See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/332430005

Minimization of Bottleneck and Workstations for Assembly Line Balancing

Problem in Power Transformer Manufacturing Using Heuristics

Conference Paper · November 2018

DOI: 10.1109/ICSIMA.2018.8688780

CITATION READS

1 246

5 authors, including:

Nurhanani Abu Bakar Mohammad Fadzli Ramli

Universiti Malaysia Perlis Universiti Malaysia Perlis
5 PUBLICATIONS   5 CITATIONS    44 PUBLICATIONS   60 CITATIONS   

SEE PROFILE SEE PROFILE

Mohd Zakimi Zakaria Tan Chan Sin

Universiti Malaysia Perlis Universiti Malaysia Perlis
48 PUBLICATIONS   111 CITATIONS    24 PUBLICATIONS   32 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A Review on Water Pricing Problem for Sustainable Water Resource View project

Solving assembly line balancing problem using heuristic: A case study of power transformer in electrical industry View project

All content following this page was uploaded by Tan Chan Sin on 19 April 2019.

The user has requested enhancement of the downloaded file.

 Proc. of the 2018 IEEE 5th International Conference on Smart Instrumentation, Measurement and Applications (ICSIMA 2018)
29-30 November 2018, Songkhla, Thailand

Minimization of Bottleneck and Workstations for

Assembly Line Balancing Problem in Power
Transformer Manufacturing Using Heuristics
First Nurhanani Abu Bakar#1, Mohammad Fadzli Ramli#2, Mohd Zakimi Zakaria*3, Tan Chan Sin*4, Hafiz Masran#5
#
Institute of Engineering Mathematics, Universiti Malaysia Perlis, Pauh Putra Main Campus, 02600 Arau, Perlis, Malaysia
1nurhanani143@gmail.com, 2mfadzli@unimap.edu.my, 5hafizmasran@unimap.edu.my

*
School of Manufacturing Engineering, Universiti Malaysia Perlis, Pauh Putra Main Campus, 02600 Arau, Perlis, Malaysia
3zakimizakaria@unimap.edu.my, 4cstan@unimap.edu.my

2second.author@second.com

Abstract— This project deals with a case study of SALBP in
power transformer manufacturing. Some workstations in
assembly line experienced bottlenecks, while the others suffered
with high idle time. Therefore, four heuristic methods are chosen
in order to provide solutions for minimizing number of
workstations and improve the bottlenecks simultaneously. In this
study, LCR, RPW and LPT are successfully minimized number of
workstation from 19 to 16 workstations. Consequently, these
solutions affect the layout. On the contrary, SPT able to improve
the bottleneck by reducing number of workstations from 19 to 17
without affect the line layout. Therefore, this has created an option
for engineer to decide which decision should be implement to the
assembly line due to enhance the line efficiency.
SALBP can be split into two problem types which are SALBP-
1 [10] (minimize number of workstations, given cycle time) and
SALBP-2 (minimize cycle time, given number of workstation)
[5], [11], [12].
There are three types of assembly line [13], which are single
model [14], batch model [15], [16] and mixed model assembly
line [17], [18]. Besides, there are some assembly line designs
such as U-shape [2], [19], [20], parallel station [20], [21] straight
line [13], [22], loop and rectangular assembly line. This paper
only considers the single model assembly line with straight-line
design of SALBP-1.

Keywords—Assembly line balancing problem; heuristic method; II. PROBLEM DESCRIPTION

