Edward J. Williams
Marcelo Zottolo
Production Modeling Corporation
Three Parklane Boulevard, Suite 1006 West
Dearborn, Michigan 48126
United States
{ewilliamsmzottolo}@pmcorp.com

KEYWORDS

Industrial processes, Performance analysis, Discrete simulation, Simulation interfaces, Process-oriented.

ABSTRACT

Simulation has long been recognized as an analytical tool of high power and wide applicability when applied to the improvement of manufacturing processes. Indeed, historically, manufacturing applications were the first major purview of simulation usage by large companies. As manufacturing processes increase in complexity, both operational and economic, the capabilities of simulation increase both in importance and in difficulty of duplication by alternative analytical methods. In this paper, we examine in detail the application of simulation to an automotive stamping plant. Simulation identified bottlenecks in material handling, pointed the way to increasing utilization of a costly stamping press and quantified the relationships between specific capital investments under consideration and the throughput increase to be expected.

INTRODUCTION

Manufacturing industries and their complex processes have long been justifiably recognized as the earliest and still among the most frequent users of simulation for productivity improvement (Miller and Pegden 2000). In turn, the automotive industry, among the largest participants within the manufacturing sector in many countries (the United States, in this case) and hence a bellwether of the national economy, has traditionally maintained and enhanced its competitiveness with the help of simulation analyses to increase efficiency within its extensive manufacturing operations (Ülgen and Gunal 1998).

Here, we examine a case study of simulation application to productivity improvement within one stamping plant owned and operated by a major automotive company in the United States. This client wished to increase throughput in response to changing and increasing market demand while simultaneously increasing utilization of the expensive stamping press (indeed, the keystone of the process) and reducing congestion among the material-handling vehicles responsible for delivering panels to, and taking them away from, this stamping press. After providing an overview of the manufacturing process, we describe the construction, verification, and validation of the simulation model used in the analysis of this process. We then outline the results and recommendations obtained from the model and indicate the directions of expected follow-up analyses. Numerous precedents of this type of study appear in the literature. For example, (Hoff et al. 1997) focused attention on automated guided vehicle systems; (Mosca, Queirolo, and Tonelli 2002) focused on job-sequencing in a system, like this one, dominated by an expensive machine; and (Mossa and Mummolo 2002) examined a manufacturing system characterized by intermixed automatic and manual operations.

THE MANUFACTURING PROCESS AND ITS CONCERNS

Since the press is the centerpiece of the production operations within the scope of the study, the manufacturing process is conveniently viewed and described as partitioned into one portion upstream from the press (input side) and one downstream (output side) from the press. On the upstream side, an SGV [self-guided vehicle] (Heizer and Render 2004) fetches a pallet, loaded with many (potentially more than 200; the specific quantity a system variable) steel blankets, from the input supply dock. This pallet is transferred first to the loaded pallets stand and then, via forklift and operator, to the empty destacker. After the forklift operator removes all dunnage from the steel blankets, the loaded pallet is shuttled to the press, which empties it by withdrawing one blanket at a time for the stamping operation. Subsequently, the empty pallet is shuttled out of the press via the destacker and transferred, again via forklift and operator, to the empty pallet stand and thence to an SGV. The SGV returns the empty pallet to the dock and returns with a full pallet, thereby completing the upstream operational cycle of supply to the press. These SGVs are powered by lead-acid storage batteries, which must periodically be recharged; to operate safely and effectively, an SGV must have between 20% and 80% of the theoretically maximum charge. Therefore, each SGV must be monitored and routed to one of the charging spurs in the system when its charge level approaches the lower threshold. An SGV exhausts its charge much more rapidly when moving than when standing still. The complexities introduced into the system and its model representation by these considerations resemble the complexities examined for the same reason in a study by (Norman 2002).

On the downstream side, the press feeds three vertical conveyance systems [VCSs] via two conveyor belts, as depicted in Figure 1.

