CHAPTER-v

FUZZY ANALYTICAL HIERARCHICAL PROCESS APPROACH

IN LEAN PRODUCT DESIGN

5.1 INTRODUCTION

This chapter presents h a y based analytic hierarchical Process (Fuzzy-AHP)

to efficientlytackle both quantitative and qualitative decisions involved in the design
of a product, which satisfies more customer needs. A fuzzy logic-based approach to
handle the vague, imprecise and uncertain information and fuzziness preferences
available regarding a module or subsystem, which are used for assembly, is presented.
Fuzzy-AHP is the combination of two concepts: fUzzy set theory and the AHP. The
AHP weights of both criteria and modules with respect to criterion are t r a n s f d to
Triangular Fuzzy Numbers, which is input to Fuzzy logic approach. The painvise
comparison matrices help in deciding the synthetic extent value of each comparison
and finally, priority weight of each module is used for making input membership
functions in fuzzy approach. Hence, the proposed method that comb~ningAHP and
Fuzzy logic for evaluating customer preferences is termed as Fuzzy AHP.
The membership functions are built h m the priority weights, instead of
using fuzzy set theory to reduce more time in computation of fuzzy weights. Hence,
this method leads to reduction in product design wastages like repetition and over
processing of customer information. Therefore, this type of approach is more suitable
for Lean product design. The chapter focuses on the integration of Data Envelopment
analysis (DEA) to tizzy logic mult analysis for ranking or computing efficient and
inefficient assembled systems for Bench marking. Hence, this is an attempt to
integrating DEA with fuzzy AHP process for computing very clarity nsults than
fuzzy approach, and the new integration termed as Fuzzy DEAHP.
The organization of this chapter is divides into five sections. Introduction and
importance of fuzzy logic approach in product design and relevant literature is
presented in section 5.1. Section 5.2 discusses about Fuzzy Analytical Hierarchy
Process. Section 5.3 deals the use of Fuzzy AHP for the customs preference
d u d o n in L .product design, and Section 5.4 of this chapter integrates DEA
method into Fuzzy AHP for computing efficiencies of different assembled syetmre.
Section 5.5 v t s brief summary and conclusions of integrated Fuzzy DEAHP
method.
5.2 FUZZY ANALYTICAL HlERARCBlCAL PROCESS

A fUzzy decision-making method is usined for selecting an optimum design,

from various proposed design alternatives. The selection of different modules fbr
computer assembly is taken as a case study. The evaluation objectives are arrangad in
a hierarchical structure with several levels. The relative contribution (weighting
function) of each objective to the overall value of the solution and the rating or degree
of approximation of a solution with respect to a given objective are quantified with
the membership functions of a fuzzy set. A computer programme built with a
weighted generalized mean method is used to calculate the fuzzy probability from
level by level from the lowest-level objectives. After the fuzzy, expected
values of the top-level objectives are calculated and they are then used to make a
decision quantitatively on selecting the optimal design alternative.

Bellman et al. (1995) discussed the importance of product concept design

stage, and during this period, the product knowledge is vague, imprecise, uncertain,
ambiguous, inexact or probabilistic in nature. Human thinking and reasoning
frequently involve fuzzy information, possibly originating From inherently inexact
human concepts and matching of similar rather than identical experiences.

Temponi et al. (1999) demonstrate a fuzzy-based theory and methodology in

order to represent the imprecise and conflicting customers' requirements and facilitate
the communication between multidisciplinary team members while implementing
quality hnction deployment (QFD) process for product development projects.

Shehab and Abdalla (2001) illustrate the fuzzy logic based methodology for
estimating the manufacturing cost of a product during concept development stage that
could not be handled with traditional parametric and generative-based models because
of unclear and subjective natuie of the data.

Rajasekaran et al.,(2005), presented a timy set as a class of objecta with a

membership function ranging between zero and one, and mostly p r d d tiom AHP
weights. It was specifically desigped to mathematically represeat d t y and
vagueness. Fuzzy set theory implements groupinp of data with boundaries that are
not sharply defined, Any methodology or theory implementing "crisp" definitions
such as classical set theory, arithmetic, and programming, may be "fimifiad" by
generalizingthe concept of a'crisp set to a fuzzy set with unclear boundaries.

The literature discussed by Yasemin et al.(2006), explained at the benefit of

extending crisp theory and analysis methods to fUzzy techniques is the strength in
solving real-world problems, which inevitably entail some degree of imprecision in
the variables and parameters measured and processed for the application.

5.2.1 Fuzzy Set theory

Fuzzy set theory has proven advantages within vague, imprecise and uncertain
contexts and it resembles human reasoning in its use of approximate information and
uncertainty to generate decisions (Gungor et a]., 2000). It was specially designed to
mathematically represent uncertainty and vagueness and provide formalized tools for
dealing with the imprecision intrinsic to many decision problems (Kristin, 1992).
Fuzzy set theory implements classes and grouping of data with boundaries that are not
sharply defined (i.e Fuzzy). Fuzzy set theory includes the fuzzy logic, fuzzy
arithmetic, k y mathematical programming, fuzzy graph theory and fuzzy data
analysis, usually the term fuzzy logic is used to describe all of those. The major
contribution of fuzzy set theory is its capability of representing vague data.