multi-objective optimization; workstation; idle time; bottleneck
I. INTRODUCTION
An assembly line is defined as a set of works (operations)
that assigned to a set of workstations. They linked together by
the transport mechanism which determined by the flow process
from one station to another [1]. Authors from [2] described
briefly the statement above and stated that an assembly line
contains a sequence of works that connected by a conveyor
moving at constant speed. At the same time, [3] defined the
assembly line consist of several workstations linked by certain
transport mechanisms. A task in the assembly line is the smallest
element of work and the way it works, based on the precedence
diagram. The time required to perform a task is called working
time or processing time. The cycle time that also known as
station time is the maximum working time for each station.
Generally, the cycle time is determined based on the demand for
the product over time. In other words, the cycle time is the
production rate required by the industry.
The improvement of assembly line is important in
manufacturing industry. The best way to improve the assembly
line is by achieving the line balancing. The line balancing is
referred to the reducing idle time between workstations [4]. This
have attracted more researchers like [5] - [8] to study about
Simple Assembly Line Balancing Problem (SALBP) which also
known as Assembly Line Balancing Problem (ALBP) [9]. The
Assembly line is a kind of production system that is widely
used in manufacturing companies [23], [24]. Many
manufactured products including automobile [25], [26],
computer, electronic equipment [27] and others are
manufactured by the assembly line system. However, most of
the manufacturing industries faced with imbalanced among
workstations due to the idle time and bottleneck [28]. This
problem can be seen clearly in one of the electrical industry.
This industrial site uses an assembly line to manufacture one
type of power transformer model called PM 17-P.
This study deals with the problem in factory that
manufacture the power transformer, which consist of 19
workstations. The industrial engineers have set 29 seconds for
cycle time as a target for each workstation. The efficiency rate
of assembly line during this study is 67%. This rate is consider
poor, as there are other competing companies that capable to
produce the same product with better performance and less
number of workstations.
The SALBP in this electrical factory have shown that
current system is less efficient. This leads to the higher costs
and low productivity. Graph in Fig. 1 shows the current
performance SALBP of PM 17-P. Fig. 1 clearly shows that
workstation 10 is facing with the bottlenecks problems. While
other workstations are suffering with idling problems, where
operators are forced to wait for the next task from the previous
workstation.
 TABLE I. TASK AND WORKSTATION TIME FOR SALBP OF PM 17-P
SALBP of PM 17-P Working Workstation Precedence
40 Workstation Task
Time (s) Time (s) Constraint
35 1 1 17.95 17.95 -
Workstation time (s)

30 cycle time, C 2 2 18.53 18.53 -

25 3 3 26.72 26.72 2
20 4 4 18.12 18.12 3
15 5 5 20.91 20.91 4
6 6 22.26 22.26 5
10
7 7 11.28 11.28 1
5 8 8 8.64 8.64 6
0 9 5.50 7. 8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 9 14.05
10 8.55 9
Workstation 11 18.05 10
12 8.24 11
10 37.96
13 2.40 12
Fig. 1. Graph for current system SALBP of PM 17-P 14 9.27 13
15 19.83 14
11 29.11
16 9.28 15
Therefore, the engineers believe that the number of 17 10.55 16
workstations and idle time should be minimize and reduce in 12 14.34
18 3.79 17
order to make the assembly line more balanced. Hence, an 13 19 13.71 13.71 18
approach should be taken to improve the efficiency of the 14 20 17.39 17.39 19
15 21 22.50 22.50 20
existing assembly line. This step should be done to reduce costs 22 11.02 21
and increase production productivity. Thus, this paper is 16 15.83
23 4.81 22
solving SALBP by reducing idle time, minimizing number of 17 24 10.85 10.85 23
workstation and balancing workload each workstation. 25 21.36 24
18 23.02
26 1.66 25
19 27 24.11 24.11 26
III. METHODOLOGY
Heuristic is a popular methods [29] and often used as a of weight is calculated by summing up the total task time with
method for solving SALBP. This method was chosen because it previous weights value. A task that has the highest weights
suggests a solution based on the rules and procedures compared value is selected first to assign to the workstation. Then,
to others complex mathematical approaches. Solution using followed by a smaller RPW value. The task assignment should
mathematical methods can be too complex especially when be considered by the previous constraint.
involving with bigger problems. Some authors like [30] stated
C. Shortest Processing Time (SPT)
that heuristic is the procedure for finding optimal solution.
So, Largest Candidate Rule (LCR), Ranked Positional Shortest Processing Time (SPT) method uses the criteria of
Weight (RPW), Shortest Processing Time (SPT) and Longest selecting the shortest processing time or working time to assign
Processing Time (LPT) are used to determine whether the to the workstation. In this method, the processing time for node
number of workstations in the assembly line can be reduced or without previous constraint and independent node are compared
not. Together, overcome the bottlenecks and high idling times and the shortest processing time will be selected.
between workstations. The comparison of these methods with
current system are analyzed in order to identify the best decision D. Longest Processing Time (LPT)
to optimize the assembly line. Table I shows the current system
of assembly line that consists of 19 workstations and 27 tasks. Longest Processing Time (LPT) method uses the criteria of
Some parameters and formulas have been used in this paper selecting the longest processing time or working time to assign
in order to identify the best method that give the optimal to the workstation. In this method, the processing time for node
solution. Terminologies and formulas are showed as below: without previous constraint and independent node are compared
and the longest processing time will be selected.
A. Largest Candidate Rule (LCR)
In this method, task (𝑖 = 1,2, … , 𝑛) are arranged in IV. RESULT AND DISCUSSION
descending order of workstation time from highest to lowest Theoretically, heuristic methods such as LCR, RPW, SPT
value. The assignment of tasks into workstations are sort from and LPT are able to improve the assembly line balancing
top to bottom based on previous constraint. When one task is problem of PM 17-P. It is shown that these four heuristic
selected to assign into the workstation, the next task selection methods succeed in balancing the assembly line better than the
will be repeated from the top list. This process are repeated until current system. Table II represents a comparison of analysis
all the tasks are assigned to the workstations. between heuristic methods and current system.
Based on Table II, calculations of heuristic methods LCR,
B. Ranked Positional Weight (RPW) RPW, SPT and LPT have proposed two new system models
Ranked Positional Weights (RPW) is a method that sorting SALBP of PM 17-P which are model of layout changes and
task based on weights value from largest to smallest. The value model of layout without changes.
A. Model I: Layout Changes Model TABLE III. TASK SPECIFICATION FOR LAYOUT CHANGES (MODEL I)

