Professional Documents
Culture Documents
Department of Mechanical Engineering, Polytechnic University of Madrid, Jos Gutirrez Abascal, 2, 28006 Madrid, Spain
Department of PLM, Process and Tool Solutions, AIRBUS Group, 41020 Sevilla, Spain
a r t i c l e
i n f o
Article history:
Received 13 November 2015
Received in revised form 12 April 2016
Accepted 21 April 2016
Keywords:
Assembly process design
Assembly line denition
Process evaluation
a b s t r a c t
The decision-making process during the conceptual design of aircraft nal assembly lines requires of
software applications to assist industrialization engineers. This work proposes a method, a supporting
model and ad-hoc software development to address that issue. The proposed method divides the problem
of the assembly process denition into four parts: logistic plan denition, assembly line denition, layout
design and evaluation of complete solutions. The proposed approach is implemented in a prototype
software application that is integrated within a commercial software system widely used in the aerospace
sector. Feasible design solutions, which are created by the user interactively, are evaluated and ranked
with a Fuzzy Logic based multi criteria utility. A case study shows the feasibility of the proposed approach.
2016 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
1. Introduction
Literature shows the relevance of the assembly process within
the aircraft lifecycle. In aircraft manufacturing projects, the assembly has a big impact on the production volume and cost, where
the latter is estimated as nearly as the 30% of the total cost
[14]. Ultimately, the execution of assembly processes, manually
or automatically, requires work instructions and/or robot programs. To be able to generate them, designers work collaboratively
along the functional and industrial design process [510]. Collaborative working procedures are developed to integrate teams and
processes. Although, still there are issues related to project management and software tools capabilities and interoperability, the aim
is to create a common industrial Digital Mock Up (iDMU) [512].
The denition of the assembly process can be structured into
three main top-level tasks: create conceptual assembly process,
dene assembly process and develop detailed assembly process [4].
The latter task deals with the denition of the basic assembly tasks,
the documentation to be used by the assembly personnel and the
programs to be used in the automatic machines. The execution of
the two prior top-level tasks is necessary to dene the lowest level
assembly information. Precisely, the create conceptual assembly
process task denes the scope of this work.
http://dx.doi.org/10.1016/j.jmsy.2016.04.002
0278-6125/ 2016 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
38
[16,19]. Assembly operations are performed manually in a high percentage [24], so the number of assigned worker and their hourly
cost are key factors in manufacturing process costing [29,30]. In
addition, a FAL involves parts and subassemblies with large dimensions, which requires a careful distribution of spaces within the
assembly line facilities [10,11].
Fig. 1 shows the method proposed in this work. Four main
tasks are considered to create an assembly process conceptual
design solution: denition of the product assembly structure (or
as-planned structure), denition of the FAL, denition of the FAL
layout and denition of the logistic plan. The last three activities
must be completed following an iterative loop, as it is shown in
Fig. 1. This ensures the feasibility of the solutions, since the different
parts of the problem are coupled. A FAL conguration implies some
resources requirements that have an impact on the FAL layout. The
FAL layout determines space requirements has to be fullled by any
available plant to be part of the network that denes the logistic
plan. On the other hand, the logistic plan represents a big percentage of the total cost [3134]. Therefore, it can be optimized in rst
place, and distributing the industrial workload into a set of plants.
Then, a FAL conguration and layout can be dened considering the
available surface of such plants as space constraints.
A feasible conceptual solution for the FAL is dened each time
the tasks within the loop are completed at least once. Then, the
solution can be evaluated using a multi-criteria Fuzzy Logic-based
system. Several solutions, or variants of a single one, can be generated by executing the previously mentioned tasks. Then, they can
be compared by the evaluation system to nd the best one in terms
of costs, time and efciency in resources usage. Section 4 details
the proposed method and its implementation.