A k y set is characterized by membership function with assigns to each

object a grade of membership ranging between 0 and 1. In this set the general terms
such as 'large', 'medium', and 'small' each will be used to capture a range of
numerical values. If nl, n2 and n3 respectively denote the smallest possible value, the
most promising value and the largest possible value that describe a fuzzy event then
the triangular fuzzy number (TFN) can be denoted as a triplet (nl, n2 and n3). A k z y
number N express the meaning of 'about N'. A TFN is shown in Fig. 5.1
Assumptions in F u w AHP model

1. Fuzzy Logic Mamdani module input variables membership functions arc built,
based on AHP weights obtained in Chapter-I11
2. Triangular Membership !?unctionis assumed for input and output variables.
 Fig. S.l(a) A Triangular Fuvy Number

Definition: The relationship function of a TFN which associated with a real

number in the interval [O,l]can be defined as :

0 ....otherwise

A fuzzy number is also represented by the corresponding left and right

representation of each degree of membership:

Here I(y) and r(y) are the left and right side representation of fuzzy number.

Fuzzy numbers are naturally easy to deploy in the expression maker's

assessment of the qualitative situations.
5.2.2 Representation of Fuzzy Sets:
If U is a collection of objectivesdenoted generally by x then Fuzzy set A in U
can be represented by
 Where pA(x) is called the membership function or grade of membership (also

degree of compatibility or degree of truth) and the terms -

(x ) k o t e hc mlation
Xi

between xi and its membership function p ,( x i ) instead of fractions.

5.2.3 Fuzzy Matrices

Let R be fuzzy relation on X x Y and X has m elements and Y has n elements,

then R can be defined by the following relational matrix

Whose elements are given by

r, = P ~ ( X , * Y ] ) *X X Y

Such matrix is called a fuzzy relational matrix or fuzzy matrix in brief

Operations for Fuzzy sets
5.2.4 Classical Set Operations:
Union Operation: The membership function pdx) of the union D = A II B is
pointwisedefinedby r,(x)= max (y,(x), r,(x))x€X
Intersection Operation: The membership function pdx) of the union
C = A n B is point wise defined by ~~(x)=mi&~(x)&(X)x € X

In a Complicated system, not only a number of objectives are included, an

objective may also contain one or more next lower level objectives. The multi level
decision-making model should be wed to calculate the membership grade level by
level tiom the lowest level objectives. The final judgment for determining the optimal
alternatives done based on the membership grade of the top-level objectives.
5.25 Application of fuzzy operator in mtaedent

Once the inputs have been fUzzified, the degree to which each part of the
antecedent has been satisfied for each rule is detnmined. If the a n t d e n t of a given
rule has more than one part or input, the most common way to Combine IF- pad of
each input variable is with a pair wise minimum or min T-norm operator (Jang et al.,
1997);
pBl = m i n b 4 ( x ) , ~ 4 ( x ) 7(X),+.---..~-P~(X)\
~r

Where n is the number of inputs and 1 represents the rule number. The min
combiner or T-norm operator is a conjunctive or AND operator. We can also use
anather T- norm operator known as 'product' in place of mint-norm operator to
combine IF-part of fuzzy mles. As a matter of fact, unlike in the min T-norm operator,
the product T-norm oppator does not ignore any information. The product T-norm
operator can be expressed as (Jang et al., 1997);
5.2.6 Aggregate all outputs

Aggregation is when we unifl the outputs of each rule by joining the parallel
threads. It's just a matter of taking all the fUzzy sets that represent the output of each
rule and combining them into a single fuzzy set in preparation for the final step of
fuzzy logic called defuzzification. Aggregation only occurs once for each ouput
variable. The input of aggregation process is the list of truncated output functions
returned by the implication process for each rule. The output of the aggregation
process is one fuzzy set for each output variable. Three built-in methods arc supported
Mamdani module of fuzzy logic software such as max (Minimum), p b o r
(Probablistic) and sum (simply the sum of each rules output set)

The Fig. 5.l(b) shows two rules have been placed together to show how the
output of each rule is combined or aggregated, lnto a single fuzzy set for the overall
output.

5.2.7 Defuzzification

The input for the defuzzification process is a fuzzy set (the aggregate ouput
fizzy set) and the output is single number- crispness recovered from fuzziness at last,
As much as fuzzinesshelps the rule evaluation during the intermediate steps, the final
output for each variable is generally a single crisp number known as Defuzzification.
Aggregation only occurs for each output variable. So given a hzzy set that
encompasses a range of output values, we need to return one number, therby moving
from a fuzzyset to a crisp output. The most popular defuzzification method is the
centroid calculation, which returns the center of are under the curve. There are five
built-in methods supported mamdani that are centroid, bisector, middle of maximum
( the average of the maximum value output set), largest of maximum and smallest of
maximum.

