Int. J.

Production Economics 239 (2021) 108188

Contents lists available at ScienceDirect

International Journal of Production Economics

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

Fuzzy multi-objective model for assembly line balancing with ergonomic

risks consideration
Rifat Ozdemir *, Ilkan Sarigol, Sarah AlMutairi, Sarah AlMeea, Abrar Murad, Aseel Naqi,
Noor AlNasser
College of Engineering and Technology, American University of the Middle East, Kuwait

A R T I C L E I N F O A B S T R A C T

Keywords: The main motivation behind the improvement of manufacturing systems is that they are competing for a larger
Assembly line balancing share and better image in the market. Many manufacturing systems, particularly those with repetitive products,
Ergonomic risk utilize assembly lines for the final stage of their production systems. Most of the assembly lines still rely on
Ergonomic simulation
human worker performance since higher flexibility can be attained with manual systems. The performance of
Fuzzy multi-objective
manual assembly line systems depends on how ergonomic factors are involved with balancing the lines.
Therefore, the main focus of this study is to develop a model to design assembly lines with ergonomic risk
consideration. Ergonomic risks are the factors of weight, size, and shape of the components in addition to the
posture of the workers performing the tasks with the components. This study also conducts ergonomic risk
analysis for assembly tasks using simulation software and incorporates the results into the developed model.
There are two conflicting objectives in the problem, thus the required solution model is constructed as multi-
objective. Since the objectives of the problem have vagueness, the proposed model is developed adopting
fuzzy set theory. The developed fuzzy multi-objective model is implemented in a real-life case problem gathered
from a refrigerator company.

1. Introduction human factors might escalate the effect of those risk factors, thus, result
in the increase of work-related musculoskeletal disorders (WMSDs).
Most of the production systems face the problem of reconfiguration, Ergonomic risk consideration in production systems targets particularly
based on the changes needed to modify their production lines. Due to WMSDs. Therefore, ergonomics should be integrated with assembly line
their modular nature, assembly lines are the usual target for these balancing (Sgarbossa et al., 2020; Battini et al., 2011). Main expecta­
changes. Configuration of the line and task assignment in stations for the tions from ergonomic risks consideration in assembly line balancing are
manual assembly systems is easier to change than other manufacturing the reductions in WMSDs and the improvements of the comfort in work
lines or automated assembly systems. This is the reason why the problem conditions with a significant decrease in costs associated with job
of assembly line balancing is still a widely researched topic in literature absenteeism, medical treatments, worker rehabilitation, errors in
(Scholl, 1999; Amen, 2001; Becker and Scholl, 2006; Barathwaj et al., manual operations, and decrease in quality (Battini et al., 2011).
2015). More importantly, improvements on assembly systems directly A poor ergonomic design of an assembly line generates not only
affect the productivity of the production systems, since improvements WMSDs but also economic losses for companies due to a large number of
mean decreased input (i.e., less worker time, a smaller number of sta­ sick leaves. The Fourth European Survey on Working Conditions showed
tions, etc.) or increased output (i.e., higher throughput rate, larger de­ that 35% of machine operators and assemblers experience backache and
livery rate, etc.). muscular pains (Schneider et al., 2010). The cost of these disorders
The activities in assembly lines involve many risk factors such as reached 2% of the Gross National Product in the EU (Schneider et al.,
repetitive motions, awkward postures, heavyweights, and other stress 2010). Improving ergonomic conditions might reduce the sick leaves
factors. Pressure on workers for higher efficiency without considering and the defect rate (Eklund, 1995; González et al., 2003). Since there is a

* Corresponding author.
E-mail addresses: [email protected] (R. Ozdemir), [email protected] (I. Sarigol), [email protected] (S. AlMutairi), Sarah.
[email protected] (S. AlMeea), [email protected] (A. Murad), [email protected] (A. Naqi), [email protected] (N. AlNasser).

https://doi.org/10.1016/j.ijpe.2021.108188
Received 13 June 2020; Received in revised form 15 April 2021; Accepted 22 May 2021
Available online 27 May 2021
0925-5273/© 2021 Elsevier B.V. All rights reserved.
R. Ozdemir et al. International Journal of Production Economics 239 (2021) 108188

