Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Evaluating weapon system using fuzzy

analytic hierarchy process based on


entropy weight
Don-Lin Mon decision making problems. The traditional
multiple critcria programming is not flexible
enough in building the system evaluation model
because i ) f limited experimental data and human
experience. Nor it can accommodate qualitative
requirements (such as securil!. , reliability.
maneuverability). human inspiration and tech-
nological advances in the process of design and
:\hrfr(irr: 'The performance evaluation of wcapiw ryctciiis :IT,: manufacture of the weapon system.
multiple criteria decision inaking problenis. The descriptions
and judgements on weapon systems arc u s u a l l v linguistic and
During the last decade the AHP has become
fuzzy. The traditional method of Analytic Ilierarchy one of the most widely used methods for the
Process (AIIP) are mainly used in crisp (non-furzy) decision pr:ictical solution of numerous ranking problems
iipplications with a very uiih;ilanced scale n l judFements. Tu in different areas of human needs and interests.
,ivcrcoane these prohlems. wc will propow a ne\v and gcncrd Saaty's AIIP [9, IO] i s a systematic process to
dccisian making method for evaluating weapon systems using
fuzzy A H P h a x d on entropy weight. We will use symmetric
represent the elenxiit :)f a problem hierarchi-
triiingular fuzzy numher i to d to indicate the rel:ilive cally and i d u d e s procedures and principles
ctrcngth [jf thc demcnts iii the hierarchy. and huild f u n ) . which are used to synthesize the various
judgcmcnt vectors (matrix) through comparison of the judgements to derive priorities among criteria
performance scores. We will derive the priority among the and subsequently for an alternative solution.
;alternatives by the entropy weight through the use of interval
arithmetic. a-cuts. and index of optimism (the degree o f
However, Saaty's A H P creates and deals with
optimism of the decision maker) to estimate the degree of a very unbalanced scale of estimations, such
satisfaction of the judgement. In this way it does not req:iire as 4,A,. . . , f , 4, 1, 2, 3., . . . , S. 9. For Saaty's
J series of painvise comparison judgements as requireci h) matrices with reciprocal elements. half of the
[he traditional A H P metbod. Ry multiplying the fuzzy nondiagonal elements are in the range from 2 to
suhjective weight vector W with the corresponding fuzzy
judgement matrix ,? (the orderly list of fuzzy judgement
9 and another half in the range from d to
vectors i , for every criterion C,). a fuzzy problem can he which are smaller compared with the former (2
solved and a weapon system selection example will be used to 9). In part of the reciprocals. the range is
to illustrate our method and compare it with other methods. about 4-8 = 0.4 compared with 9 - 2 = 7 in the
integers part. Though the use of the discrete
K e y w w d r : Military application; tactical missile system scale of 1 to 9 has the advantage of simplicity, it
(T.M.S).; fu7z.y numher; a-cuts; analytic hierarchy process
( A H P ) ; interval arithmetic; entropy weight.
does not take into account the uncertainty
associated with the mapping of one's perception
(or judgement) to a number. In addition, we can
1. Introduction list the shortcoming of the AHP as follows:
(1) The AHP are mainly used in nearly crisp
The performance evaluation and optimal decision applications.
design of weapon systems are multiple criteria (2) Ranking of the AIW is rather not precise.
Correspondence I O : D r . Don-Ljii Mon. Department of
(3) The subjective judgement, selection and
Weapon System Engineering. Chung Chrnc. Institute of preference of decision makers have large
Technology, Tahsi 33509. Taiwan. influence on the AHP result, i.e.. if the

