Professional Documents
Culture Documents
A Fuzzy Multi-Attribute Approach To Select The Welding Process at High Pressure Vessel Manufacturing
A Fuzzy Multi-Attribute Approach To Select The Welding Process at High Pressure Vessel Manufacturing
Technical Paper
a r t i c l e i n f o a b s t r a c t
Article history: Metal welding process selection calls for extensive know-how which deals with a huge amount of knowl-
Received 18 June 2011 edge. Hence, the automation of knowledge through a knowledge-based system will greatly enhance the
Received in revised form 11 October 2011 decision-making process. Fuzzy analytical hierarchy process (AHP)-TOPSIS investigates the relative weld-
Accepted 12 October 2011
ing process selection factors and it can compare and evaluate different welding processes between 0 and
Available online 27 December 2011
1.
This paper describes a knowledge-base system developed for identifying the most appropriate welding
Keywords:
processes to suit specific circumstances. In the present study, nine important welding processes were
High Pressure Vessel
Fuzzy Analytical Hierarchy Process
used in the system. Ten parameters namely alloy class, material thickness, deposition rate, design appli-
TOPSIS cation, joint configuration, operator factor, capital cost, equipment portability, filler metal utilization and
welding position, are used to determine the best selection among competitive welding processes for high
pressure vessel production. Also the sensitivity analysis was carried out for five cases.
© 2011 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
1526-6125/$ – see front matter © 2011 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.jmapro.2011.10.006
M. Jafarian, S.E. Vahdat / Journal of Manufacturing Processes 14 (2012) 250–256 251
Table 1
Nine welding processes used in this research.
SAW Submerged arc welding GMAW Gas metal arc welding OFW Oxy fuel welding
PAW Plasma arc welding FCAW Flux-cored arc welding EBW Electron beam welding
GTAW Gas tungsten arc welding SMAW Shielded metal arc welding LBW Laser beam welding
Table 2
The guideline to select a fusion welding process [1].
Design application
Primary structural B B B B A A C A A
Secondary structure A A A B B B B B B
Noncritical A A A B C C B B B
Dissimilar metal B B B C A A C A A
Joint configuration
Butt A A A A A A A A A
Tee A A A B B B B B B
Edge B B B C A A B A A
Corner B B B C B B B C C
Lap
Welding positions A A A C A A B C B
Equipment portability 3 3 3 3 3 3 4 1 1
Capital cost 1 2 2 3 2 2 1 4 4
Operator factor 1 3 3 4 2 2 2 4 4
Deposition rate 2 3 3 4 1 1 1 1 1
Filter metal utilization 1 2 3 4 4 4 3 4 4
Thickness of parts
0.03–0.5 D D D D B B D A A
0.5–1.3 C C B D A A B A A
1.3–2.5 B C B D A A B A A
2.5–6.4 B B A C A A B A A
6.4–12.7 A A A B B B B A B
12.7–25 A A A B C C B A C
25–64 A A A A C C C A C
64–over A A A A C C C A D
Thick to thin B B B C A A B A A
Alloy class
Plain low carbon steel A A A A B B B C B
Low carbon steel B B B B B B C B B
High strength steel B B B B B B D B C
Stainless steel 300 B B B B A A C A A
Aluminum D B B D A A C A C
p
where x denotes the element of the set X, Ã (x) is called the mem- Ã1 a1 am ao
= o , 1m , 1p (7)
bership function or the degree of membership of x in à that maps Ã2 a2 a2 a
2
X to the membership space [0, 1].
Definition 2. If supx à (x) = 1, the fuzzy set à is called normal.
Definition 3. The membership function C̃ (x) of the intersection Definition 6. Given two fuzzy numbers Ã1 and Ã2 , the truth value
C̃ = Ã ∩ B̃ is defined by: of the fuzzy number Ã1 is greater than or equal to the fuzzy number
Ã2 expressed as in the following:
C̃ (x) = min{Ã (x), B̃ (x)}, x∈X (2)
n
n
n
à (x) and à (y) (see Fig. 2).
