You are on page 1of 20

International Journal of Production Research

ISSN: 0020-7543 (Print) 1366-588X (Online) Journal homepage: https://www.tandfonline.com/loi/tprs20

Empirical study to explore the impact of


ergonomics on workforce scheduling

Matteo M. Savino, Carlo Riccio & Marialuisa Menanno

To cite this article: Matteo M. Savino, Carlo Riccio & Marialuisa Menanno (2019): Empirical study
to explore the impact of ergonomics on workforce scheduling, International Journal of Production
Research, DOI: 10.1080/00207543.2019.1591645

To link to this article: https://doi.org/10.1080/00207543.2019.1591645

Published online: 26 Mar 2019.

Submit your article to this journal

Article views: 17

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tprs20
International Journal of Production Research, 2019
https://doi.org/10.1080/00207543.2019.1591645

Empirical study to explore the impact of ergonomics on workforce scheduling


a∗
Matteo M. Savino , Carlo Ricciob and Marialuisa Menannoa
a Department of Engineering, University of Sannio, Benevento, Italy; b LMP Costruzioni Elettromeccaniche, Naples, Italy
(Received 12 October 2018; accepted 28 February 2019)

This work is focused on the workforce scheduling problem, investigating how the ergonomic exposure of a worker may
affect the results of workforce allocation and the relative impact on production performances. A research methodology has
been developed within a case study of an assembly flow with the twofold objective of (i) setting up a system able to consider
the ergonomic parameters of workers within the solution of a workforce scheduling problem and (ii) finding the possible
impacts that ergonomic postures can have on workforce allocation parameters. The ergonomic analysis, joint with the devel-
opment of a constraint optimisation problem, resulted in a mathematical depiction of the ergonomic exposures and in a
workforce ergonomic scheduling model. The study assessed how ergonomics may impact on workforce scheduling and the
relative production capacity. Then, an experimental campaign joint with a tuning activity of the model resulted in a work-
force scheduling configuration able to front the apparently contrasting objectives of production capacity optimisation and
ergonomic stress lowering. The empirical results allowed also to quantify the trade-off between production performances,
in terms of production capacity and idle time, versus ergonomic stress of the workers.
Keywords: workforce scheduling; ergonomics; postural analysis; ergonomic exposure; constraint optimisation problem

1. Introduction
In production systems, the allocation of the type and number of workers at the right time and on the right workstation is
commonly known as workforce scheduling (WS). This problem usually occurs when the available workforce is less than the
number of workstations, or some type of workforce’s feature is not suitable for some kind of job, thus generating potential
working shortage (Thompson and Goodale 2006). Typically, a WS problem fronts the assignments of each worker on certain
workstation(s) for a certain time period. Employees’ planning and scheduling problems have been extensively studied in
the last decades, and further research has been carried out on a number of decisional variables (Battaia et al. 2015). Past
works have concerned areas like general services, nurse, health care and call centres, as well as flow shops (Bergh et al.
2013; Neubert and Savino 2009).
Recently, WS was linked to ergonomics for the concurrent improvement of productivity and ergonomics within the
design of assembly systems (Battini et al. 2011), while some last findings attempt to consider ergonomics in terms of energy
expenditure assembly line balancing (Battini et al. 2016).
The present study addresses the problem of ergonomic exposures of the workers – intended as the set of variables relative
to the body postures of the workers during their working activities – within a flow shop assembly system of electromechani-
cal components. With the aim to fill some gap experienced in the extant body of literature, a combined approach is proposed
with the twofold objective of (i) appraising the impact of ergonomics on WS solutions and (ii) investigating how WS can be
optimised also in regard to ergonomic parameters.
The paper has the following remainder. After the literature review, Section 3 fronts the research questions, outlining the
research methodology. Section 4 provides the WS model developed within the study, along with the description of the flow
shop production system. Section 5 analyses the results throughout two different application cases, while Section 6 discusses
the data. Section 7 concludes the work, giving its limitations and future outlooks.

2. Literature review and research questions


An early attempt to front the WS problem in industrial operations is the one of Quan et al. (2007) who used the evolutionary
algorithms to solve a multi-objective jobs’ scheduling problem with the purpose of minimising workforce costs. Then,

*Corresponding author. Email: matteo.savino@unisannio.it

© 2019 Informa UK Limited, trading as Taylor & Francis Group


2 M.M. Savino et al.

Hytonen, Niemi, and Toivonen (2008) used the discrete event simulation to optimise the allocation of the workforce in
assembly lines for customised products. On this line, the approach of Neubert and Savino (2009) studied the distribution
of workforce on workstations with a centralised scheduling able to find the best workforce allocation with the aim to
optimise Throughput, Makespan and Work in Process. As regards to maintenance operations, Safaei, Banjevic, and Jardine
(2009) fronted the assignment of the appropriately skilled maintainers of assets in a certain timeframe with multi-objective
programming model to minimise workforce’s cost and maximise plants’ availability.
Within production management, Alwadood, Kassim, and Rani (2010) developed a workforce schedule model to min-
imise the total lead-time of service and to maximise the total number of completed jobs. Their work was based on
Grouped-Efficiency, with which the average working time is reduced and the total number of completed jobs is increased.
On a similar line, Valls, Pérez, and Quintanilla (2009) studied the Skilled Workforce Project Scheduling Problem (SWPSP),
considering the start and end times of the of jobs including penalties for delays, client-priorities, generalised precedence rela-
tionships and percentage time lags. SWPSP produced the first model considering also the variable of worker’s efficiency.
Castillo, Joro, and Li (2009) proposed an approach to generate a large number of plausible schedules by the simultaneous
evaluation of workforce costs and service levels.
Within assembly lines, the most common WS goal is to ensure that all tasks are performed within a certain deadline,
sizing the workforce with respect to the specific skills required for each task. Towards these objectives, Yazgan et al. (2011)
explain the concept of common job, proposing a heuristic to assign more workers to the same task within an assembly line.
Blum and Miralles (2011) fronted the so-called Assembly Line Worker Assignment and Balancing Problem (ALWABP)
with a beam search heuristic to minimise cycle time with a fixed number of workers. On this research line, Araújo, Costa,
and Miralles (2012) extended the ALWABP with the minimisation of the number of stations, subjected to an upper bound
value of the cycle time. The analysis starts from a lower limit value; then, if the desired solution for the number of stations
considered is not found, the cycle time is increased and the solution is searched again. More recently an extension of the
ALWABP to minimise the expected cycle time under uncertain worker availability was proposed by Ritt, Costa, and Miralles
(2015). In contrast, the minimisation of workforce is fronted by Battaia et al. (2015), through an integer linear programming
with conventional and randomised heuristics.
Artificial intelligence has been used in WS problems mainly to front unforeseen situation with real-time re-scheduling. In
this context, Savino, Brun, and Mazza (2014) faced the problem of real-time workforce balancing with a multi-agent system
to reschedule workers in case of unforeseen events. More recently, Genetic Algorithms (GA) were used by Mas, Poveda-
Bautista, and Garzon-Lea (2017) for the optimisation of workstation layouts, showing to be effective approaches to find
solutions for WS problems where exact methods are less effective (Awwad and O’Kane 2010; Cowling et al. 2006; Urlings,
Ruiz, and Serifoglu 2010; Zhang, Gao, and Shi 2011). GAs have also been applied to problems combining scheduling and
routing (Algethami and Landa-Silva 2017; Pinheiro, Algethami, and Landa-Silva 2016).
As regards to workers’ capacities, an example of WS problem solution is the one of Kara et al. (2014), who fronted
the problem of assigning workers to workstations with respect to their skill levels. These authors used a binary linear
programming formulation to minimise overall costs associated with workers and to improve ergonomic working conditions.
WS has been widely fronted in the service sector as well. In this research context, the cross-training strategy has been
widely used for managing workforce’s flexibility (Fowler, Wirojanagud, and Gel 2008; Thannimalai et al. 2013).
In Ergonomics, a high level of relevance has been given to the problem of workforce rotation, with many researchers
who fronted it in their studies as a solution to balance workload and reduce ergonomic risks among operators. Just to cite
a few, Nanthavanij, Yaoyuenyong, and Jeenanunta (2010) through a heuristic approach develop a model to plan workers
by maximising production, while Wongwien and Nanthavanij (2012) extended this approach considering the limitations of
workers and the requirements of activities for daily planning using less workers. On this research stream, Moussavi, Mahd-
joub, and Grunder (2016) analyse the worker’s assignment based on the ergonomic adequacy between the characteristics of
the jobs and the features of the operators. The problem of multi-objective ergonomic workforce scheduling (MO-EWSP),
has been addressed by Wongwien and Nanthavanij (2016) with a mathematical model considering workforce’s size, pro-
ductivity and job satisfaction. Then, they developed another model considering costs and productivity in a combined way
(Wongwien and Nanthavanij 2017). Battini et al. (2016) include the ergonomic aspect to solve the problem of balancing
the assembly line with the development of a multi-objective model in which time and energy are considered. Mossa et al.
(2016) proposed a dual approach to the problem of the WS through a mixed integer programming model that maximises
production by reducing and balancing the human workload and the ergonomic risk assessed by the Occupational Repetitive
Actions (OCRA) checklist. In Table 1 we classify the most relevant contributions to this domain, with a classification on the
approaches and relative goals, along with the relative industrial areas and the approaches used.
Based on the analysis of these previous studies, we experienced a lack of methods to optimise workforce scheduling and
balancing with regard to Ergonomics.
International Journal of Production Research 3