The input of the aggregation proms is the list In the general, there are five
methods of defuzzification, but centroid of area is the one that is most commonly used
(Jang et al., 1997). So, in this chapter, we used the centroid of area (L)
as a
defuzzifier, which is
Where p ,[z ] is the aggregated output membership function (MF).

5.3 CUSTOMER PREFERENCES EVALUATION IN PRODUCT DESIGN-

FUZZY LOGIC APPROACH

Custom? preferences evaluation methods discussed in chapter 111, and deal

with customer needs that are expressed in terms of numbers or ratio. To find the
optimum requirements of the customer, AHP method is applied to evaluate needs. It is
relatively simple and straightforward procedures, for the assessment of local
priorities, based on pairwise comparisons. If customer expressed prefmces about
the alternatives in comparative fashion, it is difficult to accommodate in traditional
AHP method. Hence, the standard AHP Eigen value prioritization approach cannot be
used, when the complex and uncertain preferences expresses the comparison
judgments as uncertain ratios, such as "about two time more important", "between
two to four times less important", etc. A natural way to cope up with such uncertain
judgments is to express the comparison ratios as fuzzy sets or b y numbers, which
reflect better in reduction of vagueness of customer thinking.

Fuzzy clustering:

Jiao and Tseng(1999), employed the fuzzy cluster analysis to evaluate the
similarities of customer needs by applying c-means clustering analysis. In the same
way, Moon et al. (2006) wed fuzzy c-means clustering to determine initial clustag
representing modules and to identify the platform and its modules by a platform level
membership function and classification. Jiao and Zhang (2005) adopted a fUzy
clustering approach to aeate a,hierarchicaldecomposition of the given sat of objects,
and to form groups in different levels of similarity. Zha et al. (2004) developed a
knowledge intensive support scheme a comprehensivesystematic fuay clustering and
ranking methodology for product family design evaluation a selection.
 Fuzzy goal programming has been adopted to determine the fulfiIhent llevds
of the engineering design requirements, where the d c i e n t s in these models are
also fuzzy in order to expose the firzziness of the linguistic information (Chan and
Weng, 2006) and to simultaneously optimize multiple objectives for product
modularization.

Fuzzy multiple'attribute decision-making:

The consideration of multiple attributes during the decision making praccss

has been considered an important issue to make accurate decisions, Jiang and Chi-
Hsing (2001) used fuzzy logic decision model and tiuzy multiple attribute decision-
making model to construct the goal decision and activity decision spaces respectively
into the proposed manufacturability evaluation decision model. Shipley ct al. (2004)
used fuzzy set based multi criteria decision- making process to determine the
distributions of effort directed toward technical changes. Kuo et al. (2006) used fuzzy
multi-attribute decision-making techniques to develop a method for green fuzzy
design analysis, which involves simple and efficient procedures to evaluate product
design alternatives based on environmental consideration to select the most desirable
design alternative.

Fuzzy inference:

Fuzzy inference has been significantly used for numerous purpow such as
determination of the priority of customer demands (Chen et a1 2004), to accommodate
the possible imprecision and vagueness during the interpretation of the voice of the
customer during the interpretation of the qualitative and sometimes imprecise
customer requirements (Fung @ al., 1999). The process new product idem into the
product evaluation process by using hzzy interference system to adjust the
membership function to d a n c e their systematic t i n y clustering and ranking model
by adopting a neural network technique( Zha et al., 2004), to perform the learning
process of the fuzzy inference system by using adaptive nmo-fuzy inference
systems.
Fuzzy Numbers

Fuzzy m ~ b e r have
s been widely applied for different purposes. Vanegas and
Labib (2001) used h y numbers to represent the imprecise nature of the judgements,
and to define more appropriately the relationships between engineering characteristics
and customer attributes in QFD. They also developed a new firzzy weighted average
during the engineering design evaluation process trying to reduce the obtained
imprecision during such process. Vanegas et a1.(2005) capture the relative importance
of the considered criteria and performance levels of the different alternatives in the
evaluation process for engineering design. Chen et al. ( 2006) express and represent
the input data inorder to calculate the importance of the technical attributes in the
fuzzy QFD. Other applications include Lin and Chen (2004) used fuzzy numbers to
describe the criteria ratings and their corresponding importance in the proposed
method for new product screening. Buyukozkan et al. (2004) to represent the
performance of different ideas into the fuzzy preference relation to express the
assessment of the decision makers into the fuzzy analytical hierarchy process and
applied fuzzy triangular membership functions to represent the customer attribute and
engineering characteristic into the rule based fuzzy logic system to examine their
relationships.