positive association between employee health risks and system perfor­ precedence and cycle time constraints, a GALB problem contains further
mance, considering ergonomic factors in the assembly line might in­ characteristics like cost functions, equipment selection, paralleling,
crease the productivity of manufacturing and quality of products U-shaped line layout, and mixed-model production (Becker and Scholl,
(Battini et al., 2016; Di Benedetto and Fanti, 2012; Mutlu and Özgörmüş, 2006). SALB and GALB problems have been extensively studied in the
2012; Neumann and Dul, 2010; Otto and Battaïa, 2017). literature for decades (Amen, 2000, 2001bib_Amen_2000bib_A­
The prior literature has considered the issue as minimization of er­ men_2001; Battaïa and Dolgui, 2013; Becker and Scholl, 2006; Erel and
gonomic risks by using well-known energy expenditure equations or Sarin, 1998; Ghosh and Gagnon, 1989; Kriengkorakot and Pianthong,
ergonomic risk assessment tools such as the NIOSH (National Institute of 2007; Kumar and Mahto, 2013; Malakooti, 1991; Scholl, 1999). The
Occupational Safety and Health) lifting equation, REBA (Rapid Entire SALBP has two types of classical objective functions. While SALBP of
Body Assessment) matrix, and RULA (Rapid Upper Limb Assessment) type 1 (SALBP-1) minimizes the number of stations for the given cycle
tools, (Hignett and McAtamney, 2000; McAtamney and Corlett, 1993; time, SALBP of type 2 (SALBP-2) minimizes the cycle time for the given
Waters et al., 1993). Recently, significant improvements have been number of stations.
established in human factor engineering and ergonomic risk assessment The next subsections show a literature review on ergonomic ALB and
by introducing Digital Human Modeling (DHM) and Ergonomic Simu­ risk assessment tools.
lation Modeling (ESM) (Cheshmehgaz et al., 2012; Ding and Hon, 2013;
Enomoto et al., 2013; Krüger and Nguyen, 2015; Longo and Mirabelli, 2.1. Ergonomic assembly line balancing
2009; Shewchuk et al., 2016). However, most of the studies focused on
DHM and ESM primarily for workplace and work tool designs, while few Integrating ergonomic factors into ALBP is a maturing (Grosse et al.,
have also worked on risk assessment by using ergonomic simulation. 2017) research field. Proper assignment of tasks to the workstations
There have been limited studies incorporating digital human mod­ might affect the ergonomic risk level while keeping cycle time, the
eling/ergonomic simulation with assembly line balancing problems number of workstations, and profitability unchanged (Otto and Scholl,
(Bortolini et al., 2017). 2011). Improving ergonomic conditions might reduce the sick day
The main contribution of this study to the ergonomic assembly line leaves, and the defect rate (Eklund, 1995; González et al., 2003).
balancing problem is introducing a new risk assessment index by using Moreover, ergonomics assembly line designs might increase the effi­
DHM and ESM, particularly the Siemens Tecnomatix Jack simulation ciency and profitability of manufacturing. Carnahan et al. (2001) stud­
software. The proposed risk assessment tool assumes that each assembly ied the assignment of gripping tasks and their effect on fatigue and
task consists of lifting, twisting-hip, twisting-wrist, and squatting work recovery of workers. They balanced the assembly line with minimum
elements with different risk levels. The study also proposes a Jack cycle time and physical workload and compared different solution ap­
simulation model for accomplishing lower back force analysis to proaches. Colombini and Occhipinti (2006) decreased the ergonomic
determine risk for lifting work element, the strength of joint analysis for risk without changing the cycle time and number of workstations, by
twisting-wrist and twisting-hip work elements, and recovery time re-balancing the assembly line. Costa and Miralles (2009) defined the
analysis for squatting work element. Simulation results are then incor­ problem of workload assignment to stations and workers with different
porated into a fuzzy multi-objective model by using newly developed abilities. Otto and Scholl (2011) suggested applying the ergonomic
risk assessment equations in which the risk values of the work elements objective function and developed Ergo-ALBP depending on the ergo­
obtained from simulation runs are used. nomic risk aggregation. They described modeling ergonomic factors
The joint problems of assembly line balancing and ergonomic risks with various possible objective functions in the assembly line balancing
consideration, which has two conflicting objectives of minimizing cycle problem. The authors adapted several risk assessment tools to measure
time to make the line more efficient and minimizing imbalance of er­ ergonomic risks. They studied the minimization of the average ergo­
gonomic risks between workers to make the line more humane. The nomic risks, the number of stations with high ergonomic risks, and de­
achievement degree of the objectives is challenging to define, which is viations from acceptable levels of station physical loads. Despite most of
more precisely due to the lack of knowledge towards the significance of the risk functions being nonlinear; Xu et al. (2012) measured the risk
their contribution to the organizational goals. Therefore, the utilization values with linear functions. Bautista et al. (2016) defined Time and
of a fuzzy multi-objective model was carried out with two objectives. Space Constrained Assembly Line Balancing Problem (TSALBP) by
This study falls into the category of SALBP-2 with ergonomic risks and including space constraint and measuring ergonomic risk at stations as a
the major contributions of this paper to the literature are threefold. First, sum of task-specific parameters. Their objective was to minimize either
it introduces a new structural framework of ergonomic simulation the maximum ergonomic risks for a station or absolute deviations be­
modeling for risk assessment of assembly tasks with digital human tween the ergonomic risks of stations.
modeling. Second, it presents new risk assessment equations for incor­ Researchers mostly preferred the weighted sum method to integrate
porating the results of the Jack ergonomic simulation model with the ergonomic and economic objective functions. Jaturanonda et al. (2013)
fuzzy multi-objective mathematical model. Third, it guides the practi­ define the objective function as a weighted sum of variances of station
tioners from the field since it has a focus on the real case problem of a times and ergonomic risks among stations. Barathwaj et al. (2015)
manual refrigerator assembly line with real ergonomic design data. minimized the weighted sum of the number of stations, the average
The remainder of this paper is structured as follows; section 2 re­ deviation of the station idle times, and the total ergonomic risks with the
views the related literature, and section 3 describes the methodology for given cycle time. Whereas, Amid et al. (2011) developed a weighted
the problem and the solution. The case study, results, and discussion are max-min fuzzy model for dealing with imprecision in a multi-objective
presented in section 4. Section 5 concludes the paper and presents di­ decision-making environment. Later, Mutlu and Özgörmüs (2012)
rections for future research. applied fuzzy numbers for the expert evaluations and presented a fuzzy
linear model limiting the physical load of each station.
2. Literature review
2.2. Ergonomic risk assessment
The Assembly Line Balancing Problem (ALBP) has been attracting
interest since it was first formulated by Salveson, 1955. The purpose of Numerous researches suggested applying expert scores for ergo­
ALBP is to optimize the partitioning assembly tasks among the stations, nomic risks instead of common risk measurement methods. Choi (2009)
with respect to the specified objective function. The problem is classified defined parameters counting the physical load for each task. After the
into Simple Assembly Line Balancing (SALB) and Generalized Assembly experts evaluated parameters with an ordinal scale, the summation of
Line Balancing (GALB) problems. While a SALB problem simply involves the particular parameters provided an index value for each category of

2
R. Ozdemir et al. International Journal of Production Economics 239 (2021) 108188

ergonomic risks. considering both the workers’ motion and ergonomic assessments. Some
Several methods such as direct, observational, subjective, and psy­ commercial software incorporates the time and ergonomic features in
chophysiological are provided in the literature to measure the ergo­ the ALBP (Cheshmehgaz et al., 2012). Siemens, Jack, and Dassault
nomic factors (Otto and Battaïa, 2017). Some of the methods have a Delmia are the most common software which uses DHM to simulate the
limited scope and they either target only specific work systems or manufacturing environment and assembly line. The Jack mannequin has
consider only specific risk factors. The NIOSH (National Institute of a high-fidelity model resembling the human skeletal structure. Manne­
Occupational Safety and Health) lifting index and OCRA (Occupational quins can be designed using the posture library of Jack. More impor­
Repetitive Actions) index are two well-known members of this category. tantly, certain behavioral controls can be specified to restrict the
NIOSH lifting index as implied by the name has a focus on lifting and allowable set of motions of the mannequins in order to reflect the real
handling work systems. It calculates the risk value as a ratio between work system more realistically. (Raschke and Cort, 2019). The tasks and
actual weight and recommended weight. NIOSH considers factors such operators are virtually demonstrated and evaluated with a wide range of
as posture while carrying and lifting, the coupling of weight, frequency performance indicators. It is possible to examine the operator task from
of the task, and angle of motion in carrying the weight. Whereas a the ergonomic (NIOSH, REBA index, etc.) aspect and the time (MTM
handling work system with high repetition is the main concern of the motion analysis, MOST, etc.) viewpoint. However, the deficiency of
OCRA index. This measure considers the repetition of work, force, and software is the optimization part that should be discussed together with
posture of upper limbs. Musculoskeletal disorders related to upper limbs software results (Bortolini et al., 2017).
are in the scope of the OCRA index. The RULA (Rapid Upper Limb
Assessment) is also used for risk assessment of particularly upper ex­ 3. Problem definition and methodology
tremities and considers the postural effects of the neck, trunk, upper
limbs, and legs on the risk score. RULA can evaluate one moment of The current study addresses the problem of assembly line balancing
work content at a time, so determining the extreme high-risk posture with ergonomic risks consideration and defines it according to a specific
requirement of the task is one challenging step in the application. REBA set of requirements observed in a real-life case, particularly, a specific
(Rapid Entire Body Assessment) evaluates musculoskeletal disorder segment of a refrigerator assembly line. The specific segment of the
risks of the entire body by considering the combination of the trunk, assembly line evaluated in this study has already been established. The
neck, legs, upper arms, lower arms, and wrist postures. REBA also previous records of WMDSs of the company show that the assembly line
considers the weight and coupling of components for material handling requires improvements in working conditions from an ergonomic point
work systems. The challenge, same as RULA, for deciding on the correct of view. A simple observation of the station loads indicates that the ef­
snapshot of the task exists for REBA as well. The EAWS (Ergonomic ficiency can be improved through with reduction in cycle time. Thus, the
Assessment Work Sheet) assesses postures and forces for manual mate­ study defines the problem of assembly line balancing with ergonomic
rial handling, particularly in the manufacturing environment. It con­ risk consideration as follows: assignment of a given number of tasks of a
siders whole-body risk factors and repetitive loads of the upper limbs. single product to a given number of stations for given task times while
The EAWS calculates the aggregate risk points for the whole body and incorporating ergonomic risk levels of tasks in order to minimize cycle
upper limbs (Otto and Battaïa, 2017). time and imbalance of accumulated ergonomic risks between stations.
This problem statement places the addressed problem into the SALBP-2
2.3. Ergonomic simulation modeling class with ergonomic risks.
The sub-sections included in this section illustrate the problem that is
The main disadvantages of the above-mentioned risk assessment being addressed, based on an example, to explain ergonomic risk
methods are low sensitivity to the variety of motions, anthropometric assessment methods and to propose a fuzzy multi-objective formulation.
differences between workers, and lack of consideration of interactions This will lead to finding a compromised solution between minimum
between workplace design and worker performance. DHM (Digital cycle time and minimum accumulated ergonomic risk imbalance among
Human Modelling) and ergonomic simulation are emerging areas to workers (stations).
evaluate human-machine systems and it is expected to provide impor­
tant contributions to ergonomic risk assessment (Krüger and Nguyen, 3.1. Illustration of SALBP-2 with ergonomic risks
2015). The DHM is used in industry, army, and academia to evaluate
systems in the virtual environment for many purposes such as product An assembly line with four tasks is assumed, as given in Fig. 1, to
design, workplace safety, performance analysis, etc. Mannequins in a illustrate the problem of assembly line balancing with ergonomic risk
virtual environment have been improved significantly since they have consideration. Ergonomic risks are denoted with initial letters of each
been commercialized and have started to be utilized in the industry. type as being L for lifting risk, Tw for twisting-wrist risk, Th for twisting-
Nowadays, there are digital mannequins able to mimic most of the hip risk, and S for squatting risk.
critical joints of the human body with high fidelity. Mannequins are Fig. 1 also represents the relationship between assembly tasks and
created in a virtual environment by using ergonomic simulation soft­ ergonomic risk factors. A typical assembly line consists of work stations
ware libraries or by motion capturing technology. The most crucial with sets of tools, equipment, components, and worker(s) to perform a
advantage of working with DHM is to have the animation capability of a set of tasks, each consisting of a set of work elements such as lifting,
work system. This ability helps a designer understand how realistically twisting-wrist, twisting-hip, and squatting. These work elements may
the ergonomic simulation model mimics the real system. After vali­ cause ergonomic risks with respect to some work-related factors such as
dating the simulation model, all measurements (e.g., risk assessment, weight of component, duration, and frequency of the task, the posture of
force analysis, injury analysis, etc.) related to the real work system can the worker, angle of motion. Therefore, the ergonomic risks to be
be conducted with high accuracy. Numerous researchers implemented considered are as follows: (1) Lifting risk, (2) Twisting-wrist risk, (3)
DHM by considering both ergonomic risks and economic factors. Longo Twisting-hip risk, (4) Squatting risk. Typically, an assembly line is
and Mirabelli (2009) applied DHM by including work measurements, defined with task times and precedence relationships between tasks, and
line balancing, and ergonomic factors and developed an assembly line SALBP-2 is an assignment of tasks to a given number of stations subject
for heater production. Enomoto et al. (2013) studied the evaluation of to precedence relationships for minimizing cycle time. The problem
ergonomic risks with simulation. Similarly, Ding and Hon (2013) addressed in this study is defined with ergonomic risk levels associated
applied a DHM to simulate assembly lines and selected the one with the with each task, in addition to task times and precedence relationships, as
least ergonomic risks. Shewchuk et al. (2016) provided an approach to shown in the illustrative example (Fig. 1). Assessment of ergonomic risk
simulate workers while they were performing physical tasks, by levels is one of the most critical stages in the proposed approach.

