Journal of Manufacturing Systems 34 (2015) 66–73

Contents lists available at ScienceDirect

Journal of Manufacturing Systems

journal homepage: www.elsevier.com/locate/jmansys

Technical Paper

Lean production system design for fishing net manufacturing using

lean principles and simulation optimization
Taho Yang a , Yiyo Kuo b,∗ , Chao-Ton Su c , Chia-Lin Hou a
a
Institute of Manufacturing Information and Systems, National Cheng Kung University, Tainan 701, Taiwan
b
Department of Industrial Engineering and Management, Ming Chi University of Technology, New Taipei City 243, Taiwan
c
Department of Industrial Engineering and Engineering Management, National Tsing Hua University, Hsinchu 300, Taiwan

a r t i c l e i n f o a b s t r a c t

Article history: Value stream mapping (VSM) is a useful tool for describing the manufacturing state, especially for dis-
Received 13 September 2012 tinguishing between those activities that add value and those that do not. It can help in eliminating
Received in revised form 1 October 2014 non-value activities and reducing the work in process (WIP) and thereby increase the service level. This
Accepted 24 November 2014
research follows the guidelines for designing future state VSM. These guidelines consist of five factors
which can be changed simply, without any investment. These five factors are (1) production unit; (2)
Keywords:
pacemaker process; (3) number of batches; (4) production sequence; and (5) supermarket size. The five
Lean manufacturing
factors are applied to a fishing net manufacturing system. Using experimental design and a simulation
Simulation
Value stream mapping
optimizing tool, the five factors are optimized. The results show that the future state maps can increase
Fishing nets service level and reduce WIP by at least 29.41% and 33.92% respectively. For the present study, the lean
principles are innovatively adopted in solving a fishing net manufacturing system which is not a well-
addressed problem in literature. In light of the promising empirical results, the proposed methodologies
are also readily applicable to similar industries.
© 2014 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.

1. Introduction MTO manufacturing need not be a pull production system—a

The common challenges that companies face are market com-
petition, increased pressure on inventory, increased service levels,
and reduced work in process (WIP). Lean manufacturing is one of
the approaches that can help companies respond appropriately to
these challenges [11,18]. The focus of the approach is on cost reduc-
tion, by eliminating activities that do not add value by linking and
balancing work stages, so that products from one stage are con-
sumed directly by the next stage, until the end of the production
line is reached [4,5,7,18].
Applications of lean manufacturing have spanned many sec-
tors, including the automotive industry, electronics, white goods
and consumer goods [1,21]. However, there is no evidence of work
applying lean manufacturing principles to fishing net production.
Fishing net manufacturing is a high make-to-order (MTO) environ-
ment, because net size and type change according to the ocean
environment, fish kinds and ship size.
∗ Corresponding author. Tel.: +886 2 29089899x3118; fax: +886 6 29085900.
E-mail addresses: tyang@mail.ncku.edu.tw (T. Yang), yiyo@mail.mcut.edu.tw
(Y. Kuo), ctsu@mx.nthu.edu.tw (C.-T. Su), clhoua@tsmc.com (C.-L. Hou).
lean design principle. Usually, there are many (or some) simulta-
neous orders in the manufacturing process. These orders may be
different in size and specifications. Thus, there is no guarantee that
the system will follow a first-in-first-out rule. In fact, it is often quite
messy when production is loaded with work-in-process. Thus, it
may be effective to adopt lean principles to regulate the flow and
to control the work-in-process level.
Higher WIP and longer cycle times always result in a lower
service level and produce lower customer satisfaction in MTO
environments. Moreover the fishing net manufacturing system is
distinctive. Through the production procedure the production units
become bigger and bigger and the processing time in most steps are
quite long (between 3 and 10 days). This provides the motivation
for this research, to design a lean manufacturing system for this
industry.
Value stream mapping (VSM) is a visual tool that facilitates the
process of lean production by helping to identify the value-adding
steps and eliminating the non-value adding waste [22]. In recent
years, VSM has emerged as the superior method to implement lean
production in factories, and has been used to identify where waste
occurs [3,12,14,15,17,24,25,28]. VSM creates a pictorial represen-
tation and common language for the production process, enabling