Table 1. Workforce scheduling domains and main goals.


Approach Domain/area Main goal References
Integer/mixed lin- Assembly lines Minimise the expected cycle Ritt, Costa, and Miralles (2015)
ear/nonlinear time
programming
Minimise the overall cost Kara et al. (2014)
associated with operating costs
of workforce and resources
Automotive industry, Minimise the number of workers Battaia et al. (2015)
assembling large products
Home health care, Minimise the costs and the routes Laesanklang, Landa-Silva,
semiconductor of the operators and Castillo-Salazar (2015);
manufacturing Fowler, Wirojanagud, and
Gel (2008)
Manufacturing processes Minimise makespan and Huq et al.(2004)
workforce sizing
Automotive assembly line Minimise the daily production Moussavi, Mahdjoub, and
cycle time Grunder (2016)
Industrial production Minimise the number of workers, Wongwien and Nanthavanij
maximise person-job fit (2016)
scores, minimise worker-job
changeovers, minimise
worker-job dissatisfied
Minimise the number of workers, Wongwien and Nanthavanij
maximise the total worker-task (2017)
fit score, minimise the total
worker-task changeover
Automotive assembly line Minimise individual penalty Hochdörffer, Hedler, and Lanza
costs (2018)
Maximise production rate and Mossa et al. (2016)
minimise ergonomic risk
Simulation Logistics processes in Minimise labour cost, matching Maccarone (2017)
production facilities deadlines of industrial plans
Plant maintenance Minimise the average total time Alwadood, Kassim, and Rani
of service and to maximises (2010)
the total number of completed
jobs
Assembly lines Increase the line efficiency Hytonen, Niemi, and Toivonen
(2008)
Genetic algorithms Home health care Minimise operational costs Algethami and Landa-Silva
and travelling distance and (2017); Pinheiro, Algethami,
maximise customers and and Landa-Silva (2016)
workers satisfaction
Manufacturing systems Minimise makespan time Zhang, Gao, and Shi (2011);
Urlings, Ruiz, and Serifoglu
(2010)
Call centre Minimise time lags Valls, Pérez, and Quintanilla
(2009)
Industrial production Minimise the importance of the Cowling et al. (2006)
tasks scheduled and minimise
the travel time
Cross-training strategy Call centre Increase efficiency Thannimalai et al. (2013); Gel,
Hopp, and Van Oyen (2007)
Sales and customer service Minimise overall costs Robbins, Medeiros, and
Harrison (2010)
Multi-agent system Electromechanical assembly Maximise workers production Savino, Brun, and Mazza
line capacity, minimise buffers (2014)
level
(Continued)
4 M.M. Savino et al.

Table 1. Continued.
Approach Domain/area Main goal References
Assembly centres Minimise operational costs and Sabar, Montreuil, and Frayret
personnel dissatisfactions (2009)
Alternative multidimensional Call centre Minimise costs and maximise the Castillo, Joro, and Li (2009)
paradigm level of service
Decomposition based Network maintenance Minimise travelling of workers Goel and Meisel (2013)
algorithm and network assets downtimes
Beam search algorithm Assembly lines Minimise cycle time with a fixed Blum and Miralles (2011)
number of workers
Multi-response Taguchi Automotive Minimise assembly station, Yazgan et al. (2011)
method number of workers and
maximise productivity
Multi-objective simulated Plant maintenance Minimise the workforce cost and Safaei, Banjevic, and Jardine
annealing algorithm maximising the equipment (2009)
availability
Multi-objective models Assembly line Optimise throughput, makespan, Neubert and Savino (2009)
work in process
Assembly line and workplace Minimise cycle time and number Araújo, Costa, and Miralles
design of stations (2012)
Minimise the number of workers Wongwien and Nanthavanij
(2012)
Minimise cycle times Battini et al. (2016)
Evolutionary algorithm Plant maintenance Minimise labour costs Quan et al. (2007)
Heuristic algorithms Assembly line Increase workforce efficiency Noack and Rose (2008)
Industrial production Maximise competency, minimise Nanthavanij, Yaoyuenyong,
the number of workers, and Jeenanunta (2010)
maximise the productivity

Nowadays the objectives of Ergonomics can be identified as the usability and safety of systems where the operator is
considered as a user and an integral part (Vignais et al. 2013, 567). Recent studies assessed Ergonomics as the study and
design of complex systems whose effectiveness is determined by its sustainability in terms of technological, economic and
social features (Savino, Mazza, and Battini 2016, 49). Within Ergonomics, the Ergonomic Exposure (Bao et al. 1997) deals
with the postural variables of the worker observed during his activities (e.g. neck angle, trunk rotation, arms extension, legs
flexion, etc.).
The present study aims to extend the previous results on WS, by appraising the impact of ergonomics over production
performances. The research methodology and the relative model developed aims to quantify the ergonomic exposure, and
then to constrain it to the production capacity.
The results are then used to investigate the impact on workforce scheduling and balancing, thus filling the gap of the
previous works that do not consider the influence of ergonomic exposures among the variables to optimise in workforce
scheduling.
Based on the extant body of literature, the research attempts to answer some basic questions regarding the potential
influence of the ergonomic exposure on the assignment of the worker to each task/workstation.
The first research question aims to address the basic viability of the workforce scheduling considering ergonomic
parameters as well.

