Balancing U-Shaped Assembly Lines by Considering Human
Factors
Mehmet Kursat Oksuz
Management Faculty
Industrial Engineering Department
Istanbul Technical University
Macka, Istanbul 34367, Turkey
Sule Itir Satoglu
Istanbul Technical University
Faculty of Management
Industrial Engineering Department
Macka, Istanbul 34367, Turkey
Abstract
Assembly lines are mass production system which improve productivity, flexibility and production quality in
industrial systems. Compared to traditional straight lines and parallel lines, U-shaped assembly lines have lots of
advantages in terms of performance of workers and production system. Especially, by means of U-shaped layout,
improved balancing of assembly lines can be achieved. Moreover, the implementation of Lean Manufacturing and
Just in Time Production (JIT) systems entail U-shaped assembly line (Chiang, 2006). One-piece flow which
significantly reduces manufacturing lead times is facilitated by this type of facility layout. Multi skilled workers can
handle more than one machine in U-shaped assembly lines. An important issue in U-shaped line balancing problem
is the consideration of human factors. There are a small number of studies about human factors in the literature.
Some of the factors are the learning effect (Biskup, 1999; Toksari et al., 2008) and task time variability due to
human factors (Becker and Scholl, 2006; Chiang and Urban, 2006). The purpose of this study is developing a
methodology for balancing U-shaped assembly lines while considering the human factors to improve productivity of
the production system.
Keywords
U-shaped line balancing, just in time production, human factors
1. Introduction
Assembly lines are flow-oriented mass production systems where the productive units or stations perform the
operations. An assembly line consist of a sequence of m stations through which the product units proceed. Each
station performs a subset of the n operations necessary for manufacturing the products. The design of an assembly
line requires task to be grouped into stations such that the precedence relations among the tasks are satisfied and
some performance measures are optimized. This problem known as Assembly Line Balancing Problem (ALBP). As
a result of several simplifying assumptions, this problem was called Simple Assembly Line Balancing Problem
(SALBP) (Baybars,1986).
Traditionally assembly lines are arranged in a straight line. However, U-shaped lines are used especially in lean
production environments (Miltenburg, 2001). The associated line balancing problem is called UALBP which was
introduced by Miltenburg and Wijngaard (1994). Compared to the traditional straight lines and parallel lines, Ushaped assembly lines offer several advantages in terms of performance of workers and production system
effectiveness. Especially, by means of U-shaped layout, improved balancing of assembly lines can be achieved
(Scholl and Klein, 1999). This is due to the fact that precedence constraints of the U-shaped lines are more relaxed
than those of the the straight lines, such that, a task can share a station with any of its predecessors and/or any of its
189
sucessors (Scholl and Klein, 1999). This provides solution of the problem with a better line efficiency and lower line
imbalance.
Moreover, the implementation of Lean Manufacturing and Just in Time (JIT) production systems entail U-shaped
assembly lines, since one-piece flow can be implemented which significantly reduces manufacturing lead times is
facilited by this type of facility layout. Due to high capital requirements when installing or redesigning a line, its
configuration planning is of great relevance for practitioners. Cheng, Miltenburg, and Motwani (2000) proposed the
following factors that enhanced the wider acceptance of U-shaped lines.
U-shaped line is preferred to a straight line because of its volume flexibility. By increasing or decreasing the number
of operators on the line, a company can adjust the production rate as required. This level of volume flexibility is
harder to obtain with a straight line. Besides, since walking distance is shorter in a U-shaped than on a straight line,
it is easier for an operator to oversee several work station.
Another advantage of U-shaped lines is that the number of workstations required is never more than that required by
a straight line. There are more possibilities for grouping tasks into workstations on a U-line. Moreover, a U-line
eliminates the need for special material-handling equipment such as conveyors and other special material-handling
operators those are necessary in straight line. Instead, production operators move products from machine to
machine.
Moreover, U-shaped lines offer organizational and social advantages. The operators re expected to become multiskilled and they are able to manage several tasks which provides job-enrichment. This prevents the daily activities of
the operators from becoming routine. Since the operators can see the end product, they are motivated to achieve a
higher quality level (Scholl and Klein, 1999). Moreover, a better visibility, communication and team-work can be
achieved by means of the U-shaped lines. This facilitates a sense of belonging, and increases responsibility and
ownership compared to a straight line.