To support the proposed method, an object-oriented Product,
Process, Resources (PPR) meta-model was required. The created
model (see Fig. 2) takes as antecedent, a prior model created to support industrialization conceptual design [4]. This antecedent PPR
model was modied to fulll the requirements of the proposed
method and to be implemented in the commercial software system CATIA/DELMIA V5. Rules were dened to estimate costs and
times parameters. As input initial data, a library of resources and
processes is required for each assembly design specic case. Such a
library is called processes and resources dictionary (PRD). Section
2 details the created models.
There are four main topics involved in this work. The rst topic
is assembly planning. Assembly planning deals with the denition
of the product structure showing the main sub-assemblies to be
built and the execution sequence. The second topic is assembly line
39
Fig. 2. PPR meta-model with Products (up and left part), and Resources and Processes (right and down part).
of the class MfgAttribute. This approach is necessary for the metaclasses to support that properties may differ from one activity or
resource type to another. The unit of measurement of the attributes
(e.g. D /h, Tm) are also included in the meta-model. The following
two sections describe the main parts of the PPR information model.
2.1.1. Product
The part of the model related to Product supports information
about individual components, industrial nodes and the join of components. The class PhysicalProduct represents a component. The
class Joint has two components of the class PhysicalProduct and
an attribute related to the type of joint. The class SubAssembly
represents components set in groups. Each dened subassembly
is manufactured in a separated assembly line, which is represented
by an IndustrialNode object. The IndustrialNode class supports the
as-planned product structure.
The Joint class is modeled to support the as-prepared product structure, where the precedence in the execution of joints is
specied. Each Joint object is linked to some others, one Next,
one Previous and a number of Simultaneous. These associations
40
Table 1
Specic simple and compound activities from a processes and resources dictionary (PRD).
ID
Type
Var.
Req. res.s
Skills
Att.s
Involved operations
Op. type
Op. var.
Subject
R1
Riveting
Manual
Worker
LS
M. oper.
Rivet applier
LS
Auto CNC
machine
Cost
Time
U cost
U time
Nb elem.s
Cost
Worker
Rob. oper.
Time
...
Cost
Drilling
Open
N
Y
2
12
13
Generic
Riveting
Drilling
Manual
Open
Open
Y
Y
Y
14
15
1
Counter-boring
Riveting
Pos. at jig
Open
Auto
Base comp.
Pos. at jig
Added comp.
Assembling
CC
10
Pos. at jig
Added comp.
Auto
R2
MJ3
Assembling
CC
Time
U cost
U time
JT4
Join-up
FF
Specic jig
Cost
Time
U cost
U time
...
Assemble
...
Retire from jigs
Heaviest
(H)
Not H
Set
Table 2
Resources types from a PRD.
ID
Type
Assignment
Addition
Attributes (name-type-unit)
Lifter
Activity
D/A
Worker
Activity
D/A
Capacity
Cost rate
Assignment to activity
Name
Skills()
Speciality
Mean speed
Cost rate
Assignment to activity
Name
Real
Real
Real
String
Enumeration
String
Real
Real
Real
String
Tm
D /h
step/s
D /h
(VAR), RequiredSkills (SKILLS) and Subject (SUBJECT) are properties of this kind. Other properties are dened as objects of the class
MfgAttribute. For instance, Addition (A), for controlling how adding
new resources affect to the activity: Linear at speed (LS) when
more resources nish the job faster, Linear at capacity (LC) when
they sum their capacity or Not (N) when only one resource of that
type can be assigned to the activity. The property is Separable (S)
denes if a child activity can be separated from its parent activity.
The property Priority (P) denes the precedence order of an activitys children. The Attributes (ATT.s) column contains properties
that are specic to each PRD.
Conceptually, Resources are modeled in a similar way to Processes. Table 2 shows two examples of resource types (lifter and
worker). The column ASSIGNMENT represents the attribute that
determines the kind of activities to which those resources can be
Table 3
A resource model in a PRD.
