Description
Innovative new products are the fuel for the most powerful growth engine you can connect to. You can grow without new products--AT&T sold essentially the same telephones for decades while becoming the world's largest telecommunications concern--but most small companies will find it difficult to grow at all, much less rapidly, without a constant stream of new products that meet customer needs.
Study on Decision making in Conceptual Phase
of Product Development
Table of Contents
Heading Page
Abstract 1
Chapter 1: Introduction 5
Section 1.1: Product Development
Section 1.2: Decisionmaking in General
Section 1.3: Decisionmaking in Conceptual Phase of
Product Development
Section 1.4: Thesis Structure
Chapter 2: Mathematical Settings
Section 2.1: Fuzzy Number
Sections 2.2: Operations on Fuzzy Number
Section 2.3: Linguistic Likelihood
Section 2.4: Linguistic Truth Value
Section 2.5: Information Content
Section 2.6: Monte Carlo Simulation of Discrete
Events
Chapter 3: Customer Needs Assessment
Section 3.1: Customer needs data collection
Section 3.2: Selection of reliable answers
Section 3.3: Monte Carlo simulation of unknown
answers
Section 3.4: Determination of truth values of product
feature Kano-evaluation
Section 3.5: Determination of truth value of product
feature status
Section 3.6: Determination of information content of
product feature status
Section 3.7: Determination of coherency measures of
product feature status
Section 3.8: Making final decision
Chapter 4: Sustainability Assessment
Section 4.1: Decision-relevant information
Section 4.2: Computational framework
i
5
7
9
9
13
13
14
15
19
21
25
29
30
34
35
38
39
42
43
47
51
52
54
Table of Contents continues
Section 4.3: Results
Chapter 5: Creativity Assessment
Section 5.1: C-K theory
Section 5.2: Differentiating creative and non-creative
concepts
Section 5.3: Results and discussions
Chapter 6: Concluding Remarks
Bibliography
Appendix A: Customer Needs Data Collection
List of Research Achievements
Acknowledgments
57
65
66
68
72
79
83
91
95
97
Pages 3-4, 12, 28, 50, 64, 78, 82, 90, 93-4, 96, and 98-9 are intentionally left blank.
ii
Abstract
Product Development studies the activities underlying a product life cycle in a
concurrent manner. In the conceptual phase of product development, a set of key
solutions are determined by using an appropriate decisionmaking approach. However,
decisionmaking in this phase is a difficult task to perform due to incomplete
information, lack of knowledge, and abundance of choice. This thesis describes
logical approaches for making decisions in conceptual phase of product development.
In particular, the emphasis is given on such issues as customer needs, sustainability,
and creativity. Multi-valued logic and information content from the context of
epistemic uncertainty have been used for the sake of computation. The thesis is
structured, as follows:
Chapter 2 describes the key mathematical entities used in this study such as
fuzzy number, imprecise probability, information content, and discrete event
simulation.
Chapter 3 describes customer needs assessment in conceptual phase of product
development. In this study, a set of field data has been collected from Bangladesh by
using Kano-model-based questionnaires on some features of small passenger
vehicles. The opinions obtained exhibit a high variability and controversy and simple
relative-frequency-based approaches have been found less effective in assessing the
product features, correctly. To solve this problem, a logical customer needs
assessment approach has been developed. The approach has been found effective in
classifying a product feature into the following categories: the feature "must,"
"should," and/or "could" be included in the product. The findings are useful for
developing more customer-focused passenger vehicles.
Chapter 4 describes a decisionmaking approach for assessing the sustainability
in conceptual phase of product development. One of the key sustainability issues of a
product is whether or not the product is made of environmentally friendly materials.
To assess the environmental burden of a material in conceptual phase of product
development, the information of greenhouse gas emissions (CO
2
, NO
X
, SO
X
) and
?
resource consumptions (e.g., water usage) of primary material production is needed.
This kind of information exhibits a high variability and incompleteness. To deal with
this issue, an entity called range compliance is used that logically defines the degree
of belongingness of a given numerical range to a linguistically defined class. Using
range compliance, the environmental burdens of hard materials used to produce a
grinding wheel (i.e., Alumina, Zirconia, Silicon Carbide, and Boron Carbide/Nitride)
have been identified. The findings help develop more environmentally friendly
abrasive tools (products) used in precision engineering.