0-7803-2461-7/95/$4.00 0 1995 IEEE

59 1

Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 23,2021 at 13:50:41 UTC from IEEE Xplore. Restrictions apply.
judgement has error. the decision is prohahly areas due to its practical nature. suitable for
incorrect. solving complicated and elusive decision proh-
To overcome-these problem. we will usc Fuzzy lems. In this paper we will consider the
number i to 9 [ 3 , 5 ] t o indicatc the relative problem, using the same hierarchy structure, of
strength of the elements in the hierarchy. and choosing the best weapon system among a finite
build fuzzy judgenient vectors (matrix) through number of alternatives. The proposed weapons
comparison of the performance scores. We will are tactical missile systems. The evaluation of a
derive the priority among the alternatives by the good missile system depends upon a number of
entropy weight [fi. 81 through the use of interval attributes, namely tactics, technology, main-
arithmetic, &-cuts [1.2,4], and index of tenance. economy and advancement as described
optimism , I(the degree of optimism of the in Figure 1. Thus we propose a multi-criteria
decision maker) to estiniate the degree of decision making problem which we will treat by
satisfaction of the judgement. In this way we d o using fuzzy set theory and entropy weight
n o t need a series of pairwise comparison method.
judgements as required by traditional AllP
method. By multiplying fuzzy subjective weight 3. Computational aspects of the fuzzy number
vector \i'with every element of fuzzy judgement using interval arithmetic and a-cuts
matrix X (the orderly list of fuzzy judgement
vcctors 2.; for each criterion C;). we can obtain A triangular fuzzy number can be defined by
tlic total fuzzy judgement matrix A . By this a triplet (o,,a,, a?). The membership function is
defined as
way, a fuzzy decision problem can be solved and
;I weapon system selection esaniple will he used
to illustrate our method and compare with other
methods.

2. Establish a hierarchy structure model of the


tactical missile systems (T.M.S.) x > 03.
Iri less than two dccades since its introduction. Alternatively. by defining the interval of
thc AI-IP has found its way into various decision confidence at level a., we can characterize the
triangular fuzzy number as [ 1.2,4]
VflE [O. I].
A" = [ [ I ; . ole] (2)
= [(n. - n , ) n + a , , - ( a 3 - n2)a.+ n 7 ] .
Some main operations for positive fuzzy
numbers A and B described by the interval of
confidence are
Va,. OR, h/.. P/< E R .
A , = [ a : , o i l , B , = [ h i .b i ] . O'E [O. 11,

A @ l? = [ U : , + bi, a i + b;].
A OR = [ U ; - b i , U : + /I;+], (3)
A @ h = [atb;,,u;b;jq
A 0 B = [ a T / h i ,u;/b;].
I n this paper, the computational technique is
Fig. 1. Structure modcl of evaluating 3 'I . h l . S based on the following fuzzy numbers defined in

592

Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 23,2021 at 13:50:41 UTC from IEEE Xplore. Restrictions apply.
Table 1 . Parameter that define the characteristic function of Let S,,k = 1,2,. . . , n , be the sum of the kth
the fuizy number used row, and fki be the relative frequency fkj = a k j / s k .
Then we have
Fuzzy number Characteristic (or membership) function

1 (t*1.3)
r (.r - 2, x . x + 2) for x = 3. 5 , 7
c! (7. 9.9)

Table 1. A fuzzy number 1 expresses the


meaning of 'about x ' . Here each characteristic where sf = Cy=, ski. We can use (4) to calculate
function is defined by three parameters of the the entropy:
symmetric triangular fuzzy number, the left
point, middle point and right point of the range
over which the function is defined.

4. Entropy weight

The Shannon entropy [6]. which is n measure


of uncertainty in inforrllntinn formiilated in
terms of probability thtory, \#.:ISinitinll!; rlrrived
from thermodynamics and used to descrihe the
Hence the entropy weight can be ohtaincd
irreversible phenomcnnn of a motion n r :1
through normalizing (6).
process. I t can be expressed by the functiilii
16. S]