ID
Model
Type (s)
Attributes (name-value-type-unit)
TJ5
SinCrane5X11
Name
Capacity
Mean speed
Cost
Life expectancy
CAD path
String
Real
Real
Real
Integer
C:\Doc. . .
Tm
m/s
D
days
41
Table 4
A human specialization in a PRD.
ID
Specialization
Skills
Attributes (name-value-type-unit)
SA
Systems assembler
Wage
Extra hours wage
Mean speed
Mean speed (electric systems verier)
14
20
1
0.7
Real
Real
Real
Real
D /h
D /h
1/s
1/s
Table 5
Nomenclature of the equations.
General nomenclature:
TC is cycle time, dened as the time elapsed between two consecutive
assembling products being processed at a same step.
TT is total time, dened as the time to complete the whole considered
assembly process
TNS is the total number of stations within the process
Time equations nomenclature:
NCi is the number of children of station or composed activity i.
Ti is the time, or length, of process element i (e.g. industrial node i, station i,
operation i,. . .).
TMi is the time measured for the type of a given simple operation i under
controlled conditions (i.e. for one element using a mandatory resource with
known speed)
NEi is the number of elements (e.g. number of holes, rivets to perform) of a
given simple operation i.
STi is the speed of the mandatory resource employed to measure the type of
activity i time.
SR1 is the speed of the mandatory resource nally assigned to activity i.
NP1i is the number of activity is children with priority 1 (i.e. without any
precedence constraint with other child)
NHi is the number of shifts set for the station which includes activity i.
TEj is time to end, dened as the time from the beginning of a given child
activity j to its father activity is completed.
NASj is the number of activities included in a precedence sequence after
activity j.
BNSk is the number of stations in the main branch, or maximum number of
station and transport steps of the process branches until the completion of
the operations within industrial node k.
NPrk is the number of direct precedent nodes of industrial node k.
NSkl is the number of stations to be completed within node k after the
entrance of the output product of node l.
NoSl is the number of stops set for the transport of output products from
node l.
Costs equations nomenclature:
C is the total cost of the considered assembly process.
CnoT is the total cost excluding the transport of products between nodes.
Ci is the cost of industrial node i.
CoTi is the cost of the transport from industrial node i to its subsequent node.
NSi is the number of stations within industrial node i.
CAj is the cost of station or activity j.
NRk is the number of resources directly assigned to activity k (or to activities
within node k).
CHr is the hourly cost of resource r.
Wr is the wage of human resource r.
WEr is the wage for special shifts of human resource r.
CSr is the remaining amortization total value of resource r.
ELr is the remaining life expectancy of resource r.
Efciency equations nomenclature:
E is the overall efciency.
Er is the efciency associated to resource r.
Tar is the time of an activity a, which resource r is assigned to.
Uncertainty equations nomenclature:
UTT is the uncertainty in the total process time estimation.
UTC is the uncertainty in the cycle time estimation.
UTi is the uncertainty in an activity i time estimation.
UTEj is the uncertainty in an activity j time to end of its father estimation.
UC is the uncertainty in the process cost estimation.
UUCi is the uncertainty in a process element i cost estimation.
42
the PRD, and it is adjusted with the mean speed of the key assigned
resource (i.e. the rst resource in the denition at the PRD); the
number of elements involved and the numbers of work shifts of the
workstation. Time of workstations or composed activities is given
by their critical path time, which is calculated using the variable
time to end, dened as the time from the beginning of an activity to
the end of its father activity (Eq. (3)). The time of a composed activity is the maximum value of time to end between its child activities
with Priority 1.
TC = MAXiTNS T (i)
Ti =
(1)
MAXp
TEj =
(TEp ),
NASj = 0
NASj
Tj + MAXn
f =FAL
(TEn ),
(3)
NASj > 0
) TC
MAXlNPrk (NSkl
(2)
NCi > 0
Tj ,
TT = ( BNSf
BNSk =
NCi = 0
+ BNSl + NoSl + 1)
(4)
(5)
(6)
Cost is another essential aspect in Assembly processes denition
and evaluation. A brief review of costs estimations approaches is
provided to facilitate a better understanding of the way followed
to include costing in the proposed model.
