Proceedigs of the 15th IFAC Symposium on

Proceedigs of the 15th IFAC Symposium on

Information of
Proceedigs Control Problems
the 15th in Manufacturing
IFAC Symposium on
Available online at www.sciencedirect.com
Information
Proceedigs Control
of Problems
theOttawa,
15th IFAC in Manufacturing
Symposium on
Information
May Control
11-13, 2015. Problems in Manufacturing
Canada
May 11-13, 2015.
Information Ottawa,
Control Canada
Problems in Manufacturing
May 11-13, 2015. Ottawa, Canada
May 11-13, 2015. Ottawa, Canada
ScienceDirect
IFAC-PapersOnLine 48-3 (2015) 586–591
Assembly
Assembly line balancing with ergonomics paradigms: two alternative methods
Assembly line
line balancing
balancing with
with ergonomics
ergonomics paradigms:
paradigms: two
two alternative
alternative methods
methods
Assembly line balancing with ergonomics paradigms: two alternative methods
D. Battini*,
D. Battini*, X. Delorme**,
Delorme**, A. Dolgui**,Dolgui**, F. Sgarbossa*Sgarbossa*
D. Battini*, X. X. Delorme**, A. A. Dolgui**, F. F. Sgarbossa*
D. Battini*, X. Delorme**, A. Dolgui**,StradellaF. Sgarbossa*
*Department
*Department of of Management
Management and and Engineering,
Engineering, UniversityUniversity of of Padova,
Padova, Stradella San San Nicola,
Nicola, 3 3 36100 Vicenza –
36100 Vicenza – Italy
Italy
*Department of Management and Engineering, University of Padova, Stradella San Nicola, 3 36100 Vicenza – Italy
**École*Department
Nationale of Management
Supérieure des and Engineering,
Mines, FAYOL-EMSE, University
CNRS, ofUMR
Padova,6158, Stradella
LIMOS, San
158 Nicola,
cours 3 36100 42023
Fauriel, Vicenza – Italy
Saint-Étienne
**École Nationale Supérieure des Mines, FAYOL-EMSE, CNRS, UMR 6158, LIMOS, 158 cours Fauriel, 42023 Saint-Étienne
**École Nationale Supérieure des Mines, FAYOL-EMSE, CNRS, UMR 6158, LIMOS, 158 cours Fauriel, 42023 Saint-Étienne
**École Nationale Supérieure des Mines, FAYOL-EMSE, CedexCNRS, UMR 6158, LIMOS, 158 cours Fauriel, 42023 Saint-Étienne
2, France
Cedex
Cedex 2, 2, France
France
Cedex 2, France
Abstract: In this paper, the assembly line balancing problem
Abstract: In this paper, the assembly line balancing problem when when ergonomics
ergonomics principles principles are are taken
taken intointo
Abstract:
account is In this paper,
considered. Twothedifferent
assemblyapproaches
line balancing are problem when
presented and ergonomics
discussed with principles
a numerical areexample
taken into to
account
Abstract:
account is
is In this paper,
considered.
considered. Two
Twothedifferent
assemblyapproaches
different line balancing
approaches are
are problem when
presented
presented and
and ergonomics
discussed
discussed with
with principles
aa numerical
numerical areexample
taken into
example to
toa
provide
account new insights
is considered. on Twothe field and to illustrate their use on a didactic example. The first one applies
provide
provide new
new
multi-objective
insights
insights
modelon
on thedifferent
the
based
field
field andapproaches
on andthe to
to illustrate
illustrate
energy
are presented
their
their
expenditure,
use
use on
on
used
and discussed
aa didactic
didactic
to estimate
estimate
with aThe
example.
example. numerical
The
the ergonomics
first
first one
ergonomics one example
applies
applies
level.
toa
Theaa
provide new
multi-objective
multi-objective insights
model on
model based the
based field
on and
on thethe to illustrate
energy their
expenditure,
energy expenditure, use on
used a didactic
to example.the The first one applies
level. The
second one one transforms
multi-objective
second transforms
model based the energy
the energy
on the expenditure rate in
energy expenditure,
expenditure rate restused
in aa rest timeto
used
time toininestimate
order to
estimate
order tothe reduce
the
reduce
ergonomics
ergonomics
the
level. The
the multi-objective
multi-objective
level. The
second
problem one transforms
to aatransforms
single objectivethe
objective energy expenditure
one. expenditure
The two two methodsrate
methods in a
are rest time
alternatively in order to
usabletoin reduce
inreduce
practice the multi-objective
and
second one
problem to single the energy one. The rate inare a rest time in order
alternatively usable practice theand aa numerical
numerical
multi-objective
problem
model to a single objective
demonstrates their one. The two methods are alternatively usable in practice and a numerical
foundation.
problem
model to a single objective
demonstrates one. The two methods are alternatively usable in practice and a numerical
their foundation.
model demonstrates their foundation.
© 2015,demonstrates
model
Keywords: IFAC (International
Assembly their Federation
foundation.
Balancing Problem,of Automatic Control)Energy
Ergonomics, HostingExpenditure,
by Elsevier Ltd. RestAllAllowance,
rights reserved. Pareto
Keywords: Assembly Balancing Problem, Ergonomics, Energy Expenditure, Rest Allowance, Pareto
Keywords:
Frontiers. Assembly Balancing Problem, Ergonomics, Energy Expenditure, Rest Allowance, Pareto
Keywords: Assembly Balancing Problem, Ergonomics, Energy Expenditure, Rest Allowance, Pareto
Frontiers.
Frontiers.
Frontiers.
European
European Assembly
Assembly WorkSheet WorkSheet (EAWS), (EAWS), which which grants
grants loadload
1.
1. INTRODUCTION European Assembly WorkSheet (EAWS), which grants load
1. INTRODUCTION
INTRODUCTION points
European for unfavourable
points for Assembly
unfavourable WorkSheetphysical
physical(EAWS), workload
workloadwhich and due
due to
and grants the
toload
the
In many industrial 1. INTRODUCTION
sectors, the production of high value and points
total for
score unfavourable
assigns a physical
traffic light workload
risk and
scheme due to to the
work
In many industrial sectors, the production of high value and total pointsscore
for unfavourable
assigns a traffic physical light workload
risk scheme and dueto towork the
In many industrial
customized products sectors,
is even the production
today only of high
feasible withvalue
a and total
large score(Schaub
situations assignset aal,traffic 2013). light risk scheme
However, the to work
observational
In many industrial
customized productssectors,is eventhe production
today of high
only feasible withvalue and total
a large score(Schaub
situations assignset aal,traffic 2013).light risk scheme
However, to work
the observational
customized products is evenskilled
today workforce.
only feasibleMoreover, with a large situations
methods (Schaub etEAWS
al, 2013). However, the observational
involvement
customized
involvementproductsof
of aa highly is evenskilled
highly today workforce.
only feasibleMoreover, with a large in situationsapplied
in methods (Schaub
applied by
by etEAWS
al, 2013).and
and
and the
the other
However,
other studies in the
the observational
studies in the field
field
involvement
many countriesof athehighly skilledpresence
increased workforce. of Moreover,
highly in methods
skilled are subject applied
to the by
usualEAWS limitations theofother
this studies
category, infirst
the of
field
all
involvement
many
many countries
countriesof athe
the highly skilled
increased
increased workforce.
presence
presence of
of Moreover,
highly
highly in
skilled
skilled methods
are
are subject
subject applied
to
to the
the by
usual
usualEAWS and
limitations
limitations theof
ofother
this
this studies
category,
category, in the
first
first field
of
of all
all
