You are on page 1of 7

FUNCTIONAL VS CELLULAR LAYOUTS:

USING SIMULATION AS A COMPARISON TOOL

Abdessalem JERBI *, Hédi CHTOUROU *,**and Aref Y. MAALEJ *


*
Laboratoire des Systemes Electro-Mecaniques
Ecole Nationale d’Ingénieurs de Sfax
BP W 3038, Sfax – Tunisie
**
Département de technologie
Institut Préparatoire aux études d’Ingénieurs de Sfax
Sfax -Tunisie

ABSTRACT
A number of simulation studies were carried out by several researchers in order to compare the performances of cellular
and functional layouts. The majority of these studies often lack objectivity in their methodologies and lead to paradoxical
results. The aim of this study is the development of simulation models for both layouts. These models are used for the
illustration of some objectivity lacks in one of the main comparative studies.
KEYWORDS
Functional layout - Job shop - Cellular manufacturing - simulation - model – comparison-objectivity

INTRODUCTION that none of the two layouts outperforms the other one in the
Cellular manufacturing, a facet of group technology, has totality of the tested contexts. These paradoxical results may
emerged as one of the major techniques being used for the be imputed to the different contexts in which the comparison
improvement of manufacturing competitiveness. Several test series were carried out as well as to the lack of neutrality
studies tempted to verify such an issue by comparing the of a number of these contexts.
cellular layout (CL) to the functional layout (FL). These Hence, the goal of this paper is to present simulation
studies may be classified into three main categories: models that can be used to illustrate the objectivity lack of
simulation based investigations, analytical models and the main comparative studies and eventually, to achieve an
empirical surveys. Simulation based studies represent the objective comparison methodology of the two layout types.
mainstream of the work in this domain. In this way, Morris Accordingly, the second section is dedicated to a taxonomy
and Tersine (1990), Jensen et al (1996) and Faizul huq et al of existing FL-CL comparison research. It mainly focuses on
(2001) found that the cellular manufacturing system cannot the main experimental factors and performance measures
outperform its functional counterpart even though machine used in the published simulation studies. The third section
reservation that characterize cellular layout assures a deals with the development of simulation models for both
considerable setup reduction. In contrast, Shafer and layouts using a commercial simulation tool. Finally and
Charnes (1992) proved the superiority of the cellular layout before concluding, a section is reserved to the illustration of
when using the overlapping of production operations the objectivity lack of one of the main comparison studies
combined to setup reduction. Besides, Suresh and Meredith done by Morris and Tersine (1990).
(1994), Shafer and Charnes (1995) and Li (2003) showed
TAXONOMY OF EXISTING COMPARISON studies considered very high degrees decomposability
RESEARCH leading to completely independent cells.
A total of 8 simulation-based comparative simulation studies
Batch size (q)
are considered in the present investigation. They were
In order to reduce the number of the machines setup and
published in top specialized refereed journals between 1990
transport between work stations, products are generally
and 2003. Some of the 16 main parameters used in these
manufactured and transported in batches. Many authors used
studies were kept constant throughout the simulation series
the batch size as variable factor and demonstrated that the
(e.g.: the number of cells, the number of product types, the
use of small batch sizes in combination with an efficient
number of machines, the number of departments ...). On the
scheduling rules results in the improvement of the cellular
opposite, some others such as the lot size, the setup time and
layout.
the setup reduction factor, were generally considered as
Scheduling rules
variable experimental factors. In addition, the most widely
Jobs arriving at a department or a cell may have to wait in a
used performance measures are the mean flow time (MFT)
queue until the required machine becomes available. The
and the work in process (WIP). Also, throughput and some
order in which the jobs are to be processed is generally
due-date-related performance measures are used in some of
specified by a scheduling rule or a priority rule such as “First
the studies.
Come First Served” (FCFS), “Shortest Process Time” (SPT),
MAIN EXPERIMENTAL FACTORS “Earliest Due Date” (EDD) or “Repetitive Lots” (RL). Using
the first, jobs are simply selected according to their arrival
General MS characteristics
times whereas the second rule prioritizes the job with the
The studied MSs are characterized by a number of machines
shortest processing time. Alternatively, the EDD rules aims
ranging between 12 and 30. These are either arranged into
at reducing the number of tardy jobs by prioritizing jobs
departments or else into manufacturing cells. The number of
with the earliest due dates. The limited versions (-L) of these
departments (d) ranges between 4 and 15 whereas the
three rules are often used in order to avoid the duplication of
number of cells (c) ranges between 3 and 5. Furthermore, the
machines setup for the same product type. Finally, when the
studied systems were designed for a demand pattern
RL rule is used, jobs of the same type that the one just
comprising 3 to 80 product types (t) belonging to a number
processed are always prioritized in order to avoid
of families (f) varying between 3 and 10. Each product type
unnecessary setups.
requires a number of manufacturing operations ranging
between 2 and 25. Finally, only one of the treated studies Transfer mode
dealt with the labor resources whereas the others overlooked In the FL, the interdepartmental distances are considerable.