There are two main perspectives for cost modeling and estimation: based on production volume and based on activities (ABC)
[29]. While production based models separates costs in variable
costs, proportional to the production volume, xed costs and overhead cost; ABC assigns cost to activities, based on the cost driver of
each activity, which requires estimating the usage of the resources
assigned [29]. Different research works dealing with assembly
alternatives evaluation adopted the ABC approach.
Amen [30] considers the resources hourly cost and assigns
resources with similar hourly cost to the same workstations. The
objective is to minimize the assembly line cost by minimizing the
cost of inactive times. Baud et al. [31] propose an ABC model to
evaluate solutions for the logistic strategy. Cost of transport activities is calculated using the products physic volume as a resource
cost driver, and processing time is used as cost driver for assembly
operations. Kampker et al. [32] and Tang et al. [33] include the Net
Present Value concept in their models, to evaluate, respectively, the
assembly process and the logistic chain. They both use ABC models,
but they also take into consideration the investments required at
each solution.
Some other works also implement simplied ABC models,
where cost is calculated as the sum of the hourly cost of the
resources assigned to every workstation multiplied by the cycle
time and the resource assignment percentage [25,27,34].
The proposed cost model was designed following an activitybased approach. It considers activities executed for every product
unit produced. This approach allows calculating the total cost per
day and per assembled aircraft. It also allows future improvements,
such as implementing learning curves and including additional
activity types.
The cost of the whole assembly process is estimated as the sum
of all the industrial nodes costs (Eq. (7)). The cost of each industrial
node is calculated by two different ways. Eq. (8) includes costs of
the involved workstations and the transport to the next node. The
cost of a workstation or an activity is calculated by multiplying the
directly assigned resources hourly cost by the cycle time, plus the
costs of the children activities (Eq. (9)). A resource hourly cost is
given by Eq. (10), considering the expectancy of life, the number of
shifts that each resource is assigned to and the different wage for
special shifts in the case of humans.
That way to estimate an industrial node cost is used when the
denition of its involved assembly line is still in progress, so the
designer may have an idea of the implications of his decisions (e.g.,
assigning a resource instead of another one). However, the obtained
value is inaccurate, since the sharing of resources between activities
and workstations is not included in the calculations at this point in
time. The nal set of resources units is dened at later steps and
then the cost is calculated again.
The second way estimates an industrial node cost by using costs
of the actually assigned resources, as Eq. (11) sets. This way can be
followed only when the nal resources conguration is achieved.
C=
NI
i=1
Ci
Ci = CoTi +
CAk =
NRk
CHr =
r=1
j=1
CAj
(8)
TC CHr +
NCk
l=1
CAl
Wr + (NH 1) WEr ,
human resource
NH CSr /ELr ,
other case
Ci = CoTi +
CnoT =
(7)
NSi
NRi
r=1
TC CHr
NI NRi
i=1
r=1
(9)
(10)
(11)
TC CHr
(12)
NAr
Er =
E=
E=
T
a=1 ra
TC
NR
r=1
(13)
NR
[Er CHr ]/
NR NAr
Tar CHr
r=1 a=1
CnoT
r=1
CHr
(14)
(15)
and costs computing. Eqs. (16)(19) are used to calculate uncertainty in times, according to the propagation rule.
A single activity time uncertainty requires values from the measured activities time list included in the PRD (Eq. (16) if the number
of children NCi is zero). For composed activities, a recursive procedure is used to calculate the cumulative uncertainty of the critical
path length (Eqs. (16) and (17)). Uncertainty in cycle time is considered as the maximum time uncertainty between the workstations
(Eq. (18)). Eq. (19) gives uncertainty in total time. Uncertainty
in transport times is not considered in the model, as the logistic
network is dened to t the cycle time resulting of dening the
assembly line processes.