RQ1 How is it possible to consider the ergonomic stress of workers within the problem of workforce scheduling?

The influence of other parameters within the WS has stimulated Academics and Practitioners to explore the impact of
allocation optimisation in respect of some additional workers’ requirements, such us skills (Michalos et al. 2010), fatigue
and personalities (Othman, Gouw, and Bhuiyan 2012), workers’ age (Boenzi et al. 2015) or other hexogen parameters
(Costa, Fichera, and Cappadonna 2013; Mossa et al. 2016; Wongwien and Nanthavanij 2012, 2017).
The second research question aims to investigate the impact of the ergonomic stress, defined as the set of variables
relative to the ergonomic exposures, on workforce allocation in terms of specific production parameters.
International Journal of Production Research 5

RQ2 May we optimise production capacity with regard to the ergonomic stress of workers?

The aim of RQ2 is to explore the workforce allocation under the twofold perspective of production performances and
ergonomics. Within RQ2 , the main goals pursued throughout the case study has been: (i) maximising or keeping constant
the production capacity and (ii) keeping the total value of ergonomic stress as low as possible.

3. Research methodology
The Research Methodology of Figure 1 has been purposely designed as a flexible approach to study WS within the
ergonomic domain.
Within the study, a Workforce Ergonomic Scheduling Model (WESM ) has been designed with the goal to allocate the
workers considering the ergonomic parameters. Throughout the WESM we attempted to answer to the RQ raised by the

Figure 1. Research methodology.


6 M.M. Savino et al.

analysis of the extant literature. In this regard, Figure 1 reports those steps in which these Research Questions find the
answer.
In the first step (S1) a set of ergonomic parameters is defined. In this step, we propose the Ergonomic Stress Matrix
(ESM ) to model the ergonomic assessment of the whole shop floor. Then, the ESM is used to appraise the Total Ergonomic
Stress (TES) of the current workforce scheduling configuration.
The step (S2) has been fronted with the Constraint Optimisation Problem (COP) approach (Neubert and Savino 2009),
by realising a COP model that includes the ergonomic constraints that result from S1.
In S3 the elements that compose the WESM are detailed and explained with examples extracted from the case study. In
the S4 step, we analyse the outputs of the WESM applied to the case study. In this step, a tuning of the system is conducted
to obtain an optimal scheduling with regard to the production capacity and the ergonomic stress.
The focused on practice research conducted within the case study is based on the data of the ergonomic postures and
working sheets, along with the data regarding the number of operators, cycle times and production constraints. The data
relative to the ergonomic postures have been acquired through direct observations and video recordings of production
activities (Battini et al. 2011). The observations and video recording resulted in the ergonomics points found for each
worker, according to his own posture.

4. The workforce ergonomic scheduling model


The possible inclusion of the ergonomic parameters within the WS has been tested by modelling how ergonomic parameters
may affect the scheduling.
Toward this objective, we were inspired by the work of Savino, Mazza, and Battini (2016) who evolved the classic REBA,
RULA and OCRA ergonomic indicators to appreciate the ergonomic postures, and by the work of Battini et al. (2016) who
appreciated the human energy in balancing an assembly line. While the first work allows to obtain a complete postural stress
index, the second one forwards the concept that ergonomic aspects may be among the variables of an objective function
in which other parameters are also optimised. Following these concepts, the ergonomic variables of this study have been
conceived to obtain (i) a basic ergonomic assessment conducted for each worker and (ii) the possible workforce scheduling
solutions of the model with respect to the ergonomic assessment obtained. It is worth mentioning that the assessment depends
directly from the ergonomic postures, while the possible solution(s) of the workforce allocation may be affected not only
by the number of workstations visited and the time spent by the worker on each workstation, but also by the ergonomic
assessment itself.
Hence, for each operation conducted on the shop floor, we define the ergonomic stress (es) of the worker as a function of
(i) a score obtained from worker’s ergonomic assessment, and (ii) the time spent on each operation assigned to this worker
within the working shift.
Based on the above basic principles, we conceived an extended version of the Overall Ergonomic Score (OES) defined
by Savino, Mazza, and Battini (2016), computed with the partition of the tasks into Elementary Operations (EO). In this
study, the OES is extended towards the inclusion of the ergonomic assessment for each (EO) of the task through a Partial
Ergonomic Score (PES). Thus, the new OES is expressed by the relation (1):
N
i=1 PESi
OES =  (1)
n × NEI × Ni=1 EOi

where n is the number of domains, Ni=1 PESi is the sum of each score obtained for a single elementary operation, NEI is
the number of ergonomic indicators, and EOi is the elementary operation.
In Table 2 we report an example of PES and OES calculation relative to a task composed by seven ergonomic indicators
and relative to the Upper Limb Assessment. In Table 2, the ergonomic indicators of the worker are identified with the
numbers [1 . . . 7]. These numbers correspond to the description of each ergonomic indicator in the rows of Table 2. Then,
the evaluation is iterated for each EO.
As a further example, for a task featured by EO1 and EO2 whose scores are 11 and 22, respectively, and considering the
upper limb domain, the OES is computed as follows:
n
i=1 PESi 11 + 22
OES = N = = 2.06
n × NEI × i=1 EOi 1 ×8×2

Then, OES is normalised with respect to its maximum value. For this example, OESMAX = 3, we obtain

OESN = OES/OESMAX = 0.69


International Journal of Production Research 7

Table 2. Example of computing PES for a single operation.

Ergonomic indicators
Levels (L)
Description L1 L2 L3 Value Level Score PES (EO1 )
1. Trunk bending angle > 30° 30° ÷ 15° < 15° 16° L2 2 11
2. Waist rotation angle > 45° 45° ÷ 15° < 15° < 15° L3 1
3. Arm height Over shoulders At shoulders Waist height Waist height L3 1
4. Neck bending or rotation > 20° and rotation > 20° 0° ÷ 20° < 20° L3 1
angle
5. Forearm rotation angle > 90° and crossed > 0° 0° ÷ 90° > 90° L2 2
6. Wrist rotation with respect > 50% 50% ÷ 0 0 < 50% L3 1
to maximum torsion
potential
7. Wrist bending angle > 15° 15° ÷ 0° 0° < 15° L2 2
8. Operation time with respect > 2/3 Tt (2/3 ÷ 1/3) Tt < 1/3 Tt < 1/3 Tt L3 1
to the total task time Tt
Note: The score is assigned as follows: L1 = 3, L2 = 2, L3 = 1 (Savino, Mazza, and Battini 2016).

The es for the whole shop floor is modelled by the Ergonomic Stress Matrix (ESM), purposely conceived as a methodological
tool for the portion of the research. The ESM considers j workers [j = 1 . . . M ] scheduled on i operations [i = 1 . . . N] as
follows (2).

⎡ wk ⎤
oeswk
op
1
oesop1j oeswk
op1
M

⎢ wk11 wk ⎥
ESM = ⎣oesopi oesopij opi ⎦
oeswk M (2)
wk
oeswk
opN
1
oesopNj oeswk
opN
M

wk
where wkj is the j worker [j = 1 . . . M ], opi is the i operation [i = 1 . . . N], and oesopij is the OESN of wkj for the opi .
According to WS principles (Savino, Mazza, and Battini 2016), each wkj spends a certain amount of time on the i
wk
operation (opi ) scheduled within the working shift. Hence, for each opi the es of the worker is a relation of (i) oesopij and (ii)
the time in which that opi is carried out by that worker during the shift time. This relation can be expressed by (3).

wk tS wkj
wk oesopij ∗ t=0 zopi (t)
esopij = (3)
TST

wk
where zopij (t) is the execution time of wkj for the opi , TST is the total shift time, and tS is the number of time slots of the
shift.
8 M.M. Savino et al.