Fuzzy Optimization:

Some important applications of fiuzy optimization include Vanegas and Labib

(2001), employed fuzzy optimum selection in the reasoning process, w h m the
constraint satisfaction and fuzzy optimum selection interact to search the optimum
solution. Fung et al. (2002) applied a limy -linear optimization model for QFD
planning to obtain a set of feasible solutions to support more practical and coi-
effective QFD planning under resource constraints. Chen et al. (2004) applied fuzy
optimization theory with symmetric or non-symmetric triangular limy coefficients to
model the relational functions between engineering characteristics and customer
requirements in QFD methodology.

Fuzzy outranking: Wang (1 999), proposed a new fuzzy outranking approach and an
out ranking decision model to select the critical design q u i r e m a t s for product
development in the imprecise and uncertain design enviromnant in the QFD plamh8
process. Focusing on the application of the outranking approach, Gungor et al. (2000)
used the outranking approach to model an imprecise preference st~chtrein a project
selection problem. Buyukozkan et al. (2004) applied the out ranking concept into the
pseudo-order fuzzy prefence model to discriminate the set of alternatives without
information about their information.

Fuzzy Quality function deployment:

Ramasamy et al. (2004) proposed a h z y logic-quality hction deployment to

determine optimum rating of engineering characteristics by using a rulebased fuzzy
logic system. Also Shipley et al. (2004) presented a model to develop the QFD into
Fuzzy-set based multi criteria decision -making process to determine the distributions
of effort directed toward technical changes.

Fuzzy ranking:

A fuzzy ranking'methodology by employing the fuzzy preference relation to

model the fuzziness in conceptual design evaluation in configuration design for mass
customization was developed by Jiao (1998). Jiao and Tseng (1999) developed a
fuzzy ranking approach and methodology using information-content measure for
solving the multi- attribute design evaluation problem. More recently, focusing on the
PFD process Zha et al. (2004) developed a ranking methodology for the product
family design evaluation and selection.

Fuzzy regression

Chen (1999) developed a fuzzy regression applying nonlinear programming to

solve the fuzzy ranking problem. Kim et al. (2000) employed fuzzy regression to
consider mathematically the inherent k i n e s s during the estimation of the functional
relationship between customer' requirements and engineering characteristics in the
QFD application. Chen et al. (2004) considered the fuzzy linear ~grwsionwith
systematic triangu1ar fuzzy coefficients to model the relational functions between
engineering chderistics and customer requirements considered traditionally in
QFD methodologies.
Fuzzy weighted average

Vanegas et al. (2001) developed a new fuzzy weighted average to produce

fuzzy numbers as a better basis for making decisions more credible, and with less
imprecision. Fuzzy weighted average has been used for different purposes such as
ranking of projects in new product development process. The aggngation of fiuzy
numbers into the product rating process to calculate the overall performance of the
alternativesconsidered in the evaluation of designs (Vanegas and Labib, 2005).

Fuzzy analytical hierarchy process

Fuzzy-AHP is the fiuzy modified form of AHP. It has the ability to extract the
merits of both approaches to efficiently and effectively tackle the multi-attribute
decision making problems. The AHP is one of the extensively used multi-criterion
decision making methods but it has generally been criticized because of the use of a
discrete scale of one to nine, which cannot handle the uncertainty and ambiguity
present in decision priorities of different attributes. The relative importance of
different decision criteria involves a high degree of subjective judgement and
individual preferences. The linguistic assessment of human feelings and judgements
are vague and it is not reasonable to represent in terms of precise numbers. It feels
more confident to give interval judgements than fixed value judgement. To improve
the AHP method, this chapter discusses fuzzy modified AHP approach using
triangular fuzzy numbers to represent decision maker's comparison judgements and
fuzzy synthetic extent analysis method to decide the final priority of different decision
criteria.

In chapter-111, we used only AHP Technique for finding best combination of

various modules for assembly of computer, but in AHP, it is difficult to accommodate
two different discipline criteriri as first Level is a physical criterion, and third Level is
functional Criteria. Hence, solution for such a complicated hierarchies, we proposed a
method of integrating AHP with Goal Programming (GAHP) (Venkatamuni and Rao,
2010). An addition for the GAHP is Fuzzy AHP and it is discussed in the forthcoming
section.
53.1 New Filzzy AHP method proposed for Lean Product design,

In Lean Product design for the customer satisfaction, we consider following

important factors in the Product design: (1) Decreasing product Assembly cost, (2)
Increasing Reliability and (3) Ease of assembly. In this chapter, we have proposed a
new method of Fuzzy AHP method, each alternative under every module is a one
input membership function and hierarchical of modules and these alternatives are
shown in Fig.5.2.