UTi =
MAXp
UTEj =
UTj2 + UTEk2 ,
UTC = UTl ,
UTT =
(UTEp ),
l = s,
BNSf
f =FAL
NCi = 0
(16)
NCi > 0
MAXsNAS (UTEs )
k = s,
(17)
MAXsNS (Ts )
(18)
UTC
(19)
NI
UC =
i=1
UCi2
(20)
Ci = CoTi + TC C
(21)
UCi =
UCi =
CoTi2 + C2 UTC 2
C CoT
i
i
TC
(22)
UTC
(23)
UCj = UTC
NRj
r=1
(CHr PSrj )2
(24)
Eq. (15) is needed to obtain the equation to calculate the uncertainty in the total efciency. Applying the propagation rule, and
considering that a resource hourly cost uncertainty is null, then Eq.
(25) is obtained.
UE =
2
[E UC] +
r=1 a=1
NR
NAr
[UTar ]2 [CHr ]
CnoT
(25)
43
44
45
Fig. 4. Conceptual ow diagram, process structures created and Computer Aided views of the assembly line denition.
Fig. 5 shows the GUI form designed for this stage of the method.
The left list box shows the parts and sub-assemblies that have to
be joined. Joints can be moved up and down within the sequence.
The joint information box shows the intermediate products that
Fig. 5. Graphical user interface to dene the assembly sequence or as-prepared product structure.
46
47
After this initial automatic step, activities having resources pending of allocation are displayed with the tag NAY!! (for Not
Assigned Yet) and activities without dened variant display the tag
Gen!!.
Fig. 7 shows the GUI designed to assist the designer with the
variant selection and the resources assignment. The list in the left
side allows exploring the process structure. For every activity, the
variant can be selected, and the resource models or specializations
assigned from one list. Properties of activities and resources can be
checked to evaluate different scenarios.
The designer can automatically assign the best resource, the best
activity conguration (i.e. the set of assigned resources and the variant) or the best conguration of the whole process, by clicking on
the three respective buttons. The criteria that denes best can be
switched from time to cost with the buttons (D and t) located in
the left side below the activity list.
The dened conguration is checked when Apply is clicked. As
a result of the checking, when different assigned resource models
are found, for activities that could share such a resource (i.e. the
activities are in the same workstation and not overlapped in time),
the designer is asked to conrm if a single model can be selected
and shared. The dened rule is: choose the cheapest resource
model that fullls the capacity requirements. The FAL conguration created by the designer with the assistance of the developed
application is stored in a CATIA/DELMIA V5 le of type CATProcess.
The designer can generate several FAL conguration alternatives
and each one is stored in a CATProcess le.
Table 6
Rules table of the FL based system.
Time
Cost
Eff.
A
A
A
A
A
B
9
8
D
D
C
C
B
C
6
1
E
E
E
E
D
E
1
1
48
the interval [ai ,ei ] for the inputs of all the evaluated alternatives,
for a better differentiation between them.
domain of each input parameter is dened by ve domain parameters (i.e. a, b, c, d and e), as there are ve sets of triangular geometry
(Fig. 8). Sets A and E are different, with a constant membership
value of 1 for input values smaller than a or greater than e. Since
the best efciency is the highest one, the distribution of the fuzzy
sets along the input axis changes for parameter efciency, following the sequence: E, D, C, B, A; instead of the opposite sequence used
for cost and time.
