Proceedings of the 1998 Winter Simulation Conference
D.J. Medeiros, E.F. Watson, J.S. Carson and M.S. Manivannan, eds.
SIMULATION AND PRODUCTION PLANNING FOR MANUFACTURING CELLS
Shahram Taj David S. Cochran
James W. Duda
University of Detroit Mercy Jochen Linck
College of Business Administration
Detroit, MI 48219-0900, U.S.A. Massachusetts Institute of Technology
Department of Mechanical Engineering
Cambridge, MA 02139, U.S.A.
ABSTRACT one part to go through the entire process) and elimination
of the need to repeatedly sort the parts within the process.
Simulation is used to verify the feasibility of the design of The conversion to a lean cellular manufacturing
manufacturing cells. The cell design, which combines new system could help improve quality, eliminate waste, and
and existing machines in a component manufacture, is reduce total costs. An ideal, lean cell (shown schematically
presented, showing the difficulties that can result with such in Figure 1) would have all machines needed to process a
a system. Simply changing the layout (arranging machines part located very close together, single-piece flow of parts
into cells) could provide some benefits, but these benefits between operations, and operators running multiple
were offset by a high level of required investment. The machine types. Such a system would improve the ability to
reasons for the increased costs include poorly matched trace defects back to the machine that caused them and
cycle times, machine downtimes, complex material would eliminate nearly all of the transportation and storage
handling and long changeovers. Improvements in machine of parts between operations.
and material handling designs were found to be necessary
in order to increase cell performance and reduce Station
4
investment to a feasible level.
Station Worker Station
5 1 3
1 INTRODUCTION Station Station
6 2
The purpose of this study is to show how simulation can be Station
7
Worker Station
1
2
used to examine the feasibility of converting an existing
system to a cellular manufacturing system at a component Out In
manufacturer for a major automotive company. An existing Part flow Operator Station Machine
2
manufacturing system was already in place and had been
operating for several years. The original motivation for
redesigning the existing system was the need to upgrade Figure 1: Lean Manufacturing Cell
some of the operations. Several of the machines in use
were rather old (up to 40 years old) and management 1.1 Product Information
desired to replace the old equipment with machines using
more modern technology (CNC) to reduce cost and ensure The product being manufactured at this factory is a
better part quality. Because some areas of the factory matching set consisting of one part called P and one part
would require significant changes in terms of layout and called R for use in a sub-assembly. The existing system
material handling when the machines are replaced, this was produces these parts in five different sizes (or families)
seen as a potential opportunity to redesign the entire with 2-7 ratios per family for a total of about 20 different
system to improve efficiency and eliminate waste. It was varieties of sets. Total production volume of the system is
also desired that the new system be designed with as much approximately 11,000 sets per day, or about one set every 6
volume and product mix flexibility as possible, as the seconds. A typical R part is about 9 inches in diameter, 2.5
factory is facing increased diversity and uncertainty in inches thick, and weighs about 15 lbs. A typical P part is
demand for its products. Other specific goals included the about 8 inches long, 4 inches in diameter, and weighs
reduction of throughput time (defined as the time it takes about 8 lbs.
973
Taj, Cochran, Duda and Linck
1.2 Existing Manufacturing System cellular system significantly more complex (Wemmerlöv
and Johnson 1997). Specifically, in this case we will see
The existing system has a departmental, process layout, that issues of machine design and reliability as well as
with machines grouped by operation (i.e., all OP 40 material handling become critical in designing a successful
machines are located together in one area) rather than by cellular system. Much less literature exists relating these
product flow. The existing machines were originally problem areas to cell design, although some work has been
selected so as to optimize performance in each department done showing the importance of machine reliability and
(machine utilization, uptime, low operating costs, etc.). maintenance (Kelly et al. 1997, Banerjee and Flynn 1987,
Buffers, in the form of tubs full of parts, are kept and Flynn 1989). As more and more implementation
between some subsequent departments. Tubbing parts has factors become important, the problem of cell design
the advantage of keeping the parts sorted by ratio, but it becomes more and more difficult to study analytically, and
also results in increased inventory, increased throughput research tends to focus more on developing guidelines to
time (the tubs are not transported to the next department help designers make good decisions where there is a lack
until they are full) and the need for fork-truck drivers to of scientific analysis. Studies such as (Edwards 1996 and
move the tubs around. In other cases, parts are manually Charles 1997) provide guidelines for designing and
loaded to and from an overhead conveyor running from developing “lean” cellular systems that will be able to meet
one department to the next. The use of such conveyors a factory’s requirements.