H(p1, 1 7 2 , . . . P,,) = -
1 c p,
i= I
Ine, p , (4 ) 5. Methodology and algorithm

where p i is a relative frequency. Saaty's AHP method is known as the


The entropy H ( p , , p 2 , . . . , p " ) has the eigenvector method. Here we use another
unique form if it satisfies the followins three concept of entropy weight to solve the decision
reasonable and compatible conditions, i.e., when making problem. After invoking comparison of
M ( p , , p z , . . . , p,,) ff(l/tz, 1/11, . . . , 1/11), the performance scores, we will w e symmetric
triangular fuzzy numbers to implement the
H(Pi, p 2 , . . . * p.) = H(p1, p z , . . . .pit, 0 ) - scaling scheme in the judgement vector (matrix),
H ( A B )= H ( A )+ N ( D / A ) , and interval arithmetric to find the total fuzzy
judgement matrix and the entropy weight. The
then computational procedure of this decision making
methodology is summarized as follows:
H ( p l pz. . . . p n ) =
I -2 P , log2 pi.
i=l Step 1 . To compare the performance score.
we
- - can - -use - symmetric triangular fuzzy numbers
Now, we propose the entropy weight com- 1 , 3 , 5 , 7 , 9 to indicate the relative strength of the
putational method. Let A be a crisp judgement elements in the hierarchy (judgement vector o r
matrix, matrix).
Step 2. To establish the total fuzzy judgement
matrix A , we can multiply the fuzzy subjective
weight vector W with the corresponding column
of fuzzy jusgement matrix (the orderly list of
fuzzy judgement vectors - f , for every criterion

593

Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 23,2021 at 13:50:41 UTC from IEEE Xplore. Restrictions apply.
C , ) . Thus 6. Numerical example
KJ,x f , , I?, x f t 2 . . ' \?,, x i ] , ,
Three tactical missile system (T.M.S.) alterna-
Ell Xf,, It, X f Z 2 ' ' ' \?,, x i 2 , , tives A , B , and C will be evaluated b y fuzzy
AHP based on entropy weight in this section.
fi, x x , , k2x i , , . '. ' I?" x x.,,,, The tactical specification data of the three
missile systems and the experts' opinions have
Step 3. Perform fuzzy number multiplication been listed in Tables 2 and 3 (data source [7])
and addition using interval arithmetric and for the decision making process.
a-cuts. From (2) and (3). (7) can be simplified We have established the evaluation model of
to. for 0 < a ,C 1 and for all i. j . the missile systems in Figure I . T h e evaluation
is based on the five criteria: tactics, technology.
maintenance, economy and advancement men-
tioned before and will be described in details as
in the following.
where a;, = ir$ . x&,. U;,, = iv;, . x:,,. (1) Tactics. Tactical considerations include
Step 4 . Estimate the degree of satisfaction of effective range, flight height, flight velocity,
the judgement A with fixed n using the index of reliability, firing accuracy, destruction rate and
optimism A. The index of optimism 1 indicates kill radius. If the effective range is further than
the degree of optimism of a decision maker. A
large A indicates a higher degree of optimism.
Thus we have Tahle 3. Characteristic and experts' opinions
... -
Item A B c
...
(9) I operation condition higher general general
rcquircnient
... 2 safety good yeneral general
3 defilade general god gcnrral
where A is a crisp judgement matrix. and 4 simplicity general general gencral
6 ; = ha&,t (1 - A),;,,, VA E [O, 11 is a linear con- 5 assemhilit). general general poor
vex combination. h combat capability good general gcnrral
Step 5. The entropy can be computed firstly 7 material limitation higher general higher
8 mohility poor good general
by using the relative frequency of ( 5 ) and 9 modulization general good gcneral
entropy formula of (6). The resultant entropy IO standardization general general good
weight can be determined by normalizing (6).