Chapter 5 describes a decisionmaking approach for assessing a creative concept
in conceptual phase of product development. Creative concept is one of key
ingredients of developing a product. According to a theory called C-K theory, a
creative concept means a concept that is undecided with respect to the existing
knowledge at the point of time when it (the concept) is conceived. In this study, a
logical decisionmaking approach has been developed to assess the degree of
creativeness of a set of concepts. In particular, it has been found that for conceiving a
creative concept one should maximize the information content (in the epistemic
sense) of conceptual phase of product development. The findings help manage the
cognitive processes of conceptual phase of product development.
The last chapter, Chapter 6, provides the concluding remarks and discusses the
scope of further research opportunities.
?
?
?
Chapter 1
Introduction
This chapter describes the general background, scope, and limitation of this
thesis. For the sake of better understanding this chapter is structured into the
following sections: Product Development, Decisionmaking in General,
Decisionmaking in Conceptual Phase of Product Development, and Thesis Structure.
1.1. Product Development
Product Development is a field of study wherein the activities underlying
product lifecycle are studied in a concurrent manner (Ulrich and Eppinger 2004,
Dieter and Schmidt 2009). The lifecycle of a product can be represented in many
ways. One of the ways, which is relevant to this thesis, is schematically illustrated in
Fig. 1.1.
External Customers
(Real Customers)
Customer needs
Use
(Satisfaction)
Internal Customers
(Product Development Team)
Creativity
Conceptual Phase
(Key Solutions)
Detailed Design
Disposal
(Recycle, Downcycle, Landfill)
Manufacturing
Sustainability
Primary Materials
Production
Figure 1.1. A product development scenario.
?
A product development process first decides a set of key solutions (conceptual
phase). The key solutions are further pursued and a detailed design of the product is
produced. Based on the detailed design, manufacturing of the product is conducted.
The manufactured product is then made available to the external customers (real
customers) for use. Using the product, an external customer gets satisfied. If the
product can no longer be used, the product is disposed off. In this phase, the product
might experience recycling, down-cycling, remanufacturing, and/or land-filling.
However, the activities of product development are carried out by a team
(hereinafter referred to as internal customers) wherein a group of individuals from
different departments of an organization (or different organizations) work together to
materialize a product or a family of products. The internal customers first try to make
sure the liking-disliking of external customers (the potential real-customers who will
use the product to get satisfied). The internal customers need to be creative to suggest
many potential key solutions for satisfying the needs of external customers.
Therefore, the following questions might arise in the conceptual phase of product
development:
How to differentiate a creative key solution from a non-creative key solution?
What is the appropriate customer need model?
How to deal with the unknown customer needs?
How to classify the key solutions based on customer responses?
Nowadays, sustainability has become an important issue (Fiksel 2009).
Sustainability often means that the product is environmentally benign on top of other
desired performances. One should incorporate so-called Life-Cycle Assessment
(LCA) into the product development processes to ensure the sustainability (Donnelly
et al. 2004, Kobayashi, 2006). In addition to conventional sustainability assessment
(i.e., LCA), it is important to do scenario analysis (Umeda 2009, Fukushige et al.
2012) taking a broader perspective into the consideration. One of the remarkable
finding underlying scenario-analysis-based sustainability assessment is that the
primary production of materials used in the product plays a critical role to ensure the
sustainability (Higuchi et al. 2012). This implies the following question:
How to deal with the sustainability of materials (used in the product) in key
solution determination process?
?
However, around 80% cost of a product is decided by the key solution
determination process (in the conceptual phase) and it cannot be rectified by making
adjustments in the downstream of product lifecycle (Wood and Agonigo 1996). This
means that the decisionmaking in conceptual phase of product development is a
critical task. In addition, in conceptual phase the knowledge is very limited and there
is an abundance of choice (Dieter and Schmidt 2009). This means that the
decisionmaking in conceptual phase is a very difficult task to perform on top of its
criticalness, as mentioned before.
