston, Ontario. He teaches manufacturing strategy and is the author of a new book, "Manufacturing in Real-Time," which deals with the dilemma of people in industry and the advent of smart machines. Frontini is chairman of Free Forming Technologies Inc, a CAD/CAM metal forming enterprise. He has had an extensive industrial career in senior management including stints as vice president of technology for Alcan International and chief information officer for Alcan Inc.
Andrew Burns is a master's degree candidate at the Centre for Automotive Materials and Manufacturing, Queen's University. His thesis involves dynamic simulation, and he is an advanced user of the GoldSim software.
OR/MS Today copyright © 2003 by the Institute for Operations Research and the Management Sciences. All rights reserved.
Lionheart Publishing, Inc.
506 Roswell Rd., Suite 220, Marietta, GA 30060 USA
Phone: 770-431-0867 | Fax: 770-432-6969
Web Site © Copyright 2003 by Lionheart Publishing, Inc. All rights reserved.