The Total Stress Index (TSI) is conceived to measure the ergonomic stress for the entire working shift. The TSI is
wk
computed for each wkj , that is function of the esopij , as follows Equation (4):
⎛ wkj t wkj

opi ∗ (t)
S
N N oes t=0 z opi
TSI wkj =
wk
esopij = ⎝ ⎠ (4)
op =1 op =1
TST
i i

Then the methodology is provided of the Total Ergonomic Stress (TES) for the entire shop floor. The TES is computed
as the average value on the M workers of the TSI, according to (5).
⎛ ⎛ wkj t wkj
⎞⎞
1
M N oes opi ∗ S
t=0 z opi (t)
TES = ∗ ⎝ ⎝ ⎠⎠ (5)
M wk =1 opi=1 TST
j

wk wk
From (5) we may see that TES is function of oesopij ∈ ESM and of zopij (t).

4.1. Production system modelling


The production system of this case study is an assembly flow shop composed by 13 workstations, which layout is given in
Figure 2. The operations conducted are detailed in Table 3.
In this flow shop the assembly cycle consists of N = 13 operations (opi i = [1 . . . N]) conducted by three operators,
namely wk 1 , wk 2 , wk 3 within a shift of eight hours.
The opi have the following constraints:
(1) op4 and op8 require four hours for drying, thus they are carried out two shifts and one shift before, respectively;
(2) op5 , op6 , op7 are carried out in sequence by the same operator wk 3 , since he is able to carry out only these activities.
Thus, the are conducted the shift before to avoid the introduction of waiting times;
(3) op10 requires that op1 , op2 , op3 , op9 have been completed before. This condition implies that also op4 , op5 , op6 ,
op7 , op8 have been completed, because they are conducted in the previous shift;
(4) op11 requires that op10 is completed;
(5) op12 requires that op11 is completed;
(6) op13 requires that op12 is completed.
Table 3 shows the operations with the respective cycle times and the constraints.
According to the above constraints this flow shop needs to schedule the two workers wk 1 , wk 2 on 10 opi .
From Table 3 we may notice that for the op10 , the constraints, op5 , op6 and op7 are not reported because they are
conducted by wk 3 during the previous shift. This because to run the model we need to include only the constraints for the
operations executed in the same working shift.

Figure 2. Layout of the production system.


International Journal of Production Research 9

Table 3. Operations and constraints.


Operation Description Cycle time (sec) Constraint(s)
op1 Insulating protection 18 None
op2 Case taping 18 None
op3 Spacer caulking 9 None
op4 Coil switch coating 10 None
op5 Bridge soldering 10 Conducted by wk 3
op6 Coil soldering 45 Conducted by wk 3
op7 Coil switch soldering 45 Conducted by wk 3
op8 Coil coating 10 None
op9 Coil taping 36 None
op10 Assembling 32 op1 , op2 , op3 , op4 , op8 , op9
op11 Dielectric test and case closing 24 op10
op12 Label 36 op11
op13 Electric test and packaging 51 op12

For this portion of the study the Constraint Optimisation Problem (COP – Neubert and Savino 2009) has been used to
model workers’ states, production constraints and objectives, with the following notations:
wk = {1, . . . , wkM } is the number of available workers;
op = {1, . . . , opN } is the number of operations;
t = {0, . . . , tS } is the workday length, in S time slots;
op (t) is the
xwk

operator’s state;
1 if wk is working on the op at time t
op (t) = 0
xwk
otherwise
;
popi (t) is the pieces worked within opi at time t.
popi (t) can be expressed by the ratio between the time spent by the wkj on a workstation and the cycle time of a single
workpiece on that workstation, according to Equation (6):
 wk
wk zop (t)
pop (t) = (6)
tuop
wk
where zop (t) is the time to conduct opi by the wkj and tuop is the cycle time of the workstation.
With these variables, the objective function of the COP is given by Equation (7).
 wk
op zopN (tS )
max (7)
tuopN

where N is the total number of operations of the line.


The maximisation of (7) implies to maximise popN (tS ), namely the number of workpieces of the last operation opN at tS .
The constraints have been modelled as follows:
p10 (t) ≤ p1 (t); p10 (t) ≤ p2 (t); p10 (t) ≤ p3 (t); p10 (t) ≤ p9 (t); p11 (t) ≤ p10 (t); p12 (t) ≤ p11 (t)p13 (t) ≤ p12 (t);
pop (t0 ) = 0 ∀st. No workpiece in the workstation at the beginning of the working day;
pop (t) ≤ tn ∗wk n
tuop
∀st.
op

This last constraint models the maximum number of workpieces that can be worked in each opi is equal to the production
capacity in the optimal case, when there is no idle time.
The priority constraints are translated into constraints relative to the number of pieces that have been already worked in
each opi .
p13 (opn ) ≤ p4 (opn ),
p13 (opn ) ≤ p8 (opn ).
These operations have no priority constraints, but they must be completed within the end of the working shift.
 wk
wk zop (t0 ) = 0 ∀op. Initially all the times spent on the operations, by each operator, are equal to 0;
10 M.M. Savino et al.

op zop (t) ≤ tS ∀wk. The time that each worker can spend for all the operations is equal to the duration of the
wk

working
 wk day;
wk xop (t) ≤ 1. Each workstation can be used by a single worker or, similarly, an operation can be carried out by
asingle worker at a wk
time;  wk
z
wk op
wk
(t) − wk op (t − 1) −
z wk xop (t) = 0. The time spent on an operation can increase only if a worker has
been assigned to that operation.
Each operation needs to be completed within the tuop time according to the cycle time of Table 2.
The WESM considers a value TESh , to which the value of the possible solutions of the COP should be constrained to.
Hence, the entire WESM model can be written as follows:
 wk
op zopn (tS )
max (8)
tuopn

subjected to the following constraints:


⎛ ⎛ wkj t wk
⎞⎞
1
M N oesopi ∗ t=0
S
zopij (t)
TES = ∗ ⎝ ⎝ ⎠⎠ ≤ TESh
M wk =1 op =1 TST
j i

pop (t0 ) = 0

tn ∗ wkn
pop (t) ≤ 
op tuop


wk zop (t)
wk
pop (t) =
tuop

p10 (t) ≤ p1 (t)

p10 (t) ≤ p2 (t)

p10 (t) ≤ p3 (t)

p10 (t) ≤ p9 (t)

p11 (t) ≤ p10 (t)

p12 (t) ≤ p11 (t)

p13 (t) ≤ p12 (t)

p13 (opn ) ≤ p4 (opn )

p13 (opn ) ≤ p8 (opn )