Figure 1: Downstream Material Handling of Press Output

Downtimes, when they occur, significantly affect material routing. If shuttle conveyor #1 is down (worst case), neither VCS 1 nor VCS 2 can be fed; the press can send output only to VCS 3 via belt 3. If shuttle conveyor #2 is down, VCS 2 cannot be fed; the press will alternate its output between VCS 1 and VCS 3. If shuttle conveyor #3 is down, the press will alternate its output between VCS 1 and VCS 2 via belt 1-2. No parts can be routed to VCS n (n = 1, 2, or 3) if it is down, since accumulation of parts in either belt is forbidden.

CONSTRUCTION, VERIFICATION, AND VALIDATION OF THE MODEL

At the client’s request, the model was developed using the WITNESS® simulation software tool marketed by the Lanner Group worldwide. This software provides its own internal programming-logic language; simulation constructs such as machines, buffers, vehicles, and tracks; concurrent development of a simulation model and its accompanying two-dimensional animation, nearly automatic provision for basic statistical output reports, and ability to read input data from Microsoft® Excel workbooks (Mehta and Rawles 1999). The client had previously adopted the policy of developing almost all simulation models using one of only two software tools (one of these being WITNESS®) to achieve better return on the investment of training its own industrial, production, and/or process engineers in simulation and to make possible easier transferability of models, among these engineers and consultant engineers collectively, in different corporate departments or divisions (Williams 1996).

To encourage frequent use of the model among client engineers (not all of whom know the details of WITNESS® software) to explore various alternatives, and to make such use more convenient, more rapid, and less prone to input data error, system variables likely to change were placed into worksheets within a Microsoft® Excel workbook – as vigorously recommended by many expert practitioners of simulation (Ülgen et al. 1994) – and thence read into the model at initialization time. Examples of variables thus input were the numbers of SGVs at the input and output sides of the press, the speeds of these vehicles either loaded or unloaded, the number of blanks per pallet, the press rate, and numerous pallet transfer times within the system.

The client and consultant engineers agreed upon and documented underlying model assumptions, or restrictions of its scope, necessitated by the confluence of absence of data and project time constraints. The time constraints were imposed by client management due to exigencies of making capital-expenditure decisions and implementing production increases and improvement efficiencies on short notice. Examples of these accommodations to a tight schedule included:

1.All times to fail and times to repair are exponentially distributed.

2.The model will run only one part type at a time, without inclusion or examination of changeover times.

3.Production runs on three abutting eight-hour shifts per business day.

4.Operators and forklift trucks have zero downtime. Here too, the explicit acknowledgment and documentation of assumptions, forestalling potential later misapplication of a simulation model within a context it is unequipped to handle, represent practices vigorously expounded on the basis of experience (Musselman 1994).

Construction, verification, and validation of the model were naturally divided into two major components – the upstream side of the press, and the downstream side of the press, matching the canonical view of the manufacturing process as described in the previous section. The press itself was represented as a WITNESS® “machine;” the blankets it stamped were represented as WITNESS® “parts.” The SGVs responsible for material handling and transport were represented as WITNESS® “vehicles” running on WITNESS® “tracks.” In WITNESS® modeling methodology, a vehicle boards a track at the rear and travels to the front, and the logic specifying where (i.e., to which other track) the vehicle should go next is specified within the track, not within the vehicle. Whenever a vehicle reached the front of the track, logic checked the level of its battery charge; if this level was too low to allow the vehicle to proceed on its canonical route, it was routed to a charging spur. On such a charging spur, the vehicle spent the required recharging time loading a dummy “part” conceptually representing electrical energy. Operators were represented as WITNESS® “machines.”

Verification of the model entailed numerous walkthroughs, in which a total of four analysts participated, especially of the complex internal code specifying SGV routing logic. Additional verification techniques used included desk checking, traces, and careful examination of the animation accompanying the simulation model (Carson 2002). Validation techniques used in several meetings among the client and consultant engineers included walkthroughs of the model entities, examination of the animation, and Turing tests (Law and Kelton 2000).