1 2 M̃i = ãij ⊗ ⎣ ãij ⎦ (10)
j=1 i=1 j=1
3. Fuzzy AHP-TOPSIS
n n n
where ã
j=1 ij
and i=1
ãij are fuzzy triangular numbers
Uncertainty and vagueness are common characteristics in many j=1
n n −1
decision-making problems [5]. Since an expert’ judgment is the obtained by Eq. (4) and also i=1
ã
j=1 ij
is obtained by Eq.
verbal statement, the qualitative preferences are preferred to the (7).
point of estimation. Conventional AHP that requires the selection of Step 2: The truth value of M̃i M̃j is defined by Eqs. (8) and (9).
arbitrary values in pair-wise comparison may not be sufficient and To compare M̃i and M̃j , both the values of T (M̃i M̃j ) and T (M̃i M̃j )
uncertainty should be considered in some or all pair-wise com- are needed.
parison values [5]. Since the fuzzy linguistic approach can take Step 3: The degree possibility for a convex fuzzy number M̃j to
the optimism/pessimism rating attitude of decision-makers into be greater than n convex fuzzy numbers M̃i (i = 1, 2, . . ., n; i =/ j) can
account, linguistic values are recommended to assess preference be defined by
ratings instead of conventional numerical equivalence method [6].
As a result, the fuzzy-AHP should be more appropriate and effec- T (M̃i M̃1 , . . . , M̃i−1 , M̃i+1 , . . . , M̃n ) = T [(M̃i M̃1 ) ∧ . . . ∧
tive than conventional AHP in real practice where an uncertain (M̃i M̃i−1 ) ∧ . . . ∧ (M̃i M̃i+1 ) ∧ . . . ∧ (M̃i M̃n )], (11)
pair-wise comparison environment exists [7].
TOPSIS is a multiple attribute decision making to investigate assume that
solutions from a finite set of alternatives. The main concentration
on it is that the chosen alternative should have the shortest dis- d (Ai ) = min T (M̃i M̃j ) (12)
tance from the positive ideal solution, and the farthest distance is for j = 1, 2, . . ., n; j =
/ i, then the weight factor is given by
the negative ideal solution [8,9]. In this case, fuzzy-AHP and TOPSIS
W = (d (A1 ), d (A2 ), . . . , d (An )) ,
T
are combined to rank alternatives with qualitative and quantitative (13)
attributes. First, to obtain the weight of criteria and local weight of
where Ai (i = 1, 2, . . ., n) is the nth element.
alternative from each criterion, fuzzy-AHP was used, and later, TOP-
Step 4: the normalized weight vectors via normalization are
SIS is applied to rank the alternatives out of alternative to criterion
given by
table obtained by the fuzzy-AHP.
T
W = (d(A1 ), d(A2 ), . . . , d(An )) , (14)
3.1. Fuzzy AHP
where W is a crisp number.
In tackling MADM problems in real situations, the analyti-
cal hierarchy process (AHP) is a widely used approach [11]. AHP 3.2. TOPSIS
can handle twelve types of problems: deciding priorities, gener-
ating alternatives, choosing best alternatives, deciding demands, One of the attractive tool in multiple attribute decision making is
allocating resources, forecasting results or evaluating risks, mea- TOPSIS, where the main idea is that the best alternative is nearest
suring performance, designing systems, making systems stable, to the positive ideal solution (PIS), maximum amount of benefit
optimizing, planning, and resolving conflicts [11]. Based on expert criteria and minimum amount of the cost criteria, and farthest from
linguistic comparison, this approach lacks the ability to handle the the negative one (NIS), maximum amount of the cost criteria and
inherent uncertainty connected with quantifying decision maker’s minimum amount of the benefit criteria. Four advantages of TOPSIS
perceptions [12]. Therefore, the adoption of fuzzy set theory, Fuzzy are listed in Shih et al. [10] study: a sound logic that represents the
Analytical Hierarchy Process, is applied as a suitable tool to inte- rationale of human choice; a scalar value that accounts for both
grate input data uncertainties. The fuzzy-AHP methods introduced best and worst alternatives simultaneously; a simple computation
by these methods [13–19] seem to be systematic for the alternative process that can be easily programmed into a spreadsheet; and
selection and problem justification by using the fuzzy set theory the performance measures of all alternatives on attributes can be
and hierarchical structure analysis. visualized on a polyhedron, at least for any two dimensions.
M. Jafarian, S.E. Vahdat / Journal of Manufacturing Processes 14 (2012) 250–256 253
where wj is the weight of the ith attribute or criterion obtained by The power of importance is considered through five levels of
the fuzzy-AHP. preference shown in Table 3 by the expert weld designer. It means
Step 3: Determine the positive ideal and negative ideal solution a questionnaire is designed and three expert weld designers have
replied back. Then, a sensitivity analysis of each criterion is carried
A+ = {1+ , . . . , n+ } = {(maxj ij |i ∈ I), (minj ij |i ∈ J)}, (17) out. Weld quality is considered by the design of application factor
[1]. It means that design of application is the most important factor
A− = {1− , . . . , n− } = {(minj ij |i ∈ I), (maxj ij |i ∈ J)}, (18) in these criteria [30]. After that, the capital cost, deposition rate and
operator factor are the second more important. But the equipment
where I is set of benefit criteria, and J is set of cost criteria.
portability is not that important because the high pressure vessel
Step 4: Calculate the separation measures, using the n-
factory cannot move to another place [29]. External parameters like
dimensional Euclidean distance. The separation of each alternative
capital cost, operator factor and deposition rate are affected by the
from the ideal solution is given as follows:
investors’ and customers’ needs. So, their importance can change in
⎧ ⎫1/2 various situations. Therefore, these selections are done in the five
⎨
n ⎬ cases, shown in Table 4.