http://dx.doi.org/10.1016/j.jmsy.2014.11.010
0278-6125/© 2014 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
 T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
more purposeful decisions to improve processes. Numerous appli-
cations can be found in the literature. For example, Seth and Gupta
[22] applied VSM to establish lean production in an Indian two-
wheeler motor company. Lummus et al. [16] report on a VSM
project in a small medical clinic. The new system can increase
patient throughput and reduce patient waiting time. Abdulmalek
and Rajgopal [1] describe a steel mill case, and apply VSM to identify
the opportunities for improvement. Lian and Landeghem [14] pro-
pose a generator which automatically yields a simulation model
for VSM. The current and future state VSM of a poultry and pig
rearing equipment manufacturing system is introduced to demon-
strate the proposed generator. Barber and Tietje [2] demonstrate
the use of VSM to achieve both greater efficiency and value cre-
ation in a sales process. Seth et al. [23] identify and address various
sources of waste in the supply chain of an edible cottonseed oil
industrial process and use a VSM approach to improve produc-
tivity and capacity utilization in an Indian context. Lasa et al.
[12,13] redesigned six production systems with the VSM tech-
nique, and then analyze the causes for the limited adoption of lean
manufacturing concepts. The six production systems are manufac-
turing systems for kit furniture, water heaters, forging, detonator
systems, mechanized and stamped parts, and thermoplastic
parts.
Rother and Shook [20] proposed seven guidelines based on the
concept of lean manufacturing to construct the future VSM which
can be implemented in a reasonably short time period without any
major investment. According to the seven guidelines, there are sev-
eral factors that should be considered. Different combinations of
the factors would result in different production performance. How-
ever, the implementation of the recommendations is likely to be
risky. Simulation is a useful tool for evaluating the performance
of a new design but it cannot provide the optimal design. Com-
bining simulation with experimental design or intelligent search
has been successfully adopted for simulation optimization. Yang
et al. [27] use a commercial tool called OptQuest for optimizing an
integrated-circuit (IC) packaging system. OptQuest embeds scatter
search (SS) in a simulation tool Arena to optimize the simula-
tion models. Yang et al. [26] combine simulation with genetic
algorithm for optimizing dispatching rules in a flow shop with
multiple processors. Kuo et al. [9] integrate simulation and the
Taguchi method to optimize an integrated-circuit (IC) packaging
system.
The objective of the present study is to model a non-typical
production system – a fishing-net manufacturing system – and to
propose a lean production system design which is optimized by
simulation optimization.
The remaining sections of this paper are organized as follows.
In Section 2 the lean principles are introduced, and a number
of decision factors are highlighted. Then, the fishing-net produc-
tion system is introduced and the current state VSM is provided
in Section 3. Finally, a hybrid experimental design and intel-
ligent search approach is adopted for optimizing the decision
factors and then the future state VSM is constructed in Section
4. A summary of results and concluding remarks are presented in
Section 5.
2. The lean principles
Rother and Shook [20] propose a five phase implementation of
VSM. The phases are (1) selection of product family; (2) current
state mapping; (3) future state mapping; (4) definition of working
plan; and (5) achievement of working plan. The lean techniques
to be used are defined in the third phase which contains seven
guidelines to define the future state map [12]. The seven guidelines
are summarized below:
67
2.1. Guideline #1: produce to your takt time
“Takt time” is used to synchronize the pace of production with
the pace of sales. In general, it can be calculated by Eq. (1)
available working time per day
Takt time = (1)
customer demand rate per day
However, when the process times of different products are quite
different, the takt time calculated by Eq. (1) is not reasonable. For
example, suppose there are two types of product, A and B, and the
unit process time and demand rate of product A are 120 min and
4 units per day, respectively and the process time and demand rate
of product B are 60 min and 8 unit per day. If the available working
time is 18 h per day, then the takt time is 90 min (18 × 60/(4 + 8)).
This means that one unit should be produced every 90 min, but this
is impossible for product A. The present research uses Eqs. (2)–(4)
for calculating the takt time.
The lowest common multiple of processing time for all products
KSi = (2)
Pi
Di
NKi = (3)
KSi
Available working time per day
Takt time = N (4)
i=1
NKi
In Eq. (2) KSi indicates the kanban size of product i (i = 1, 2, . . ., N),
and Pi indicate the processing time of product i. In Eq. (3) NKi indi-
cates the number of kanbans of product i, and Di indicates the
demand for product i per day. Thus, for the same example, the kan-
ban size for product A and B are 1 and 2 units respectively, and the
number of kanbans for product A and B are both 4. The takt time is
2.25 h. This means that one unit of product A or two units of prod-
uct B should be completed every 2.25 h. Each kanban indicates one
unit of product A or two units of product B and four kanbans are
required per day for both product A and B.
It should be noted that the lowest common multiple of
processing time for all products in Eq. (2) is an approximate value
and should be as small as possible. Managers can increase the pro-
cess time of some product types to tune it. For example, if there are
three product types and their corresponding processing times are
12, 11 and 25 min, then the processing time of the first two prod-
uct types can be increased to 12.5 min, so that the lowest common
multiple of processing time will be 25 min.
Corollary 1. The takt time is dependent on the size of a production
unit.
Corollary 1a. A kanban represent the same required production
time regardless of the different product types.
2.2. Guideline #2: develop continuous flow where possible
Continuous flow refers to producing one piece at a time to
reduce the inventory of WIP and production CT. However continu-
ous flow requires a great deal of creativity to achieve and sometimes
it requires plant layout redistribution [12]. In this research, this
guideline is not applicable and is not taken into consideration in
the case study.
2.3. Guideline #3: use a supermarket to control production where
continuous flow does not extend upstream
A “supermarket” is nothing more than a buffer or storage area
located at the end of the production process for products that are
ready to be shipped [1]. When continuous flow is not possible and
batching is necessary, a supermarket can smooth the whole man-
ufacturing process. Supermarkets use a kanban system to fix the
68 T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73