MODEL RESULTS AND OUTPUT

First, the model was run with no equipment downtime. Equipment downtime can cause the parts material flow to stop. The stamping press might be forced to slow down (i.e., suffer blockage) depending on which equipment fails (location in the material handling chain) and the duration of the failure. Leaving the downtime aspect out of the experimentation allowed identifying any and all blockages of the press due to material handling capacity. Several operating conditions were evaluated, as indicated in the table below: For the experimentation, various combinations of all of these parameters were utilized. Thirty-six scenarios were run, each with a warm-up of 24 hours and a run time of 5 days. As previously stated, the stamping press operating statistics were the main performance metric under investigation.

Press starvation (idleness) represented how robustly the press input side material handling system is designed. In the same way, press blockage represented how robustly the press output side material handling system is designed.

When only one forklift truck is available at the Marketplace (the final downstream destination, beyond Figure 1, for stamping press output), increasing the number of SGVs available (at the output side of the press) does not reduce the press blockage (about 60%). This is true for VCS operator’s cycle time of 11, 15 or 20 seconds. It is evident that, when only one forklift truck is available at the Marketplace and under the given assumption of 2 minutes to load or unload an SGV cart, the number of SGVs used at the output side and the operator’s cycle time at the VCS have no impact on the press performance statistics.

1.The same effect discussed in the previous item is observed when only two forklift trucks are available at the Marketplace. Under the assumption of 2 minutes to load OR unload an SGV cart and 3 carts per SGV, regardless of the operator’s cycle time, increasing the number of SGVs does not reduce the press percentage blockage. Increasing the number of SGVs at the output side only causes the queue for the Marketplace to increase. SGVs cannot leave the Marketplace any oftener than 1 every 6 minutes.

2.For any combination of number of SGVs at the input and output side and VCS operator’s cycle times of 11 or 15 seconds, as the number of forklifts (loading/unloading stations) at the Marketplace is increased, press blockage is reduced. Increasing the number of SGVs at the output side helps reduce press blockage by allowing a more efficient delivery of stamped parts to the Marketplace. However, when the operator’s cycle time is 20 seconds, making available a third forklift truck does not help reduce the percentage block of the press. It seems that for operator’s cycle time of 15 seconds and above, increasing the number of forklift trucks stops having a positive impact on the press performance. Therefore, depending on the range of the operator’s cycle time utilized, the number of forklift trucks at the Marketplace might have a significant positive impact on the press performance or no impact at all.

3.8 to 10 trays are adequate for maximum VCS efficiency.

4.At the assumed press rate (750 strokes/hour – worst case scenario), the operators manually loading stamped parts at the VCSs receive a set of parts from the press every 14.4 seconds. When three forklift trucks are available at the Marketplace, assigning a loading cycle of 11 seconds allows this manual operation to keep up with the press rate. Under this Marketplace and operator’s cycle time conditions, 10 SGVs at the output side are enough to avoid blocking the press. However, when only 4 SGVs at the input side are allocated and under the previously mentioned conditions (11-second cycle time, 10 SGVs, and 3 forklifts), they cannot continuously feed the press and approximately 1.3% press starvation is observed. With 5 SGVs, no starvation appears. This statement holds true under any of the 36 scenarios evaluated. Therefore, minimum SGVs required are 5 and 10 at the upstream and downstream side of the stamping press, respectively.

5.Due to different downtimes (stamping press, distribution conveyors, VCSs, etc.) the average throughput is 671 strokes/hour when the press runs at 750 strokes/hour.