1.2. Decisionmaking in General
In late 1940s, Neumann and Morgenstern introduced a theory called game
theory. This theory has been accepted as a means to develop methods and tools for
rational decisionmaking. Two approaches have emerged from the game theoretic
practices. One of the approaches uses traditional settings of game theory (e.g.,
conflict/coalition analysis method using graph theory (e.g., see Fang et al. 1993,
Inohara and Hipel 2008 and the references therein)). The other approach has taken
the form of multiple-attribute utility analysis, wherein a set of attributes and their
relative weights are used to simultaneously evaluate (tradeoff) a set of given
alternatives, and, thereby, to select the optimal alternative corresponding to the
maximal utility (Saaty 1980, 1990). However, many authors have studied the
applicability of the multi-attribute utility analysis from the context of real-life
decisionmaking. Some of the salient points are briefly described below. In real-life
decisionmaking, a decision-maker often seeks a balanced alternative rather than an
optimal alternative and it is important to visualize the state of an alternative rather
than to automate the decisionmaking process (Kujawski 2005). Sometimes mental
biases of decision-makers affect the utility-based tradeoff and it is important to take
measures for reducing the biases in terms of problem statement, weights of
importance, alternative solution, evaluation data, scoring function, and combining
function (Smith et al. 2007). Sometimes the sequence of acts (i.e., bring the required
parties together, determine the needs, analyze the data, make a decision and
implement it) is important than the calculation process of tradeoff (Briggs and Little
2008). Sometimes determining the relevant set of criteria and their weights for
?
tradeoffs is a cumbersome task that involves the opinions of stakeholders (Keller et
al. 2008). Thus, in real-life settings it is not an easy task to utilize the utility based
decisionmaking approaches (i.e., rational decisionmaking approaches).
Opposed to rational decisionmaking, there is a faculty of thought of
decisionmaking called naturalistic decisionmaking (Klein 1989, Rasmussen 1993,
Hutton and Klein 1999, Klein 2008). In particular, human experts perform
naturalistic decisionmaking under the following context: time pressure,
incomplete/unreliable information, ill-defined goal, organizational constraints,
multiple decision-makers, and alike. Humans make decision under the
abovementioned context using a decisionmaking approach called recognition-primed
process (Klein 2008) that consists of the following steps: plausible goals, cues to
monitor, expectancies, and sequential action evaluation (Klein 1989, Hutton and
Klein 1999). There are three types of cognitive controls in recognition-primed
process, namely, 1) skill-based spontaneous act, 2) ruled-based conscious attention
and selection of relatively familiar action, and 3) knowledge-based conscious
attention and selection of relatively unfamiliar action (Rasmussen 1993).
Either it is a rational decisionmaking process or it is a naturalistic
decisionmaking process, the decision-relevant information may not necessarily be
crisp in nature. It might be granular in nature (Bellman and Zadeh 1970, Zadeh 1965,
1975, 1997). Zadeh and his colleagues have argued that the manifestation of human
cognitive is a set of "granular information"—imprecisely defined linguistic classes or
clusters of points—and multi-valued logic (known as fuzzy logic) is needed to
formally compute the linguistically expressed imprecise arguments (i.e., granular
information). Multi-attribute utility analysis community (i.e., rational
decisionmaking community) has integrated this idea to make the rational
decisionmaking more realistic (Yager 1978, Herrera and Herrera-Viedma 2000).
There are different models available to deal with the computational complexity of
stakeholder-driven heterogeneous formulation of decision problem and imprecisely
defined decision-relevant information (e.g., Herrera and Herrera-Viedma 2000,
Shamsuzzaman et al. 2003, Chen and Ben-Arieh 2006, Ullah 2005, Noor-E-Alam et
al. 2011). This kind of decisionmaking approach is suitable when the decision-
relevant information is dominated by personal preferences, judgments, and
?
vaguely defined alternatives, weights, and requirements.
1.3. Decisionmaking in Conceptual Phase of Product Development
As mentioned before, decisionmaking in conceptual phase of product
development decides around 80% cost of the product and the decisionmaking process
suffers lack of knowledge and abundance of choice (Wood and Agonigo 1996, Dieter
and Schmidt 2009, Ullman 2009, Ulrich and Eppinger 2004, Ullah 2005). Therefore,
the decision-relevant information in conceptual phase is predominated by personal
preferences, judgments, and vaguely defined alternatives, weights, and requirements.
As a result, granular information based decisionmaking approach is suitable for
making decisions in conceptual phase of product development (Ullah 2005).
However, decisionmaking in conceptual phase of product development requires
an explicit measure that quantifies the lack/abundance of knowledge. For example,
consider the measures called degree of certainty of knowledge in robust
decisionmaking (Ullman 2006) and certainty compliance (entropy) in
general-pinion-desire based decisionmaking (Ullah 2005). In addition, a measure is
needed to quantify the degree of fulfillment of requirement, though the requirement
might be vaguely defined or vary across the external customers. For example,
consider the measure called criteria satisfaction in robust decisionmaking (Ullman
2006) and requirement compliance (entropy) in general-opinion-desire based
decisionmaking (Ullah 2005).