Sabc (x) =
(c x)/(c b),
x [b, c]
0,
et = Ncyc0 tcycobj
(27)
at = 40%et
(28)
ec = cos tMAX
(29)
ac = 30%et
(30)
eeff = 5
(31)
aeff = 50
(32)
Usually, single deterministic values are used as input to FL systems, and memberships are given by mathematical functions, as the
one expressed in Eq. (26). However, due to the statistical nature of
most of the parameters used in this work, where the parameter
has an expected nominal value and a variance, the membership to
a fuzzy set Sabc is calculated by using Eq. (33), where Sabc is the
conventional membership expression of Eq. (26). E and are the
expected value and variance, and p is the probability density function, as Gaussian, given by Eq. (34). Using its deterministic value,
an input parameter could be member at the most of two fuzzy sets,
but considering its statistical distribution, it could belong to more
fuzzy sets (Fig. 9).
(26)
Every set has the same amplitude, so two parameters are enough
to dene a domain. Eqs. (27)(32) dene standard values for the
extreme parameters of the domains time, cost and efciency (i.e.
et , at , ec , ac , eeff and aeff , respectively). tcycobj and costMAX are initial
data, specic of each assembly case. The latter refers the maximum
allowed value of cost. tcycobj represent the objective cycle time for
an adequate production rate. Ncyc0 refers a reference value for the
total number of workstations within the process, set as if there was
one only workstation for each joint.
The value of domain parameters can be changed to improve the
rating accuracy. The extreme parameters et and ec will always adopt
the values given by Eqs. (27) and (29), but at and ac can be modied
for a better adjustment. Both parameters aeff and eeff may also vary.
It is highly recommended that every mean value Xi is included in
S
abc
(E, ) =
a
1
p(E, , x) = ( 2) e(xE/)
(33)
(34)
Fig. 9. Membership functions for parameter time and Gaussian distribution (E,) = (40, 2.5) h, being a and e 19.2 h and 48 h.
49
Fig. 10. Output fuzzy sets and resulting joint membership function.
parameters to the sets involved in the rule. For the rule described
above, if time, cost and eff. had membership values of 0.7, 0.8 and
0.3 to their fuzzy sets A, A and B, the rule would be activated with
a degree of 0.3.
Since there are three input parameters with ve sets, 125 rules
are required, but some simplications can be done. For instance,
time belonging to the worst fuzzy set always implies the worst
output set, 1, as it represents the delivery time, which would be
exceeded. That denes 25 rules at one time. The nal score is
ranked between 0 and 10 (Fig. 10). The nal joint membership
function, usually called Defuzzy function, is built by using the
Sup-Min composition principle [39]. The nal score is calculated
5. Case study
A case study was developed to prove rstly the effectiveness of
the proposed method. Secondly, the execution of the case study
allows validating the developed prototype application. The prototype application aims assisting the industrialization designer in
the decision-making process to dene aircraft FAL conceptual solutions. The following subsections illustrate the case study.
As planned
Fuselage
LW
RW
V_Stab
LW
RW
WB
H__Stab
As prepare
Co
omplex Assembly n1
LW
RW
WB
RFBL
RFBU
RFF
CFL
CFU
FFL
FFU
FFB
VSU
VSR
VSF
VSM
HSRL
HSRR
HSRF
HSRM
HSLL
HSLR
HSLF
HSLM
HSCB
H_Stab
Compllex Assembly n2
Fuselage
RFBL
RFBU
RFF
CFL
CFU
FFL
FFU
FFB
VSU
VSR
VSF
VSM
Fig. 11. As Planned view and As Prepared view of the nal industrial node, dened for the case study.
WB
HSRL
HSRR
HSRF
HSRM
HSLL
HSLR
HSLF
HSLM
HSCB
V_Stab
50
reduce the workstation cycle time. The cheapest resources congurations were adopted for both alternatives. The layouts were
designed using the developed GUI. Additionally, standard DELMIA
V5 functions were used to generate Gantt diagrams. Fig. 12 shows
the Gantt diagrams and the 3D representations of the studied
alternatives.