wk
zop (t0 ) = 0
wk


wk
zop (t) ≤ tS
op


op (t) ≤ 1
xwk
op
International Journal of Production Research 11

op (t) ≤ 1
xwk
wk


wk
zop (t) − wk
zop (t − 1) − op (t) = 0
xwk
wk wk wk

op (t) = {0, 1}
xwk ∀wk, op, tpop (t) ∈ N

wk
zop (t) ∈ N

5. Results analysis
The postural analysis has been made for the Upper Limb Domain (Savino, Battini, and Riccio 2017). This analysis has been
conducted for the opi carried out by wk 1 and wk 2 , resulting in the ESM (Equation (9)).
⎡ ⎤ ⎡ ⎤
oes11 oes21 0.44 0.44
⎢ oes1 oes2 ⎥ ⎢0.11 0.11⎥
⎢ 2 2⎥ ⎢ ⎥
⎢ oes1 oes2 ⎥ ⎢0.35 0.35⎥
⎢ 3 3⎥ ⎢ ⎥
⎢ oes1 oes2 ⎥ ⎢0.15 0.15⎥
⎢ 4 4⎥ ⎢ ⎥
⎢ oes1 oes2 ⎥ ⎢0.15 0.15⎥
⎢ 8 8⎥=⎢ ⎥ (9)
⎢ oes1 oes2 ⎥ ⎢0.42 0.42⎥
⎢ 19 9⎥ ⎢ ⎥
⎢oes oes210 ⎥ ⎢ ⎥
⎢ 10 ⎥ ⎢0.26 0.26⎥
⎢oes1 oes2 ⎥ ⎢0.48 0.72⎥
⎢ 111 11 ⎥ ⎢ ⎥
⎣oes12 oes212 ⎦ ⎣0.33 0.33⎦
oes113 oes213 0.24 0.24

From (9) we may notice that the OESN have the same values for all the operations, except for op11 . For this op11 the postural
analysis resulted in a partition of four EOs, namely: Positioning (P), Leverage handle (LH), Pressing (PR), Picking (PI). Of
them, the OES identified with P, LH and PI have the same PES for both workers. In contrast, PR values indicate differences
in the posture among the two workers, thus generating a higher PES value. This generated the values of OESN = 0.48 for
wk 1 and OESN = 0.72 for wk 2 . From the postural analysis, it resulted that the reason is mainly because PR requires a wider
arm extension, which is greater if the worker has a smaller trunk, thus forcing a higher neck extension backwards. This
matter can be noticed in Table 4 for wk 2 , on op11 . For this op11 , the same Table 4 gives the values of the ergonomic indicator
resulted from the postural analysis compared with the two workers, with the respective Level (L) of each ergonomic indicator
and the PES values.
The es values of both workers are calculated according to (3) as follows:
  S wk1 
0.48 ∗ tt=0 zop11 (t)
esop11 =
wk1
(10)
TST

 tS 
0.72 ∗ t=0 zop11 (t)
wk2
eswk2
op11 = (11)
TST
The WESM has been implemented through the Mosel language, using the F-Xpress solver. To investigate the answer to RQ1
we have tested the WESM with two different scenarios of TESh values. In the first case (Case #1) we attempted to model
the absence of ergonomic constraints by putting TESh → + ∞, while in the second case (Case #2) we adopted the value of
TESh = 0.865.
Case #1
Table 5 shows the results obtained using 29 time slots, equal to 435 min with TESh → + ∞; wk 3 is scheduled on op5 ,
op6 and op7 as per previous constraints, thus this schedule does not appear in Table 5.
The above schedule of Table 5 generates the values shown in Table 6, where the scheduled times for each opi and the
respective es values are reported.
In this Table 6, the scheduled time equal to zero means that the worker has not been scheduled on that operation for that
shift, as for the case of wk 1 for op4 .
12 M.M. Savino et al.

Table 4. op11 – Results of the postural analysis on both workers for the EO3 .
op11 – EO3 wk 1 wk 2

Ergonomic indicator
Forearm rotation angle 154° (L1) 176° (L1)
Arm height Over shoulders (L1)
Neck bending angle < 20 (L3) > 20° and rotation (L1)
Wrist rotation < 50% (L3)
Wrist bending angle < 15° (L2)
Trunk bending angle < 15° (L3)
Waist rotation angle < 15° (L3)
Operation time with respect to the total task time Tt 0.5 * Tt (L2)
PES 14 16

Table 5. WS with TESh → + ∞.


Time slot op
1 2 3 4 8 9 10 11 12 13
1 wk 1 wk 2
2 wk 2 wk 1
3 wk 1 wk 2
4 wk 2 wk 1
5 wk 2 wk 1
6 wk 1 wk 2
7 wk 1 wk 2
8 wk 2 wk 1
9 wk 1 wk 2
10 wk 2 wk 1
11 wk 2 wk 1
12 wk 2 wk 1
13 wk 1 wk 2
14 wk 2 wk 1
15 wk 1 wk 2
16 wk 2 wk 1
17 wk 2 wk 1
18 wk 2 wk 1
19 wk 2 wk 1
20 wk 2 wk 1
21 wk 1 wk 2
22 wk 1 wk 2
23 wk 1 wk 2
24 wk 1 wk 2
25 wk 2 wk 1
26 wk 1 wk 2
27 wk 1 wk 2
28 wk 1 wk 2
29 wk 2 wk 1
International Journal of Production Research 13

Table 6. Scheduled times and ergonomic stress (TESh → + ∞).


wk 1 wk 2
op Scheduled time (min) es Scheduled time (min) es
1 18 18.21 67 67.77
2 48 12.14 41 10.37
3 47 27.01 0 0.00
4 0 0.00 45 15.52
8 4 1.38 19 6.55
9 153 147.73 5 4.83
10 14 8.37 99 59.18
11 26 28.69 26 49.01
12 38 28.83 16 12.16
13 15 8.28 46 25.38

Case #2
Table 7 reports the new workforce allocation with TESh = 0.865. In Table 7, we can see how the scheduling changes as
the introduction of the ergonomic constraint occurs.
In this case, the algorithm assigned three time slots for the heaviest operation (op11 ) to the operator wk 1 who has a lower
value of es, thus obtaining a lower value of TES. In contrast, for Case #1 the algorithm assigned two time slots. Table 8
reports the values resulting from this scheduling.
For this specific case, from Table 8 we may notice that wk 1 is not scheduled in op4 and op8 .

Table 7. WS with TESh = 0.865.


op
Time slot 1 2 3 4 8 9 10 11 12 13
1 wk 2 wk 1
2 wk 1 wk 2
3 wk 2
4 wk 1 wk 2
5 wk 2 wk 1
6 wk 2
7 wk 1 wk 2
8 wk 2 wk 1
9 wk 1 wk 2
10 wk 2 wk 1
11 wk 2 wk 1
12 wk 2 wk 1
13 wk 2 wk 1
14 wk 1 wk 2
15 wk 2 wk 1
16 wk 2 wk 1
17 wk 1 wk 2
18 wk 2 wk 1
19 wk 2 wk 1
20 wk 2 wk 1
21 wk 2 wk 1
22 wk 2 wk 1
23 wk 2 wk 1
24 wk 2 wk 1
25 wk 2 wk 1
26 wk 1 wk 2
27 wk 1 wk 2
28 wk 2 wk 1
29 wk 1 wk 2
14 M.M. Savino et al.