this factor assuming that a worker is always available when Hence, products are often transferred by batches or lots in
ever he is needed. order to reduce the transfer costs. Some studies used this
transfer mode in the CL whereas others exploited the
Degree of decomposability of the part machine matrix
proximity of machines of a same cell to transfer products by
(DD)
part. The “part by part” transfer mode allows simultaneous
This factor translates the feasibility of the decomposition of
execution of several operations on the same batch called
the MS into independent cells. The more the
operations overlapping.
product/machine matrix is diagonal the more the
decomposability is feasible. The majority of the comparison
Flow direction PERFORMANCE MEASURES
A number of authors included the flow direction within a
Work in process (WIP)
cell as an experimental factor. This factor has two possible
This measure characterizes the fluidity of the part flow in the
levels: unidirectional flow or else, flow with backtracking
system. It has mainly been measured in two manners: the
allowed. The first level is obtained either by altering product
number of parts in the system and the time weighted number
type routings or also by duplicating cell machines in order to
of parts in the system. The latter is obtained by summing up
avoid backtracking.
the average setup and process times of all parts either being
Processing time (PT) and Set up time (ST) processed or waiting in the different MS queues.
Several studies modeled both times by two independent
Mean flow time (MFT)
probabilistic laws. Others formulated ST as a fraction of PT.
This performance measure constitutes with WIP the most
It is worth noting here that PT is often given for the whole
popular measures used in the FL-CL comparative studies. It
batch and not per part.
also characterizes the fluidity of the part flow in the system.
Set up time reduction factor (δ) For each part, the flow time is simply obtained by
This factor materializes one of the crucial advantages of the subtracting the exit time from the entry time.
CL and more particularly of the group technology Due date related measures
philosophy. Indeed, part types of a same family have usually Mainly, researchers used Mean Tardiness (MT) and Mean
very similar setups on the machines. Hence, if a machine is Earliness (ME) as due date related performance measures.
set up for a certain type and then it should be set for another The first is taken as the average, over all tardy jobs, of the
type of the same family, the nominal setup time for the difference between actual delivery date and the promised
second batch should be reduced by the δ factor. due date. ME is obtained by analogy for all early jobs. Some
Transfer time (TT) researchers also use the percentage of tardy jobs and the
In the FL, this parameter corresponds to the percentage of early jobs without taking into account the
interdepartmental travel time. It is often modeled using an actual amount of tardiness or earliness.
appropriate probabilistic law. On the other hand, within the Other measures
CL framework, it corresponds to the durations of intra cell The system throughput is usually considered as productivity
moves. These times are generally very small compared to measure. It is the average number of parts exiting the system
those in the FL. This constitutes the second major advantage by time unit. It is the main indicator used for detecting the
of CL. attainment of steady state in a simulation run. Also, some
Demand rate studies used the average machine utilization rate or the
The demand rate is mainly expressed by the batch inter- average PT/ST ratio as performance indicators. Maximizing
arrival times (IAT) in the MS. They are generally generated the former ensures a high degree of resource exploitation
by common probabilistic distributions. Besides, some whereas the minimizing the latter enhances the efficiency of
studies focus only on the stability of this factor without the MS piloting.
changing its mean. SIMULATION MODELS
GENERALITIES
The functional layout splits the shop into d departments Di
(i=1,…,d) Each department is formed by Mi machines of the
same type. On the other hand, in a cellular production
system, products with similar shapes or production processes FL MODEL
are grouped into f families Fj (j=1,…,f). The Nj required When entering the functionally structured shop, the batches
machines to the manufacturing of every product of the are routed to the first department listed in their routing (Fig.
family Fj are clustered into the independent manufacturing 1). Then, they are queued until a machine becomes
cell Cj. No labor constraints are considered in both models. available. The department queue can be governed by priority
Manufacturing orders are launched by batches of constant rules such as first come first served (FCFS) or smallest
size and the batches inter arrival times are governed by an processing time (SPT) or repetitive lot (RL) or any other
appropriate probabilistic law. As soon as the batches are needed rule. The RL rule is the most frequently used one in
created, they are assigned a number of attributes such the order to minimize overall set up times. When it is used,
routing, the processing times and set up times on each work every time a machine becomes available the queue is
station included in this routing. searched for batches of the same type than the one just
The commercial simulation software Arena 7.0 was used in processed and the first batch found is prioritized. Just before
this study. It provides an integrated framework for building being processed, all batches are split into individual parts.
simulation models in wide variety of applications. It These parts are regrouped after processing. A processed
integrates all the functions needed for simulation including batch is then routed toward the next step of its routing or, if
animation, analysis of input and output data. processing is completed, toward the system exit (Fig.2).