The analysis of LCR, RPW and LPT bring up Model I that Workstation Task Task Description
Working Workstation
gives the same result for all three methods. However, the Time (s) Time (s)
Secondary
changes in the task specification of the operator and balancing 1 2 18.53 18.53
winding (sec 1)
result also affect the machine layout. Where the winding Secondary
machine currently at workstation 1 had to be moved to 2 3 26.72 26.72
winding (sec 2)
workstation 6. Whereas the change in task specification involve Secondary
3 4 18.12 18.12
at workstations 6, 7, 8, 9, 10, 14 and 15. Table III shows the winding (sec 3)
change in specification for this model. Secondary
4 5 20.91 20.91
winding (sec 4)
This model only involves small layout changes without 5 6
Secondary outer
22.26 22.26
creating the complicated flow process. In addition, layout wrap
changes model uses 16 workstations and 17 operators in 1 Primary winding 17.95
6 Secondary bobbin 26.59
assembly line compare to current system which consists of 19 8 8.64
soldering
workstations and 19 operators. Primary bobbin
7 11.28
B. Model II: Layout Without Changes Model soldering
7 9 Bobbin assembly 5.50 25.33
This model is a result of SPT heuristic analysis that only
10 Side tape 8.55
involves changes in task specifications on operators. This
model does not change the machine layout and it uses 17 11 E-core insert 18.05
workstations with 17 operators only. Table IV shows the 8 12 Hi-pot test 8.24 28.69
specification of task on the operator for this model. 13 Gluing 2.40
Referring to Table IV, there are changes in the task 14 I-core insert 9.27
specification on operator at workstations 8, 9, 10, 11 and 15. 9 29.10
15 Welding 19.83
The change of assignment task occurred in workstations located
16 Copper foil insert 9.28
at the center of the assembly line, which is at workstation 10 is
proposed to overcome the bottleneck problems. 10 17 Core shield insert 10.55 23.62