System

Dell 17" (A,,) PIV 3.0 core Zebronics lntex

Samsung 21" PIV 3.06 GHz

7-1 LG Digital 17"

Fig. 5.2 Computer System Modules m d rvrihble rlttrnrtlvet (RefChapter-III)

Input membership functions generated based on AHP weights of alternatives
under each module instead of customer's triplet data about alternative. Hence,
repeated data collection for fuzzy process is avoided. Therefore, this type of analysis
reduce the waste involved in Product design, like o v a processing, repetition of
collecting customer preferences and so on. The assumed triplet data based on ANP
weights for different alternatives is presented in Table 5.1. A triangular fizzy number
is the special class of fuzy number whose membership is defined by three real
numbers, expressed as (1, m, u) (Kristin et al., 1992). The triangulat membership has
the advantage of simplicity and is commonly used in the modeling of product
development analyses (Shehab and Abdalla, 2001).

Table-5.1 Input membership functions of Merent alternatives based on ABP lnlgM

(Ref. Chapter-III, Fig. 3.20)

Similarly, the output of the model is minimizing cost, reliability, and ease of
assembly measured with linguistic terms uses five generic levels such as, 'very low',
'low', 'moderate', 'high', and 'very high'. And triplet numbers of different linguistic
t m s are presented in Table 5.2
 Table 5.2 Output Membership functions of M u e a t linguistic numbers for three
output variables

The relative contribution (weighting function) of each objective to the overall

value of the solution and the rating or degree of approximation of a solution with
respect to a given objective are quantified with the membership functions of a fuzzy
set. Matching the customer's needs and various characteristicsof the designed product
together is the key point to develop a successfirl product. A decision-making method
by quantifjmg the correlations between human sensations, and the physical
characteristics of products using fUzzy logic analysis methodology is shown in Fig.
5.2(a). This technique will assist designers to create products that satisfy the needs of
customers. The Product designer now has to decide which of the many product
varieties available are best suited to customa, in line with predefined targets. These
targets may include High Reliability, Low cost, and Easy of Assembly.
The proposed integrated methodology for analyzing the perfomm~tof
product architecture for selecting the components is largely based on the evaluation of
qualitative product information using &y logic. The crisp outputs of the fuzy logic
models will be incorporated in the decision-making process like Maximization of
reliability, ease of Assembly, and Cost of assembly.
The basic steps of the methodology are grouped into three phases as follows;
(a) knowledge acquisition process, (b) fuzzy inference process, and (c) product
selection process. Fig. 5.4 shows the flow diagram of the proposed methodology to
evaluate and utilize the metrics for product selection.
Knowledge Acquisition Process
As in any other system, the quality of the fuzzy approximation depends very
much on the acquired level of precision of the subjective knowledge of the product
under study. Therefore, it is highly desirable to have a knowledge acquiring team of
experienced people. The other factors that affect the accuracy of the proposed fuzzy
logic-based assessment of performance of the prospective modules are level of
manufacturing knowledge of the designer, availability of the information about
similar products, and the proper design and definition of fuzzy set membership
functions.
Fury nk buc H l l t ~ l l&u
l A~lysb

Fig. $.2(a) Overvim of fuzy logic model for perfonnrnce index &&nation
 General analysis of M u c t (Computer Assembly).

I Identified Input and output liguistic variables 1

I
*
AHP weights

Fuzzy Memebrship Functions

J
I

I Development of Evaluation Guidelines I

I Fuzzy rules framed with the user experience 1

I Train fuzzy logic Mamdhani module with rules I

I Obtained Output for specific configurationiCompute

Output for selective input I
I Graphical representation of large number of
confieurations I
Fig. 5 3 @) Fuzzy AHP Flow Chart

Let the computer assemble is made from four modules, i.e Monitor, Processor,
Sound Card and Graphic card with different alternatives called input variables. Final
assembled computer system is checked through three parameters i.e reliability, Cost
index, easy of assembly, called output variables, When MATLAB is opened, then the
GUI (Oraphid User Intarface) is opened by typing "fuzzy", the nsult is shown
below Fig. 5.3

Fig. 53 CUI of fuzzy toolbox

The input variable is named, say, monitor and the same is entered in the
variable name box. The mouse is moved to the main edit menu and click on edit and
add other input variables viz., Processor, Graphic Card and Sound card. Similarly the
Output variable is activated in GUI tool box, and add output variables such as
Reliability, Cost and Easy of addembly shorn in Fig. 5.4 below.
 Fig. 5.4 GUI Fuzy logic Tool box for application of ~lelefHonof modnles

Double-click on an input variable icon (box), immediately jump to the

Membership function.

5.3.2 Membership Functions for Fuzzy AHP

A membership function (MF) is a curve that defines how each point in the
input space is mapped to a membership value (or degree of membership) between 0
and 1. The Fuzzy Logic Toolbox includes 11 built-in-membership function types.
These 11 functions are in tum, built from several basic fundiom such as linear
functions, the Gaussian distribution function, the sigmoid curve and Quadratic and
polynomial curves.

The simplest membership functions are formed using straight lines. Of &use,
the simplest is the triangular membership function, and it has function m e Code
"trimf'. It's nothing more than a collection of three points forming a triangle. The
straight line membership functions have the advantage of simplicity and easily
ammunodate AHP global weigh& for built the membership functions. Hence, an
AHP weight minimizes the computational process of fwzy logic membership
functionsparameters.