elderly workers not easily substitutable with robots or large
are
large amount
subject
amount to of
the
of time
usual
time required
limitations
required to
to reach
of this
reach a
a precise
category,
precise estimation
first of
estimation all
many
elderly
elderly countries
workers
workers the
not
not increased
easily
easily presence
substitutable
substitutable of highly
with
with skilled
robots
robots or
or large amount of time required to reach a precise estimation
machines requires the development of new strategies capable and
large
and strong
amount
strong dependence
of
dependencetime on
required
on the
the analyst
to reach
analyst subjective
a precise
subjective perception
estimation
perception
elderly
machines
machines workers
requires
requires not
the easily
development
the development substitutable
of new
of new with
strategies
strategies robots
capable
capableor
and
duringstrong the dependence
observation on the phaseanalyst and subjective
input perception
to
to integrate
machines
integrate assembly
requires
assembly system
system design
the development design with
of new ergonomics
withstrategies
ergonomics and
and during
capable and strong thedependence
observation on the phaseanalyst input parameters
andsubjective perception
parameters
to integrate assembly system design with ergonomics and during
evaluation. the observation phase and input parameters
safety
to science.
integrate
safety science. In
In this
assembly context,
this system aa poor
context, design poor withergonomic
ergonomics
ergonomic design
designand of during theIn
of evaluation.
2014,
2014, the
In observation the authors phasehave
authors haveandfirstly input
firstly introduced
parameters
introduced the
the
safety science.
assembly and In this context,systems
manufacturing a poor can ergonomic
generate design
a of evaluation.
large ergo-balancing In 2014,
approach the by authors
applying havethe firstly
energy introduced
expenditure the
safety science.
assembly In this context,systems
and manufacturing a poor can ergonomic
generatedesign a largeof evaluation.
ergo-balancing In 2014,
approach the by authors
applying havethe firstly
energy introduced
expenditure the
assembly
amount ofand
sick manufacturing
leaves due systems
to can generatedisorders
musculo-skeletal ergo-balancingmethod
a large computation approachand by applying
a the energyoptimization
multi-objective expenditure
assemblyofand
amount sickmanufacturing
leaves due to systems can generatedisorders
musculo-skeletal ergo-balancingmethod
a large computation approachand by applying the energyoptimization
a multi-objective expenditure
amount of
(MSDs). sickit leaves
Thus, is due to musculo-skeletal disorders computation
approach (Battini methodet al, and a The
2014). multi-objective
novelty of this optimization
approach is
amount of
(MSDs). is even
sickit leaves
Thus, even due more
moretonecessary to
to design
musculo-skeletal
necessary efficient
designdisorders computation
efficient approach
approach (Battini
(Battini methodet al, and a The
al, 2014).
etintegration
2014). multi-objective
The novelty
novelty of of onthisoptimization
approach
thisALBP
approach is
is
(MSDs). Thus,
assembly systems it isand even more necessary
enhance work to design
related efficient
satisfaction of the analysis of the of ergonomics based
(MSDs).
assembly
assembly Thus,
systems
systems it isand
andeven more
enhance
enhance necessary
work
work to
related
related design efficient
satisfaction
satisfaction of
of approach
the
the analysis
analysis (Battini
of
of the
the et al, 2014).
integration
integration The
of
of novelty
ergonomics
ergonomics of this
on
on approach
ALBP
ALBP based
based is
human operators. As also highlighted in the new European on
the
on energy
analysis
energy expenditure
of the
expenditure with
integration
with two
of
two alternative
ergonomics
alternative methods.
on ALBP
methods. Using
based
Using
assembly
human systems
operators.
human operators. and
As
As also enhance
also work
highlighted
highlighted related
in the
inofthe satisfaction
new
new European
European of on energy expenditure with two simplifies
alternative the methods. Using
Horizon 2020 Program (i.e. the Factory the Future calls), energy
on energy
energy expenditure
expenditure
expenditure formulations
with two simplifies
formulations alternative the problem
methods.
problem and
Using
and
human operators.
Horizon
Horizon 2020
2020 Program
ProgramAs also (i.e.
(i.e.highlighted
the
the Factory
Factory inof
ofthe
the
the new
Future
Future European
calls),
calls),
this will require new thinking both on scheduling of work and energy
speed up expenditure
the solving formulations
process (Battinisimplifies
et al, the
2104). problem
Moreover and it
Horizon
this 2020 Program
will require new thinking (i.e. theboth Factory of the Future
on scheduling of work and speed
calls), energyupexpenditure
the solvingformulations
process (Battini simplifies
et al, 2104). the problem
Moreover and it
this
design will ofrequire
attractivenew thinking
and safe both on
workplaces, scheduling
taking of
into work and
account speed
permits up the
to solving
convert process
the (Battini
ergonomics et al, 2104).
evaluation Moreover
into restit
this willofrequire
design attractivenewand thinking both on scheduling
safe workplaces, taking into of workaccount speed up to
and permits the convert
solving process (Battini et al,
the ergonomics 2104). Moreover
evaluation into rest it
design
the of attractive
ageing workforce andneeds.
safe workplaces,
In fact, in taking
the Europeaninto account
Union permits
allowance to(Rohmert,
convert 1973), the ergonomics
thus reducing evaluation
the into rest
multi-objective
design
the of attractive
ageing workforce andneeds.
safe workplaces,
In fact, in the taking
Europeaninto account permits to(Rohmert,
Union allowance convert 1973), the ergonomics
thus reducing evaluation into rest
the multi-objective
the ageing workforce needs. In diseases
fact, in the European Union allowance
problem (Rohmert, 1973), thus reducing the multi-objective
(EU),
the
(EU), over
over 38%
ageing of
of occupational
workforce
38% needs. In diseases
occupational fact, in the are related
related to
are European to MSDs problem to
Union allowance
MSDs
aa single-objective
to(Rohmert, 1973), thus
single-objective one.
one. In
In this
reducing
this thenew work,
work, we
newmulti-objective we aimaim
(EU),
(EHSAW over2010)
38% of withoccupational
a cost up diseases
to 2% of are
the related
Gross to MSDs to
National problem
discuss to by a single-objective
a methodological one.
pointIn of
thisviewnewtwo work, we aim
alternative
(EU), over2010)
(EHSAW 38% of withoccupational
a cost up to diseases
2% ofare therelated
Gross to MSDs to
National problem
to discuss
discuss to bybya single-objective
aa methodological
methodological one.
point
pointIn of
thisview
of newtwo
view work,
two we aim
alternative
alternative
(EHSAW
Product in 2010)
the withBevan
EU. a cost(2012)
up to 2% of the Gross
estimates more National
than 200 methods applicable in order to obtain an optimized assembly
(EHSAW
Product
Product in 2010)
the
infor withBevan
EU.
thedirect
EU. a cost(2012)
Bevan up to 2%
(2012) of the Gross
estimates
estimates more
more Thus, National
than
than 200
200 to discuss
methods
methods by a methodological
applicable
applicable in
in order
order to point of
to obtain
obtain an
an view two alternative
optimized