inventory level. If the number of kanbans in a supermarket is too

high, it causes higher inventory cost. On the other hand, the down-
stream production process will be subject to delays if the number
of kanban in the supermarket is too low. Therefore, the number of
kanbans in a supermarket is an importation factor when designing
the future VSM.

Corollary 2. A supermarket is required for the implementation of

lean-pull system.

2.4. Guideline #4: try to send customer scheduling to only one

production process

In Guideline Number 1, for calculating the takt time, the cus-

tomer demand rate is based on customer scheduling. The process
time is set by one of the production processes. That process is
called the “pacemaker process”. The pacemaker process synchro-
nizes the pace of the entire manufacturing process and there are
no supermarkets downstream of the pacemaker process. Therefore,
selecting different workstations as the pacemaker process will have
an important influence on the performance of the entire manufac-
turing system.
Corollary 3. A pacemaker process is required for the implemen-
tation of lean-pull system.
Fig. 1. The composition of a rope.
decreasing the production unit and so on. It has been addressed in
Corollary 4.
Lean management is well recognized as a management philos-
ophy rather than a generic tool set. In other word, its principles are
generic but its solution methodology is often case specific. For the
present study, we adopt the lean principles from the existing liter-
ature as an exemplar to identify the five corollaries as the basis of
our proposed methodology development which will be discussed
in details in the following sections.

2.5. Guideline #5: level the production mix

3. Fishing-net manufacturing and the current state map

Leveling the product mix means dividing the volume of all the
The case study was adopted from Hsieh et al. [8]. They
product types based on their kanban size, and then producing them
introduced the fishing net manufacturing process in detail and then
evenly over a time period. The more level the product mix is, the
developed a hierarchical rough-cut capacity planning model and
greater is the ability to respond to different customer requirements
demand management system. While they focused on the strategic
with a short lead time. For example, if there are three product types
planning decision, the production control decision is not addressed.
and the production sequence can be A-A-A-B-B-B-C-C-C or A-B-
The application of lean principle to the fishing-net manufacturing
C-A-B-C-A-B-C. The latter is more level and allows the process to
is not found in literature. The solution to this specific application
proceed smoothly with smaller supermarkets, but it would cause
could be applicable to similar industries such as textile industries,
more changeovers.
variable product-size production, etc. Since the fishing-net man-
Due to make-to-order (MTO) production environment, an order
ufacturing is rarely seen in literature, we discuss the production
split to a certain number of production batches should be consid-
process forehead in details as follows.
ered to achieve a leveling production. The use of a larger number of
There are various fishing nets, such as gill nets, lift nets, drag
batches with fewer products in each batch results in a more level
nets, surrounding nets (purse seine nets), set nets (trap nets), cov-
process. Therefore, the number of batches is also an importation
ering nets, and fish breeding nets. Each net has a unique application
factor in designing future VSM. Moreover, the production sequence
which is based on the ocean environment, fish type and ship size.
also influences the performance of the manufacturing system. That
Therefore, the same customer could order nets of different types
means the managers have to decide between A-B-C-A-B-C-A-B-C
or different sizes. Each order requires different raw material types,
and A-C-B-A-C-B-A-C in the example of the production sequence
twine sizes, mesh sizes, mesh depths and mesh lengths. The raw
above.
materials for fishing nets include nylon, trawl, nylon filament, nylon
Corollary 4. An order splitting strategy is able to improve a lean- multi-filament and polyester. Twine size is determined by the com-
pull system performance. position of a rope as shown in Fig. 1 [8]. Raw material filaments are
twisted into yarn, then twined yarns are made into strands. Strands
Corollary 5. The production sequence decision could impact a are twisted into twine, and then braided twines into rope. Mesh size
lean-pull system performance. is a special unit of a net; different mesh depth and mesh length will
result in a different shape and area of fishing net. The relationship
2.6. Guideline #6: level the production volume between mesh size, mesh depth and mesh length is show in Fig. 2
[8].
Leveling the production volume is related to Guideline number
1. It means that production should be based on a fixed pace, the
takt time. It has been addressed in Corollaries 1, 1a, and 4.