For solving the UALBP, Miltenburg and Wijngaard (1994) employed a dynamic programming formulation to solve
small problems with up to 11 tasks and proposed a heuristic to solve larger problems. Urban (1998) presented an
integer linear programming formulation to solve small to medium sized U-line balancing problems with up to 45
tasks by using standart mathematical programming software CPLEX. Scholl and Klein (1999) developed a branch
and bound procedure called ULINO to solve problems with up to 297 tasks. Becker and Scholl (2006) extended the
problem by integrating practical and relevant aspects, like parallel lines or processing alternatives. In spite of these
efforts, there is a wide gap between the academic discussion and practical applications (Boysen et al., 2007).
Since U-shaped lines are concerned with manual assembly operations, human factor is a vital aspect. However, there
is a limited literature on human factors in line balancing. Ergonomic aspects were introduced in line balancing such
as physical effort of operators and the risks that they face (Choi, 2009; Otto and Scholl, 2011; Xu et al., 2012),
learning effect (Costa and Miralles, 2009; Toksari et al., 2008; 2010), fatigue (Digiesi et al., 2009), and workers’
skills (Miralles et al., 2007, 2008; Wong et al., 2006). Avikal et al (2013) proposed a CPM based heuristic approach
for UALBP, and observed that this approach yields solutions with higher labor productivity.
Learning affect must be considered in line balancing, because as a worker becomes more experienced in performing
a specific task, in other words his/her cumulative production quantity increases, his/her task time per one piece of
product decreases in a non-linear sense. This fact can significantly cause actual task times differ from the
preassumed standard times. However, most researhers assumed that the time of task is independent from learning of
worker(s) for repetition tasks. Mosheiov (2001) determined this phenomenon. It has been examined on assembly
line balancing problems by only a few researchers so far (Chakravarty and Shtub, 1988; Cohen and Dar-El, 1998;
Cohen et al., 2006, Toksari et al., 2008).
In this study, a heuristic methodology that intends to assign tasks to the operators of a U-shaped assembly line is
developed. Since the workers have different skills and competence levels in performing assembly tasks due to the
learning effect, workers’ task performance times may vary from the standard times. An experienced worker can
execute a task in a short time, while an inexperienced one can perform the same task in a time longer than the
standard duration. Therefore, the learning effect and skills of workers affect the balancing of manual assembly lines.
Thus, in this study, the learning effect that is a significant human factor was reflected to the U-shaped assembly line
190
balancing problem (UALBP). This is a unique aspect of this study. To the best of the authors knowledge, there is no
past study that considers learning effect while balancing the U-shaped assebly lines.
The rest of the paper is organized as follows: The proposed heuristic methodology is explained in Section 2. Later, it
was exemplified in Section 3. Finally, the concluding remarks and future studies are explained in the last section.
2. Proposed Heuristic Methodology
The proposed heuristic intends to balance single model U-shaped assembly lines while considering the operators
level of competence in performing each task. Competence level of each worker was classified into four levels,
namely very competent, competent, developing, inexperienced. The capability of the workers in performing each of
the task are known in advance. These levels affect the performance index and thus the task performance time of each
worker, provided that the task is assigned to that worker.
In addition, the authors employed ULINO methodology of Scholl and Klein (1999) while assigning the tasks to the
stations/workers. The notation including the indices, sets and parameters of the methodology is presented below.
Notation:
C : Cycle time
k :
index of worker/workstation
i
:
index of tasks
S :
the set of tasks that can be assigned to workstations
the set of tasks unassigned to workstations
Su :
the station time of workstation k
Tk :
Standard time of task-i
Ti :
Performance index of worker-k for performing task-i.
Pik :
The performance index levels and their corresponding values are assumed as shown in Table 1. For instance, Pik =1
implies that this worker-k is competent in performing task-i and it can complete the task within the standard time of
the task (ti), as ti*Pik=1*ti= ti. However, a worker-k having Pik =0.75 for a task-i is very competent and can
complete this task during the time as long as ti*Pik=0.75ti. This means that this worker can complete the task less in
a time that is less than the standard task time.