Table 8. Scheduled times and ergonomic stress (TESh = 0.865).


wk 1 wk 2
op Scheduled time (min) es Scheduled time (min) es
1 44 44.50 23 23.27
2 16 4.05 63 15.93
3 24 13.79 19 10.92
4 0 0.00 44 15.17
8 0 0.00 47 16.21
9 47 45.38 89 85.94
10 73 43.64 15 8.97
11 36 39.73 13 21.52
12 22 16.69 29 22.00
13 38 20.97 17 9.38

6. Discussion
Table 9 shows the values of the maximum daily production capacity, the idle times of each operator and the normalised TES
(TESN ) for both cases.
From Table 9 we may see how for Case #1 the WESM resulted in a traditional COP solution, with a balanced schedule
of both workers. In contrast, in Case #2 the WESM allowed to obtain the same number of pieces produced by unbalancing
the workload among the two workers.
Analysing the results of Table 8, we see that due to the ergonomic constraint of wk 2 for op11 , the WESM resulted to
schedule this worker only on the op11 . In contrast, wk 1 is scheduled for 36 min on op11 , whose ergonomic parameters allows
him to work on this op. In addition, again according to the postural analysis, the scheduling made by the WESM does not
allocate wk 2 on the op with high es values, thus keeping low the ergonomic stress for this worker. This first empirical result
is consistent with the main objective of this study to investigate the possibility to consider the ergonomic factors in the WS.
Additionally, the tuning activity applied to the WESM resulted in a minimisation of 14% of the TES as regards to Case # 1.
Thus, from these results we may answer to RQ1 by arguing that it is possible to include the ergonomic stress into a WS
problem with an approach that (1) integrates a COP model with the results of postural analysis, expressed in terms of a the
ESM and then (2) by constraining the COP model with the threshold value of the TES variable, function of the ESM.
A possible limitation for this answer may be that, even if the workforce allocation considers the ergonomic constraints,
allowing at the same value of daily production, it does not allow the balance of the working times among the workers.
According to RQ2 , the second part of the study has been finalized to verify if it is possible to optimise the production
parameters with respect to the ergonomic stress. This investigation required to analyse the possible correlations between
the es, expressed by the TES index, and the production parameters such as production capacity and idle time (Neubert and
Savino 2009; Savino, Brun, and Mazza 2014).
For this portion of the study, Figure 3(a,b) reports the WESM functions for production capacity and idle time,
respectively, when the TESh decreases.
From Figure 3(a,b) we may argue the second empirical result of this study. The decreasing of the TESh causes the
lowering of the daily production and a concurrent increasing of the idle time. In particular, the production capacity function
of Figure 3(a) shows a shape similar to a step function, whose steps increase with regard to the increasing of the TES
threshold. Figure 3(a) also shows that the threshold of ergonomic stress can be reduced up to TES = 0.86, while keeping
constant the value of the daily production capacity. Then, from Figure 3(b) we may see how this value of TES corresponds
to an increase of the values of idle time for both workers.
Thus, the WESM is able to suggest to the production manager how to decrease the ergonomic stress up to the most
feasible value as regards to the production capacity.

Table 9. WS summary (TESh → + ∞, TESh = 0.865).


Case #1 Case #2
Parameter wk 1 wk 2 wk 1 wk 2
Production capacity 52 pieces 52 pieces
Idle time 72 min 71 min 135 min 76 min
TESN 0.999 0.860 < TESh
International Journal of Production Research 15

Figure 3. (a) Daily production capacity versus ergonomic stress. (b) Idle time versus ergonomic stress.

Another relevant aspect resulting from the test campaign is the possibility to set a threshold for the production capacity
as regards the ergonomic stress. In this way the production manager can indeed choose to reduce the daily production
capacity in case there may certain requests of reducing the ergonomic stress, thus having a useful tool able to manage these
two apparently contrasting aspects.
To deep RQ2 and its answer, a tuning activity has been conducted on the WESM by setting the TESh = 0.812, that is the
lowest limit of TES for the production capacity function of Figure 3(a).
Tables 10 and 11 report the relative scheduling scenario. From Table 10, and by comparing the three scheduling cases,
we may notice that decreasing the TES threshold results in the exclusion of op8 and op11 from the tasks scheduled to the
wk 2 , similarly for wk 1 on op10 . From Table 11 we may also see that, through the tuning activity the production manager is
able to decide the most feasible level of the production capacity with a concurrent balancing of the workload between the
two workers.
Table 12 shows the comparison between the results of the three scheduling cases.
From Table 12 we can see how the value of the daily production capacity remains constant if the TESh is decreased up to
the first step of the TES function in Figure 3(a). Hence, with the TES tuning we have experienced (i) a reduction of the TES
of 19% and (ii) an increase of 53% of total idle time of the workers. Figure 3(a,b) along with the data of Table 12, may better
answer to RQ2 in terms that the WESM is a system allowing (i) to decide which can be the optimal production capacity
in respect to the allowable values of ergonomic stress for each worker and (ii) to keep the threshold of the es as lowest
16 M.M. Savino et al.

Table 10. Scheduled times and ergonomic stress (TESh = 0.812).


wk 1 wk 2
op Scheduled time (min) es Scheduled time (min) es
1 39 39.45 22 22.25
2 50 12.64 48 12.14
3 21 12.07 16 9.20
4 19 6.55 24 8.28
8 45 15.52 0 0.00
9 28 27.04 89 85.94
10 0 0.00 87 52.00
11 47 51.87 0 0.00
12 35 26.55 17 12.90
13 42 23.17 22 12.14

Table 11. WS summary (TESh = 0.812).


Parameter wk 1 wk 2
Production capacity 52 pieces
Idle time 109 min 110 min
TESN 0.809 < TESh

Table 12. Comparison between the scheduling cases.


Parameter Case #1TESh → + ∞ Case #2TESh = 0.865 Tuning caseTESh = 0.812
Production capacity 52 pieces 52 pieces 52 pieces
Total idle time 143 min 211 min 219 min
TESN 0.999 0.860 0.809

as possible, while keeping constant the production capacity. Thus, the system can be a useful tool for those production
managers who aim to have an optimal workforce scheduling under the twofold perspective of ergonomics enhancements
and productivity optimisation. As the last consideration, it is worth mentioning that short-term time slots require operators to
alternate more frequently between operations. This may result in an increased moving stress caused by the several changes
of workstation requested to the worker. Hence, within this research stream, a further development may regard also the
correlation between the duration of the time slots as regards to the ergonomic stress suffered by the operators.

7. Conclusion
The present work was mainly prompted by the need to link the possible WS solutions with the potential ergonomic stress
that the worker may suffer during the working shift. In particular, the research methodology investigated the possibility of
establishing a relationship between the sequence of operations assigned to the workers and the ergonomic conditions in
which these operations are conducted, hence assessing the impact of the ergonomic load as regards to the main production
parameters.
This portion of the study has been fronted through a preliminary ergonomic analysis, resulting in a basic ergonomic score
for each worker. This score, combined with the time spent by the worker on each workstation, allowed to define (i) how the
workforce scheduling impacts on the ergonomic stress of the workforce and (ii) which WS policies may be adopted based
on the apparently contrasting objectives of maximising production capacity and minimising the physical load for workers.
Although the results obtained, the WESM developed in the present study is indeed affected by two main technical
limitations and by an ethical one.
The first limitation concerns the lack of correlation between the level of ergonomic stress and the idle time period of
each worker provided by the scheduling sequence. It is worth mentioning that idle times of the workstations allow operators
to have rest, thus reducing the ergonomic stress suffered. Further works might investigate the impact that this parameter
may have in the WS solutions.
The second limitation is that the WESM does not allow to completely exclude from the WS those operations with
OES > 0.7, that in our case study is the threshold beyond which the op is considered critical from an ergonomic point of
view.
International Journal of Production Research 17