2.7. Guideline #7: develop the ability to make “every part every
day” (then every shift, then every hour or pallet or pitch) in
fabrication processes upstream of the pacemaker process

This guideline is related to Guideline number 5. In order to divide

the volume of all product types into more batches, some production
methods have to be changed, such as reducing the changeover time, Fig. 2. The mesh size.
 T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
Fig. 3. The current state map.
Net cage is one kind of fishing net which is the main product of
the case study company. The main function of the net cage is to be
fixed under the sea and cultivate salt water fish in it. According to
the size and type of the cultivated fish, the corresponding specifi-
cations of net cages are also different. The general diameter varies
from 10 m to 30 m, depth varies from 4 m to 10 m, while mesh varies
from 2 mm to 30 mm.
In the case study company there are six workstations for net cage
manufacturing. They are (1) Twisting, (2) Braiding, (3) Net knitting,
(4) Dyeing, (5) Heating and (6) Suturing. All manufactured net cages
follow the sequence of workstation above, from 1 to 6.
First the filament is twisted to form the yarn or strand as the
twine in the Twisting workstation. Then the rope is braided from
three twines in the Braiding workstation. The rope size is deter-
mined by the request of customers. The thicker the rope size, the
thicker twine is required, and consequently more twisting jobs are
required. In the Net knitting workstation the ropes are knitted to
form nets in standard sizes.
A knitting machine processes a net continuously. Theoretically,
there is no limit to the length of a net, but in practice, it is a multiple
of 10 m. There two options for the width of a net: 1 or 2 m. According
to the specification of the knitting machines in the case company
the standard size can be 10 m × 1 m or 10 m × 2 m. Every knitting
machines can knit one single 10 m × 2 m net or two 10 m × 1 m nets
simultaneously. In the Dyeing workstation, a number of nets in
standard size can be dyed together, but the total weight of the nets
cannot exceed the capacity of the dyeing machine. After the Dyeing
workstation the net will be heated in the Heating workstation. One
single 10 m × 2 m net or two 10 m × 1 m nets can be heated simul-
taneously. However, the processing time is determined by the rope
size. The thicker the rope size, the longer processing time is. Finally,
the nets with standard size are sutured to form net cages accord-
ing to the request of the customers. In other words, a final net is
a combination of a certain number of unit nets—either 10 × 2 or
10 × 1 in size. It is the suture process that combines the required
number of unit nets into the final product. Because the operators
cannot start suturing an order until all the required nets of standard
69
size are available, the waiting time for nets of standard size in this
workstation is very long.
The current stat map of the case company is shown in Fig. 3. It
should be noted that although the fishing net manufacturing envi-
ronment is MTO, every workstation has a weekly schedule, because
of the long processing time. For example the processing time in the
Suturing workstation is 10 days.
In the figure, the small boxes represent the processes (worksta-
tions) and the numbers inside the boxes present some information
about the corresponding processes, such as process time (PT), num-
ber of machines, working time (WT) per day, number of shifts
and so on. Each process receives its schedule from the depart-
ment of production control, which is represented by arrows from
the department of production control to all the processes. The
timeline at the bottom of the map shows value-adding and non-
value-adding time. The processing time is calculated by adding the
time for each process (value-adding time) in the map. The lead-time
is obtained by summation of the waiting time (non-value-adding
time) for each WIP triangle before each process and process time.
The resulting value-adding and non-value-adding times are 25.08
and 19.78 days, respectively. The lead-time is 44.86 days. Thus the
value-adding time is 56% of the lead-time. Note that fishing-net
manufacturing is a flow shop production line; thus, all products
traverse through the same six processing workstations. The process
times and lead-times represent their mean values.
According to the current state map, reducing the waiting time
in front of the each workstation is an opportunity to reduce the
lead-time and, consequently, to increase the service level. This is of
interest because the focus of the present study is to explore a lean