Tahle 2 . Tactical specification data of the three missile systems

Item A R C
- ~

I effective range ( k m ) 43 36 38
flight hcight ( m ) 25 20 23
flight velocity (M No) (1.72 0.8 0.75
4 fire rate (round/min) 0.6 n.6 0.7
5 reaction time (min) i .2 1.5 1.3
6 missile scale (cm)(l X d-span) 5 2 1 x 3s - 13s 381 x 34 - in5 445 x 35 - 120
7 firing accuracy (76) 67 70 63
8 destruction rate (Ir) 84 88 86
9 kill radius (m) 1s 12 18
10 anti-jam (%) 68 7s 70
11 reliability (%) RO 83 76
I? system cost (ten thousand) XIK) 755 785
13 system life (year) 7 5 5

594

Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 23,2021 at 13:50:41 UTC from IEEE Xplore. Restrictions apply.
Table 4. Tactical criteria a n d scores for the 3 Table 6. Maintenance criteria and scores for the
T.M.S. 3 T.M.S.

Tactics A R C Maintenance A B C

effective range 1 0.5 0.5 operation condition I 0.5 U.5


flight height 0.5 I 0.5 requirement
flight velocity 0.5 I 0.5 safety I 0.5 fl.5
reliability 1 I 0.5 defilade 11.5 1 0.5
firing accuracy I I 0.5 simplicity 0.5 0.5 0.5
destruction rate 11.5 I I assembility 11.5 0.5 0
kill radius 1 0.5 1 total score 3.5 3 2
total score 5.5 6 45

tion condition requirement, safety, defilade.


simplicity and assembility, etc. If they are good
40 km. the flight height smaller than 20 m, the
o r higher. their scores are 1, otherwise, the
flight velocity greater than 0.8 Mach number.
scores are 0.5. The score results are in Table 6 .
reliability greater than 8076, firing accuracy
the fuzzy judgement vector is ;ISfollows:
greater than 65% destruction rate greater than
85% kill radius greater than 15m. then the
corresponding score of the subitem is 1.
Otherwise the score is 0.5. Table 4 shows the
total scores of the tactical consideration. (4) Economy. The evaluated subitems contain
Using these scores, the decision makers can be system cost, system life and material limitation.
guided through an approximate comparison to etc. The score results are in Table 7, the fuzzy
make jtdgements and rank the alternatives for judgement vector is
each criterion. In this way, we can get a fuzzy
judgement vector such as the following for the
first criterion:
( 5 ) Advancement. This property is judged on
modulization, mobility and system st::tidard-
ization. The score results are in Table 8, with
(2) 'Tcchnology. The sutrit~ms relate4 to the fuzzy judgement vector
technology contain missile scale. fire rate,
c - = A- B C
- - ,
reaction time, anti-jam and combat capahility.
Scores are listed in Table 5 , and the fuzzy
[I 7 51
judgement vector can be written as Table 7. Economy criteria and scores for the 3
T.M.S.

Economy A B C

(3) Maintenance. The subitems contain opera- system cost 0.5 i 0.5
system life I 0.5 0.5
material limitation 0.5 I 0.5
totai score 2 2.5 1.5
Table 5. Technology criteria and scores for
the 3 T.M.S.
Table 8. Advancement criteria and scores for the 3
Technology .4 R c T.M.S.
-
missile scale 0.5 I 0.5 Advancement A B c
reaction time 0.5 0.5 0.5
fire rate 0.5 1 1 modulization 0.5 1 1).5
anti-jam 0.5 I 1 mobility n I 0.5
combat capability I 0.5 0.5 system standardization 0.5 (1.5 I
total score 3 4 3.5 total score 1 2.5 2

595

Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 23,2021 at 13:50:41 UTC from IEEE Xplore. Restrictions apply.
I
Table 9. F u u y judgement matrix for
systems A , B . and C based on criteria
c,,cz. c,. c,, c, 0 37
038
P

(6) If the priority of the criteria is in the order


5
3,034 -

- 033
-
U
-
-
.
.
.
-
---a
--.___
-

-1
1. -*
G , A

__I
I Q SvscmR
+ Svrlem C

of tactics, advancement, economy. technology


and maintenance according to the experts'
opinion, then we can use a fuzzy vector such as
(15) to represent a weighting vector among these
criteria. 000 010 020 040 060 080 100
n

(15) Fig. 2. Total arrangement order of all levels for moderate


decision maker.