2
di+ = (ij − 1+ ) , i = 1, . . . , m, (19) A graph depicts the hierarchy for welding process selection
⎩ ⎭
j=1 factors that is used in pressure vessel shown in Fig. 3. It shows
that the welding process selection factors have the complex rela-
Similarly, the separation from the negative ideal solution is tionships among each other which are considered in fuzzy AHP.
given as follows: This means that, for choosing a favorite welding process, it is
⎧ ⎫1/2
⎨
n ⎬
2 Table 4
di− = (ij − 1− ) , i = 1, . . . , m, (20)
⎩ ⎭ Description of the five cases.
j=1
Condition Description
Step 5: Calculate the relative closeness to the ideal solution. The 1 It is done for the general design. It means there are limitations
relative closeness of the alternative Ai with respect to A+ is defined for capital cost, deposition rate and operator factor.
2 There is no limitation only for capital cost. It means the
as follows:
investor has a lot of money for production.
− 3 There is no limitation only for deposition rate. It means there is
di
Ri = − + , i = 1, . . . , m, (21) no force to produce pressure vessels.
di + di 4 There is no limitation only for operator factor. It means there
are a lot of skills for all of the welding process listed in Table 1.
Since di−
≥ 0 and di−
≥ 0, then, Ri ∈ [0, 1]. 5 There are no limitations for capital cost, deposition rate and
Step 6: Rank the preference order. operator factor.
254 M. Jafarian, S.E. Vahdat / Journal of Manufacturing Processes 14 (2012) 250–256
In fuzzy-AHP, first the criteria and alternatives’ importance Step 4: Calculate the normalized weightWM̃ =
1
weights must be compared. For this reason, there must be linguis- W
M̃1
tic terms and their equivalent fuzzy numbers denoting comparison W +W +···+W
= 0.1256
M̃1 M̃2 M̃9
measures. To construct the pair-wise comparison of alternative for
After calculating the local weight of alternatives according to
each criterion, we have used experts’ opinions introduced in SME
each criteria and the weight of criteria, TOPSIS method is used
handbook [1]. The guideline to select a fusion welding process is
to rank the alternatives. Table 7 shows distance from the positive
shown in Table 2. There the typical scale for pair-wise compari-
ideal solution (D+) and negative ideal solution (D−) for case 1 and
son was proposed which may be used for preparing the pair-wise
also show the amount of ranking of each alternative. The general
comparison matrix elements for each criterion in Table 5 regarding
welding process selection methodology for high pressure vessel is
this guideline. One of the tables related to the pair-wise compari-
shown in Fig. 4.
son and local weight of alternatives with regard to the criterion of
equipment portability shown in Table 6.
In this case, the stages of fuzzy AHP are presented to obtain the 5. Results and discussion
local weight of alternatives with regard to the criterion of equip-
ment portability in the The ranking of each alternative in different conditions is shown
nfirst case.n n in Table 8. According to Fig. 5 the favorability changed in various
Step 1: Calculate j=1 ãij and i=1 j=1 ãij and M̃i from Eq. (10).
cases. The factor’s weights are calculated through research cases
n
(see Tables 5 and 6), and also shown in Table 9. Joint configuration,
ã1j = (1, 1, 1) + (1, 1, 1) + · · · + (3, 5, 7) = (12.2, 16.33, 21)
welding position, equipment portability and filler metal utilization
j=1
factors are not effective on the welding process selection for high
pressure vessel production because the production will take place
n
in the factory and through weld design as specified before. The
ãij = (12.2, 16.33, 21) + (12.2, 16.33, 21)
result is as follows:
i=1 j=1
+ · · · + (2.97, 3.34, 4.2) = (96.13, 137.13, 183.4)
1. GTAW and PAW are the most appropriate welding processes
for high pressure vessel production in all research cases but the
M̃1 = (12.2, 16.33, 21) × (96.13, 137.13, 183.4)−1 favorability changed (see Fig. 5 or Table 8). Because design appli-
cation, capital cost, operator factor, deposition rate, thickness
= (0.066, 0.119, 0.218) of parts and alloy class are the effective parameters in various
weights for welding process selection, these processes can sat-
Table 5 isfy mentioned parameters (see Table 9 and Table 2).