Table 1: Performance index levels of workers
Index Level Explanation
Pik
Very competent
Competent
Developing
Inexperienced
0.75
1
1.25
1.50
The flowchart of the proposed methodology is illustrated in Figure 1. The methodology assumes that at each
workstation, one worker is employed. The set of unassigned tasks are established, and the first temporary station is
created. The tasks that are eligible for assignment to the U-shaped line according to the precendence constraints are
determined. In other words, the tasks are eligible provided that either all predecessors or all sucessors are assigned
before or there are no predecessors or sucessors of them. This is the precedence constraint of U-shaped assembly
lines, as mentioned by Scholl and Klein (1999). Then, a worker whose performance index of Pik is known in
advance is picked from the set-W. For the selected worker possible taskgroup-worker assignments are determined
such that the cycle time constraints as denoted in (1) are satisfied.
(1)
𝑇𝑘 + 𝑡𝑖 𝑃𝑖𝑘 ≤ 𝐶; ∀𝑘
In the left hand side of (1), ti*Pik is the standard time of task-i (ti) multiplied by Pik that is the worker’s performance
index of performing task-i. As mentioned above, each worker’s task performance time changes based on his/her
competence level and thus the performance index. Having a performance index of 1.25 or 1.5 increases the task
performance time of the worker by 25 or 50%, respectively. However, a performance index of 0.75 reduces the task
performance time of the worker. Here, cycle time constraints imply that the current station time summed with the
task performance time of the worker-k must be smaller than or equal to the cycle time.
191
Initialization;
Pik, Su=S, k=1
Create a new temporary
station.
Determine tasks eligible
for assignment.
Take a worker from W &
enumerate possible task
group assignments to
the worker such that;
Tk+ti*Pik≤ C
NO
NO
All workers in W
considered for
assignment?
YES
Is there
any task group that
is a subset of
another one?
YES
Eliminate taskgroups that are
subset of others
NO
Convert temporary
station to permanent.
All tasks have been
assigned?
YES
Determine the workertask group assignment
with least total number
of workers.
STOP
Figure 1: Flowchart of the proposed heuristic methodology
192
If all of the other workers are not considered yet, another worker is selected, and other alternative task-worker
assignments are made for the current temporary workstation. Among the alternative task-worker alternatives, if there
is a task group that is a subset of another, that one is eliminated at that stage. So, the alternative task-worker
assignments are achieved for the first permanent station.
If there is still any unassigned tasks, the number of stations is increased by one, and the heuristic returns to the initial
step where another temporary station is created. Next eligible tasks that have not been considered yet are
determined. The workers not employed in the previous station are determined, and all of the worker-task group
assignmenta are enumerated. As long as there are unasssigned tasks, this procedure is repeated. If all tasks are
assigned to the workers and stations, all feasible worker-task group assignents are achieved. Among the alternaive
assignments, the ones with the least total number of workers is selected as the best solution.
3. An Application of the Methodology
In this section, the methodology is exemplified by using a case partially adapted from (Scoll and Klein, 1999). Cycle
time of the model is 10 minutes/product. In Figure 2, precedence diagram of the tasks is presented. Here, at the
upper right hand side of the circles the standard task times (ti) are presented. For simplification, performance indices
of each worker for different tasks are assumed to be the same. Namely, Pi1=1, Pi2=0.75, Pi3=1.5.
1
7
4
5
4
5
4
2
6
3
Figure 2: Precedence diagram of the example
At the first iteration, the eligible tasks are attempted to be assigned to the first station. The calculations are
summarized in Table 2.