Click the input variable block twice with mouse to open the membership function
window. First define the following values for each membetship function. For example
firstmembership function in Monitor input variable is entered by following values.
Range: 0 to 1.
Name of Membershipfinction = Alternative Name (All)
Type = ~ r i m (select
f from list)
Parameters = [0.757 0.874 0.9901 (From the Table 5.1)
Next, choose fiom Edit, Add MFs, for the monitor five membership functions as
equal to number alternatives considered as shown in Fig. 5.5, And, close the
membership function editor. Similarly double click on processor, graphic card and
sound card input variables in FIS Editor for adding membership functions. The input
variables membership functions are shown from Fig.5.5 to 5.8,

Fig.55 Modtor Input memberhip fanaton Plob

 5.6 Proweor Input membershipfunction Ploh

PIS Variables

Fig. 5.7 CnpMc Cuul Input membenhipfundion PI&

 Fig. 5.8 Sound Card Input membership function Plot8

Reliability, Cost and Easy of assembly the impact of architecture on

performance of prospective module in linguistic term is captured in five generic levels
such as very low' ,'low', 'moderate', 'high', and 'very high'. The Table 5.2
represents membership function parameters, developed by System Administrator for
this study. The membership functions for each output variable are required for
Fuzzification and entered into to FIS Editor as similar to input variables and
explained earlier, The reliability, cost and easy of assembly membership functions are
shown from Fig. 5.9 to 5.1 1.
Fig. 5.9 Relhbillty output membership h c t i o n Plot8
 Fig. 5.11 Assembly output membership fundon Ploh
53.3 Development of Fuzzy "if-then Rules" to Relate Input to Output Variable6
In the field of artificial intelligence, there are different ways to represent
knowledge. Perhaps, the most common way to represent human knowledge is to form
it into natural language expression of type (Bimal et al., 2005). If premise
(antecedent), then conclusion (consequent). The form of expression commonly
referred to as the IF- THEN rule based form. It typically expresses an i n f m c e such
that if we know a fact (premise, hypothesis, antecedent), then we can infer or derive
another fact called, a conclusion (consequent). The rule based system is distinguished
from classical expert systems in the sense that the rules comprising a debased
system might derive from sources other than human experts and, in this context are
distinguished from expert systems. A fuzzy 'if-then rule' is expressed as If x is A then
y is B.

Where A and B are the linguistic values &finad by fuzzy sets on universe of
d i m w X and Y. The 'if-pait' of the rule 'x is A' is called the antedent or
premise, while the 'then-paa' of tha rule 'y is B' is called the consequent or
conclusion. All the rules tbat have any truth in their antedent will train and
contribute to the fuzzy conclusion set. If the antecedent is true to some degree of
membership, then consequent is also hue to that same degree (Bellman and Zadeh,
1995). This point also leads a natural way to combine multiple asswsments
(BuyuKozkan, and Feyzioglu, 2004).

The above Open view menu in FIS editor file and click Edit lules. Then
following display opens

Fig. 5.12 Rule Editor (Sample use)

The left-hand side contains the membership functions of input, Monitor. The right-
hand side has the membership tinctions of the output, Reliability. If the input side has
several variables, which are connected either by "and" or "or", the connection block is
in the lower left hand corner. In the above figure shows only one input variable, so the
connective is not used. The weight factor (default value =I) indicates the importance
of the rule and Fig. 5.13 shows the rule editor having five rules relates input (Monitor)
and output (Reliability).

Fig.5.l3 Rule Editor with five ntlcr (Sample cuc)

161
For the above mentioned sample m e , viewing rules in the overall picture of the
developed fUzzy system. Fmm the main FIS editor, choose h m 4ew menu view
rules.

Fig. 5.14. The rule base displayed graphically (sample came)

The vertical (red colour) line on the left indicates the value of the input, H.P
17" Monitor (Al3). Similarly, the vertical sttip (red colour bar) on the right hand side,
indicates the output, the reliability value, 0.917 (Very High). In the right-hand side
lower corner is the result of fuzzy reasoning. At this point, it is a fuzzy net. Applyin&
defuzzification method, in the Fig. 5.14, centre of gravity has been chosen, a crisp
value is obtained. Different input values can be tried by moving the red, vertical line
on the left-hand side (Fig. 5.14). Therefore, the output value is correct as per fuzzy
rules. Hence, fuzzy logic Mamdani module is suitable for solving critical problems
like four input variables with different levels (number of alternatives) and three output
variables of constant number of levels.
Data collection