Conversion of experts comments to preferred value. 2. For GTAW and PAW, the highest value belongs to case 5 because
4 3 2 1 they have the lowest value of deposition rate, capital cost and
A B C D operator factor relatively. So when they are not limited, the value
4 A (1,1,1) (1,3,5) (3,5,7) (5,7,9)
of welding process will rise rapidly.
3 B (1/5,1/3,1/1) (1,1,1) (1,3,5) (3,5,7) 3. OFW and SAW have the lowest value in all research cases because
2 C (1/7,1/5,1/3) (1/5,1/3,1/1) (1,1,1) (1,3,5) they do not have sufficient design applications, and also they
1 D (1/9,1/7,1/5) (1/7,1/5,1/3) (1/5,1/3,1/1) (1,1,1) cannot satisfy the thickness of parts and alloy class.
M. Jafarian, S.E. Vahdat / Journal of Manufacturing Processes 14 (2012) 250–256 255
Table 7
7
7
7
7
7
7
9
1
1
Positive and negative ideal solution distance case 1.
0
(D+) (D−) R Ranking
5
5
5
5
5
5
7
1
1
LBW
SMAW 0.02684 0.16383 0.66067 2
FCAW 0.01983 0.14082 0.60582 4
3
3
3
3
3
3
5
1
1
GMAW 0.01983 0.14082 0.60584 3
SAW 0.00816 0.09036 0.31940 8
7
7
7
7
7
7
9
1
1
GTAW 0.03667 0.19149 0.68845 1
PAW 0.03667 0.19149 0.68845 1
OFW 0.01491 0.12210 0.40136 7
0
5
5
5
5
5
5
7
1
1
EBW 0.02623 0.16197 0.51382 5
EBW
0.24648
0.3
0.3
0.3
0.3
0.3
0.3
0.1
0.1
1
OFW
0.1
0.1
1
0.12559
0.2
0.2
0.1
0.1
the expert
1
1
1
1
1
1
1
0.3
0.3
1
1
1
1
1
1
5
Fuzzy AHP
0.1
0.1
1
1
1
1
1
1
1
by Fuzzy AHP
0.12559
0.2
0.2
1
1
1
1
1
1
3
SAW
The pairwise comparison of alternatives with regard to the criterion of equipment portability.
by TOPSIS
0.3
0.3
1
1
1
1
1
1
5
and SAW decrease rapidly (see Fig. 5 or Table 8). EBW and LBW
have the best design application level but OFW and SAW have
0.3
0.3
1
1
1
1
1
1
5
the least design application level (see Table 2). In this case, design
0.12559
0.2
0.2
1
1
1
1
1
1
3
FCAW
0.1
0.1
1
1
1
1
1
1
1
0.3
0.3
1
1
1
1
1
1
5
0.12559
0.2
0.2
1
1
1
1
1
1
3
SMAW
0.1
0.1
1
1
1
1
1
1
1
Weight
GMAW
SMAW
GTAW
FCAW
Table 6
PAW
OFW
SAW
EBW
LBW
C4
Table 8
Ranking of each alternative in different cases.
Table 9
The weight of factors for the all research cases.
application and alloy class are the effective factor for welding [8] Abo-Sinna MA, Amer AH. Extensions of TOPSIS for multi-objective large-scale
process selection (see Table 9). nonlinear programming problems. Applied Mathematics and Computation
2005;162(1):243–56.
5. GTAW, PAW and EBW are the most appropriate welding pro- [9] Jahanshahloo GR, Lotfi FH, Izadikhah M. Extension of the TOPSIS method for
cesses for high pressure vessel production in the research case 5 decision-making problems with fuzzy data. Applied Mathematics and Compu-
(see Table 8 or Fig. 5). This is because they enjoy the same design tation 2006;181(2):1544–51.
[10] Shih H-S, Shyur H-J, Stanley Lee E. An extension of TOPSIS for group decision
application value and alloy class value (see Table 2). making. Mathematical and Computer Modelling 2007;45(7–8):801–13.
[11] Saaty TL. The analytic hierarchy process. New York: McGraw-Hill; 1980.
[12] Tseng ML, Lin YH. Selection of competitive advantages in TQM implementa-
6. Conclusions tion using fuzzy AHP and sensitivity analysis. Asia Pacific Management Review
2008;13(3):583–99.