3
R. Ozdemir et al. International Journal of Production Economics 239 (2021) 108188

Fig. 1. A precedence diagram of four tasks and their relationships with ergonomic risks.

However, it is assumed that the ergonomic risk levels of the tasks are imbalance results in higher ergonomic risks in one station. On the other
given in this illustrative example. Every task in the illustrative example hand, solution (b), given in Fig. 2, offers a lower imbalance of ergonomic
consists of all four types of ergonomic risk with different levels. The risks between stations, however, an increase in cycle time is inevitable
assignment of tasks determines the work content of the stations, not only according to the given task times of the illustrative example. In real life,
in time units (e.g., minutes) but also in ergonomic risk levels (e.g., high, this conflict between two objectives (i.e., cycle time minimization and
medium, or low). Task time is an important factor affecting the ergo­ ergonomic risk imbalance minimization) can be observed more
nomic risk levels; however, it is not the only factor used to determine frequently due to a larger number of tasks and a larger variety between
risk levels. Therefore, the risk levels and the task times do not have to be these tasks. Therefore, the proposed solution approach should handle
directly correlated. Hence, one solution to the assignment of the tasks to this matter to find a compromise between objectives.
the stations may provide low cycle time with high ergonomic risks in Multi-objective problem formulation is a well-known approach in
some stations; however, another solution may provide lower ergonomic literature to handle conflict between objectives if there is more than one
risks in those stations with higher cycle time. This conflict between objective (Amid et al., 2011). In this case, the achievement degree of
objectives of minimum cycle time and minimum imbalances of ergo­ these objectives is also challenging to determine precisely in a
nomic risks is illustrated in Fig. 2. compromised solution, because their significance of contribution from a
As seen in Fig. 2, solution (a) provides a cycle time of 5 min managerial perspective is unclear and somewhat vague. Fuzzy sets are a
(maximum station-load in minutes) and high ergonomic risk in station 1 well-known theory to model this sort of uncertainty. This section also
while relatively much lower ergonomic risks in station 2. The difference presents the proposed methodology (i.e., fuzzy multi-objective model)
in ergonomic risk levels between stations is called imbalance, and higher to solve the problem of assembly line balancing with ergonomic risk

Fig. 2. Solution alternatives to the illustrative example with different objectives.