production system design for fishing net manufacturing using lean
principles and simulation optimization. However, reducing WIP
and its associated non-value-adding time, while maintaining the
required system performance, is not straightforward. In fact, it is
quite challenging and is the main concern for any scheduling deci-
sion and shop floor control system. There may be other noise factors
that cause non-value-adding time but they are not the objective of
the present study and are not covered here.
70 T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
Fig. 4. Simulation interface.
Two performance indicators, average WIP level and average
service level, are taken into consideration in this case study. WIP
level is defined as the total volume of semi-finished goods in the
production system. Service level is defined as the percentage of
orders that are completed before their corresponding due date. As
the company did not collect data on service level and WIP, this
research developed a simulation model to evaluate it. Commercial
software, Arena [29], is adopted as the simulator for the present
study. The resulting example of the model interface, in which all
the operation logics are verified by the managers of the case study
company, is shown in Fig. 4.
To evaluate the performance of each scenario, a run length of
750 days with 3–12 replications was adopted for the simulation
model. The number of replications depends on 10% of the mean
for a 95% confidence level. The estimated service level and WIP are
68% and 63,971 kg respectively. The results of the simulation run
are validated by the managers of the case study company.
4. Future-state map optimization
Although the case factory has already streamlined its production
line prior to the present study, its distinctive manufacturing process
causes a prolonged lead-time and high work-in-process level as
discussed in the previous section. Because of the effectiveness of
lean system design in lead-time reduction, it is the objective of the
present study to use the lean principles in an innovative way to
solve the case problem. In light of the five corollaries identified in
Section 2, we identify five key controllable factors as: (1) production
unit size, (2) the pacemaker process, (3) the number of batches for
an order split, (4) the production sequence, and (5) supermarket
size, to optimize the proposed fishing-net manufacturing system.
The production units in net cage manufacturing process have
changed and are becoming bigger and bigger. In the beginning,
the filaments change to twines and the twine size depends on
the customer’s requirements. Then the twines change to ropes.
In the Net knitting workstation the production units become nets
with standard size. Finally the net cages are sutured or stitched. A
manufacturing system with different production units requires a
complicated production strategy and is difficult to control. Accord-
ing to the guidelines for designing the future state map, the
production unit should be as small as possible to level the product
mix. However, in the net cage manufacturing process, reduc-
ing the production unit will not only cause a greater number
of changeovers but will also increase the processing time in the
Suturing workstation. In this research, the production unit of each
workstation depends on the standard net size. For different cus-
tomer orders, the specifications such as the net weight and rope
size, are also different. Thus the processing times of the standard
net size could be different in every workstation.
As regard the pacemaker process, this research selects Knitting
workstation, Dyeing workstation and Suturing workstation as the
candidates. Knitting workstation usually has the highest machine
utilization. Dyeing workstation is batch processing. Its capacity
 T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73 71

Table 1
The design factors for lean manufacturing.

Factors Description (unit) Factor levels

Level 1 Level 2 Level 3

A Production unit (m) 10 m × 1 m 10 m × 2 m

B Pacemaker process Suture workstation Knitting workstation Dyeing workstation
C Production sequence EDD FIFO SPT
D Number of batches 1 2 3

is batch-size dependent. Suturing workstation is the most labor ratio means better performance for combinatorial parameters [9].
intensive workstation. Its throughput is limited by its staffing level. The S/N ratio can then be defined as
For the sake of cost saving, it is often the company policy to staff  
1 1
r
as few as possible, and thus, often makes it as the bottleneck of
j = −10 × log (5)
the line. In summary, the potential bottleneck could shift among r v2
k=1 jk
the three workstations which are thus considered as the candidate
pacemaker locations. or
 