optimized assembly
assembly
billion € and indirect costs of MSDs. the system
methods
system by
by taking
applicable
taking into
in
into consideration
order to obtain
consideration anboth
both ergonomics
optimized
ergonomics assemblyand
and
Product
billion
billion €
€in the
for
for EU.
direct
direct Bevan
and
and (2012)
indirect
indirect estimates
costs
costs of
of MSDs.
MSDs.more than
Thus,
Thus, 200
the
the system by taking into consideration both ergonomics and
ergonomics of the system needs to be deeply analyzed and productivity
system
productivityby paradigms.
taking
paradigms.into The
consideration
The first
first one
both
one applies
ergonomics
applies a
a multi-
and
multi-
billion
ergonomics
ergonomics € forof
of direct
the
the and
system
system indirect
needs
needs costs
to
to be
be of MSDs.
deeply
deeply Thus,
analyzed
analyzed the
and
and productivity
objective paradigms.
optimization The first one applies a multi-
coupled
ergonomics
coupled with
withofthe
thethetime
time and
and methods
system needs toanalysis
methods be deeply
analysis in
in order
analyzed
order to
to obtain productivity
and objective
obtain paradigms.approach
optimization The first
approach in
in order
one to
order distinguish
applies
to a multi-
distinguish the
the
coupled
aacoupled
final with the time
assembly system and really
methods analysis in
performant in order
the to obtain
short and objective optimization
function linked approach
with thein assigned
order to work distinguish
load the
(the
with the time
final assembly system and really
methods analysis in
performant inorder and objective optimization
to obtain
the short function linked approach
with the in assigned
order to work distinguish
load (thethe
a final assembly
medium term. system
Recent really of
research performant
the authors in (Battini
the shortet and al., objective
time functionfrom
dimension) linked
the with
objective the assigned
function work load
linked with (the
the
a final assembly
medium term. Recent system really of
research performant
the authors in (Battini
the shortet and al., objective
time dimension) functionfrom linked
the with
objective the assigned
function linked work load with (the
the
medium
2011) hasterm. Recent research
demonstrated a linkof between
the authors (Battini et and
productivity al., time dimension)
ergo-quality level from
of the workstations
the objective function (thus linked
the with the
ergonomics
mediumhas
2011) term. Recent research
demonstrated a linkof between
the authors (Battini et and
productivity al., time dimension)
ergo-quality levelfromof the the workstations
objective function (thus the linked with the
ergonomics
2011) has demonstrated
ergonomics in a link between productivity and ergo-quality
dimension). level of the workstations
Multi-objective optimization (thus
has the
been ergonomics
applied in
in assembly
2011) has demonstrated
ergonomics assembly systems. a link between
systems. In
In particular, some
some recent
productivity
particular, recent ergo-quality
and dimension).
dimension). level of the workstations
Multi-objective
Multi-objective optimization
optimization (thus
has the
has beenergonomics
been applied
applied in
in
ergonomics
studies are in assembly
closely related systems.
and In particular, tosome
complementary our recent
work: many fields of science, where optimal decisions need to be
ergonomics
studies
studies are
are in assembly
closely
closely related
related systems.
and
and In particular,
complementary
complementary some
to
to our
our recent
work:
work: dimension).
many
many fields
fields Multi-objective
of
of science,
science, where
where optimization
optimal
optimal has
decisions
decisions been applied
need
need to
to be
in
be
Kazmierczak et al. (2007), Otto and Scholl (2011), Bautista taken
taken in
in the
the presence
presence of
of trade-offs
trade-offs between
between two
two or
or more
to
morebe
studies
Kazmierczak
Kazmierczak are closely
et al. related
(2007),
etCheshmehgaz
al. (2007), Otto and
Otto complementary
and Scholl
and(2012)
Scholl and to
(2011),
(2011), our work:
Bautista
Bautista many fields of science, where optimal decisions need
et al. (2013), et al. Otto (2014) taken in theobjectives.
conflicting presence of trade-offs ergonomics
Minimizing between tworisk or while
more
Kazmierczak
et al. (2013), etCheshmehgaz
al. (2007), Otto et and(2012)
al. Scholl and (2011),Otto Bautista
(2014) taken in theobjectives.
conflicting presence of trade-offs ergonomics
Minimizing between tworisk or while
more
et al. (2013), Cheshmehgaz et al. (2012) and Otto (2014) conflicting
maximizing objectives.
productivity Minimizing
is ergonomics risk while
introduce
et al. (2013),
introduce new constraints
constraints and
new Cheshmehgaz andet define
al. (2012)
define ergonomics-objective
and Otto (2014) maximizing
ergonomics-objective conflicting objectives.productivity is possible
Minimizing possible if
if we
ergonomics we involverisk while
involve the
the
introduce in
functions newthe constraints
ALBP and definesupporting
formulations, ergonomics-objective
the analysis maximizing
necessary input productivity
data right is possible
from the if we ofinvolve
beginning the systemthe
introduce in
functions newtheconstraints
ALBP formulations, and definesupporting the analysis necessary
ergonomics-objective maximizing input productivity
data right from is possible
the beginning if we ofinvolve the system the
functions
with in the ALBP
traditional formulations,evaluations
ergonomics supporting the analysis necessary
techniques design
necessary phase.inputIndata
inputIndata
right problems,
practical from the beginning
right problems, there
from the beginning can of
be the
more system
than
functions
with in the ALBP
traditional formulations,evaluations
ergonomics supporting the analysis design
techniques phase. practical there can of be the more system
than
with traditional ergonomics evaluations techniques design
two phase. InBy
objectives. practical
keeping problems,
the therefunctions
can be more than
(observational
with traditional
(observational methods).
methods). A
A large
ergonomics large consensus
evaluations
consensus is
is also
also given
given in
techniques in two design phase. InBy
objectives. practical the objective
keepingproblems, objectivetherefunctions
can be more separated
than
separated
(observational
the automotive methods).
industry A large consensus
community on the is also given
application of the two objectives.
in during the By keeping
analysis it’s the objective
possible to functions
define separated
how an
(observational
the automotive methods).industry communityA large consensus is also given
on the application of the two
in during objectives.
the By
analysis keeping it’s the objective
possible to functions
define separated
how an
the automotive industry community on the application of the during the analysis it’s possible to define how an
the automotive industry community on the application of the during the analysis it’s possible to define how an
2405-8963 © 2015, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Copyright © 2015 IFAC 618
Peer review©under
Copyright 2015 responsibility
IFAC of International Federation of Automatic
618Control.
Copyright © 2015 IFAC 618
10.1016/j.ifacol.2015.06.145
Copyright © 2015 IFAC 618
INCOM 2015
D. Battini et al. / IFAC-PapersOnLine 48-3 (2015) 586–591 587
May 11-13, 2015. Ottawa, Canada