DEPARTMENT 1
CREATE 1 ASSIGN 1
……..
CREATE 2 ASSIGN 2 DEPARTMENT i
BATCH ROUTE STATION DISPOSE
……………………...
……..
CREATE N ASSIGN N DEPARTMENT d

PRODUCT ARRIVALS DEPARTMENTS SYSTEM EXIT

Fig 1: Overview of the functional layout model

DEPARTMENT QUEUE

STATION Di HOLD

STATION 1 SEPARATE MACHINE 1 BATCH

PICKSTATION STATION 2 SEPARATE MACHINE 2 BATCH ROUTE


…………………………………………………….………………

STATION Mi SEPARATE MACHINEMi BATCH

Fig 2: Overview of department i sub model


CL MODELS options are available and a specific model is developed for
As for the CL model, it is characterized by the same batch each of them. In the first version (Fig 3), parts are re-
creation and attribute assignation mechanisms. However, in congregated into batches before being routed to the
addition to the cell general queue, each machine has its subsequent manufacturing step. In the second CL version
individual queue. Moreover, the routings in such a model are (Fig 4), every processed part is dispatched toward the next
expressed in terms of specific machines and not in terms of machine of its routing without waiting for the other batch
departments as it is the case in the FL model. Hence, each components. This is called operations overlapping, since a
batch waits in the general queue until the first machine in its same batch could be simultaneously processed by several
routing becomes available. Pre-manufacturing batch splitting machines.
is always required. However, after manufacturing, two
CELL ARRIVALS CELL EXIT

CREATE 1 ASSIGN 1 CELL QUEUE

CREATE 2 ASSIGN 2 DISPOSE


BATCH ROUTE STATION Cj HOLD Cj
……………………...

CREATE N ASSIGN N

STATION 1 HOLD 1 SEPARATE MACHINE 1 BATCH


STATION

PICKSTATION STATION 2 HOLD 2 SEPARATE MACHINE 2 BATCH ROUTE

……………………………………………………………………...

STATION Nj HOLD Nj SEPARATE MACHINE Nj BATCH

MACHINES

Fig 3: Overview of the cellular layout model with no overlapping

CELL ARRIVALS CELL EXIT

CREATE 1 ASSIGN 1 CELL QUEUE


CREATE 2 ASSIGN 2 DISPOSE
BATCH ROUTE STATION Cj HOLD Cj
……………………...

CREATE N ASSIGN N
BATCH

STATION 1 HOLD 1 MACHINE 1


STATION

SEPARATE PICKSTATION STATION 2 HOLD 2 MACHINE 2 ROUTE

……………………………………………...