reduces inventory and material handling, but since only
one conveyor is used, the parts get mixed up and operators 3 CELL DESIGN
at the downstream department often have to search the
conveyor to find the right part. We began by first determining the number of machines to
be included for each operation, i.e., how to balance the
2 LITERATURE REVIEW cycle times of the machines in the cell. This proved to be a
difficult problem, as the existing machines had a wide
Although a fairly large body of research exists on group range of cycle times, ranging from 6 sec/part for the fastest
technology and lean manufacturing, the situation described operation all the way up to over 200 sec/part for the
above was found to be somewhat unique relative to the slowest. The ideal case would be for each cell to contain
literature in terms of cell design and evaluation issues. exactly one of each machine for each operation (so that
Most previous work in these areas falls into one of two each part passing through the cell would go through each
main categories. On the one hand, there are several books machine), thus providing a very simple and clear material
presenting high-level ideas and concepts regarding how to flow. Since, for example, OP 50 currently requires 14
design a “lean” manufacturing system (Ohno 1988, machines running in parallel this would mean that there
Monden 1983, Shingo 1989, Womack 1997, etc.) On the would have to be 14 cells, and 14 of each type of machine.
other hand, several studies have also been done to examine However, the fastest operation, OP 20, currently requires
a few very specific issues involved in designing a cellular only 3 machines, which would mean that an “ideal”
system. cellular layout would require the purchase of 11 new
Several articles have been written on cell design machines for this operation.
techniques (e.g. Beaulieu et al. 1997, Christy 1986, Greene A number of alternative cell concepts were examined
and Sadowski 1984, and Rajagopalan and Batra 1975), but and from these, a medium sized cell was selected for
such studies generally treat the cell design process purely further investigation. A schematic of this cell is shown in
as the selection and grouping of machines into cells. In the Figure 2, showing the general layout including the number
case presented here the grouping of machines is very of machines for each operation (8 machines for OP 60 for
simple, as all parts require the same operations regardless example). Figure 3 shows the operator standard work chart
of family and ratio. Consequently, each cell must contain at for a medium sized cell.
least one of each type of machine, regardless of what
family of parts it is to produce. The only factor preventing
each cell from making each part type (each combination of
family and ratio) would be long changeover times on some
of the machines.
In some cases (see Levasseur et al. 1995, for example)
simply rearranging existing machines into a cellular
configuration is sufficient to provide significant benefits
such as reductions in inventory, throughput time, scrap,
labor, and floor space. In many other cases, several other
implementation issues arise that making the switch to a
974
Simulation and Production Planning for Manufacturing Cells
4.1 Modeling
OP 20
Cycle = 6 sec. Part P
PCs/Cycle = 1
Family C/O = 15 minutes
MTTF = 286 hours
The new cell design was modeled in Witness. The LABOR
MTTR = 2.69 hours
OP 30
element in Witness is used to represent the operator. In
Part R
Cycle = 45 sec.
PCs/Cycle = 4 Witness, LABOR is a resource like a human operator or
Family C/O = 30 minutes
MTTF = N/A
Operator
tools which may be required by other elements for
work loop processing, setting up, repair or loading. It should be noted
OP 40
Cycle = 22 sec.