6.At a press rate of 750 strokes/hour, one part every 4.8 sec is stamped. Considering a rack capacity of 20 pieces, it takes 96 seconds to fill a rack. SGVs transport three trays (with one rack in each), therefore, an SGV is needed at the VCSs every 4.8 minutes (4.8 × 20 × 3/60). As a consequence, SGVs should be released from the Market Place every 4.8 minutes to stay abreast of the stamping press rate. The chart below shows how press performance is affected by the time it takes to service an SGV (unload it and reload it) at the Marketplace. Below 9 minutes it is no longer a system constraint. The first three bars assume 8 minutes to service an SGV at the input dock. The fourth bar assumes 4 minutes to service an SGV at the input dock.

7.Simulation analysis improved locations for SGV opportunity chargers in front side of the press.

CONCLUSIONS AND FUTURE WORK

The simulation study described here has provided significant help and value to the client by confirming the throughput of the system under a variety of operating conditions, identifying proportions of time capital-intensive resources (e.g., machines, material-handling equipment, and operators) spend in states of starvation, busy, blocked, idle, or down, and by assessing the sensitivity of the upstream side of the system to the number and location of recharging spurs available to the SGVs used there. In meetings held with the client engineers and managers as the current phase of the study drew to a close, the simulation analysts presented extensive written documentation, both internal and external to the model itself, to assist all parties in implementing planned extensions to the model; the importance of such documentation, relative to achieving both model credibility and model usability over a lifetime equal to that of the process under study, has been extensively discussed in the literature, particularly in the seminal paper (Oscarsson and Moris 2002). These planned extensions to this study currently include implementing the ability to run multiple part types, thus entailing incorporation of die changeover times into the model, incorporation of operator break times within the model, and assessing the sensitivity of the system to a third belt (“belt 4” if the nomenclature above is extrapolated) on the downstream side of the system. That belt currently exists but is used only as a backup material-handling device.

ACKNOWLEDGMENTS

The authors extend warm appreciation to Dr. Onur M. Ülgen, president of Production Modeling Corporation and senior professor at the University of Michigan – Dearborn, for his mentoring, support, and guidance throughout this project. Additionally, an anonymous reviewer provided valuable suggestions to improve this paper.

REFERENCES

Carson, John S., II. 2002. “Model Verification and Validation.” In Proceedings of the 2002 Winter Simulation Conference, Volume 1, eds. Enver Yücesan, Chun-Hung Chen, Jane L.Snowdon, and John M. Charnes, 52-58.

Heizer, Jay, and Barry Render. 2004. Operations Management,7th edition. Upper Saddle River, New Jersey: Pearson Education, Incorporated.

Hoff, Axel, Holger Vogelsang, Uwe Brinkschulte, and Oliver Hammerschmidt. 1997. Simulation and Visualization of Automated Guided Vehicle Systems in a Real Production Environment. In Proceedings of the 9th European Simulation Symposium, eds. Winfried Haha and Axel Lehmann, 391-395.

Law, Averill M., and W. David Kelton. 2000. Simulation Modeling and Analysis, 3rd edition. Boston, Massachusetts: The McGraw-Hill Companies, Incorporated.

Mehta, Arvind, and Ian Rawles. 1999. “Business Solutions Using WITNESS.” In Proceedings of the 1999 Winter Simulation Conference, eds. Phillip A. Farrington, Harriet Black Nembhard, David T. Sturrock, and Gerald W. Evans, 230-233.

Miller, Scott, and Dennis Pegden. 2000. “Introduction to Manufacturing Simulation.” In Proceedings of the 2000 Winter Simulation Conference, eds. Jeffrey A. Joines, Russell R. Barton, Keebom Kang, and Paul A. Fishwick, 63-66.

Mosca, Roberto, Filippo Queirolo, and Flavio Tonelli. 2002. Job Sequencing Problem in a Semi-Automated Production Process. In Proceedings of the 14th European Simulation Symposium, eds. Alexander Verbraeck and Wilfried Krug, 343-347.

Mossa, Giorgio, and Giovanni Mummolo. 2002. Simulation Modelling of both Automated and Human-Based Manufacturing and Assembly Line. In Proceedings of the International Workshops on Harbour, Maritime, & Multimodal Logistics Modelling and Simulation; Modelling & Applied Simulation 2002, eds. Agostino G. Bruzzone, Yuri Merkuryev, and Roberto Mosca, 103-111.