Regarding the ethical limitation, in case of critical operations for the same worker(s) the WESM, by striving to not
decrease the production capacity, tends to unbalance the workload among them. This feature might generate potential situa-
tions in which some worker may be completely discharged from some operations, thus resulting with low saturation values,
while others may be saturated up their maximum availability. This under an ethical point of view may not be viable, thus
generating potential complains of the highly saturated workers.
As regards to these limitations, a possible outlook of this study may be an evolution of the WESM oriented (i) to
withdraw the most highly critical operations from the list of tasks for which a specific more vulnerable worker may be
scheduled, and (ii) to find the best workers’ scheduling with the minimal ergonomic stress.
Another possible future development of this work towards ergonomic evaluation may be the study of an ergonomic
score that is function not only of the duration of the task, but also of the time slot in which task is allocated and a recovery
coefficient function of the idle time of the operator.
Albeit the unavoidable limitations, the empirical results got within the work may give two main messages for Academics
and Practitioners as regards ergonomics’ problems versus the optimal allocation of the workforce within a shop floor. First,
that optimal ergonomic conditions may not be completely in contrast with production performances optimisation. Secondly,
production managers may agree with workers a threshold of ergonomic stress and constrain it to the production capacity.
Hence, by constraining workers’ allocation and corresponding working loads with regard to this threshold, it is possible to
have a fair production management policy that respects worker’s wellness.
By exploring some aspects of workforce scheduling under ergonomic perspective, this study got new findings to front
these apparently contrasting objectives.

Acknowledgements
The authors are grateful to the Editor in Chief, Prof. Alexandre Dolgui, and to the five anonymous reviewers for their insightful and
constructive comments throughout the review process.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This paper has been developed in the framework of the research activities of the projects funded by the EU Commission for the Project
Links in Europe and Asia for Education, Enterprise and Research – LEADER, agreement number EACEA IT 0861 001 001 within the
Key Action 4 relative to Environment and Safety.

ORCID
Matteo M. Savino http://orcid.org/0000-0002-7266-8849

References
Algethami, H., and D. Landa-Silva. 2017. “Diversity-Based Adaptive Genetic Algorithm for a Workforce Scheduling and Routing
Problem.” IEEE Congress on Evolutionary Computation, 1771–1778.
Alwadood, Z., I. Kassim, and R. Rani. 2010. “Maintenance Workforce Scheduling Using Arena Simulation.” Second International
Conference on Computer Research and Development, 517–521.
Araújo, F. B., A. M. Costa, and C. Miralles. 2012. “Two Extensions for the ALWABP: Parallel Stations and Collaborative Approach.”
International Journal of Production Economics 140 (1): 483–495.
Awwad, A. S., and J. F. O’Kane. 2010. “Agile Capabilities in Uncertain Business Environments: An Empirical Study of the Jordanian
Manufacturing Sector.” International Journal of Operations and Quantitative Management 16 (2): 155–172.
Bao, Shihan, Jorgen Winkel, Svend Erik Mathiassen, and Houshang Shahnavaz. 1997. “Interactive Effects of Ergonomics and Production
Engineering on Shoulder-Neck-Exposure – A Case Study of Assembly Work in China and Sweden.” International Journal of
Industrial Ergonomics 20: 75–85.
Battaia, O., X. Delorme, A. Dolgui, J. Hagemann, A. Horlemann, S. Kovalev, and S. Malyutin. 2015. “Workforce Minimization for a
Mixed-Model Assembly Line in the Automotive Industry.” International Journal of Production Economics 170: 489–500.
Battini, D., X. Delorme, A. Dolgui, A. Persona, and F. Sgarbossa. 2016. “Ergonomics in Assembly Line Balancing Based on Energy
Expenditure: A Multi-Objective Model.” International Journal of Production Research 54 (3): 824–845.
Battini, D., M. Faccio, A. Persona, and F. Sgarbossa. 2011. “New Methodological Framework to Improve Productivity and Ergonomics
in Assembly System Design.” International Journal of Industrial Ergonomics 41: 30–42.
18 M.M. Savino et al.

Bergh, J., J. Beliën, P. De Bruecker, E. Demeulemeester, and L. De Boeck. 2013. “Personnel Scheduling: A Literature Review.” European
Journal of Operational Research 226: 367–385.
Blum, C., and C. Miralles. 2011. “On Solving the Assembly Line Worker Assignment and Balancing Problem via Beam Search.”
Computers & Operations Research 38 (1): 328–339.
Boenzi, F., S. Digiesi, G. Mossa, G. Mummolo, and V. A. Romano. 2015. “Modelling Workforce Aging in Job Rotation Problems.”
IFAC-PapersOnLine 48 (3): 604–609.
Castillo, I., T. Joro, and Y. Li. 2009. “Workforce Scheduling with Multiple Objectives.” European Journal of Operational Research 196
(1): 162–170.
Costa, A., S. Fichera, and F. A. Cappadonna. 2013. “A Genetic Algorithm for Scheduling Both Jobs Families and Skilled Workforce.”
International Journal of Operations and Quantitative Management 19 (4): 221–247.
Cowling, P., N. Colledge, K. Dahal, and S. Remde. 2006. “The Trade-off Between Diversity and Quality for Multi-Objective Workforce
Scheduling, in Evolutionary Computation in Combinatorial Optimization.” Lecture Notes in Computer Science 3906: 13–24.
Fowler, J., P. Wirojanagud, and E. S. Gel. 2008. “Heuristics for Workforce Planning with Worker Differences.” European Journal of
Operational Research 190 (3): 724–740.
Gel, E. S., W. J. Hopp, and M. P. Van Oyen. 2007. “Hierarchical Cross-Training in Work-in-Process-Constrained Systems.” IIE
Transactions 39 (2): 125–143.
Goel, A., and F. Meisel. 2013. “Workforce Routing and Scheduling for Electricity Network Maintenance with Downtime Minimization.”
European Journal of Operational Research 231 (1): 210–228.
Hochdörffer, J., M. Hedler, and G. Lanza. 2018. “Staff Scheduling in Job Rotation Environments Considering Ergonomic Aspects and
Preservation of Qualifications.” Journal of Manufacturing Systems 46: 103–114.
Huq, F., K. Cutright, and C. Martin. 2004. “Employee Scheduling and Makespan Minimization in a Flow Shop with Multi-Processor
Work Stations: A Case Study.” Omega 32 (2): 121–129.
Hytonen, J., E. Niemi, and V. Toivonen. 2008. “Optimal Workforce Allocation for Assembly Lines for Highly Customised Low-Volume
Products.” International Journal of Services Operations and Informatics 3 (1): 28–29.
Kara, Y., Y. Atasagun, H. Gokcen, S. Hezer, and N. Demirel. 2014. “An Integrated Model to Incorporate Ergonomics and Resource
Restrictions Into Assembly Line Balancing.” International Journal of Computer Integrated Manufacturing 27: 997–1007.
Laesanklang, W., D. Landa-Silva, and J. Castillo-Salazar. 2015. “Mixed Integer Programming with Decomposition to Solve a Workforce
Scheduling and Routing Problem.” International Conference on Operations Research and Enterprise Systems, Lisbon, Portugal,
January 10–12, 112–121.
Maccarone, L. 2017. “A Simulation-Optimization Framework to Solve the Workforce Scheduling Problem in Complex Manufacturing
and Logistic Contexts.” Winter Simulation Conference, 4608–4609.
Mas, J. A., R. Poveda-Bautista, and D. Garzon-Lea. 2017. “Using RGB-D Sensors and Evolutionary Algorithms for the Optimization of
Workstation Layouts.” Applied Ergonomics 65: 530–540.
Michalos, G., S. Makris, L. Rentzos, and G. Chryssolouris. 2010. “Dynamic Job Rotation for Workload Balancing in Human Based
Assembly Systems.” CIRP Journal of Manufacturing Science and Technology 2: 153–160.
Mossa, G., F. Boenzi, S. Digiesi, G. Mummolo, and V. A. Romano. 2016. “Productivity and Ergonomic Risk in Human Based Production
Systems: A Job-Rotation Scheduling Model.” International Journal of Production Economics 171: 471–477.
Moussavi, S. E., M. Mahdjoub, and O. Grunder. 2016. “Reducing Production Cycle Time by Ergonomic Workforce Scheduling.” IFAC-
PapersOnLine 49 (12): 419–424.
Nanthavanij, S., S. Yaoyuenyong, and C. Jeenanunta. 2010. “Heuristic Approach to Workforce Scheduling with Combined Safety and
Productivity Objective.” International Journal of Industrial Engineering: Theory Applications and Practice 17 (4): 319–333.
Neubert, G., and M. M. Savino. 2009. “Flow Shop Operator Scheduling Through Constraint Satisfaction and Constraint Optimisation
Techniques.” International Journal of Productivity and Quality Management 4 (5–6): 549–568.
Noack, D., and O. Rose. 2008. “A Simulation-Based Optimization Algorithm for Slack Reduction and Workforce Scheduling.” Simulation
Conference, 1989–1994.
Othman, M., G. J. Gouw, and N. Bhuiyan. 2012. “Workforce Scheduling: A New Model Incorporating Human Factors.” Journal of
Industrial Engineering and Management 5: 259–284.
Pinheiro, R. L., H. Algethami, and D. Landa-Silva. 2016. “A Genetic Algorithm for a Workforce Scheduling and Routing Problem.” In
IEEE Congress on Evolutionary Computation (CEC), 927–934.
Quan, G., G. Greenwood, D. Liu, and S. Hu. 2007. “Searching for Multi-Objective Preventive Maintenance Schedules: Combining
Preferences with Evolutionary Algorithms.” European Journal of Operational Research 177: 1969–1984.
Ritt, M., A. M. Costa, and C. Miralles. 2015. “The Assembly Line Worker Assignment and Balancing Problem with Stochastic Worker
Availability.” International Journal of Production Research 54 (3): 907–922.
Robbins, T. R., D. J. Medeiros, and T. P. Harrison. 2010. “Cross Training in Call Centers with Uncertain Arrivals and Global Service
Level Agreements.” International Journal of Operations and Quantitative Management 16 (3): 307–329.
Sabar, M., B. Montreuil, and M. M. Frayret. 2009. “A Multi-Agent-Based Approach for Personnel Scheduling in Assembly Centers.”
Engineering Applications of Artificial Intelligence 22 (7): 1080–1088.
Safaei, N., D. Banjevic, and A. Jardine. 2009. “Multi-Objective Maintenance Workforce Scheduling in a Steel Company.” IFAC
Proceedings Volumes 42 (4): 1049–1054.
Savino, M. M., D. Battini, and C. Riccio. 2017. “Visual Management and Artificial Intelligence Integrated in a New Fuzzy-Based Full
Body Postural Assessment.” Computers & Industrial Engineering 111: 596–608.
International Journal of Production Research 19