4
R. Ozdemir et al. International Journal of Production Economics 239 (2021) 108188

assessment with two conflicting objectives of minimum cycle time and together. Thus, mannequin design and component design are changed
minimum imbalance of ergonomic risks between stations. The devel­ by feeding body posture angle, component weight, and coupling data
oped model will be explained in another subsection as well. into the Jack simulation analysis tool. The Jack simulation results in
weaker strength of joint for increased body posture angle and increased
3.2. Ergonomic risk assessment methods weight of the component is required to perform the task. Thus, the
minimum required joint strength is adopted as a reference value in a way
This study used the Jack ergonomic simulation software to assess the that risk level is zero at the natural posture of the hip and a nominal
ergonomic risks in an assembly line. As mentioned earlier in this study, weight of a component in task performance. The risk level of this
each assembly task is comprised of four ergonomic risk elements, as element is then calculated as:
follows: lifting risk, twisting-wrist risk, twisting-hip risk, and squatting rhk ​ = ​ (MRSOJ − H) ​ / ​ (SOJ − H)k
risk. One of the powerful features of Jack is the flexibility provided for
the user to select different risk analysis tools for different risk factors. Where;
Each risk element may cause different musculoskeletal disorders; thus,
different analysis tools are used for the assessment of these ergonomic rhk = Twisting-wrist risk level of task k, (SOJ-H)k = Strength of joint
risk elements in this study. for hip while performing task k (resulted from Jack)
Lifting risk assessment is conducted by using Lower Back Analysis (MRSOJ-W) = Minimum required strength of joint for hip (calcu­
(LBA), which requires virtual workplace design, mannequin design, and lated required strength for performing tasks of the assembly line with
component design. There are slight differences between workstations natural posture of hip and a nominal weight of component)
and workers of the assembly line evaluated in this study. Therefore, the
same workplace design and mannequin design are used while changing Squatting risk assessment is conducted by using fatigue and recovery
component design by inputting different weights and coupling to time analysis for the selected shift length which determines simulation
determine the lifting risk level of each task. The Jack simulates the horizon, cycle time to find the repetition of the risk element, and work-
lifting activity by incorporating a workplace, mannequin, and compo­ element duration. The static body posture of squatting is assumed, using
nent design with other factors such as task time and allowed recovery the same workstation design for the other risk assessments and the same
between activities. The simulation results in the average force applied equipment design, therefore the workplace, mannequin, and component
on the lower back of the worker when they perform the lifting element of designs are kept constant while determining risk levels of different tasks.
the assembly task. The risk level of this element is then calculated as The only parameter changed is the duration of squatting while per­
follows: forming the task. Jack simulates the fatigue and calculates the needed
recovery time before the next performance. Available recovery time is
rlk ​ = ​ (LBCF)k ​ / ​ RFL
calculated as the remaining time after subtracting squatting duration
Where; from the cycle time. The longer duration of squatting requires a longer
recovery time. However, the longer duration of squatting results in less
rlk = Lifting risk level of task k, available recovery time before the next squatting work element, which
(LBCF)k = Lower back compression force while performing task k increases the risk exponentially for increased durations of the element.
(resulted from Jack) The risk level of this element is then calculated as:
RFL = Recommended force limit (occupational safety and health rsk ​ = ​ (RTN)k ​ / ​ (ART)k
standard)
Where;
Twisting-wrist risk assessment depends on Joint Strength Requirement
Analysis (JSRA) for the selected hand posture with a main parameter of rsk = Squatting risk level of task k, (RTN)k = Recovery time needed
angle. Mannequin design is changed by inputting wrist angle for per­ before performing task k (resulted from Jack)
forming each task. Since the difference is negligible between equipment (ART)k = Available recovery time after performing task k (calculated
used by hand, the same parameters for workstation design and equip­ as the remaining time after subtracting squatting duration from cycle
ment design were set to evaluate risk levels of different tasks. The Jack time)
simulates the strength of the wrist at the given joint angle, incorporating
task repetition and weight of equipment. It results in weaker average The next section presents the fuzzy multi-objective model to deter­
strength joint value for the increased joint angle required to perform the mine the task assignment into the stations by incorporating risk levels
task. Thus, it requires a different reference value to convert the simu­ for minimizing cycle time and minimizing imbalance of accumulated
lation results into risk levels. The minimum required joint strength is ergonomic risk levels between workers subject to precedence relation­
adopted as a reference value so that the risk level is zero at the natural ship requirements.
posture of hands in task performance. The risk level of this element is
then calculated as:
3.3. Proposed fuzzy multi-objective model
rtk ​ = ​ (MRSOJ − W) ​ / ​ (SOJ − W)k
As mentioned earlier, the problem addressed in this study is a multi-
Where; objective concept due to the conflict between the efficiency of the line
and the ergonomic risk imbalance in the line. Therefore, the proposed
rtk = Twisting-wrist risk level of task k, (SOJ-W)k = Strength of joint approach for the problem is a multi-objective model with the following
for the wrist while performing task k (resulted from Jack) two objectives: (1) to maximize the membership degree of cycle time
(MRSOJ-W) = Minimum required strength of joint for wrist (calcu­ minimization, (2) to maximize the membership degree of ergonomic risk
lated required strength for performing tasks of the assembly line with imbalance minimization by distributing the workload between workers
natural posture of hands) as equal as possible in terms of the ergonomic workload dose. Fuzzy
membership functions are defined for both objectives by using the linear
The twisting-hip risk assessment is similar to the twisting-wrist membership function with pre-specified intervals of the objective values
assessment since JSRA is utilized. However, this risk element is and the equivalent objective values as a membership value in the in­
affected by body posture angle and component weight and coupling terval [0, 1]. The proposed membership functions are given in Figs. 3

5
R. Ozdemir et al. International Journal of Production Economics 239 (2021) 108188

and 4.
The formulation of the proposed multi-objective is given in the
equations below, where the objective function of the fuzzy multi-
objective model maximizes the membership function values and is
formulated as shown in Equation (1).
∑
n
Max Z = wj μ(x)j (1)
j=1

As seen in Equation (1), each membership function value is weighted

based on the decision maker’s preference or based on the importance of
the objectives (weights are assigned to make the sum equal to 1). Fig. 4. Membership function of the objective for minimum ergonomic
Membership functions for both objectives are then formulated as: risk imbalance.

Cmax − z1
μ(x)1 = (2) The rest of the model is formulated using the notation:
Cmax − Cmin
{
1, if task i is assigned to station (worker) j
RVmax − z2 Xik =
μ(x)2 = (3) 0, otherwise
RVmax
ti = task time of task i.
Where.
Pi = Predecessors of task i.
Cmax = Upper bound of the cycle time (total work content of all rli = Lifting risk level of task i with respect to Jack simulation results.
tasks). rti = Twisting-wrist risk level of task i with respect to Jack simulation
Cmin = Lower bound of the cycle time (theoretical minimum cycle results.
time). rhi = Twisting-hip risk level of task i with respect to Jack simulation
RVmax = Upper bound of the ergonomic risk imbalance value (z2 of results.
the solution with w1 = 1, w2 = 0). rsi = Squatting risk level of task i with respect to Jack simulation
results.
The values of both objectives are defined as follows: Lk = Accumulated lifting risk levels assigned to worker k.
z1 = C (4) Tk = Accumulated twisting-wrist risk levels assigned to worker k.
Hk = Accumulated twisting-hip risk levels assigned to worker k.
∑
m ∑
m
( ) Sk = Accumulated squatting risk levels assigned to worker k.
z2 = L+ − + − + − + −
kh + Lkh + Tkh + Tkh + Hkh + Hkh + Skh + Skh (5)
Hard constraints of the assembly line balancing with ergonomic risks
k=1 h=k+1

where. consideration are formulated as:

∑
m

C = Realized cycle time with respect to line balance. Xik = 1, ∀i (6)

L+kh = Positive lifting risk difference between worker k and worker h.
k=1

L-kh = Negative lifting risk difference between worker k and worker h. ∑

n
T+kh = Positive twisting wrist risk difference between worker k and ti .Xik ≤ C, ∀k (7)
worker h. i=1

T-kh = Negative twisting wrist risk difference between worker k and

∑
m ∑
m
worker h. k.Xik ≤ h.Xlh , ∀iεPl (8)
H+ kh = Positive twisting hip risk difference between worker k and k=1 h=1

worker h.
H-kh = Negative twisting hip risk difference between worker k and ∑
n
rli. Xik = Lk , ∀k (9)
worker h. i=1
S+kh = Positive squatting risk difference between worker k and worker
h. ∑
n
rti. Xik = Tk , ∀k (10)
S-kh = Negative squatting risk difference between worker k and i=1
worker h.
∑
n
rhi. Xik = Hk , ∀k (11)
i=1

∑
n
rsi. Xik = Sk , ∀k (12)
i=1

Lk − Lh = L+ −
kh − Lkh , ∀k, h (13)

+
Tk − Th = Tkh −
− Tkh , ∀k, h (14)

+
Hk − Hh = Hkh −
− Hkh , ∀k, h (15)

+
Sk − Sh = Skh −
− Skh , ∀k, h (16)
Fig. 3. Membership function of the objective for the minimum cycle
time value.