TABLE II. COMPARISON OF ANALYSIS BETWEEN HEURISTIC METHODS 18 Bracket insert 3.79
AND CURRENT SYSTEM 11 19 Bracket clamping 13.71 13.71
Heuristic Methods Current 12 20 Functional test 17.39 17.39
Analysis
LCR RPW SPT LPT System 13 21 Varnish 22.50 22.50
Minimum number of
16 16 17 16 19 22 VMI 11.02
workstations, 𝑆𝑚𝑖𝑛
Number of operators, 𝑤 17 17 17 17 19 23 Labelling 4.81
14 26.68
Total idle time, 𝑇𝑖𝑑 Touch-up
96.72 96.72 125.72 96.72 173.72 24 10.85
(sec/unit) soldering
Efficiency line, 𝐸𝑏 (%) 79 79 74.5 79 67
25 Functional test 21.36
Balance delay, 𝑑 (%) 21 21 25.5 21 33 15 23.02
Line balancing loss, L 26 Packing 1.66
21.1 21.1 25.5 21.1 49
(%) 16 27 QC checking 24.11 24.11
Layout changes Yes Yes No Yes -
Change of operator Theoretically, the heuristic method implemented on the
Yes Yes Yes Yes -
specification work
SALBP of case studies succeeded in balancing the assembly
line. The assembly line of 17-P PM should be improved to
Comparison of workstation time between current system,
Model I and Model II
overcome the bottleneck condition and reduce the idling time
between the workstations. These two problems had caused the
Current system Model I Model II imbalanced assembly line and unstable production.
Solutions using heuristic methods such as LCR, RPW, SPT
40 and LPT are aimed to minimize the number of workstations in
35
Workstation time (s)