1 2
The number of batches to be allowed will decide the trans- r
portation volume between workstations. More batches allow for j = −10 × log vjk (6)
a more level product mix, but will cause more changeovers. The r
k=1
changeover times between different products are different.
An appropriate production sequence can reduce the total Eq. (5) is used for the ‘larger-the-better’ responses and Eq. (6) is
changeover time. This research selects earliest due date (EDD), first used for the ‘smaller-the-better’ responses.
in fist out (FIFO) and shortest process time (SPT) as alternative An L18 (21 × 37 ) orthogonal array has the least number of treat-
criteria for designing the future state map. ments to account for one two-level and more than three three-level
As regard the supermarket size, this research uses the Arena control factors. Therefore, an L18 (21 × 37 ) orthogonal array was
embedded optimizer, OptQuest, as the optimization tool which used to collect the experimental data. The experimental scenar-
embeds scatter search in a simulation model to optimize the super- ios are shown in columns 2–5 of Table 2. The two levels and three
market size for all lean manufacturing strategies. OptQuest is a levels of all factors were denoted as 1–2 or 1–3 (the lower, the mid-
proven tool that can avoid entrapment in a local optimal solution dle and the upper levels). For each scenario, OptQuest was used to
and is not sensitive to its initial solution. Glover et al. [6], Laguna search for the best supermarket size to maximize the service level.
and Martí [11] give details of scatter search and the OptQuest algo- The service level and WIP for each scenario with the best super-
rithms. Thus, four factors (production unit, pacemaker process, market size are shown in columns 6 and 7 of Table 2. Service level
number of batches and production sequence) are taken into consid- is to be maximized, and the WIP is to be minimized. Let j be the
eration in the experimental approach to designing the future state S/N ratio of scenario j and let vjk be the simulation result for sce-
map. For each scenario, OptQuest is used to find the best super- nario j, in the kth replication. r is the total number of replications.
market size and the corresponding performances are viewed as the
performance of each scenario. The design levels of each factor are
shown in Table 1.
The Taguchi method aims to find an optimal combination of
parameters that have the smallest variance in performance. The
signal-to-noise ratio (S/N ratio, ) is an effective way to find sig-
nificant parameters by evaluating minimum variance. A higher S/N

Table 2
The experimental scenarios and results.

Scenario L18 Mean value S/N ratio

A B C D Service level WIP (kg) Service level WIP Fig. 5. The effect of S/N ratio of service level.
1 1 1 1 1 0.83 34,887.11 19.99 −89.94
2 1 1 2 2 0.85 36,725.24 20.23 −88.64
3 1 1 3 3 0.80 35,266.78 19.71 −91.08
4 1 2 1 1 0.82 34,578.18 19.91 −90.59
5 1 2 2 2 0.82 35,939.26 19.90 −89.84
6 1 2 3 3 0.77 36,267.69 19.39 −91.21
7 1 3 1 2 0.81 35,765.61 19.74 −91.11
8 1 3 2 3 0.78 39,411.85 19.41 −91.95
9 1 3 3 1 0.84 37,659.27 20.06 −90.03
10 2 1 1 3 0.85 41,036.15 20.16 −92.29
11 2 1 2 1 0.83 38,989.49 20.01 −90.14
12 2 1 3 2 0.89 36,979.23 20.61 −88.73
13 2 2 1 2 0.82 35,661.37 19.92 −89.78
14 2 2 2 3 0.84 37,201.86 20.10 −90.19
15 2 2 3 1 0.82 39,286.82 19.92 −91.91
16 2 3 1 3 0.84 40,285.11 20.05 −92.12
17 2 3 2 1 0.83 39,868.90 19.97 −89.56
18 2 3 3 2 0.83 38,042.93 20.00 −91.65
Fig. 6. The effect of S/N ratio of WIP level.
72 T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73