Combining (10)-(14), we can obtain Tablc 9.


Multiplying the ith element of (15) to the C, From Table 10, system B is the best choice when
column of Table 9. we can obtain the total fuzzy a = 0.05, A = 0.5.
judgement matrix A as Similarity, letting a = 0, 0.05, 0.1, 0.15, 0.2,
. . . , I and A = 0.5, 1. 0, we obtain Figure 2. 3,
c, c, c, c
4 cs and 4 respectively.

1
gxj j x i i x i 5x3 ?xi From Figures 2-4, this method reflects the
Bx? 5x5 ix.5 5 x 5 i x ? uncertainty associated with one's subjective
judgement in human thinking. However, system
Bxi j x j ixi 5xi ix5 (I6) B is the best selection for all the degree of
From (2), we have optimism h.
For the sake of comparison, we will use a
i, = [ I . 3 - 2 4 , 3, = [ I +?@. synthetic evaluation method and Saaty's AHP
5* = [3 + 2 a , 7 - ?a], 7 , = 1s + ?a.,Y - 2 4 . method to treat the same problem
B,, = [7 + 2 a , 91. (i7)
Substituting (17) into (16). and using Step
3-Step 5 from Section 5 , we can derive entnipy
weight for the corresponding a level.
0 36
Let cr = 0.05, 1 = 0.5 (for a moderate tlccisiori
maker), (16) ends to Table 10 (column 2-6). R v
package Lotus 123, we can obtain the entropy
weight of the alternativcs as shown in column 7
of that table.

Table 10. Total judgement matrix and entropy weight for


(Y = 0.05. I = 0.5
..
Entropy 031 -
Cl c, c, C, C, weight I
03 -
A 40.25 5.85 13.65 15 13.65 0.3368 000 0 10 020 040 060 080 100
n
B 56.35 IS 9.75 25 49 0.3392
C 15.6975 9 3.8015 9.75 35 0.3241 Fig 3 Total arrangement order of all levels for optimistic
decision maker

596

Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 23,2021 at 13:50:41 UTC from IEEE Xplore. Restrictions apply.
038 I
1 3 7 113 3
037 - C? = 113 1 5
P
117 l / 5 1 1/3 1/5 1
0 36 ;
/‘ I
I (18)
1 117 115
Cs = 7 1 3
5 113 1
0331 1
1 7 9
117 1 3
o= 119 113 1
1/5 3 5
113 5 7 3 1

The procedure of the A H P is wcll established


[ 5 ] , we will only give the resulting outcomes as
in Table 11. where system B is the best choice.
We can see from the above analysis that the
7. Synthetic evaluation method ranking of the AI-IP is the most rough This is
because A H P creates and deals with a very
Here we use a simple nonfuzzy additive unbalanced scale of estimations. The matrices
weighting approach t o treat the evaluation with recripocal elements such as in (18) have one
problem. From Tables 4-8, the relation matrix half of the nondiagonal elements in the range
,f (whose elements are the performance scores) from 2. to 9 and another half in the range frofn 4
and the weighting vector cc.’ (normalizing to 4 which are smaller compared with the former
[9.3. 1,5,7]) are (3-9). In the part of the reciprocals, the range is
about 4-4 0.4 compared with 9 - 2 = 7 in the

L 1
5.5 3 3.5 2 1 integer part. We can say that AHP does not
x = 6 4 3 2.52.5, reflect the uncertainty associated with the
4.5 3.5 2 1.5 2 mapping of one’s perception to a real number.
The synthetic evaluation method, o n the other
= [O.36 0.12 0.04 0.2 0.281. hand. is neither flexible nor can show the degree
Therefore the weighting matrix = , f o W‘= of optimism of the decision makers.
[3.16 3.96 2.98)‘ where 0 denotes multiplication
and t denotes transposition. From these
calculated weight values, system B is the best
choice. 9. Conclusion