The explanation refers to the fact that a two-dimensional decision measure is
needed for making decisions in conceptual phase of product development. One of the
coordinates of the measure should measure the degree of certainty of knowledge and
the other should measure the degree of fulfillment. However, it would be convenient
if the decision measure is directly related to some of the important principles of
systems design. In this case, general-opinion-desire based decisionmaking is a
desirable one because the certainty entropy and requirement entropy (Ullah 2005) are
directly related to general systems design principles (i.e., information axiom of
axiomatic design of systems) (Suh 1990, 1998, Ullah 2005b).
1.4. Thesis Structure
?
The remainder of this thesis is organized as follows:
Chapter 2 describes the key mathematical entities used in this study namely,
fuzzy number, range compliance, linguistic likelihood, information content, and
discrete event simulation.
Chapter 3 describes customer needs assessment in conceptual phase of product
development. In this study, a set of field data has been collected from Bangladesh by
using Kano-model-based questionnaires on some features of small passenger
vehicles. The opinions obtained exhibit a high variability and controversy and simple
relative-frequency-based approaches have been found less effective in assessing the
product features, correctly. To solve this problem, a logical customer needs
assessment approach has been developed. The approach has been found effective in
classifying a product feature into the following categories: the feature "must,"
"should," and/or "could" be included in the product. The findings are useful for
developing more customer-focused passenger vehicles.
Chapter 4 describes a decisionmaking approach for assessing the sustainability in
conceptual phase of product development. One of the key sustainability issues of a
product is whether or not the product is made of environmentally friendly materials.
To assess the environmental burden of a material in conceptual phase of product
development, the information of greenhouse gas emissions (CO
2
, NO
X
, SO
X
) and
resource consumptions (e.g., water usage) of primary material production is needed.
This kind of information exhibits a high variability and incompleteness. To deal with
this issue, an entity called range compliance is used that logically defines the degree
of belongingness of a given numerical range to a linguistically defined class. Using
range compliance, the environmental burdens of hard materials used to produce
agrinding wheel (i.e., Alumina, Zirconia, Silicon Carbide, and Boron Carbide/Nitride)
have been identified. The findings help develop more environmentally friendly
abrasive tools (products) used in precision engineering.
Chapter 5 describes a decisionmaking approach for assessing a creative concept
in conceptual phase of product development. Creative concept is one of key
ingredients of developing a product. According to a theory called C-K theory, a
creative concept means a concept that is undecided with respect to the existing
knowledge at the point of time when it (the concept) is conceived. In this study, a
??
logical decisionmaking approach has been developed to assess the degree of
creativeness of a set of concepts. In particular, it has been found that for conceiving a
creative concept one should maximize the information content (in the epistemic
sense) of conceptual phase of product development. The findings help manage the
cognitive processes of conceptual phase of product development.
The last chapter, Chapter 6, provides the concluding remarks and discusses the
scope of further research opportunities.
??
??
Chapter 2
Mathematical Settings
Chapter 2 describes the mathematical entities used in this thesis, namely, fuzzy
number, range compliance, linguistic likelihood, information content, and discrete
event simulation. For the sake of better understanding, this chapter is organized into
the following five sections: Fuzzy Number, Operations on Fuzzy Numbers,
Linguistic Likelihood, Information Content, and Monte Carlo Simulation of Discrete
Events.
2.1 Fuzzy Number
A fuzzy number A is a special type of fuzzy subset (Zadeh 1965) wherein the
universe of discourse X is a segment of real-line ? and the maximum membership
value of the membership function ·
A
(x) is equal to 1 (condition of normality). In
addition, ·
A
(x) fulfills the conditions of convexity, continuity, and compactness
(Dubois and Prade 1978, Ullah and Harib 2006). Fuzzy numbers are suitable for
bringing the linguistically defined qualitative entities into a formal computation.
1
Comfortable
0 .7 5
0.5
Cold
0 .2 5
0
0
5
Hot
10 15 20 25 30 35 40 45 50 55 60
Temperature x, (°C)
Figure 2.1. Example of fuzzy number.
For the sake of understanding, consider the three fuzzy numbers labeled Hot,
??
(
.