improvement in one of them could affect the others and
viceversa. The second method here discussed transforms the
original multi-objective problem into a simpler single
objective optimization algorithm by transforming the energy
expenditure amount during assembly task into a rest time in
order to express the balancing and sequencing problems in a
time-unit base. The two methodological approaches works
well in practical environments and can be applied
alternatively in accordance to the input data available. While
the first one could often bring to a set of different feasible
solutions and drive the designer towards a unique final
solution only after an interactive analysis, the second one
tries to simplify the problem to a unique objective function.
2. METHODS DESCRIPTION
2.1 Method n°1: multi-objective problem based on energy
expenditure
The first approach proposed by the author applies a multi-
objective problem with two objective functions: the solution
is not a single optimum but instead it is represented by the set
of balancing designs belongings to the Pareto frontier. The
methodology could be easily extended to different kind of
ALBP models and different objective functions, according to
Scholl, A. (1999). The procedure is made up by 4 steps:
Step 1) Calculate the mean working time for each task by a
times and method analysis technique.
systems based on oxygen or metabolic consumption, where
each movement features specific energy expenditure. Such
approach seems to be very useful in this context because it
quickly calculates the ergonomics level of different assembly
tasks without leave the evaluation to the subjective analyst
interpretation. According to Garg et al. (1978) the net
metabolic energy expenditure is influenced by: gender, body
weight, load weight, vertical heights of lifting/lowering,
lateral movements of arms in horizontal plane, speed of
walking and carrying load, postures and time duration of the
task. To help an easy and fast computation of the tasks’
energy values a set of parametric tables have been recently
provided by the authors in Battini et al. (2014). Thus, it is
possible to define a generic SALBP-2 model that aims to
jointly minimize two objective functions, the smoothness
index in time and the smoothness index in energy expenditure
between the different work stations. Based on the binary
linear model (Scholl, 1999) in the single-model assembly
line, the binary variable called x jk is used to indicate the
assignment and Bk is the set of task assignable to the station
k (within a set of workstations from 1 to K). We introduce the
first objective function expressed in time, the Time
Smoothness Index ( SX  T ) to measure the equality of
work distribution among the stations:
2
N
 