A second study was executed to evaluate the impact of selecting
automated or manual variants for the drilling activities. Table 10
shows the properties of both compared alternatives; such data
(i.e. resource mean speed, resource cost and worker cost) are
included in the case PRD. The next section presents and discusses
the obtained results.
6. Results and discussion
As time related hypothesis, the FAL is synchronous, consequently, all the workstations have the same value as cycle time
target. The target for the workstation cycle time is equal to the
takt time. The FAL cycle time is the sum of the cycle times of all
the workstations. Every workstation that is precedent of another
workstation includes a transportation operation as last task.
In the rst part of the case study, from the results (Table 7), it
can be observed that including a third workstation causes a reduction on the workstation cycle time. Consequently, the takt time
would be also reduced. Fig. 12 shows both Gantt diagrams, with the
workstations highlighted for each alternative. Obviously, an extra
workstation has an impact in the resources needed, some resources
has to be duplicated, some new jigs might be required and the FAL
51
Table 7
Results for rst studied alternatives.
Alternative
Surf. (m2 )
Cost (D /cycle)
UCost
tcyc
tproc
2 Workstations
3 Workstations
6052
11 996
171 018.63
234 087.20
10 162D
13 453D
10.07 h
8.53 h
20.14
25.59
UTime
EffT
2.52 h
0.5 h
18.4%
28.6%
Table 8
Activated fuzzy rules function for the rst alternative of the rst study.
A
A
A
A
A
A
A
A
A
A
A
A
A
A
B
B
B
B
B
B
B
B
A
A
A
A
A
B
B
B
B
C
C
C
C
C
A
A
A
A
A
B
B
B
A
B
C
D
E
A
B
D
E
A
B
C
D
E
A
B
C
D
E
A
B
C
9
8
6
6
5
8
7
6
5
7
7
6
5
4
9
8
7
5
5
8
7
7
0.001
0.039
0.039
0.039
0.039
0.001
0.064
0.094
0.054
0.001
0.064
0.094
0.094
0.054
0.001
0.039
0.039
0.039
0.039
0.001
0.064
0.454
B
B
B
B
B
B
B
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
B
B
C
C
C
C
C
A
A
A
A
A
B
B
B
B
B
C
C
C
C
C
D
E
A
B
C
D
E
A
B
C
D
E
A
B
C
D
E
A
B
C
D
E
5
5
7
7
5
5
4
7
6
5
5
4
6
5
5
4
3
5
5
4
4
3
0.427
0.054
0.001
0.064
0.172
0.172
0.054
0.001
0.039
0.039
0.039
0.039
0.001
0.064
0.108
0.108
0.054
0.001
0.064
0.108
0.108
0.054
Table 9
Final scores obtained with different evaluation parameters congurations.
Conguration
Parameters
Alternative
Final score
et = 50 h; ec = 300 000D
at = 0.2et ; ac = 0.2ec
aeff = 0.5; eeff = 0.05
et = 50 h; ec = 300 000D
at = 0.3et ; ac = 0.3 ec
aeff = 0.4; eeff = 0.05
et = 50 h; ec = 300 000D
at = 0.4et ; ac = 0.4 ec
aeff = 0.3; eeff = 0.05
et = 50 h; ec = 300 000D
at = 0.5et ; ac = 0.5ec
aeff = 0.2; eeff = 0.05
et = 50 h; ec = 300 000D
at = 0.5et ; ac = 0.5ec
aeff = 0.5; eeff = 0.05
et = 50 h; ec = 300 000D
at = 0.5et ; ac = 0.7ec
aeff = 0.5; eeff = 0.05
et = 50 h; ec = 300 000D
at = 0.2et ; ac = 0.4ec
aeff = 0.4; eeff = 0.05
2 Workstations
5.1192
3 Workstations
2 Workstations
4.2108
5.8986
3 Workstations
2 Workstations
5.0679
6.2770
3 Workstations
2 Workstations
5.8938
7.9756
3 Workstations
2 Workstations
6.6820
5.8552
3 Workstations
2 Workstations
5.1599
5.8784
3 Workstations
2 Workstations
6.2545
5.7975
3 Workstations
4.993
7
Fig. 13. Membership to fuzzy set of process parameters for the rst alternative of
the study.