6
R. Ozdemir et al. International Journal of Production Economics 239 (2021) 108188

Xjk = {0, 1}, ∀j, k (17) labeled from T01 to T29, as being the first task of placing the compressor
unit in the line is T01 and the last task of installing accessories is T29.
Lk , Tk , Hk , Sk , L+ − + − + − + −
kh , Lkh , Tkh , Tkh , Hkh , Hkh , Skh , Skh ≥ 0, ∀k, h (18) Four ergonomic risk elements can also be seen next to each task and each
task involves different levels of risks. For instance, T01 (placing
Constraint sets given in Equation (6) ensures that each task is
compressor unit) has high-risk levels for the lifting work-element and
assigned to only one worker. Constraint sets given in Equation (7) en­
twisting-hip work-element, however, it has a low-risk level for the
sures that the sum of task times cannot exceed the realized cycle time for
squatting work-element and it does not involve twisting-wrist work
each worker. Constraint sets given in Equation (8) ensures that pre­
element.
decessors of each task should be assigned to an earlier station in the line.
As mentioned in section 3.2., Siemens Tecnomatix Jack 8.4 is used to
Constraint sets given in Equations (9)–(12) calculate total ergonomic
model human factors virtually to determine the level of risks related to
risk values of each posture for each worker. Constraint sets given in
each assembly task. The human model design, workplace design, and
Equations (13)–(16) calculate the workload difference between the
component design are required before starting the simulation for risk
workers in terms of the ergonomic risk value of each posture. Constraint
assessment. This study utilizes three different analysis tools of the Jack
sets given in Equations (17) and (18) define the decision variables of the
(i.e., 1- lower back analysis, 2- joint strength requirement analysis, and
model.
3-fatigue and recovery time analysis) to evaluate ergonomic risk levels
of four work-elements. Fig. 5 presents the structure of the developed
4. Case problem and numerical results
simulation model for ergonomic risk analysis of the case problem.
As seen in Fig. 5, the Jack simulation modeling starts with a work­
The case problem focuses on a specific segment in a manually
place, mannequin, and component design, as the structural elements for
operated refrigerator assembly line. The segment accomplishing
evaluating different work-elements with different risk factors. As
compressor, electrical supply units, and some final assembly tasks,
explained in section 3.2, lifting risk assessment of the tasks is conducted
before the gas station, consists of six workstations, each operated by one
with lower back analysis by varying component weights while keeping
worker, and 29 manual tasks. The complete task list, along with their
other design factors constant. Twisting-wrist and twisting-hip risk as­
descriptions, is presented in Table 1. The total work content of the entire
sessments are accomplished with joint strength analysis by changing the
line is 21.3 min, which makes the theoretical minimum cycle time 3.55
joint angle of the related body part. Squatting risk assessment is con­
min, by distributing total work content equally to six stations. The task
ducted with fatigue and recovery time analysis by manipulating the
times are gathered from the line supervisors and assumed that they do
duration of the work element while keeping the posture of the manne­
not change from one worker to another since all six workers have an
quin constant. For more detail on risk assessment by Jack simulation
identical experience. Precedence relationships between tasks are also
software, refer to section 3.2. As shown in Table 1, there are 29 tasks,
gathered from the line supervisors, and they are derived based on the
each of which consists of four work elements (i.e., lifting, twisting-wrist,
technological difficulties for changing the order of the tasks. Tasks are
twisting-hip, and squatting). The complete combinational matchings

Table 1
Risk levels (H: High, M: Medium, L: Low), task times, and predecessors of tasks.
Task index, Task Description Time, ti Lifting risk, Twisting-wrist risk, Twisting-hip risk, Squatting risk, Immediate Predecessor,
i (min) rli rti rhi rsi Pi

T01 Placing compressor unit 0.3 H – H L –

T02 Assembling the compressor unit 0.3 – H – M T01
T03 Preparing fan and evaporator 0.3 H – H – –
T04 Placing fan and evaporator 0.3 – H – – T03
T05 Wiring upper and lower units 1.5 L H L – T02, T04
T06 Welding input/output 0.5 L H – – T05
connections
T07 Installing interior lights 0.25 L H M – –
T08 Wiring interior lights 1 L H L H T07
T09 Welding input/output 0.5 L H M M T08
connections
T10 Installing door switch 0.2 L H M – T09
T11 Wiring door switch 1 L H L H T10
T12 Welding input/output 0.5 L H M M T11
connections
T13 Installing thermostat control 0.2 L H M – –
T14 Wiring thermostat control 1 L M L – T13
T15 Welding input/output 0.5 L H M – T14
connections
T16 Installing electrical supply unit 0.3 L H M M T11, T14
T17 Wiring electrical supply unit 1 L H M H T16
T18 Welding input/output 0.5 L H M M T17
connections
T19 Fixating wires with the base unit 1.1 L H M – T06, T17
T20 Covering wires with rubber 1.5 L H M – T19
T21 Installing wire compartment 1.5 L H H M T20
T22 Covering upper unit with latten 1.1 L H M – T21
T23 Covering bottom unit with latten 1.1 L H M H T22
T24 Installing shelf holders 1.1 L M L – T23
T25 Assembling door 0.5 H – H – –
T26 Drilling holes on the door 0.25 L H L – T25
T27 Installing door stops 0.5 L M L – T26
T28 Installing magnetic gasket on 1.5 L H L – T26
door
T29 Installing accessories 1 L M L – T26

7
R. Ozdemir et al. International Journal of Production Economics 239 (2021) 108188

Fig. 5. The Jack simulation modeling structure constructed for the risk assessment.

between tasks and work elements constitute 116 simulation models to system. Lifting risk is categorized as low when weight is less than or
run (29 tasks × 4 work-elements). However, after several trials for equal to 9 kg, medium when weight is between 9 and 20 kg, and high
several selected tasks, it is observed that each work element can be when weight is more than 20 kg. Twisting-wrist risk is categorized as
assigned with one of the three risk categories (i.e., high, medium, or low when the angle is less than 30◦ , medium when the angle is between
low). Since each task is comprised of these four work elements; tasks are 30◦ and 60◦ , and high when the angle is more than 60◦ . Twisting-hip risk
assigned with the corresponding risk categories of the work elements. is categorized based on the joint angle as well. The categories are
Therefore, instead of developing 116 simulation models, we determined as low when the angle is less than or equal to 45◦ , medium
construct one simulation model for each risk category and each work- when the angle is between 45◦ and 90◦ , and high when the angle is more
element; thus, we reduce the simulation design load from 116 models than 90◦ . Squatting risk is categorized based on the duration of the work
to 12 (4 work-elements × 3 risk categories) models by classifying work- element, which ranges between 10 s and 45 s. Risk categories are
elements into three risk categories. Each risk category has boundaries in determined as low when the duration is less than or equal to 15 s, me­
terms of variable risk factors associated with work elements. This study dium when the duration is between 15 s and 30 s, and high when the
utilizes the highest risk values of the risk categories obtained from the duration is more than 30 s.
Jack simulation. The categories are then used to determine the risk levels of each work
Next, we run these 12 models by manipulating variable risk factors element of each task based on the corresponding variable risk factors in
(i.e., component weight, joint angle, and work-element duration) each task. We realized that some risk categories of some work-element
around the upper bounds of risk categories to determine the values of do not exist in any task. For instance, there is no task in the medium
the selected measures (i.e., LBCF, lower back compression force; SOJ-W, lifting risk category and most of the tasks involve low lifting risk since
the strength of joint for the wrist, SOJ-H, the strength of joint for hip and most of them are related to a component weighing less than 9 kg, as can
RTN, recovery time needed). The Jack determines LBCF in N for lifting be observed from Table 1. In addition to that, there is no task in the low
risk, SOJ-W in N for twisting-wrist risk, SOJ-H in N for twisting-hip risk, twisting-wrist category and most of the tasks have work elements
and RTN in seconds for squatting risk. The following figure shows the requiring joint angle for the wrist larger than 60◦ due to the poor design
values for the risk categories resulted from the Jack simulation for all of the product (base part). This situation can also be observed in Table 1,
work elements in the assembly line. which shows a high-risk level for twisting-wrist in most of the tasks.
The component weights vary between 1.5 and 25 kg; the wrist angles Reference values for risk measures (i.e., RFL, reference force limit;
range between 0 and 90; the hip angles range between 30 and 180; MRSOJ-W, the minimum required strength of joint for the wrist;
squatting durations change between 10 s and 45 s for performing all MRSOJ-H, the minimum required strength of joint for hip and ART,
tasks in the line. Fig. 6 represents the risk categorization of the Jack with available recovery time) are also needed for accomplishing risk assess­
respect to the given design parameters of the current assembly line ment of the tasks. The following table represents the value of each