(1)
m in S X  T  m in   c r   x jk  t j 

k 1  j B k 

Step 2) Calculate the energy expenditure to be assigned to

each task by using formulas provided by Garg et al. (1978)
and parametric maps provided by Battini et al. (2014).
where c r  m ax  j B k
x jk  t j | k  1...m ,  Ej is the

earliest station of task j and L j is the latest station of task j

Step 3) Express the same objective function in time and
(Scholl, 1999), with the following constraints:
energy, thus defining two different objective functions.

Step 4) Solve the bi-objective problem by defining and  k   E j , L j 

x jk  1 for j = 1…n (2)
mapping in a graph all the feasible solutions and the
balancing designs belonging to the Pareto frontier.
 j B k
x jk  t j  Tk  c for k = 1…K (3)
In order to give an example of the methodology we here
consider the so-called SALBP-2 method, which allows the
minimization of the cycle time for a defined number of  k   E h , Lh 
k  x hk   i  E j , L j 
i  x ji
stations in order to increase the productivity of the assembly
line when constrained by a trial cycle time c , known and for ( h , j )  A and L h  E j (4)
fixed (Scholl, 1999). In the example following reported an
assembly process consisting of n = 17 tasks denoted A, B,..,Q The second objective function expressed in energy, the
is used. This example represents a real case regarding small Energy Smoothness Index ( SX  E ), in order to reduce
cleaning equipment. Some simplifications, such as tasks risks among the stations distributing physical load among
grouping, have been made without any loss of information. workers at a similar low level (as underlined also by Otto and
Let’s consider a set A of tasks to be assigned to work stations, Scholl, 2011), is defined as follows (with the same
the assembly times ( t j ) and the direct predecessors of the constraints as the previous one):
tasks are given. For each task j also the energy expenditure (
2
e j ) is computed and coupled with a task node in the N
 