)
(
x
)
·
Cold, and Comfortable in the universe of discourse X = [0°C,60°C]. The membership
functions are as follows:
·
A
: X = |0,60| ÷ |01| ,
x ·
A
(x) = ÷ x
25 ÷
0
A = Cold
x ·
A
(x) = x ÷ 15 , 35 - x (2.1)
25 ÷ 15 35 ÷
25
A = Comfortable
x ·
A
(x) = x ÷
60 ÷
25
A = Hot
Note that ·
A
(x) is called the membership value of x with respect to A. It, ·
A
(x), is
also called the degree of belief of x in terms of A. The Truth-Value (TV) of the
proposition "x is A" is also equal to ·
A
(x).
For example, let x be 20°C. According to equation (2.1), the membership value of
x = 20°C is equal to 0.2 for A = Cold, 0.5 for A = Comfortable, and 0 for A = Hot.
This means that the TV of a proposition "20°C is a Cold temperature" is equal to 0.2, of
a proposition "20°C is a Comfortable temperature" is equal to 0.5, and of a
proposition "20°C is a Hot temperature" is equal to 0.
In this thesis, fuzzy numbers are used to define such entities as linguistic
likelihoods, linguistic truth values, and classes like high, low, moderate, and alike for
the sake of formal computation.
2.2 Operations on Fuzzy Number
There are many operations applied on a fuzzy number or on a set of fuzzy
numbers. The operations relevant to this thesis are Alpha-cut (Dubois and Prade
1978), Expected Value (Dubois and Prade 1978), and Range Compliance (Ullah
2008).
An Alpha-cut of a fuzzy number A denoted by A
o
is the crisp range for all ·
A
(x) >
o, o e (0,1]. Thus,
A
o
= {x | ·
A
(x) > o e (0 1|} , (2.2)
For example, A
0.5
= [20°C,30°C], if A = Comfortable (see equation (2.1)). The largest
alpha-cut is called Support (Supp(.)). Thus,
Supp(A) = (a,b) {a,b}= |a,b| max(A
o
| ¬o e (0,1|) = (a,b) (2.3)
For example, Supp(A) = [15°C,35°C], if A = Comfortable (see equation (2.1)).
??
Expected value of a fuzzy number A is the first moment of the shape of its
membership function ·
A
, denoted as E(A). Thus,
E ( A) =
} · (x)xdx A
(2.4)
} · (x)dx A
For example, E(A) = 25°C, if A = Comfortable (see equation (2.1)).
Range compliance of a numerical range L is its average membership value with
respect to a fuzzy number A, denoted as R(L,A).
}·
A
(x)dx
R(L, A) =
L
L?
(L? _ L). (L? _ Supp(A))
(2.5)
Here, L? is the segment of L that belongs to Supp(A).
For example, R(L,A) = 0.583, if L = [10°C,30°C] and A = Comfortable.
Range compliance as defined in equation (2.5) will be used in Chapter 4 for
sustainability assessment. Other operations will be used in Chapters 3 and 5.
2.3 Linguistic Likelihood
In real-life cases, probability of an event is difficult to known accurately
(O'Hagan and Oakley 2004). One of the useful representations of real-life probability
is called imprecise probability (Pr
imp
) wherein both upper and lower limits of
probability are used to define the probability of an event (Walley 1991). Alternatively,
the imprecision associate with the probability of an event can be represented by a set
of fuzzy numbers defined in the universe of discourse [0,1] (i.e., possible values of
probability). The labels of the fuzzy numbers are called Linguistic Likelihoods (LLs)
(most likely, less unlikely, and alike). It is useful in determining the probability of an
event given the relative frequency (fr) of the event (Ullah and Harib 2006, Ullah and
Tamaki 2011). It is needless to say that the membership functions of LLs denoted as
·
LLi
(.), i = 1,...,n, are defined in the universe of discourse [0,1].
However, the linguistic counterpart of fr corresponds to an LLj, j e{1,...,n} that
corresponds to the maximum membership value for fr. This means that if
max(·
LLi
(fr) i = 1,...,n) = ·
LLj
(fr) then Pr
imp
(fr) = LLj, j e {1,...,n}. Later the fuzzy
number corresponding to LLj can be used for having a crisp probability for the sake
of calculation. In this thesis, three cases of LL are considered.
??
qu su ns sl ql 1
0.75
0.5
0.25
0
0 0.2 0.4 0.6 0.8 1
fr
Figure 2.2. Membership functions of five LL (Case 1).
1
0.75
0.5
0.25
0
mu
mu
qu qu
su su ns ns
sls
l
ql
ql
ml ml
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
fr
Figure 2.3. Membership functions of seven LL (Case2).
eu
1
eu
mu
u m
qu
qu
su
su
ns
ns
sl
sl
ql
ql
ml
ml e
el
l
0.75
0.5
0.25
0
0 0.125 0.25 0.375 0.5 0.625 0.75 0.875 1
fr
Figure 2.4. Membership functions of seven LL (Case 3).