8
R. Ozdemir et al. International Journal of Production Economics 239 (2021) 108188

Fig. 6. Ergonomic risk categories determined by Jack simulation.

measure for the associated risk category and the reference value corre­ rli = LBCFhigh/RFL = 7948/6400 = 1.24, for all i with high lifting risk.
sponding to each risk measure. rli = LBCFlow/RFL = 3329/6400 = 0.52, for all i with low risk lifting
As seen in Table 2, there are 10 values to be used for risk assessment risk.
of the tasks, since no task involves medium lifting risk and low twisting- Twsiting-wrist risk levels for the tasks are determined as follows:
wrist risk. C denotes the cycle time and is used for calculating the rti = MRSOJ-W/SOJ_Whigh = 0.09/0.075 = 1.2, for all i with high
reference value of the ART (available recovery time). C is set to 7 min twisting-wrist risk.
Because one refrigerator to be made every 7 min is adequate for satis­ rti = MRSOJ-W/SOJ_Wmedium = 0.09/0.108 = 0.83, for all i with med.
fying the current demand rate. Resulted values are the average values of twisting-wrist risk.
the simulated work condition for the specified risk indicator and the Twsiting-hip risk levels for the tasks are determined as follows:
predetermined parameter values. Risk assessments of the tasks are then rhi = MRSOJ-H/SOJ_Hhigh = 15.9/15.31 = 1.04, for all i with high
conducted by using the values provided in Table 2 and the equations twisting-hip risk.
presented in section 3.2. as given below. rhi = MRSOJ-H/SOJ_Hmedium = 15.9/18.52 = 0.86, for all i with med.
Lifting risk levels for the tasks are determined as follows: twisting-hip risk.
rhi = MRSOJ-H/SOJ_Hlow = 15.9/21.24 = 0.75, for all i with low
Table 2 twisting-hip risk.
The values to be used for risk assessment of the tasks. Squatting risk levels for the tasks are determined as follows:
rsi = RTNhigh/ARThigh = 463/375 = 1.23, for all i with high squatting
Risk Risk Categories Reference Value for Risk Assessment
Measure risk.
Low Medium High
rsi = RTNmedium/ARTmedium = 335/390 = 0.86, for all i with med.
LBCF 3329 n/a 7948 RFL = 6400 N (NIOSH standard) squatting risk.
N N rsi = RTNlow/ARTlow = 74/405 = 0.19, for all i with low squatting
SOJ-W n/a 0.108 N 0.075 MRSOJ-W = 0.09 N (simulation
risk.
N result)
SOJ-H 21.24 18.52 N 15.31 MRSOJ-H = 15.9 N (simulation The above-given risk levels can also be interpreted as the applied
N N result) dose of ergonomic load on the worker while performing the tasks. For
RTN 75 s. 335 s. 463 s. ARTlow = 405 s. (C – 15 s) instance, the risk level of lifting depends on the force applied on the
ARTmedium = 390 s. (C – 30 s)
lower back, and the highest risk level is 1.24 doses (24% excess of
ARThigh = 375 s. (C – 45 s)

9
R. Ozdemir et al. International Journal of Production Economics 239 (2021) 108188

normal dose or 24% higher force applied on the lower back over the the function is positive semidefinite or positive definite at all points in
maximum allowed force of 6400 N). The above-calculated risk levels are the set (Marler and Arora, 2010). The weighted sum multi-objective
numerical values of the risk categories displayed in Table 1, for instance, model in this study provides that the Hessian matrix for the con­
all cells with H denote high risk of the corresponding work element, straints and objective functions are positive semidefinite. Fig. 7 shows a
thus, each H in the fourth column (lifting risk) is input as 1.24 to the subset of optimal solutions from the Pareto front in terms of membership
multi-objective model and so on. degree and actual values of both objectives.
The total work content of all tasks is the sum of all task times in the The solutions provided in Table 1 and Fig. 7 show that there are
line, which is 21.3 min. Theoretical minimum cycle time is obtained as many opportunities for improvement of the current system. As can be
total work content divided by the number of stations (21.3 min/6 = observed from the solutions, management can use any of the given so­
3.55 min). Theoretical minimum cycle time and total work content are lutions to improve their system, however, defining the economic sig­
set as Cmin and Cmax, lower bound and upper bound of cycle time nificance of the difference between solutions require a broader study on
minimization membership function, respectively. RVmax, the upper cost analysis for the improvements of ergonomic risk imbalance and
bound of ergonomic risk imbalance minimization membership function benefit analysis for the improvement of cycle time. For instance, cycle
is obtained as 74.4 doses by solving the model given in section 3.3. with time is improved by 0.2 min while ergonomic risk imbalance is wors­
w1 = 1 and w2 = 0. ened by 2.96 doses when we change from solution no. 9 to 10. However,
After having risk values, lower and upper bounds of the objectives, the economic significance of the difference between these solutions can
task times, and precedence relationships, all the data are input to the not be defined without knowing the cost of ergonomic risk imbalance
proposed multi-objective model, and a set of the compromised solution and the benefit of cycle time improvement.
alternatives for the problem is obtained by manipulating the weights of Moreover, the results are based on the assumption that there is no
the objectives. Table 3 shows the complete set of all alternative solutions difference between workers in terms of age, gender, body mass, height,
found by using Lingo 14.0. or level of aerobic exercises. However, these characteristics of the
The largest objective function value is obtained when the objective of workers might affect the risk analysis results (Finco et al., 2019). Thus,
minimum cycle time is dominant with w1 = 0.99. This solution results in considering workers’ characteristics might improve the scope of the
the best-realized cycle time of 3.7 min, however, it results in the worst model.
outcome for ergonomic risk imbalance value of 55.38 ergonomic doses
for all workers. The smallest objective function value is found when the 5. Conclusion
objective of minimum ergonomic risk imbalance is dominant with w2 =
0.99. This solution results in the best outcome of ergonomic risks This study addresses the problem of assembly line balancing with
imbalance value of 22.5 ergonomic doses, while shows the worst real­ ergonomic risk consideration. Most of the assembly tasks involve ergo­
ized cycle time of 5.2 min. The current assembly system has very poor nomic risk factors such as repetitive motions, awkward postures,
performance for both objectives as having realized cycle time of 6.63 heavyweights, which result in work-related musculoskeletal disorders
min and ergonomic risks imbalance of 85.1 doses in total for all workers. (WMSDs). A poor ergonomic design of an assembly line causes WMSDs
As seen in Table 3, all solution alternatives dominate the performance of that might result in economic losses for companies due to job absen­
the current system for both cycle time and ergonomic risk imbalance. teeism, medical treatments, worker rehabilitation, errors in manual
The set of alternative solutions for both objectives by manipulating operations, and a decrease in quality. Assembly line balancing with er­
weights is shown in the following figure. As seen in the following figure, gonomic risk consideration is for the improvements of the work condi­
the minimum cycle time solution will naturally result in 99% achieve­ tions with a significant decrease in costs associated with WMSDs.
ment for the first objective since it is very close to the lower bound of The problem addressed in the paper is based on a real case problem
cycle time (3.55 min), 23% achievement for the second objective. that has a certain number of stations to accomplish all tasks. Therefore,
However, minimum ergonomic risk imbalance solution will result in the problem is defined as SALBP-2 with ergonomic risks. The proposed
69% achievement for the second objective and 91% achievement for the study, thus also provides guidance for practitioners who might be
first objective. willing to design a better working environment and enhance economic
The non-dominated solutions found with the weighted sum method performance using ergonomic simulation tools and adopting optimiza­
are in the Pareto optimal set, provided that, the feasible solution space tion approaches.
(all constraints) and all of the objective functions are convex in the The current study proposes a new ergonomic risk assessment tool by
multi-objective problem, (Athan and Papalambros, 1996; Geoffrion breaking the tasks into the common work elements as being lifting,
1968; Koski 1985). A function is convex, as long as the Hessian matrix of twisting-wrist, twisting-hip, and squatting. The study also introduces a