OP 10
Cycle = 15 sec.
that LABOR in Witness is not an active element, it can not
PCs/Cycle = 1
PCs/Cycle = 1
Family C/O-New = 15 minutes
Family C/O = 15 minutes
MTTF = N/A
pull/push part or walk from station to station. To model
Family C/O-Old = Major
MTTF-New = 559 hours
MTTR-New = 1.90 hours
MTTR = N/A
walking or removing parts, another type of Witness
MTTF-Old = 40 hours
MTTR-Old = 4 hours element, MACHINE is used. These dummy MACHINES
then use LABOR. To control the movement of operators
OP 50
Cycle = 45,50,75 sec.
OP 50
Pallet
and prevent them from bypassing stations an ATTRIBUTE
PCs/Cycle = 1
Family C/O = 4 hours
MTTF = 20.31 hours
RR RR
PPPP
called position was assigned to them.
MTTR = 2.77 hours
When the simulation starts, the operator position is
OP 60 OP 60
zero. For Operator Task 1 (Figure 3) a dummy MACHINE
Cycle = 180, 200 sec.
PCs/Cycle = 1 pair pulls two part R from tub. An operator with position zero is
Family C/O = 4 hours
Gantry
MTTF = 216 hours
MTTR = 8 hours
required for this machine during the cycle time. At the end
Machine Load/Unload =
45 Sec of cycle, the operator's position is set to 1. Walking to part
MTTF = 29.22 hours OP 60 OP 60
MTTR = 1.31 hours P loading station and taking two P parts from tray and
OP 60 OP 60 walking to OP20 (Task #2) is modeled using another
dummy MACHINE needing the operator with position 1.
OP 60 OP 60
This dummy machine can only cycle when it obtains the
operator with the proper position. At the end of the cycle,
Figure 2: Schematic of a Medium Sized Cell Layout the operator position is then set to 2. Other operator tasks
are modeled in the same fashion except for Task #5. In
4 SIMULATION ANLYSIS Witness a MACHINE can cycle when it has obtained all its
needed PARTS and LABORS. In Task #5, the operator
Simulation analysis was used to verify the cell design in loads parts to cooling buffer and then walks without parts
terms of production feasibility. Process flows, actual and to pallet loading station. In this case, there are no parts for
estimated cycle times, change over times, maintenance operator to carry. To model this walking with no parts, a
repair data, machine downtimes, and work patterns were dummy MACHINE needing the operator (with proper
used for the foundation of the simulation model. Standard position) is used to model the walk that pulls a dummy
work charts detailing the required operator activities were PART. After operators finishes Task #6, its position-
also developed. attribute is then reset to zero.
975
Taj, Cochran, Duda and Linck
PART: 2 sets of R + P Operators: 1
PROCESS TIME (secs)
# OPERATION Man Walk Auto 20 40 60 80 100 120
1 OP 10 8 6 30
2 2 3 0
3 OP 20 18 5 12
4 OP 30 15 4 23
5 OP 40/50 4 1 25
6 OP 60 25 7 25
# Operator Tasks
1 Take 2 part R from tub and load into OP 10 [8 sec.]
Walk to part P loading station [6 sec.]
2 Take 2 P parts from tray [2 sec.]
Walk to OP 20 [3 sec]
3 Unload/load (2 P parts) OP 20 [18 sec.]
Walk to OP 30 [5 sec.]
4 Unload/load (2 P parts) OP 30 [15 sec.]
Walk to cooling buffer input [4 sec.]
5 Load 2 P parts to cooling buffer input [4 sec.]
Walk to pallet loading station [1 sec.]
6 Load 2 P parts and 2 R parts pallet [25 sec.]
Walk to part R loading station [7 sec.]