Table 3 is the best design for the future state map to minimize the WIP.
The analysis of variance in service level.
However, in Table 3 it can be seen that the contribution of fac-
Factors Sun of squares Degrees of freedom Mean square F tor C (production sequence) to the variance of service level is very
A 0.320 1 0.320 6.065 small, and in Table 4 it can be seen that the contribution of fac-
B 0.259 2 0.130 2.455 tor A (production unit) to the variance of WIP is very small. This
C 0.002 2 0.001 0.018 research, therefore, chose the level of factor C based on the effect
D 0.215 2 0.107 2.037 of S/N ratio on WIP level (Fig. 6) and chose the level of factor A
Error 0.527 10 0.053 based on the effect of S/N ratio on service level (Fig. 5). Thus the
Total 1.324 17
A2 B1 C2 D2 is the design of future state map. The best supermarket
size for A2 B1 C2 D2 is found using the simulation optimization tool,
OptQuest. The corresponding future state maps are illustrated in
Table 4 Fig. 7.
The analysis of variance in WIP.
The optimizing results produced by OptQuest show that the
Factors Sun of squares Degrees of freedom Mean square F design of A2 B1 C2 D2 increases service level from 68% to 90% and
A 0.218 1 0.218 0.257 reduces WIP from 63,971 kg to 42,269.31 kg. The improvements
B 2.614 2 1.307 1.544 from the current state map are 32.35% and 33.92%, respectively.
C 20,792 2 1.396 1.648 Moreover the non-value adding time reduces from 44% (19.78 days)
D 70,387 2 3.694 4.361 to 35.46% (13.78 days).
Error 8.469 10 0.847 Note that the optimal design of production unit (factor A) is level
Total 21.481 17
2 (10 × 2 m) which is bigger than level 1 (10 × 1 m). Although the
guideline number 7 encourages the production unit to be smaller
the better. The size of a final net cage range from 120 to 900 m2 .
Eqs. (5) and (6) respectively are used for calculating the signal The smaller production unit would cause more processing time for
to noise ratio (S/N) of service level and WIP. The S/N ratios of all suturing process and would deteriorate the suturing workstation
scenarios are shown in columns 8 and 9 of Table 2. performance.
According to the S/N ratios in Table 2, the effect of each fac- There are many challenges for the present study, as discussed in
tor level on service level and WIP are calculated and shown in this paper. Although the lean principles are generic, their applica-
Figs. 5 and 6. The analysis of variance in service level and WIP are tion is usually innovative, as is the case in the present study. Since
shown in Tables 3 and 4. the results of the case study are promising and are readily appli-
Fig. 5 shows that A2 B1 C1 D2 is the best design for the future state cable to similar industries, the contribution of the present study is
map to maximize the service level. Fig. 6 shows that A1 B1 C2 D2 justified.

Fig. 7. The future state map of A2 B1 C2 D2 .