Table 3
The results of alternative solutions with different weights of objectives.
Solution no. w1 w2 Objective Function z1 (min.) z2 (dose) μ1 μ2 Elapsed runtime (seconds)

1 0.01 0.99 0.69 5.2 22.50 0.91 0.69 1738.54

2 0.10 0.90 0.71 4.7 22.64 0.94 0.68 1585.63
3 0.20 0.80 0.73 4.7 22.64 0.94 0.68 1340.09
4 0.30 0.70 0.76 4.7 22.64 0.94 0.68 1298.64
5 0.40 0.60 0.78 4.7 22.64 0.94 0.68 1535.25
6 0.50 0.50 0.81 4.4 23.52 0.95 0.67 1367.16
7 0.60 0.40 0.84 4.4 23.52 0.95 0.67 762.31
8 0.70 0.30 0.87 4.4 23.52 0.95 0.67 476.01
9 0.75 0.25 0.88 4.4 23.52 0.95 0.67 608.74
10 0.80 0.20 0.90 4.2 26.48 0.96 0.63 470.43
11 0.85 0.15 0.91 4.2 26.48 0.96 0.63 506.05
12 0.90 0.10 0.93 4.0 31.28 0.97 0.56 396.82
13 0.92 0.08 0.94 3.9 35.54 0.98 0.50 280.76
14 0.95 0.05 0.96 3.8 41.24 0.99 0.42 299.40
15 0.97 0.03 0.97 3.75 45.14 0.99 0.37 258.61
16 0.99 0.01 0.98 3.7 55.38 0.99 0.23 48.78

10
R. Ozdemir et al. International Journal of Production Economics 239 (2021) 108188

Fig. 7. A subset of optimal solutions from the Pareto front in terms of the membership degrees and actual values.

framework for the ergonomic simulation with digital human modeling reason for the high achievement of objective 1 in all solutions is due to
for determining risk values of the work elements. The structural the very large upper bound on the membership function of cycle time
framework of the developed simulation model with the Jack software, minimization.
explains the relationships between the work elements and risk factors of The current study can be improved by considering the combined
the weight of the component, angle of joint, and duration. The risk effect of accumulated risks on the same worker. This improvement re­
factors are used as parameters of the ergonomic simulation model for quires a more complicated ergonomic simulation model to be able to
accomplishing lower back force analysis, the strength of joint analysis, combine more than one work element in the same risk analysis tool. The
and fatigue and recovery time analysis. The work-elements are associ­ simulation runs show that risk categories can be improved by using
ated with specific ergonomic risk measures, the variables of the simu­ fuzzy clustering with the Jack. Thus, we recommend that the interested
lation model. This study also presents the association between the researcher should study more complicated Jack simulation models that
specified risk measures and the work-elements as being lifting risk is can simulate a series of work elements together, and they should
associated with lower back compression force, twisting-wrist and consider worker-related factors for risk assessment such as age, gender,
twisting-hip risks are associated with joint strength, and squatting risk is and experience level.
associated with recovery time needed. The approach for converting the
ergonomic simulation results into the numerical risk levels of the work References
elements is another outcome of this study. New equations for risk as­
sessments of the work-elements derived by using the upper bound values Amen, M., 2000. An exact method for cost-oriented assembly line balancing. Int. J. Prod.
Econ. 64 (1–3), 187–195.
on the risk measures obtained from simulation runs and newly intro­ Amen, M., 2001. Heuristic methods for cost-oriented assembly line balancing: a
duced reference values for these measures were applied with an inno­ comparison on solution quality and computing time. Int. J. Prod. Econ. 69 (3),
vative approach. 255–264.
Amid, A., Ghodsypour, S.H., O’Brien, C., 2011. A weighted max–min model for fuzzy
The developed ergonomic risk assessment tool provides inputs to a multi-objective supplier selection in a supply chain. Int. J. Prod. Econ. 131 (1),
fuzzy multi-objective model, developed in this study for assembly line 139–145.
balancing with two fuzzy objectives. The objectives specified for the Athan, T.W., Papalambros, P.Y., 1996. A note on weighted criteria methods for
compromise solutions in multi-objective optimization. Eng. Optim. 27 (2), 155–176.
improvement of the current assembly line are minimizing cycle time and Barathwaj, N., Raja, P., Gokulraj, S., 2015. Optimization of assembly line balancing using
minimizing imbalance of ergonomic risk levels between workers. These genetic algorithm. J. Cent. S. Univ. 22 (10), 3957–3969.
objectives are defined with fuzzy membership functions since the sig­ Battaïa, O., Dolgui, A., 2013. A taxonomy of line balancing problems and their solution
approaches. Int. J. Prod. Econ. 142 (2), 259–277.
nificance of their actual values in terms of contribution to the
Battini, D., Faccio, M., Persona, A., Sgarbossa, F., 2011. New methodological framework
improvement on the current assembly line is imprecise. Thus, we also to improve productivity and ergonomics in assembly system design. Int. J. Ind.
introduce a lower bound and an upper bound for calculating the mem­ Ergon. 41, 30–42.
bership value of cycle time minimization and introduce an upper bound Battini, D., Delorme, X., Dolgui, A., Persona, A., Sgarbossa, F., 2016. Ergonomics in
assembly line balancing based on energy expenditure: a multi-objective model. Int.
for the membership value of ergonomic risk imbalance minimization. J. Prod. Res. 54 (3), 824–845.
The weighted membership function values are maximized in the Bautista, J., Batalla-García, C., Alfaro-Pozo, R., 2016. Models for assembly line balancing
objective function of the proposed fuzzy multi-objective model, subject by temporal, spatial and ergonomic risk attributes. Eur. J. Oper. Res. 251 (3),
814–829.
to precedence relationship constraints and cycle time constraints. Becker, C., Scholl, A., 2006. A survey on problems and methods in generalized assembly
Weights are manipulated to obtain alternative solutions to the line balancing. Eur. J. Oper. Res. 168 (3), 694–715.
problem and they constitute a subset of optimal solutions from the Bortolini, M., Faccio, M., Gamberi, M., Pilati, F., 2017. Multi-objective assembly line
balancing considering component picking and ergonomic risk. Comput. Ind. Eng.
Pareto front. Every solution in the set dominates the current system 112, 348–367.
performance in terms of cycle time and ergonomic risk imbalance. Carnahan, B.J., Norman, B.A., Redfern, M.S., 2001. Incorporating physical demand
Minimum cycle time, 4.7 min, is obtained when weights are set as 0.99 criteria into assembly line balancing. IIE Trans. 33 (10), 875–887.
Cheshmehgaz, H.R., Haron, H., Kazemipour, F., Desa, M.I., 2012. Accumulated risk of
for cycle time minimization and 0.01 for ergonomic imbalance mini­ body postures in assembly line balancing problem and modeling through a multi-
mization. The solution for these weights shows the largest objective criteria fuzzy-genetic algorithm. Comput. Ind. Eng. 63 (2), 503–512.
function value of 0.98. Minimum ergonomic risk imbalance, 22.5 doses, Choi, G., 2009. A goal programming mixed-model line balancing for processing time and
physical workload. Comput. Ind. Eng. 57 (1), 395–400.
is obtained when weights are set as 0.0.1 for cycle time minimization
Colombini, D., Occhipinti, E., 2006. Preventing upper limb work-related musculoskeletal
and 0.99 for ergonomic imbalance minimization. The solution for these disorders (UL-WMSDS): new approaches in job (re) design and current trends in
weights shows the smallest objective function value of 0.69. The results standardization. Appl. Ergon. 37 (4), 441–450.
show that even the worst solution in terms of objective function value Costa, A.M., Miralles, C., 2009. Job rotation in assembly lines employing disabled
workers. Int. J. Prod. Econ. 120 (2), 625–632.
achieves 0.91 membership value for the cycle time minimization. The