Figure 3: Standard Work Chart for a Medium Sized Cell Run by One Operator
When the simulation starts, the operator position is one P and one R to each machine. When it reaches the end
zero. For Operator Task 1 (Figure 3) a dummy MACHINE of the row the pallet leaves the cell to go to other
pulls two part R from tub. An operator with position zero is operations (later the empty pallet returns). We have used
required for this machine during the cycle time. At the end VEHICLE and TRACK elements in Witness to represent
of cycle, the operator's position is set to 1. Walking to part the Gantry and its pathways and speed. A shortcoming of
P loading station and taking two P parts from tray and using VEHICLE to represent Gantry is modeling the
walking to OP20 (Task #2) is modeled using another downtime. In Witness, the VEHICLE does not break down.
dummy MACHINE needing the operator with position 1. A dummy MACHINE is used to model up-time and
This dummy machine can only cycle when it obtains the downtime for Gantries. During the downtime, the Gantry is
operator with the proper position. At the end of the cycle, pushed into a dummy TRACK and it stays there until the
the operator position is then set to 2. Other operator tasks dummy machine is repaired.
are modeled in the same fashion except for Task #5. In To meet forecasts of future volume, seven cells are
Witness a MACHINE can cycle when it has obtained all its required. Each was modeled in a similar way. All
needed PARTS and LABORS. In Task #5, the operator downtime and repair times were modeled using appropriate
loads parts to cooling buffer and then walks without parts statistical distributions with different pseudo-random
to pallet loading station. In this case, there are no parts for number streams.
operator to carry. To model this walking with no parts, a
dummy MACHINE needing the operator (with proper 4.2 Simulation Run
position) is used to model the walk that pulls a dummy
PART. After operators finishes Task #6, its position- Simulated manufacturing cells were executed to study the
attribute is then reset to zero. feasibility of the proposed manufacturing cells for
The pallet containing four pairs of parts P and R is production in a Just-In-Time (JIT) environment. The future
released to OP60. Two Gantries move pallets to two rows volume of about 77,000 pairs of P and R per week with a
of 4-machines (Figure 2). The Gantry unloads/load a set of complexity (mix) of 33 different types was used as a
976
Simulation and Production Planning for Manufacturing Cells
schedule for the cells. The cell performance was to meet the requirements of a cellular system: fast changeover,
customer volume and mix requirements on a daily basis. good operator access, high reliability, cycle times matched
The model was run for 10 weeks. A series of matrices was to customer takt time, etc.
generated to show the status of production for every part-
type for every day of the run and the production hours REFERENCES
required for every cell. From these matrices, descriptive
statistical analyses are easily performed. Banerjee, A., and B.B. Flynn. 1987. A simulation study of
some maintenance policies in a group technology
4.3 Findings shop. International Journal of Production Research
25:1595-1609.
The simulation results showed that not all part types can be Beaulieu, A., A. Gharbi, and Ait-Kadi. 1997. An algorithm
produced in a JIT environment due to long set ups and for the cell formation and the machine selection
downtimes. In order for a cellular system to be feasible, it problems in the design of a cellular manufacturing
was clear that certain improvements in the equipment system. International Journal of Production Research
would have to be made. For example, current changeover 35:1857-1874.
times are as long as 4 hours or more for some operations. Charles, M.D. 1997. the impact of machine and cell design
Although most cells would be dedicated to a particular on volume flexibility and capacity planning. Master's
family of parts, one or two cells would be required to thesis, Massachusetts Institute of Technology.
produce multiple part types. Poor reliability (uptimes of Christy, D.P., and U. Nandkeolyar. 1986. A simulation
approximately 70%) was also a problem for the cellular investigation of the design of group technology cells.
system as well as the current system, and efforts were In Proceedings of the Decision Science Institute, 1201-
already underway to reduce unplanned downtime in these 1203.
problem areas. Edwards, D.K. 1996. Practical guidelines for lean
Several other less obvious problems were identified as manufacturing equipment. Production and Inventory
well. Because most of the machines were designed with a Management Journal 2:51-55.