The Data has been collected from weII-experienced system administrators;

they are working in college in different Computer application Labs. Care is taken, for
developing the rule base structure in order to avoid the flaws in expressing their
opinions like reliability, cost and so on. A general rule base structure shown in Table
5.3 for developing the rules having four inputs, viz., monitor with five levels (number
of alternatives), processor with four levels, Graphic card with three levels and sound
card with three levels are considered. The levels are introduced in each input as per
number of alternatives available in each component and membership bctions framed
based on alternative weights, obtained from AHP. (Venaktamuni, Rao, 2010), And
three output variables, viz., Reliability, Cost, Easy of assembly with five levels are
developed.
Table 53 Rule base structure for eonskudingrnles
53Rule base structure for eon~tructingdes(cant,)
 e structure for ~nstrndingnles (Cont)
53~ u l batt
$3 Rule base structurefor constructingrules (Cont)
Sample Fuzp, Interface Rules:

Rulel: If @ell TIT 17" Monitor) AND (P IV 3.0 Dual core Processor) AND
(Zebronics 256 Graphic Card) ANI) ( Real talk Sound card) Then (Cost
performance index is High) AND (Reliability is Medium ) AND (Easy of
Assembly is easy).

Rule 25: If @ell TFT 17" Monitor) A N D (P N 2.8 Dual core Processor) AND
(AT1 Graphic Card) A N D (Realtalk Sound card) Then (Cost
performance index is Low)A N D (Reliability is Very Low ) AND (Easy
of Assembly is Medium).

Rule 50: If (Dell 17" Monitor) AND (P IV 3.0 Dual core Processor) AND (512
Zebronics Graphic Card) AND (intex Sound card) Then (Cost performance
index is High) AND (Reliability is High) AND (Easy of Assembly is Easy).

Rule 75: If (H.P 17" Monitor) AND (P N 3.0 Dual core Processor) AND (Zebronics
256 Graphic Card) AND (Zebronics Sound card) Then (Cost performance
index) is Very Low AND (Reliability is Very Low), AND (Easy of Assembly
is Very Easy).

Rule 100: If (Samsung 21") AND (P N 2.8 Dual core Processor) AND (Zebronics 256
Graphic Card) AND (Zebronics Sound card) Then (Cost performance index is
High) AND (Reliability is Very low), AND (Easy of Assembly is very w y ) .

Rule 150: If (L.G Digital 21" Monitor) AND (P IV 3.0 Dual Core Processor) AND
(Zebronics 512 Graphic Card) AND (Zebronics Sound card) Then (Cost
performance index is Very High) AM) (Reliability is Very High) AND (Eaec
of Assembly is Very difficult).

Rule 175: IV 2.8 Dual core Processor) AND

If (L.G Digital 21': Monitor) AND (P
(Zebronics 512 Graphic Card) AND (Real talk Sound card) Then (Cost
performance index is Modem) AND (Reliability is High) AND (Easy of
Assembly is very easy).
 53.4 Fuzzy inference process for Developed model

RULE EDITOR

N S EDITOR
MEMBERSHIPNlJCTlON

RULE VIEWER SURFACE VIEWER

Fig. 5.15.0verall picture of fuzzy AHP model developed for computing various
system configurations with respect to output variables.
Fuwjlcation of input variables:
The first step is to take the inputs and determine the degree to which they
belong to each of the appropriate fizzy sets by means of membership functions. The
input is a crisp numerical value in the form of some rating scale limited to the
universe of discourse of the input variable. The scales and membership function
identify the ranges of input values corresponding to each firzzy linguistic level. In this
study we attempted, the numb& of alternatives under each module (input variable)
itself, number of levels for respective input. For example, five monitors, four
processors, three graphic cards and three sound cards considered for this study, and
then f i input
~ variable (type of monitor) five levels, four levels for second input,
three levels for third input and three levels for fourth input. The rules entered into the
Mamdani fuzzyhterhce system rule editor as shown in Fig. 5.15(a).
 Fig. 5.15(r) Fuzzy logic process rule editor
Aggregation of consequent across the rules
Since the final decision depends upon the aggregated output of all the rules in
the fuzzy inference system, we need to combine them by some means. Aggregation is
the process by which the fuzzy sets that represents the outputs of each rule are
combined to a single fuzzy set (Gungor and Arikan, 2000). The output of the
aggregation process is one fuzzy set for each output variable (Jiang and Chi-Hsing,
2001). In this study, we are using Mamdani fuzzy inference system that employs
compositional max-min rules (Fig. 5.1 0)).It means that, the aggregation process
utilizes the max-operator or the extension principle. (Kuo et. al.,2006).
 Fig.5.16. Deffuzification of Fuzzy logic model for input and output performance $dice8

Defuzzlfication of fuzzy output

The dekzification process creates a crisp performance index from fkzy

conclusion. Thus, the dekzification process is required to decipher the meaning of
the fuzzy conclusion and their membership values, and resolve conflicts between the
differing results, which may have been triggered during the mle evaluation. As much
as fuzziness helps the rule evaluation during the intermediate steps, the final output
for each variable is generally a single crisp number. So, given a fuzzy set that
encompasses a range of output values, we need to return one number, there by moving
from a fuzzy set to crisp output. Perhaps the most popular defuzzification method is
the centroid method; this shows the center of area under the output curve, output
variables crisp values are displayed at the top of output variables as shown in Fig.
5.16. The input variable values are entered in the specified space at lower left comer
of Rule Viewer as shown in Fig. 5.16.