Table 2: Calculations of the first station
Iterations
Decision
k=1, i=1;  7*1<10
k=1, i=2; i=54*1=4<10, 4+4*1=8<10
k=2, i=2; i=14*0.75=3<10; 3+7*0.75=8.25<10
k=2,
i=2;
i=5,
i=34*0.75=3<10;
6+5*0.75=9.75<10
k=3, i=24*1.5=6<10
k=3, i=54*1.5=6<10
**eliminated**
**eliminated**
3+4*0.75=6<10,
Alternative Solution
Alternative Solution
**eliminated**
**eliminated**
So the first worker can perform only tasks 2 and 5. However, the second worker can perform tasks 2, 5, and 3 within
the cycle time. Therefore, the former task assignment group is eliminated, and the second worker should perform
either tasks 1, 2 or 2, 5, 3. In a similar manner assignment of the task 1 to the first worker is also eliminated, since
the second worker can perform 1, 2. In addition, since the third worker is inexperienced, he/she can perform only
either task 2 or 5 within the cycle time, but these are a subset of the tasks that can be performed by the second
worker. Therefore, these assignments are also eliminated. So, for the first station, the second worker is employed.
Assignments to the other workers are eliminated.
The iterations for the second station is summarized in Table 3, where it is assumed that tasks 1 and 2 are assigned to
the first station. In this stage, assignments to the third operator are eliminated. So, not the third worker but the first
worker is employed.
193
Table 3: Calculations for the second station given that (1,2) assigned to the first worker
Iterations
Decision
k=1, i=4; i=5; 6*1+4*1=10<=10
k=1, i=3; i=5; 5*1+4*1=9<10
Alternative solution
Alternative solution
k=3, i=4; 6*1.5=9<=10
k=3, i=3; 5*1.5=7.5<10
k=3, i=5; 4*1.5=6<10
**eliminated**
**eliminated**
**eliminated**
The third iteration executed for the second station is summarized in Table 4, where it is assumed that tasks 2, 5 and
3are assigned to the first station. The assignments to the thrid worker are either infeasible or eliminated. From
Tables 3 and 4, it is concluded that the first worker will be employed in the second station.
Table 4: Calculations for the second station given that (2,3,5) assigned to the first worker
Iterations
Decision
k=1, i=4;6*1<=10
k=1, i=1;7*1<=10
Alternative solution
Alternative solution
k=3, i=4; 6*1.5=9<=10
k=3, i=1; 7*1.5=10.5>10
**eliminated**
**Infeasible**
Finally, at the third station the third worker will be employed. He/she will perform either task 3 or task 4. To
visually represent the ultimate solution, a diagram is drawn as shown in Figure 3. The U-shaped assembly line has
been balanced by using three stations. If the workers competence levels were higher, better results could be
achieved. Therefore, training workers to become multi-skilled and competent in their work improves the balancing
results.
1st Station
2nd Station
3rd Station
4. Conclusion
Tasks 1, 2
Tasks 2, 5, 3
2nd Worker
2nd Worker
Tasks 4,5
Tasks 3,5
1st Worker
1st Worker
Task 3
Task 4
3rd Worker
3rd Worker
Task 4
3rd Worker
Eliminated!
Task 1
3rd Worker
Infeasible!
Figure 3: Task-worker assignment results
In U-shaped manual assembly lines, the workers skills, learning effect and thus their competence level plays an
important role, because more competent workers can perform tasks in shorter times. However, this learning effect is
usually neglected in assembly line balancing procedures, and workers are assumed to be able to complete the tasks
within the standard task times. In this study, a U-shaped assembly line balancing methodology where tasks are
assigned to workers having different competence levels is proposed. The limitaton of the methodology is that the
line balancing results are sensitive to the workers experience and skills levels. In future studies, the methodology can
be applied for well known data sets for different worker competence levels of performing tasks, to present its utility.
194
Moreover, a mathematical model of the U-shaped line balancing problem can be formulated where learning effect is
considered.
References
Avikal, S., Jain, R., Mishra, P. K., & Yadav, H. C. (2013). A heuristic approach for U-shaped assembly line
balancing to improve labor productivity. Computers & Industrial Engineering, 64(4), 895-901. doi:
10.1016/j.cie.2013.01.001
Baybars, I. (1986). A survey of exact algorithms for the simple assembly line balancing problem. Management
Science, 32, 900-932.
Becker, C., & Scholl, A. (2006). A survey on problems and methods in generalized assembly line balancing.
European Journal of Operational Research, 168(3), 694-715. doi: 10.1016/j.ejor.2004.07.023
Biskup D. (1999). Single-machine scheduling with learning considerations. European Journal of Operational
Research, 20(4), 618-627.