11
R. Ozdemir et al. International Journal of Production Economics 239 (2021) 108188

Di Benedetto, R., Fanti, M., 2012. An integrated tool to support engineers for WMSDs risk Longo, F., Mirabelli, G., 2009. Effective design of an assembly line using modelling and
assessment during the assembly line balancing. Work 41 (Suppl. 1), 2329–2333. simulation. J. Simulat. 3 (1), 50–60.
Ding, Z., Hon, B., 2013. Constraints analysis and evaluation of manual assembly. CIRP Malakooti, B., 1991. A multiple criteria decision making approach for the assembly line
Annals 62 (1), 1–4. balancing problem. The Internatıonal Journal Of Productıon Research 29 (10),
Eklund, J.A., 1995. Relationships between ergonomics and quality in assembly work. 1979–2001.
Appl. Ergon. 26 (1), 15–20. Marler, R.T., Arora, J.S., 2010. The weighted sum method for multi-objective
Enomoto, A., Yamamoto, N., Suzuki, T., 2013. Automatic estimation of the ergonomics optimization: new insights. Struct. Multidiscip. Optim. 41 (6), 853–862.
parameters of assembly operations. CIRP Annals 62 (1), 13–16. McAtamney, L., Corlett, E.N., 1993. RULA: a survey method for the investigation of
Erel, E., Sarin, S.C., 1998. A survey of the assembly line balancing procedures. Prod. work-related upper limb disorders. Appl. Ergon. 24 (2), 91–99.
Plann. Contr. 9 (5), 414–434. Mutlu, Ö., Özgörmüş, E., 2012. A fuzzy assembly line balancing problem with physical
Finco, S., Zennaro, I., Battini, D., Persona, A., 2019. Workers’ availability definition workload constraints. Int. J. Prod. Res. 50 (18), 5281–5291.
through the energy expenditure evaluation. In: Proceedings-25th ISSAT Neumann, W.P., Dul, J., 2010. Human factors: spanning the gap between OM and HRM.
International Conference on Reliability and Quality in Design, pp. 1–3. Int. J. Oper. Prod. Manag.
Geoffrion, A.M., 1968. Proper efficiency and the theory of vector maximization. J. Math. Otto, A., Battaïa, O., 2017. Reducing physical ergonomic risks at assembly lines by line
Anal. Appl. 22 (3), 618–630. balancing and job rotation: a survey. Comput. Ind. Eng. 111, 467–480.
Ghosh, S., Gagnon, R.J., 1989. A comprehensive literature review and analysis of the Otto, A., Scholl, A., 2011. Incorporating ergonomic risks into assembly line balancing.
design, balancing and scheduling of assembly systems. Int. J. Prod. Res. 27 (4), Eur. J. Oper. Res. 212 (2), 277–286.
637–670. Raschke, U., Cort, C., 2019. Siemens Jack. In: DHM and Posturography. Academic Press,
González, B.A., Adenso-Dıaz, B., Torre, P.G., 2003. Ergonomic performance and quality pp. 35–48.
relationship: an empirical evidence case. Int. J. Ind. Ergon. 31 (1), 33–40. Salveson, M.E., 1955. The assembly line balancing problem. J. Ind. Eng. 6 (3), 18–25.
Grosse, E.H., Calzavara, M., Glock, C.H., Sgarbossa, F., 2017. Incorporating human Schneider, E., Copsey, S., Irastorza, X., 2010. OSH [Occupational Safety and Health] in
factors into decision support models for production and logistics: current state of Figures: Work-Related Musculoskeletal Disorders in the EU-Facts and Figures. Office
research. IFAC-PapersOnLine 50 (1), 6900–6905. for Official Publications of the European Communities.
Hignett, S., McAtamney, L., 2000. Rapid entire body assessment (REBA). Appl. Ergon. 31 Scholl, A., 1999. Balancing and sequencing of assembly lines (No. 10881). In: Darmstadt
(2), 201–205. Technical University, Department of Business Administration, Economics and Law.
Jaturanonda, C., Nanthavanij, S., Das, S.K., 2013. Heuristic procedure for the assembly Institute for Business Studies (BWL).
line balancing problem with postural load smoothness. Int. J. Occup. Saf. Ergon. 19 Sgarbossa, F., Grosse, E.H., Neumann, W.P., Battini, D., Glock, C.H., 2020. Human factors
(4), 531–541. in production and logistics systems of the future. Annu. Rev. Contr. 49, 295–305.
Koski, J., 1985. Defectiveness of weighting method in multicriterion optimization of Shewchuk, J.P., Nussbaum, M.A., Kim, S., Sarkar, S., 2016. Simulation modeling and
structures. Commun. Appl. Numer. Methods 1 (6), 333–337. ergonomic assessment of complex multiworker physical processes. IEEE Transactions
Kriengkorakot, N., Pianthong, N., 2007. The assembly line balancing problem: review on Human-Machine Systems 47 (6), 777–788.
articles. Engineering and Applied Science Research 34 (2), 133–140. Waters, T.R., Putz-Anderson, V., Garg, A., Fine, L.J., 1993. Revised NIOSH equation for
Krüger, J., Nguyen, T.D., 2015. Automated vision-based live ergonomics analysis in the design and evaluation of manual lifting tasks. Ergonomics 36 (7), 749–776.
assembly operations. CIRP Annals 64 (1), 9–12. Xu, Z., Ko, J., Cochran, D.J., Jung, M.C., 2012. Design of assembly lines with the
Kumar, N., Mahto, D., 2013. Assembly line balancing: a review of developments and concurrent consideration of productivity and upper extremity musculoskeletal
trends in approach to industrial application. Global J. Res. Eng. 13 (2). disorders using linear models. Comput. Ind. Eng. 62 (2), 431–441.

12