Musselman, Kenneth J. 1994. “Guidelines for Simulation Project Success.” In Proceedings of the 1994 Winter SimulationnConference, eds. Jeffrey D. Tew, S. Manivannan, Deborah A. Sadowski, and Andrew F. Seila, 88-95.

Norman, Michael. 2002. “A Recursive Method for Traffic Management Through a Complex Path Network.” In Proceedings of the 2002 Winter Simulation Conference, Volume 1, eds. Enver Yücesan, Chun-Hung Chen, Jane L. Snowdon, and John M. Charnes, 537-541.

Oscarsson, Jan, and Matías Urenda Moris. “Documentation of Discrete Event Simulation Models for Manufacturing System Life Cycle Simulation.” In Proceedings of the 2002 Winter Simulation Conference, Volume 2, eds. Enver Yücesan, Chun-Hung Chen, Jane L. Snowdon, and John M. Charnes, 1073-1078.

Ülgen, Onur M., John J. Black, Betty Johnsonbaugh, and Roger Klungle. 1994. “Simulation Methodology in Practice – Part II: Selling the Results.” International Journal of Industrial Engineering – Applications and Practice 1(2):129-137.

Ülgen, Onur M., and Ali Gunal. 1998. “Simulation in the Automobile Industry.” In Handbook of Simulation, ed. Jerry Banks, New York, New York: John Wiley & Sons, Incorporated, 547-570.

Williams, Edward J. 1996. “Making Simulation a Corporate Norm.” In Proceedings of the 1996 Summer Computer Simulation Conference, eds. V. Wayne Ingalls, Joseph Cynamon, and Annie V. Saylor, 627-632.

AUTHOR BIOGRAPHIES

EDWARD J. WILLIAMS holds bachelor’s and master’s degrees in mathematics (Michigan State University, 1967; University of Wisconsin, 1968). From 1969 to 1971, he did statistical programming and analysis of biomedical data at Walter Reed Army Hospital, Washington, D.C. He joined Ford Motor Company in 1972, where he worked until retirement in 2001 as a computer software analyst supporting statistical and simulation software. After retirement from Ford, he joined Production Modeling Corporation, Dearborn, Michigan, as a senior simulation analyst. Also, since 1980, he has taught evening classes at the University of Michigan, including both undergraduate and graduate simulation classes using GPSS/H, SLAM II, SIMAN, Promodel, or SIMUL8. He is a member of the Institute of Industrial Engineers [IIE], the Society for Computer Simulation International [SCS], and the Michigan Simulation Users' Group [MSUG]. He serves
on the editorial board of the International Journal of Industrial Engineering – Applications and Practice. During the last several years, he has given invited plenary addresses on simulation and statistics at conferences in Monterrey, México; İstanbul, Turkey; Genova, Italy; and, Latvia. His email and web addresses are <[email protected]> and <wwwpersonal.umd.umich.edu/~williame>.

MARCELO ZOTTOLO, born in Buenos Aires, Argentina, came to the United States to finish his college studies. He was graduated from the University of Michigan - Dearborn as an Industrial and Systems Engineer in December 2000. He was awarded the Class Honors distinction and his Senior Design Project was nominated for the Senior Design Competition 2001. This project studied the improvement of manufacturing processes for the fabrication of automotive wire harnesses, ultimately proposing an automation tool leading to improvements in future designs. Additionally, he was co-author of a paper on simulation in a distribution system which earned a “best paper” award at the Harbour, Maritime, and Simulation Logistics conference held in Marseille, France, in 2001. He is currently working for Production Modeling Corporation as a Simulation Engineer. There, his responsibilities include building, verifying, validating, and analyzing simulation models in WITNESS and SIMUL8 for large corporate clients; he also presents in-house training seminars. His email address is <[email protected]>