Savino, M. M., A. Brun, and A. Mazza. 2014. “Dynamic Workforce Allocation in a Constrained Flow Shop with Multi-Agent System.”
Computers in Industry 65 (6): 967–975.
Savino, M. M., A. Mazza, and D. Battini. 2016. “New Easy to Use Postural Assessment Method Through Visual Management.”
International Journal of Industrial Ergonomics 53: 48–58.
Thannimalai, P., M. M. Kadhum, C. Feng, and S. Ramadass. 2013. “A Glimpse of Cross Training Models and Workforce Scheduling
Optimization.” IEEE Symposium on Computers & Informatics (ISCI), 98–103.
Thompson, G., and J. Goodale. 2006. “Variable Employee Productivity in Workforce Scheduling.” European Journal of Operational
Research 170 (2): 376–390.
Urlings, T., R. Ruiz, and F. S. Serifoglu. 2010. “Genetic Algorithms with Different Representation Schemes for Complex Hybrid Flexible
Flow Line Problems.” International Journal of Metaheuristics 1 (1): 30–54.
Valls, V., A. Pérez, and S. Quintanilla. 2009. “Skilled Workforce Scheduling in Service Centres.” European Journal of Operational
Research 193 (3): 791–804.
Vignais, N., M. Miezal, G. Bleser, K. Mura, D. Gorecky, and F. Marin. 2013. “Innovative System for Real-time Ergonomic Feedback in
Industrial Manufacturing.” Applied Ergonomics 44 (4): 566–574.
Wongwien, T., and S. Nanthavanij. 2012. “Ergonomic Workforce Scheduling Under Complex Worker Limitation and Task Requirements:
Mathematical Model and Approximation Procedure.” Songklanakarin Journal of Science and Technology 34 (5): 541–549.
Wongwien, T., and S. Nanthavanij. 2016. “Priority-Based Ergonomic Workforce Scheduling for Industrial Workers Performing Hazardous
Jobs.” Journal of Industrial and Production Engineering 34 (1): 52–60.
Wongwien, T., and S. Nanthavanij. 2017. “Multi-Objective Ergonomic Workforce Scheduling Under Complex Worker and Task
Constraints.” International Journal of Industrial Engineering: Theory Applications and Practice 24 (3): 284–294.
Yazgan, H. R., I. Beypinar, S. Boran, and C. Ocak. 2011. “A New Algorithm and Multi-Response Taguchi Method to Solve Line Balancing
Problem in an Automotive Industry.” International Journal of Advanced Manufacturing Technology 57 (1): 379–392.
Zhang, G., L. Gao, and Y. Shi. 2011. “An Effective Genetic Algorithm for the Flexible Job-Shop Scheduling Problem.” Expert Systems
with Applications 38 (4): 3563–3573.

Appendix. Terms and notations used within the paper


WS Workforce scheduling
WESM Workforce Ergonomic Scheduling Model
TES Total Ergonomic Stress
COP Constraint Optimisation Problem
es Ergonomic Stress
OES Overall Ergonomic Score
EO Elementary Operations
PES Partial Ergonomic Score
n Number of domains
N
i=1 PESi Sum of each score obtained for a single elementary operation;
NEI Number of ergonomic indicators.
EOi i Elementary Operation
OESN Overall Ergonomic Score Normalised
ESM Ergonomic Stress Matrix
wkj j worker [j = 1 . . . M ]
opi i operation [i = 1 . . . N]
wk
oesopij OESN of wkj for the opi
wk
zopij (t) Execution time of wkj for the opi
TST Total Shift Time
tS Number of time slots in the shift
TSI wkj Total Stress Index for each wkj
wk = {1, . . . , wkM } Number of available workers
op = {1, . . . , opN } Number of operations
t = {0, . . . , tS } Workday length, in S time slots
op (t)
xwk Operator’s state
popi (t) Pieces worked within opi at time t
wk (t)
zop Time to conduct opi by the wkj
tuop Cycle time of the workstation
N Total number of operations of the line
popN (tS ) Number of pieces worked at the last operation
TESh Constraint value of Total Ergonomic Stress
P Positioning
LH Leverage handle
PR Pressing

You might also like