(5)
m in S X  E  m in   E r   x jk ej 

k 1  
precedence diagram in Figure 2. The e j values are computed j B k

using the formulas provided the well-known study of Garg et

al (1978), by using the analytical ergonomic measurement where E r  m ax  j B k
x jk  e j | k  1 ...m 
619
INCOM 2015
588 D. Battini et al. / IFAC-PapersOnLine 48-3 (2015) 586–591
May 11-13, 2015. Ottawa, Canada

2.2 Method n°2: single-objective approach based on Rest N

  cˆ 
2

Allowances m in SX  Tˆ  m in  Tˆk (9)

r
k 1

This second approach applies a simple mathematical

formulation with a single objective function in order to
permit an easy and quick implementation. The method
where cˆ r  m ax  j B k 
Tˆk | k  1 ...m , and constraints:

applies the formulas introduced by Rohmert (1973) and it can

be easily used to solve small-size problems, since it takes into
account both the task time value and the rest time necessary  k   E j , L j 
x jk  1 for j = 1…n (10)
to the operator at each cycle time. The procedure is made up
by 4 steps :
 j B k
Tˆk  c for k = 1…K (11)
Step 1) Calculate the mean working time for each task by a
times and method analysis technique.
 k   E h , Lh 
k  x hk   i  E j , L j 
i  x ji
Step 2) Calculate the energy expenditure to be assigned to for ( h , j )  A and L h  E j (12)
each task by using formulas provided by Garg et al (1978).

Step 3) Transform task energy expenditure rate of each 3. NUMERICAL APPLICATIONS AND DISCUSSION
operator into rest time by using formulas provided by
Rohmert (1973). In the numerical example following reported an assembly
process consisting of 17 tasks denoted T1, T2,..,T17 is used.
Step 4) Minimize the single objective function expressed in
The assembly times ( t j ) and the energy expenditure rate
time and define the optimal balancing solution.
( e j ) associated to each task are given in Figure 1 and
Each operator k has the following energy expenditure rate
precedence relations in Figure 2.
index:

 x jk  e j
j B k
Ek  (6)
 x jk  t j
j B k

The rest allowance will be calculated using the following

equation introduced by Rohmert (1973):
Fig. 1. Input data used in the numerical applications
 E  
1.4

R Ak  1.9 T k  0.145
  k  1  
(7)
  E std  

if E k  E std ; 0 otherw ise

Equation 7 requires that all time functions are expressed in

minutes and the energy expenditure in kcal. The E std was
assumed equal to 2.7 kcal/min, which is reasonable if we
consider about 2 kcal/min for a normal standing activity Fig. 2. Precedence diagram in the numerical example
(Garg et al., 1978). Consequently, the total working time per
each operator per cycle Tˆ including also the rest allowance,
k
By applying the method n°1 introduced in the previous
paragraph it is possible to find all the feasible balancing
becomes as follows:
solutions according to the technical constraints provided with
a number of fixed station K = 4. The Pareto solutions that
Tˆk  Tk 1  RAk  (8) belong to the Pareto frontier (Pareto, 1971) and the solution
that minimize the first objective function (here the time
The single objective function will be expressed in time by smoothness index) and the one that minimize the second
including the rest allowance portion and the following new objective function (here the energy smoothness index), are
depicted in Figure 3 and Figure 4.
smoothness index ( SX  Tˆ ) can be defined and the problem
solution is a single optimum:

620
INCOM 2015
D. Battini et al. / IFAC-PapersOnLine 48-3 (2015) 586–591
May 11-13, 2015. Ottawa, Canada
Fig. 3. All feasible balancing solutions and Pareto solutions
by applying method n°1, with K = 4.
As shown in the graphs, time and energy functions yield
different balancing solutions: the optimal balancing decision
depends on both time smoothness and energy smoothness and
how these two objective functions evolve over the different
feasible balancing solutions.
Fig. 4. Pareto solutions and optimal solutions of the two
objective functions, with K = 4.
As shown in Figure 4, more we reduce the smoothness index
in energy (i.e. fostering the ergo-quality level of the
balancing solution) more we increase the smoothness index
in time. A small decrease in energy smoothness is achievable
only with a higher increase in time smoothness, since the
shape of the Pareto frontier is steep in the example here
provided. Of course the shape of the Pareto frontier is strictly
dependent on the numerical case under study and no
generalizations can be done. When it is steep after the time-
optimal solution, moving from right to left on the frontier
will bring to a decrease in the energy smoothness, but to a
rapid increase in the time smoothness, since the time-function
increases even faster. Otherwise, when the frontier is flat
around the time-optimal solution, with a smaller increase in
time smoothness, we could obtain a more relevant decrease in
energy smoothness. However, the method n°1 allows the
decision maker to quickly identify, in an interactive way, a
subset of Pareto optimal solutions that finally helps to reach a
single balancing solution that satisfies the subjective
preferences of one or more human decision makers
(Miettinen et al, 2008).
621
589
Fig. 5. Pareto solutions according to variations in the number
of workstations K
Figure 5 shows that the Pareto solutions do not depend on the
number of stations used as input parameter in the SALBP-2
model, in fact, the Pareto frontier shape only depends on task
times and task energy values, thus on the particular numerical
case under study. The method n°2 consists in four steps that
allow the analyst to simplify the multi-objective approach
into a single objective model capable to derive a single
optimal solution by using the concept of “rest time” in
addition to the “task time” by applying the rest allowances
formulas developed by Rohmert (1973).
In order to better compare the two methodologies here
provided and understand the potentiality of the both, we here
introduce the Energy-Time ratio parameter ET as the rate
between the total energy expenditure per product unit and the
total working time necessary to assembly the entire product
in the assembly system, expressed in Kcal/min:
 j
ej
(13)
ET 
tj
 j 60
In the numerical example in Figure 1 the ET is equal to 2.47.
Starting from this value, we have varied the numerical
example in order to obtain different values of the ET index,
in particular four different numerical cases with ET= [2.71;
2.47; 2.22; 1.97]. For each case, following the same
precedence diagram and fixing the number of stations K = 4,
it has been possible to apply the method n°1 and method n°2
in order to compare their results. In Figures 6, 7, 8 and 9 the
Pareto optimal solutions for method n°1 and the single
optimal solution for method n°2 are graphically compared.
As shown in the following graphs, the Pareto frontier shape is
not affected by changes in the ET parameters but only its
position in the graph is moving from right to left when the ET
index decreases. This is obviously a consequence of the
reduction in energy respect to total assembly time. The Pareto
frontier shape is still deep and derives from the specific input
data used and on task time and energy variability. Moreover
when comparing the multi-objective method with the single-
INCOM 2015
590 D. Battini et al. / IFAC-PapersOnLine 48-3 (2015) 586–591
May 11-13, 2015. Ottawa, Canada

objective method, it is easy to understand that with higher

value of the ET index (2.71 and 2.47) the optimal solution
obtained by the method n°2 is quite overlapping with the one
of the optimal solutions that minimize the energy smoothing
according to the method n°1.In particular the two solutions
are perfectly matching when ET is equal to 2.71 that is the
highest value. Otherwise, when the ET index decreases, the
optimal solution derived by the method n°2 moves in the
graph from left to right by finally reaching a position quite
close to the time-optimal solution of the method n°1. In
particular, when the ET index is equal to 1.97 the optimal
solution with the method n°2 find about the same value of the
time smoothness index but a higher value of the energy
smoothness index, and this solution is not Pareto optimal
with the method n°1 . That means, in other words, that if you Fig. 8 Pareto solutions for method n°1 and single optimal
apply the method n°2 in presence of medium/high values of solution for method n°2 with a medium/low value of the ET
the ET index you will find an optimal solution that will index.
benefit the energy smoothing rather than the smoothing in
time, on the contrary with lower value of the ET index you
will find and optimal solution that will benefit only the time
smoothness rather than the energy smoothness. With ET
values ranging from 2.22 until 2.47, the optimal solution with
the method n°2 will be positioned on the middle, offering a
balancing between time and energy smoothing.

Fig. 9 Pareto solutions for method n°1 and single optimal

solution for method n°2 with a low value of the ET index.

4. CONCLUSION

Even if today we can benefit from a large and highly

Fig. 6. Pareto solutions for method n°1 and single optimal
solution for method n°2 with a high value of the ET index.
Fig. 7. Pareto solutions for method n°1 and single optimal
solution for method n°2 with a medium/high value of the ET
index.
connected literature in assembly system design and
balancing, the ergonomics issue still needs further
investigations. In this paper the authors adapted a traditional
SALBP-2 model in order to include ergonomics paradigms
into the assembly system balancing problem. Two methods
are here discussed and applied to a numerical case. The first
one apply a bi-objective optimization problem in which time
and energy objective functions are jointly investigated and
optimized thanks to the well-known Pareto theory. This
approach is based on the Energy expenditure rates
computation derived by Garg et al. (1978). Then, the authors
propose and apply an alternative method with a single
objective function and rest allowances computations in order
to transform the spent energy rate into a rest time associated
to each station. The two method here described and applied
to a simple numerical case demonstrate to be promising tools
in order to balance an assembly system not only productive in
the short term but also sustainable by the workforce in a long
term basis. Future researches will focus on the substitution of
the equations introduced by Garg et al. (1978) by new simply
equations based on muscles activities measured by
electromyography sensors. Several tests are needed to collect
a significant number of cases in order to define new

622
INCOM 2015
D. Battini et al. / IFAC-PapersOnLine 48-3 (2015) 586–591 591
May 11-13, 2015. Ottawa, Canada

formulations. Moreover, further applications to other Schaub K., Caragnano, G., Britzke B., Bruder R. (2013). The
instances with different features will be carried out to analyze European Assembly Worksheet, Theoretical Issues in
the robustness of these approaches. Ergonomics Science, Volume 14, Issue 6, pp. 616-639.