STATION Nj HOLD Nj MACHINE Nj

MACHINES

Fig 4: Overview of the cellular layout model with overlapping


ILLUSTRATION CONCLUSION
Morris and Tersine (1990) used hypothetical shop data This paper presents a taxonomy of the key factors used in
featuring 40 product types belonging to 5 families and 30 the main published FL-CL comparison simulation studies. It
machines divided between 8 process departments in the FL also presents the development of simulation models for both
and 5 cells in the CL. In both layouts, the queues were functional and cellular layouts using the ARENA commercial
governed by the RL priority rule. Besides, the performances software. Using these models, it is demonstrated that the
were assessed using the average levels of flow time and WIP findings presented by Morris and Tersine (1990), one of the
while varying the ratio of set up to process time, the transfer most important simulation studies, lack objectivity.
time, demand stability and flow work within cells. It Furthermore, these models are intended to be the key
appeared from their results that FL outperforms CL in the instruments in establishing an objective simulation-based
majority of investigated contexts. Still, when we carefully FL-CL comparison methodology. The development of such
examine their methodology, we notice that Morris and a methodology is currently pursued.
Tersine (1990) considered very low transfer times which REFERENCES
implicitly advantages the FL. In fact, one of the main Hyer, N.L. and Wemmerlov, U., “Group Technology in the
advantages of CL is the time saving by locating machines U.S. manufacturing industry: a survey of current practices”
International Journal of Production Research, 27(8), 1989,
required to process one job close to each other. Also, the lot pp. 1287-1304
size was not included in their experimental design despite Leonard, R. and Rathmill, K., “the group technology
the fact that its importance is indisputable. Finally, the myths”, Management Today, January 1977, pp. 66-9.
authors used a CL with no operations overlapping allowed. Rathmill, K. and Leonard, R., “The fundamental
limitations of cellular manufacture when contrasted with
This does not permit to perceive full CL benefits. In this efficient functional layout”, fourth international conference
work we choose to illustrate this particular shortcoming by on production research, Taylor & Francis, London 1977.
showing that if overlapping is allowed, the CL could Craven, F.W., “Some constraints, fallacies and solution in
GT applications”, Proceedings of Fourteenth International
outperform the FL. Machine Tool Design and Research Conference, 1973, pp.
Hence, a first test with no overlapping generated MFTs of 169-175.

8524 min and 11833 min for the FL and CL models Morris, J.S. and Tersine, R.J. “A simulation analysis of
factors influencing the attractiveness of group technology
respectively. These results are very close to those of the cellular layouts”, Management Science, Vol. 36 No.
studied research (Table 1). However, when overlapping was 12, 1990, pp. 1567-78.

used, the MFT of the CL model was lowered to 6815 min. Shafer, S.M. and Charnes, J.M. “Cellular versus functional
layouts under a variety of shop operating conditions”,
This confirms the major influence of overlapping, Decision Sciences, Vol. 24, No. 3, 1992, pp. 665-681.
disregarded by Morris and Tersine. Moreover, this result Suresh, N.C. and Meredith, J.R. “Coping with the loss of
pooling synergy in cellular manufacturing systems”,
confirms that the conclusions formulated by Morris and
Management Science, Vol. 40, No. 4, 1994, pp. 466- 483.
Tersine are very specific to the “no overlapping” case.
Suresh, N.C., “Partitioning work centers for group
FL CL technology: analytical-extension and shop-level simulation
Overlapping Overlapping investigation”, Decision Sciences, Vol. 23, No. 2, 1992,
not allowed allowed pp.267-90.
Morris and Shafer, S.M, and Charnes, J.M. “A simulation analyses of
8731 13101 ----------
Tersine model factors influencing loading practices in cellular
manufacturing”, International Journal of Production
Present study 11833 6815 Research, Vol. 33, No. 1, 1995, pp. 279-290.
8524
model
Jensen, J.B., Malhotra, M.K and Philipoom, P.R. “Machine
dedication and process flexibility in a group technology
Table 1: Illustration results (MFT)
environment”, Journal of Operations Management, 14,
1996, pp. 19-39.
Farrington, Ph.A and Nazametz J.W. “Evaluation of the
performance domain of cellular and functional layouts”,
Computers ind. Engng, Vol. 34, No. 1, 1998, pp. 91-101.
Faizul huq, Douglas, A.H and Zubair, M.M. “A simulation
analysis of factors influencing the flow time and through-
put performance of functional and cellular layouts”,
Integrated Manufacturing Systems, 12/4, 2001, pp. 285-
295.
Jing-Wen Li. “Improving the performance of job shop
manufacturing with demand-pull production control by
reducing set-up/processing time variability”, International
journal of Production Economics, 84, 2003, pp. 255-270.
Rathmell, J and Sturrock, D.T. “The ARENA product
family: Enterprise modeling solutions”, Proceedings of the
Winter Simulation Conference, 2002, pp. 165-172.
Kelton, W.D., Sadowski, R.P. and Sadowski. D.A.
“Simulation with Arena” 2nd. ed. 2002, New York:
McGraw-Hill.
Arena Template Reference Guide, Rockwell software, inc.
2000-2002

You might also like