52
Table 10
Results for the second study.
Alternative
Res. model/Specialization
Attributes
Driller
HDr200
Str.assembler
Auto drilling machine
ADM80
Fig. 14. Output joint membership function for the rst alternative of the rst case.
0.5
3200
280
12
2.5
8650
150
14.5
Cost (D /cycle)
Final score
57 975
171 540.27
57 923
53
[22] Dmoly F, Dutartre O, Yan X-T, Eynard B, Kiritsis D, Gomes S. Product relationships management enabler for concurrent engineering and product lifecycle
management. Comput Ind 2013;64:83348.
[23] Wanga H, Ronga Y, Xiang D. Mechanical assembly planning using ant colony
optimization. Comput Aided Des 2014;47:5971.
[24] Morato C, Kaipa KN, Gupta SK. Improving assembly precedence constraint generation by utilizing motion planning and part interaction clusters. Comput
Aided Des 2013;45:134964.
[25] Battaa O, Dolgui A. A taxonomy of Line Balancing problems and their solution
approaches. Int J Prod Econ 2013;142:25977.
[26] Rekiek B, Dolgui A, Delchambre A, Bratcu A. State of art of optimization methods
for assembly line design. Annu Rev Control 2002;26:16374.
[27] Manavizadeh N, Tavakoli L, Rabbani M, Jolai F. A multi-objective mixedmodel assembly line sequencing problem in order to minimize total costs
in a Make-To-Order environment, considering order priority. J Manuf Syst
2013;32:12437.
[28] Keshavarzmanesh S, Wang L, Feng H-Y. A hybrid approach for dynamic
routing planning in an automated assembly shop. Robot CIM-Int Manuf
2010;26:76877.
[29] Shuford Jr RH. Activity-based costing and traditional cost allocation structures.
In: Cost estimators reference manual. New York: Johan Wiley & Sons Inc; 1995.
p. 4194.
[30] Amen M. Cost-oriented assembly line balancing: model formulations, solution
difculty, upper and lower bounds. Eur J Oper Res 2006;168:74770.
[31] Baud-Lavigne B, Agard B, Penz B. Mutual impacts of product standardization
and supply chain design. Int J Prod Econ 2012;135:5060.
[32] Kampker A, Burggra P, Wesch-Potente C, Petersohn G, Krunke M. Life cycle oriented evaluation of exibility in investment decisions for automated assembly
systems. CIRP-JMST 2013;6:27480.
[33] Tang ZE, Goetschalck M, McGinnisb L. Modeling-based design of strategic supply chain networks for aircraft manufacturing. Proc Comput Sci
2013;16:61120.
[34] Moncayo-Martnez LA, Recio GG. Bi-criterion optimisation for conguring an
assembly supply chain using Pareto ant colony meta-heuristic. J Manuf Syst
2014;33:18895.
[35] Velasquez M, Hester PT. An analysis of multi-criteria decision making methods.
IJOR 2013;10:5666.
[36] Azadegan A, Porobic L, Ghazinoory S, Samouei P, Kheirkhah AS. Fuzzy logic in
manufacturing: a review of literature and a specialized application. Int J Prod
Econ 2011;132:25870.
[37] Devs GM, Ribeiro LC. A fuzzy logic approach to supply chain performance
management. Int J Prod Econ 2011;134:17787.
[38] Kumara D, Singh J, Singh OP. A fuzzy logic based decision support system for
evaluation of suppliers in supply chain management practices. Math Comput
Model 2013;58:167995.
[39] Reyero R, Nicols CF. Sistemas de control basados en lgica borrosa: fuzzy
control. Madrid: Omron Electronics; 1995.