 T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
5. Conclusions
Lean manufacturing has been applied successfully in many man-
ufacturing industries. It focuses on cost reduction by eliminating
non-value adding activity. In general, waiting is the most common
non-value adding activity. An appropriate production strategy can
effectively reduce the total waiting time. Moreover, VSM is a tool
that helps managers to visualize value adding and non-value adding
activity. In this research, a net cage manufacturing system is pro-
posed as the case study. Although net cage manufacturing operates
in an environment of MTO, because of the long manufacturing lead
time, every workstation has its weekly schedule and the material
flows between workstations are calculated using the push strategy.
These characteristics are quite different from traditional MTO man-
ufacturing systems. Moreover, the production units are changed
through the manufacturing system, and it becomes challenging to
determine the size of moving batches.
Based on the guidelines for implementing lean manufacturing,
some important production factors are selected for designing the
future state VSM. Using experimental design and a simulation opti-
mization tool, these important factors are optimized. According to
analysis based on the simulation, the future state map not only
increases service level but also reduces the WIP.
In the case study, the selected factors can be changed to any level
without extra investment. That means the case company can imple-
ment the future state map to achieve lean manufacturing without
any financial pressure. This could be the first step to achieving lean
manufacturing. Although the lean principles are often generic, the
proposed case study is rarely seen in literature and is an inno-
vative application. In light of the promising results, the proposed
methodologies are also readily applicable to similar industries.
Acknowledgments
The authors thank the anonymous company for providing the
case study. This work was supported, in part, by the National
Science Council of Taiwan, Republic of China, under grant NSC-101-
2221-E-131-043 and NSC-101-2221-E-006-137-MY3.
References
[1] Abdulmalek FA, Rajgopal J. Analyzing the benefits of lean production and value
stream mapping via simulation: a process sector case study. Int J Prod Econ
2007;107:223–36.
[2] Barber CS, Tietje BC. A research agenda for value stream mapping the sales
process. J Pers Sell Sales Manag 2008;28:155–65.
[3] Braglia M, Carmignani G, Zammori F. A new value stream mapping approach
for complex production systems. Int J Prod Res 2006;44:3929–52.
73
[4] Doolen TL, Hacker ME. A review of lean assessment in organizations: an
exploratory study of lean practices by electronics manufacturers. J Manuf Syst
2005;24:55–67.
[5] Elmoselhy SAM. Hybrid lean-agile manufacturing system technical facet, in
automotive sector. J Manuf Syst 2013;32:598–619.
[6] Glover F, Laguna M, Martí R. Scatter search. In: Ghosh A, Tsutsui S, editors. The-
ory and applications of evolutionary computations: recent trends. New York:
Springer-Verlag; 2000.
[7] Hobbs DP. Lean manufacturing implementation: a complete execution manual
for any size manufacturer. Florida: J. Ross Publishing; 2004.
[8] Hsieh HC, Yang T, Su CT, Lin CT. The development of a hierarchical rough-
cut capacity planning model and demand management system for fishing-net
manufacturing. Eur J Ind Eng 2012;6(4):422–40.
[9] Kuo Y, Yang T, Huang GW. The use of a grey-based Taguchi method for optimi-
zing multi-response simulation problems. Eng Optim 2008;40:517–28.
[11] Laguna M, Martí R. Neural network prediction in a system for optimizing sim-
ulations. IIE Trans 2001;34:273–82.
[12] Lasa IS, Castro RD, Laburu CO. Extent of the use of lean concepts proposed for
a value stream mapping application. Prod Plan Control 2009;20:82–98.
[13] Lasa IS, Laburu CO, Castro RD. Evaluation of value stream mapping in manufac-
turing system redesign. Int J Prod Res 2008;46:4409–30.
[14] Lian YH, Landeghem HV. Analyzing the effects of lean production using a value
stream mapping-based simulation generator. Int J Prod Res 2007;45:3037–58.
[15] Lu J-C, Yang T, Wang C-Y. A lean pull system design analyzed by value stream
mapping and multiple criteria decision-making method under demand uncer-
tainty. Int J Comput Integr Manuf 2011;24:211–28.
[16] Lummus RR, Vokurka RJ, Rodeghiero B. Improving quality through value
stream mapping a case study of a physician’s clinic. Total Qual Manag
2006;71:1063–75.
[17] McDonald T, Van Aken EM, Rentes AF. Utilizing simulation to enhance value
stream mapping: a manufacturing case application. Int J Logist 2002;5:213–32.
[18] Meade DJ, Kumar S, Houshyar A. Financial analysis of a theoretical lean man-
ufacturing implementation using hybrid simulation modeling. J Manuf Syst
2006;25:137–52.
[20] Rother M, Shook J. Learning to see: value stream mapping to add value and
eliminate MUDA. Brookline: The Lean Enterprise Institute; 1998.
[21] Saurin TA, Ribeiro JLD, Vidor G. A framework for assessing poka-yoke devices.
J Manuf Syst 2012;31:358–66.
[22] Seth D, Gupta V. Application of value stream mapping for lean operations
and cycle time reduction: an Indian case study. Prod Plan Control 2005;16:
44–59.
[23] Seth D, Seth N, Goel D. Application of value stream mapping (VSM) for mini-
mization of wastes in the processing side of supply chain of cottonseed oil
industrial in Indian context. J Manuf Technol Manag 2008;19:529–50.
[24] Sullivan WG, McDonald TN, Van Aken EM. Equipment replacement decisions
and lean production. Robot Comput Integr Manuf 2002;18:255–65.
[25] Yang T, Hsieh C-H, Cheng B-Y. Lean-pull strategy in a reentrant manufacturing
environment: a pilot study for TFT-LCD array manufacturing. Int J Prod Res
2011;49:1511–29.
[26] Yang T, Kuo Y, Cho C. A genetic algorithms simulation approach for the
multi-attribute combinatorial dispatching decision problem. Eur J Oper Res
2007;176:1859–73.
[27] Yang T, Kuo Y, Chou P. Solving a multiresponse simulation problem using a
dual-response system and scatter search method. Simul Model Pract Theory
2005;13:356–69.
[28] Yang T, Lu J-C. The use of multiple attribute decision-making method and
value stream mapping in solving pacemaker location problem. Int J Prod Res
2011;49:2793–817.
[29] Kelton WD, Sadowski RP, Sturrock DT. Simulation with Arena. 4th ed. New
York: McGraw-Hill; 2007.