mass-production mentality, they presented a variety of Flynn, B. 1989. Critical machines preventive maintenance
impediments to the formation of cells. Some machines policies for group technology shops,” International
were very large and this resulted in long walking distance Journal of Production Research. 27:2009-2020.
for operators expected to run multiple machines. Had the Greene, T.J., and R.P. Sadowski. 1984. A review of
machines been designed with a cellular layout in mind, cellular manufacturing assumptions, advantages and
they clearly could have been improved in this regard. design techniques. Journal of Operations Management
Another problem with some of the particularly large 4:85-97.
machines is that they require a special concrete foundation, Kelly, C. M., C.T. Mosier, and F. Mahmoodi. 1997. Impact
resulting in a tremendous amount of work being required to of maintenance policies on the performance of
relocate the machines. Plant management has recently manufacturing cells. International Journal of
placed an emphasis on purchasing only machines that can Production Research 35:767-787.
be moved more easily in recognition of the fact that layouts Levasseur, G.A., M.M. Helms, and A.A. Zink. 1995. A
are always changing, but a number of the older machines conversion from a functional layout to a cellular
do not meet these new standards. manufacturing layout at Steward, Inc. Production and
Inventory Management Journal 36:37-42.
5 CONCLUSIONS Monden, Y. 1983. Toyota production system. Norcross: IIE
Press.
This paper reviewed the investigation of creating a cellular Mosier, C.T., D.A. Elvers, and D. Kelly. 1984. Analysis of
system design at a factory with a mass production-type group technology scheduling heuristics. International
system currently in place. Although the manufacturing Journal of Production Research 22:857-875.
process is ideal for a cellular system in some ways (all Ohno, Taiichi. 1988. Toyota production system: beyond
parts require the same sequence of operations, for example) large scale production. Cambridge: Productivity
it was extremely difficult to design a feasible system due to Press.
the constraints imposed by the existing equipment. Shingo, Shigeo. 1989. A study of the Toyota production
Problems caused by the existing machines included long system from an industrial engineering point of view.
changeover times, complex and expensive automated Cambridge: Productivity Press.
material handling, unbalanced cycle times, and poor Witness. 1997. Version WIN-307, Release 8.30, Lanner
machine reliability. In order for a cellular system to be Group.
successful in this environment, it would require that at least Wemmerlöv, U., and D.J. Johnson. 1997. Cellular
some of the machines be designed or redesigned to meet manufacturing at 46 user plants: implementation
977
Taj, Cochran, Duda and Linck
experiences and performance improvements.
International Journal of Production Research 35:29-
49.
Womack, James P. 1996. Lean thinking: banish waste and
create wealth in your corporation. New York: Simon
& Schuster.
AUTHOR BIOGRAPHIES
SHAHRAM TAJ is a Professor in the College of Business
Administration at the University of Detroit Mercy. He
received a B.S. degree in Operations Research and Systems
Analysis from the College of Planning and Computer
Applications (Iran), M.S. in Industrial Engineering from
the University of Rhode Island, and Ph.D. in Industrial
Engineering and Operations Research from the University
of Massachusetts (Amherst).
DAVID COCHRAN is an Assistant Professor in the
Department of Mechanical Engineering at the
Massachusetts Institute of Technology. He received a B.S.
degree in industrial engineering from Auburn University, a
M.S. degree in manufacturing systems engineering from
Pennsylvania State University, and a Ph.D. degree in
industrial and systems engineering from Auburn
University.
JIM DUDA is a Ph.D. candidate in the Department of
Mechanical Engineering at the Massachusetts Institute of
Technology. He received a B.S. degree in mechanical
engineering from Case Western Reserve University, and he
received his M.S. degree in mechanical engineering from
at the Massachusetts Institute of Technology.
JOCHEN LINCK is a Ph.D. candidate in the Department
of Mechanical Engineering at the Massachusetts Institute
of Technology. He received a Dipl.-Ing. degree in
mechanical engineering from RWTH Aachen.
978