Finally, the input-output mapping can be observed by viewing surfam. Choose

view menu ad under it view surface. Fig, 5.17 shows the Monitor weights and
Processor weights with respect to reliability. This type of thdimensionul graphical
view can analyze the customer preferences in fast manner, View d i f b n t
combination of inputs with respect to Rliahility, cost and easy of assanbly shown in
Fig. 5.18.

Fig. 5.17 Deffuzificationof fuzzy logic model for input and output performance
indices(GrrphicalView)

The developed Fuzzy logic process computes output variable values for
different combination of input configurations. But it is difficult to rank group of
configurations, this is solved through Data Envelopment Analysis and explained in
next section.
Fig. 5.18 DtfWfiation re#* of various c o m b W n 8 with nrpeet to output
variables.
5,4 INTEGRATION OF DEA INTO FUZZY AHP

The Deffuzification of fWy AHP model is integrated with Data Envelopment

Analysis process to identify the efficient and inefficient units (Computer systems).
DEA is the tool for analysing unlimited DMU irrespective of inputs and outputs. In
the current study, 25 computer systems (DMUs) are considered for DEA process, and
input and output variable values are computed from Funy AHP is presented in the
Table 5.4. We have discussed Data Envelopment Method (DEA) in the previous
chapter. The mathematical model for the current problem is similar to equation 4.2
(Ref, Chapter-IV).

r-I
Subject to

For the current problem the following values assigned to DEA model 4.2

Number of Inputs (m) = 4

Number of Outputs (s) = 3

Number of Decision Making Units (n) = 25

V,,U, are decision variables

X - Weight of Input

Y - Weight of Output
TORA optimization software is used to solve the above mentioned problem, the
object function value is the efficiency of Decision making unit (Computer system)
and it is presented in the last but one column of the Table 5.4. The input screen of
TORA software is shown in Fig.5.19. Similarly Change objective function with
respect to DMU and compute the efficiency, rank all the DMUs, b a d on the
efficiency. The computation of efficiency for DMUl is illustrate below

Max hjl= 0.3830V11-0.5750V~2+0.4800V~~

Subject to
V,,,U,,> 0 (s= 1,2,3, m = 1,2,3,4andj = 1,2,3,.....25) >0

After solving, objective function value which is the efficiency of DMUl = 0.8630 and
all other decision variables are above zero as shown in Fig. 5.20, so the problem is
feasible for the maximization of object function. Similarly, for every configuration
(DMU), the corresponding LPP is formulated and solved. The results are shown in
Table 5.4.
Fig. 5.19. TORA Input Screen for finding effidency of DMU1

Fig. 5.20 TOM Output wrecn for eadenry of DMUI

 Table 5.4 Cowolidated DefloPfflcatlonh&a for 25 system.

I I
5 5 CONCLUSIONS:
Fuzzy logic may be applied in different Product design issues, including
generic product sttIJcturin& association methods, and optimization of customer
and it is the way to reach the mass customization benefits. For combining two
different characteristics of layers in AHP (Modules and output variables such as
minimization of cost, maximization of reliability, and easy of assembly), we have
used Goal Programming (GP) with AHP (GAHP) and it is discussesed in Chapter-111.
Even though we could integrate GP, Pair wise comparison consistent judgement
matrices and computations takes lot of time and more complication. A method to
improve the above incorporate customer's vague data is discusses in the current
chapter. The buy-AHP model discussed in this chapter is proved to be simple, less
time taking and having less computational expense. Since we introduced a new
method of constructing membership functions of input and output variables from
AHP weights instead of separate data collection from the customer, the use of fizzy
AHP does not involve cumbersome mathematical operation and so it is easy to handle
the multi-attribute decision making problems like present case. It is ability to capture
the vagueness of human thinking style and effectively solving multi-attribute
decision-making problems. The illustrative example shows the flexibility, and
efficiency of the proposed model to directly tap the subjectivity and cwtomer
preferences.
Data Envelopment Analysis (DEA) is integrated with Fuzzy-AHP for better
defuzzifcation results than Fuzzy logic approach. The solution obtained from Fuzzy
AHP has to be presented in graphical form and is difficult to interpret. This limitation
is overbme by converting the fuzzy logic result to a DEA model. Since DEA is a
non-parametric approach for multiple inputs and outputs of homogeneous units, we
collected input and output variable indices from Fuzzy Logic. . Solved the
DEA model and the assumed alternatives (Computer systems). So we
conclude that DEA is most suitable for i3nalySiS the fuzz)' result, and is a new Fuzzy
DEAHP integration method for L a Product design.