cycle time, C
30 the assembly line, overcoming the bottleneck condition and the
25
20 idling gap. This step is taken to make ALBP of PM 17-P more
15 efficient and productive. Hence, two proposed models to
10
5 improve the balancing of the assembly line PM 17-P is
0 presented in this paper. In this case, the management team of
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
this electrical industry have some idea to look at the best
Workstation
approach to encounter the problem.
Model I is the solution by changing the position of the
Fig 2 Comparison of workstation time between current system, Model I winding machine and change the task specification on the
and Model II operator. While Model II only involves changes in the task
specification on the operator.
 TABLE IV. TASK SPECIFICATION FOR LAYOUT WITHOUT CHANGES Therefore, two models are proposed Model I and Model II
(MODEL II)
to represent the results. The LCR, RPW and LPT are
Workstation Task Task Description
Working Workstation categorized as Model I that involves changes in task
Time (s) Time (s) specification of operators and layout. Meanwhile, the SPT is
1 1 Primary winding 17.95 17.95 categorized as Model II that only involves changes in task
Primary bobbin specification of operators. Both model capable to reduce
2 7 11.28 11.28
soldering
Secondary number of workstations and number of operators. Therefore,
3 2 18.53 18.53 this has created an option for engineer to decide which decision
winding (sec 1)
4 3
Secondary
26.72 26.72
should be implement to the assembly line due to enhance the
winding (sec 2) line efficiency.
Secondary
5 4
winding (sec 3)
18.12 18.12 Hence, for further studies, it is suggested to verify and
Secondary analyze the effects of layout changes. Simulation of changes in
6 5 20.91 20.91
winding (sec 4) layout and task specifications on operators can be proposed to
7 6
Secondary outer
22.26 22.26 recognize the effects of this action.
wrap
Secondary bobbin ACKNOWLEDGMENT
8 8.64
soldering
8 9 Bobbin assembly 5.50 22.69 The authors would like to acknowledge the support from
Universiti Malaysia Perlis (UniMAP) and the Fundamental
10 Side tape 8.55
Research Grant Scheme (FRGS) under a grant number of
11 E-core insert 18.05
FRGS/1/2016/TK03/UNIMAP/03/8 from the Ministry of
9 12 Hi-pot test 8.24 28.69 Higher Education Malaysia and
13 Gluing 2.40 FRGS/1/2016/STG06/UNIMAP/03/6 from the Ministry of
14 I-core insert 9.27 Higher Education Malaysia.
10 29.10
15 Welding 19.83
REFERENCES
16 Copper foil insert 9.28
[1] İ. Baybars, "A survey of exact algorithms for the simple assembly line
11 17 Core shield insert 10.55 23.62 balancing problem", Management Science, vol. 32, no. 8, pp. 909-932,
1986.
18 Bracket insert 3.79
[2] A. Scholl and R. Klein, "Balancing assembly lines effectively – A
12 19 Bracket clamping 13.71 13.71 computational comparison", European Journal of Operational Research,
13 20 Functional test 17.39 17.39 vol. 114, no. 1, pp. 50-58, 1999.
14 21 Varnish 22.50 22.50 [3] K. Fleszar and K. Hindi, "An enumerative heuristic and reduction
methods for the assembly line balancing problem", European Journal of
22 VMI 11.02 Operational Research, vol. 145, no. 3, pp. 606-620, 2003.
15 23 Labelling 4.81 26.68 [4] A. Baykasoğlu and T. Dereli, "Simple and U-type assembly line balancing
Touch-up by using an ant colony based algorithm", Mathematical and
24 10.85 Computational Applications, vol. 14, no. 1, pp. 1-12, 2009.
soldering
25 Functional test 21.36 [5] G. Nicosia, D. Pacciarelli and A. Pacifici, "Optimally balancing assembly
16 23.02 lines with different workstations", Discrete Applied Mathematics, vol.
26 Packing 1.66 118, no. 1-2, pp. 99-113, 2002.
17 27 QC checking 24.11 24.11 [6] O. Battaïa and A. Dolgui, "Reduction approaches for a generalized line
balancing problem", Computers & Operations Research, vol. 39, no. 10,
Fig 2 illustrates the comparison of workstation time between pp. 2337-2345, 2012.
current system, Model I and Model II, where the current system [7] M. Azizoğlu and S. İmat, "Workload smoothing in simple assembly line
balancing", Computers & Operations Research, vol. 89, pp. 51-57, 2018.
consists of 19 workstations. However, number of workstation
[8] D. Giglio, M. Paolucci, A. Roshani and F. Tonelli, "Multi-manned
can be reduced to 16 workstations by using Model I and 17 assembly line balancing problem with skilled workers: A new
workstations by using Model II. mathematical formulation", IFAC-PapersOnLine, vol. 50, no. 1, pp.
1211-1216, 2017.
V. CONCLUSION [9] M. Gansterer and R. Hartl, "One- and two-sided assembly line balancing
problems with real-world constraints", International Journal of
The SALBP is the most common problem faced by Production Research, vol. 56, no. 8, pp. 3025-3042, 2017.