Boysen, N., Fliedner, M., & Scholl, A. (2007). A classification of assembly line balancing problems. European
Journal of Operational Research, 183(2), 674-693. doi: 10.1016/j.ejor.2006.10.010
Chakravarty A., & Shtub A. (1988). Modeling the effects of learning and job enlargement on assembly systems with
parallel lines. International Journal of Production Research, 26(2), 267-281.
Cheng C. H., Miltenburg J., & Motwani J. (2000). The effect of straight and U-shaped line on quality. IEEE
Transaction on Engineering Management, 47(3), 321-334.
Chiang, W.-C., & Urban, T. L. (2006). The stochastic U-line balancing problem: A heuristic procedure. European
Journal of Operational Research, 175(3), 1767-1781. doi: 10.1016/j.ejor.2004.10.031
Choi G. (2009). A goal programming mixed-model line balancing for processing time and physical workload.
Computers & Industrial Engineering, 57(1).
Cohen Y., & Dar-El M. (1998). Optimizing the number of stations in assembly lines under learning for limited
production. Production Planning and Control, 9(3), 230-240.
Costa A. M., & Miralles C. (2009). Job rotation in assembly lines employing disabled workers. International
Journal of Production Economics, 120, 625-632.
Gokcen, H., & Agpak, K. (2006). A goal programming approach to simple U-line balancing problem. European
Journal of Operational Research, 171(2), 577-585. doi: 10.1016/j.ejor.2004.09.021
Miltenburg J. (2001). U-shaped production lines: A review of theory and practice. International Journal of
Production Economics, 70, 201-214.
Miltenburg J., & Wijngaard J. (1994). The U-line line balancing problem. Management Science, 40, 1378-1388.
Miralles C., Garcia Sabater J., Andres C., & Cardos M. (2007). Advantages of assembly lines in sheltered work
centres for disabled. A case study. International Journal of Production Economics, 110(1-2), 187-197.
Miralles C., Garcia Sabater J., Andres C., & Cardos M. (2008). Branch and bound procedures for solving the
assembly line worker assignment and balancing problem: application to sheltered work centres for
disabled. Discrete Applied Mathematics, 156(3), 352-367.
Mosheiov G. (2001). Scheduling problems with a learning effect. European Journal of Operational Research, 132,
687-693.
Otto, A., & Scholl, A. (2011). Incorporating ergonomic risks into assembly line balancing. European Journal of
Operational Research, 212(2), 277-286. doi: 10.1016/j.ejor.2011.01.056
Scholl, A., & Klein, R. (1999). ULINO: Optimally balancing U-shaped JIT assembly lines. International Journal of
Production Research, 37(4), 721-736. doi: 10.1080/002075499191481
Toksari, M. D., İsleyen, S. K., Guner, E., & Baykoc, Ö. F. (2008). Simple and U-type assembly line balancing
problems with a learning effect. Applied Mathematical Modelling, 32(12), 2954-2961. doi:
10.1016/j.apm.2007.10.007
Urban T. (1998). Note: Optimal Balancing of U-Shaped Assembly Lines. Management Science, 44(5), 738-741.
Wong W., Mok P., & Leung S. (2006). Developing a genetic optimization approach to balance an apparel assembly
line. International Journal of Advanced Manufacturing Technology, 39(9), 1009-1015.
Xu W., Ko J., Cochran D. J., & Jung M. C. (2012). Design of assembly lines with the concurrent consideration of
productivity and upper extremity musculoskeletal disorders using linear models. Computers & Industrial
Engineering, 62, 431-441.
195
Biography
Mehmet Kursat Oksuz is a research assistant of Industrial Engineering at Istanbul Technical University. He
currently continues his graduate studies at the same department. His current research interests include assembly line
balancing and human factors.
Sule I. Satoglu is an Associate Professor of Industrial Engineering at Istanbul Technical University. She earned her
MSc and PhD degrees from Istanbul Technical University Industrial Engineering Department in 2002, and 2008,
respectively. Some of her research areas are design of production systems, lean production and logistics, supply
chain management, system simulation and mathematical modelling.
196