The welding process selection factors have the complex rela- [13] Van Laarhoven PJM, Pedrycz W. A fuzzy extension of Saaty’s priority theory.
Fuzzy Sets System 1983;11:1–3, 229–241.
tionships between each other. Due to this complexity, FAHP-TOPSIS [14] Buckley JJ. Fuzzy hierarchical analysis. Fuzzy Sets System 1985;17(3):233–47.
approach investigated the relative welding process selection fac- [15] Chang DY. Applications of the extent analysis method on fuzzy AHP. European
tors according to guidelines for the selection of a fusion welding Journal of Operational Research 1996;95(3):649–55.
[16] Cheng CH. Evaluating naval tactical missile systems by fuzzy AHP based on the
process presented by Society of Manufacturing Engineering hand- grade value of membership function. European Journal of Operational Research
book. With regard to the preference value derived by TOPSIS, GTAW 1997;96(2):343–50.
and PAW are the most appropriate welding process and OFW and [17] Deng H. Multicriteria analysis with fuzzy pair-wise comparison. International
Journal of Approximate Reasoning 1999;21(3):215–31.
SAW are not appropriate for high pressure vessel production in [18] Leung LC, Cao D. On consistency and ranking of alternatives in fuzzy AHP.
all research cases. Also when the external parameters like capital European Journal of Operational Research 2000;124(1):102–13.
cost, operator factor and deposition rate are not limited as case 5, [19] Mikhailov L. A fuzzy approach to deriving priorities from interval pair-
wise comparison judgments. European Journal of Operational Research
the favorability of EBW and LBW rise and the favorability of OFW
2004;159(3):687–704.
and SAW go down rapidly. Therefore, generally, GTAW and PAW [20] Lootsma FA. Fuzzy logic for planning and decision making. London: Kluwer
are the most appropriate welding process for high pressure vessel Academic Publishers; 1997.
[21] Kahraman C. Fuzzy multi-criteria decision making, theory and applications
production.
with recent developments. New York: Springer; 2008.
[22] Hsu YG, Tzeng GH, Shyu JZ. Fuzzy multiple criteria selection of gov-
ernment sponsored frontier technology R&D projects. R&D Management
References 2003;33(5):539–50.
[23] Hwang CL, Yoon K. Multiple attribute decision making. Berlin: Springer-Verlag;
[1] Veilleux Raymond F, Petro Louis W. Tool and manufacturing engineers hand- 1981.
book. Manufacturing Management, vol. 5, 4th edition Society of Manufacturing [24] Milani AS, Shanian A, Madoliat R. The effect of normalization norms in multi-
Engineers; 1998, ISBN 978-0-872633-06-3. ple attribute decision making models: a case study in gearmaterial selection.
[2] Darwish SM, Al Tamimi A, Al-Hadban S. A knowledge base for metal weld- Structural Multidisciplinary Optimization 2005;29(4):312–8.
ing process selection. International Journal of Machine Tools and Manufacture [25] Yoon KP, Hwang CL. Multiple attribute decision making: an introduction. Thou-
1997;37((July) 7):1007–23. sand Oaks, CA: Sage Publishing; 1995.
[3] Zadeh LA. Fuzzy sets. Information Control 1965;8:338–53. [26] Steuer RE. Multiple criteria optimization: theory, computation, and application.
[4] Zimmermann HJ. Fuzzy set theory and its application. 3rd edition Kluwer Aca- NY: John Wiley; 1986.
demic Publishers; 1996. [27] Berberian SK. Fundamentals of real analysis. NY: Springer; 1999.
[5] Yu CS. A GP-AHP method for solving group decision-making fuzzy AHP prob- [28] Jones DF, Mardle SJ. A distance-metric methodology for the derivation of
lems. Computers and Operations Research 2002;29:1969–2001. weights from a pairwise comparison matrix. Journal of the Operational
[6] Liang GS, Wang MJ. Personnel selection using fuzzy MCDM algorithm. European Research Society 2004;55:869–75.
Journal of Operational Research 1994;78:22–33. [29] Chen SJ, Hwang CL. Fuzzy multiple attribute decision making: methods and
[7] Lee AHI, Chen W-C, Chang C-J. A fuzzy AHP and BSC approach for evaluating applications. Berlin: Springer; 1992.
performance of IT department in the manufacturing industry in Taiwan. Expert [30] Megyesy E. Pressure vessel handbook. 14th edition PV publishing Inc.; 2008,
Systems with Applications 2008;34(1):96–107. ISBN 978-0-914458-24-1.