manufacturing companies. The bottleneck condition and the [10] Z. Li, I. Kucukkoc and Z. Zhang, "Branch, bound and remember
idle gap between workstations are two issues that can be seen algorithm for U-shaped assembly line balancing problem", Computers &
in SALBP. In this study, four heuristic methods are addressed Industrial Engineering, vol. 124, pp. 24-35, 2018.
to solve the SALBP of PM 17-P power transformer at electrical [11] S. Karabatı and S. Sayın, "Assembly line balancing in a mixed-model
industry. The heuristic methods able to balance the assembly sequencing environment with synchronous transfers", European Journal
of Operational Research, vol. 149, no. 2, pp. 417-429, 2003.
line and improve the efficiency of the production. However, the
[12] M. Fathi, D. Fontes, M. Urenda Moris and M. Ghobakhloo, "Assembly
results from LCR, RPW and LPT has led to the changes in line balancing problem", Journal of Modelling in Management, vol. 13,
layout and task specification against operators. Different from no. 2, pp. 455-474, 2018.
SPT that reveals a solution, which are affecting to the operator’s [13] M. Groover, Automation, production systems, and computer-integrated
task specification without change the line layout. manufacturing. Upper Saddle River, N.J.: Prentice Hall, 2008.
[14] L. Borba, M. Ritt, and C. Miralles, “Exact and heuristic methods for
solving the robotic assembly line balancing problem,” European Journal
of Operational Research, vol. 270, no. 1, pp. 146–156, 2018.
[15] M. Kabir and M. T. Tabucanon, “Batch-model assembly line balancing:
A multiattribute decision making approach,” International Journal of
Production Economics, vol. 41, no. 1-3, pp. 193–201, 1995.
[16] S. Qu and Z. Jiang, “A memetic algorithm approach for batch-model
assembly line balancing problem of sub-block in
shipbuilding,” Proceedings of the Institution of Mechanical Engineers,
Part B: Journal of Engineering Manufacture, vol. 228, no. 10, pp. 1290–
1304, 2014.
[17] K. Ağpak and S. Zolfaghari, “Mathematical models for parallel two-sided
assembly line balancing problems and extensions,” International Journal
of Production Research, vol. 53, no. 4, pp. 1242–1254, 2014.
[18] L. Tiacci, “Mixed-model U-shaped assembly lines: Balancing and
comparing with straight lines with buffers and parallel
workstations,” Journal of Manufacturing Systems, vol. 45, pp. 286–305,
2017.
[19] M. Li, Q. Tang, Q. Zheng, X. Xia, and C. Floudas, “Rules-based heuristic
approach for the U-shaped assembly line balancing problem,” Applied
Mathematical Modelling, vol. 48, pp. 423–439, 2017.
[20] I. Kucukkoc and D. Z. Zhang, “Balancing of parallel U-shaped assembly
lines,” Computers & Operations Research, vol. 64, pp. 233–244, 2015.
[21] P. McMullen and G. Frazier, "Using simulated annealing to solve a
multiobjective assembly line balancing problem with parallel
workstations", International Journal of Production Research, vol. 36, no.
10, pp. 2717-2741, 1998.
[22] M. H. Alavidoost, M. H. F. Zarandi, M. Tarimoradi, and Y. Nemati,
“Modified genetic algorithm for simple straight and U-shaped assembly
View publication stats
line balancing with fuzzy processing times,” Journal of Intelligent
Manufacturing, vol. 28, no. 2, pp. 313–336, 2017.
[23] M. Yuan, H. Yu, J. Huang, and A. Ji, “Reconfigurable assembly line
balancing for cloud manufacturing,” Journal of Intelligent
Manufacturing, pp. 1–15, Dec. 2018.
[24] Y. Li and D. Coit, “Priority rules-based algorithmic design on two-sided
assembly line balancing,” Production Engineering, vol. 12, no. 1, pp. 95–
108, 2017.
[25] A. Kalra, S. Marwah, S. Srivastava, and R. Bhatia, “Productivity
improvement in assembly line of automobile industry by reducing cycle
time of operations,” International Journal of Engineering Research and,
vol. 5, no. 5, pp. 28–31, Mar. 2016.
[26] M. Hamzas, S. Bareduan, M. Zakaria, S. Ghazali and S. Zairi,
"Implementation of ranked positional weight method (RPWM) for
double-sided assembly line balancing problems", In 3rd Electronic and
Green Materials International Conference, Krabi, Thailand, 2017, pp.
020183.
[27] K. Park, S. Park, and W. Kim, “A heuristic for an assembly line balancing
problem with incompatibility, range, and partial precedence
constraints,” Computers & Industrial Engineering, vol. 32, no. 2, pp.
321–332, 1997.
[28] N. T. Lam, L. M. Toi, V. T. T. Tuyen, and D. N. Hien, “Lean line
balancing for an electronics assembly line,” Procedia CIRP, vol. 40, pp.
437–442, 2016.
[29] M. F. M. A. Hamzas, S. A. Bareduan, and M. Z. Zakaria, “A review on
multi-objective optimization using evolutionary algorithms for two-sided
assembly line balancing problems,” ARPN J. Eng. Appl. Sci., vol. 11, no.
18, pp. 10882–10887, 2016.
[30] R. Schonberger and E. Knod, Operations management. Chicago: Irwin,
1997