We have mentioned in the introduction


8. Saaty’s method (Section 1) that the traditional AHP have some

Saaty’s AHP method is known as the


Table 1 I . Total arrangement order weights of hierarchy
eigenvector method. From Tables 4-8 and (15),

c, =
[ 1 1:3
3
‘1,
we can obtain pairwise comparison judgement
matrices C , , C2, C, C,. C5, and 0.

1/5 117 1
cz=[ l5
3 113
1 1!3]
1/5
Criteria

0
.4
B
c
I 2 3

0.5128 0.0634 0.0334


0.2790 0.1047 0.6191
0.6191 0.6370 0.2790
0.0719 0.2583 0.0719
J 5

0. 1289 0.2615
Weight

0.2513 0.0719 0.2235


0.6370 0.6491 0.6344
0.1047 0.2790 0.1421

597

Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 23,2021 at 13:50:41 UTC from IEEE Xplore. Restrictions apply.
shortcomings. namely: References
(1) The AHP are mainly used in nearly crisp [ I ] A . Kaufmann and M.M. Gupta. Irirrodrrction to Frizzy
decision applications. Arithmetic Tlieon and Applicariori (Van Nostrand
(2) Saaty's AHP creates and drill< with a very Reinhold. New York. 1991).
[2) A . Kaufmann and M.M.Gupta. Fuzz?. Mathematical
unbalanced scale of estimations.
Models in Engineering and .Management Science
(3) Saaty's AHP does not take into account (North-Holland. Amsterdam. 1988)
the unccrtainty associated with the mapping of 131 C.H. Juang and D.H. Lee, A fuzzy scale for measuring
one's perception (or judgement) t o a numher . weight criteria in hierarchical structures. lnternarionnl
(4) Ranking of the AHP is rather not precise. F I I Z ZEngineering
~ Symposium (1991) 415-421
[4] C . H . Cheng and D.L. Mon. Fuzzy system reliability
( 5 ) The subjectivc /litleetiicnt. selcction and
analysis by interval of confidence, FUZZ;. S E f s and
preference of decision makers havr large Sysrem 58 (1993) 29-35,
influence on the AIIP. [SI C.H.Cheng and D.L. Mon. Evaluating weapon system
To improve the AFIP method and to smooth using fuzzy analytical hierarchy process. to appear in
the evaluation process. we have propused a Defence Science Jourria/.
decision making method, using the Sam? [h] G.L. Klir and T A . Folger. Fuzzy Sets. U n r e r r n i n t ~arid
.
Informarion (Prentice Hall, Singapore. 1988)
hierarchical structure, ba4ed on the entropy [7]J . H . Wen, Guided Missile System Andysir arid Derign
weight. We have formulated the algorithms and (Beijing Natural Science & Engineering Clniversity.
applied it to a weapon system selection problem. Beijing, 1989) (in Chinese).
This problem has also h c m treated by the [RI Tang Ke, Target decision hy entropv weight a n d fuzzy.
Swrcrn Enpinwring Theory and Practice 5 (1992) (in
traditional methods for comparison. We have
Chinese) .
seen the thoughtfulness. flexibility and efficiency 191 T . L . Saaty, A scaling method for priorities in
of the proposed method, especially in terms of hierarchical structure, Journal of Mathematical Psycho/-
subjectivity and preference of the decision og;. 3 (1979) 243-281.
makers. [ 101 T.L. Saaty. The Analytiral Hierarchy Proccss (McGraw
Hill. New York. 1980).

598

Authorized licensed use limited to: Universitas Indonesia. Downloaded on January 23,2021 at 13:50:41 UTC from IEEE Xplore. Restrictions apply.

You might also like