REFERENCES
Appendix A. NUMERICAL EXAMPLE
Battini, D., Faccio, M., Persona, A., & Sgarbossa, F. (2011).
Here below, we report the application of method n°2 for the
New methodological framework to improve productivity
numerical example introduced in section 3, with K = 4
and ergonomics in assembly system design. International
stations. We calculate the smoothness indices with equations
Journal of Industrial Ergonomics, 41(1), 30-42.
(1), (6) and (10), with the related optimal balancing solutions
Battini D., Persona A., Sgarbossa F., (2014), Ergo-Balancing
as reported in the following tables.
in Assembly Lines based on Energy Expenditure Rate,
In: IFORS 2014 - 20th Conference of the International
Federation of Operational Research Societies. Barcelona, Table 1. Time-based optimal solution
Spain, 13-18 July 2014.
Assigned
Bautista Valhondo, J., Batalla García, C., Alfaro Pozo, R., & k
Task
Tk Ek Ek R Ak Tˆk
Cano Pérez, A. (2013). Extended models for TSALBP
T1, T2, T3,
with ergonomic risk constraints. IFAC Conference on
1 T4, T7, T8, 144 7.06 2.9 18.95 162.95
Manufacturing Modelling, Management, and Control. T9
"Preprints (CD)". Saint Petersburg: 2013, p. 869-874. T5, T6, T10,
Bevan, S. (2012). The Impact of Back Pain on Sickness 2 145 7.63 3.2 47.05 192.05
T12
Absence in Europe. The Work Foundation, Lancaster. 3 T11, T14 150 6.25 2.5 0 150
European Agency for Safety and Health at Work (EASHW) T13, T15,
4 146 3.13 1.3 0 146
(2010). OSH in figures: Work-related musculoskeletal T16, T17
disorders in the EU - Facts and figures. DOI: SX  T SX  E SX  Tˆ
10.2802/10952. 8.77 4.74 68.81
Garg, Arun, Don B. Chaffin, and Gary D. Herrin. "Prediction
of metabolic rates for manual materials handling jobs."
The American Industrial Hygiene Association Journal
39.8 (1978): 661-674. Table 2. Energy-based optimal solution
Kazmierczak, K., Winkel, J., & Westgaard, R. H. (2004). Car Assigned
disassembly and ergonomics in Sweden: current situation k
Task
Tk Ek Ek R Ak Tˆk
and future perspectives in light of new environmental T1, T2, T3,
legislation. International Journal of Production Research, 1 137 6.88 3.0 25.97 162.97
T8, T9, T10
42(7), 1305-1324. T4, T5, T6,
Miettinen, K., Ruiz, F., Wierzbicki, P., 2008. Introduction to 2 127 6.80 3.2 47.81 174.81
T10
multiobjective optimization, interactive approaches. In: T11, T12,
3 170 7.15 2.5 0 170
Branke, J. et al. (Eds.), Multiobjective Optimization. T13,
Springer, Heidelberg. T14, T15,
4 151 3.22 1.3 0 151
Otto, A., & Scholl, A. (2011). Incorporating ergonomic risks T16, T17
into assembly line balancing. European Journal of SX  T SX  E SX  Tˆ
Operational Research, 212(2), 277-286. 57.44 3.96 27.03
Otto, A. (2014). Minimizing Risks for Health at Assembly
Lines. In Operations Research Proceedings 2013 (pp.
341-346). Springer International Publishing.
Rajabalipour Cheshmehgaz, H., Haron, H., Kazemipour, F., Table 3. Time with rest allowance-based optimal solution
& Desa, M. I. (2012). Accumulated risk of body postures Assigned
in assembly line balancing problem and modeling k
Task
Tk Ek Ek R Ak Tˆk
through a multi-criteria fuzzy-genetic algorithm. T1, T2, T3,
Computers & Industrial Engineering, 63(2), 503-512. 1 123 6.60 3.2 46.98 169.98
T5, T6
Pareto, V., 1971. Manuale di economia politica, società T4, T7, T8,
2 141 7.08 3.0 26.79 167.79
editrice libraria. Milano, Italy: MacMillan Press (the first T9, T10
edition in 1906), (translated by A. S. Schwier as Manual T11, T12,
3 170 7.15 2.5 0 170
of Political Economy) T13
Rohmert, W. (1973). Problems of determination of rest T14, T15,
4 151 3.22 1.3 0 151
allowances Part 2: Determining rest allowances in T16, T17
different human tasks. Applied Ergonomics, 4(3), 158- SX  T SX  E SX  Tˆ
162. 58.40 3.97 19.13
Scholl, A. (1999). Balancing and sequencing of assembly
lines. Heidelberg: Physica-Verlag.

623