Case 1 corresponds to five LLs, namely, quite unlikely (qu), some unlikely (su),
not sure (ns), some likely (sl), and quite likely (ql). Case 2 corresponds to seven LLs,
namely, most unlikely (mu), quite unlikely (qu), some unlikely (su), not sure (ns),
some likely (sl), quite likely (ql), and most likely (ml). Case 3 corresponds to nine
??
µ
(
.
)
(
f
r
)
µ
µ
(
.
A
)
(
f
r
)
µ
(
.
)
(
f
r
)
LLs, namely, extremely unlikely (eu), most unlikely (mu), quite unlikely (qu), some
unlikely (su), not sure (ns), some likely (sl), quite likely (ql), most likely (ml), and
extremely likely (el).
Figure 2.2 illustrates the membership functions corresponding to Case 1. The
definitions of the membership functions illustrated in Fig. 2.2 are as follows:
·
qu
( fr) = max 0,min 1, 0.3 ÷ fr
0.3 ÷ 0
(2.6.1)
·
su
( fr) = max 0,min fr ÷ 0 , 0.5 ÷ fr
0.3 ÷ 0 0.5 ÷
0.3
(2.6.2)
·
ns
( fr) = max 0,min fr ÷ 0.3 , 0.7 ÷ fr
(2.6.3)
0.5 ÷ 0.3 0.7 ÷ 0.5
·
sl
( fr) = max 0,min fr ÷ 0.5 , 1 ÷ fr
(2.6.4)
0.7 ÷ 0.5 1 ÷ 0.7
·
ql
( fr) = max 0,min 1, fr ÷ 0.7
1 ÷
0.7
(2.6.5)
Figure 2.3 illustrates the membership functions corresponding to Case 2. The
definitions of the membership functions illustrated in Fig. 2.3 are as follows:
·
mu
( fr) = max 0,min 1, 0.1÷ fr
0.1÷
0
(2.7.1)
·
qu
( fr) = max 0,min fr ÷ 0 , 0.3 ÷ fr
(2.7.2)
0.1÷ 0 0.3 ÷ 0.1
·
su
( fr) = max 0,min fr ÷ 0.1 , 0.5 ÷ fr
(2.7.3)
0.3 ÷ 0.1 0.5 ÷ 0.3
·
ns
( fr) = max 0,min fr ÷ 0.3 , 0.7 ÷ fr
0.5 ÷ 0.3 0.7 ÷
0.5
(2.7.4)
·
sl
( fr) = max 0,min fr ÷ 0.5 , 0.9 ÷ fr
0.7 ÷ 0.5 0.9 ÷ 0.7
(2.7.5)
·
ql
( fr) = max 0,min fr ÷ 0.7 , 1÷ fr
0.9 ÷ 0.7 1÷
0.9
(2.7.6)
·
ml
( fr) = max 0,min 1, fr ÷ 0.9
1÷
0.9
(2.7.7)
Figure 2.4 illustrates the membership functions corresponding to Case 3. In the
conceptual phase of product development, a set of key solutions are determined by
using an appropriate decisionmaking approach.
??
·
eu
( fr) = max 0,min 1, 0.125 ÷ fr
0.125 ÷
0
(2.8.1)
·
mu
( fr) = max 0,min
fr ÷ 0 , 0.25 ÷ fr
0.125 ÷ 0 0.25 ÷
0.125
(2.8.2)
·
qu
( fr) = max 0,min fr ÷ 0.125 , 0.375 ÷ fr
(2.8.3)
0.25 ÷ 0.125 0.375 ÷ 0.25
·
su
( fr) = max 0,min
fr ÷ 0.25 , 0.5 ÷ fr
(2.8.4)
0.375 ÷ 0.25 0.5 ÷ 0.375
·
ns
( fr) = max 0,min fr ÷ 0.375 , 0.625 ÷ fr
(2.8.5)
0.5 ÷ 0.375 0.625 ÷ 0.5
·
sl
( fr) = max 0,min
fr ÷ 0.5 , 0.75 ÷ fr
(2.8.6)
0.625 ÷ 0.5 0.75 ÷ 0.625
·
ql
( fr) = max 0,min fr ÷ 0.625 , 0.875 ÷ fr
0.75 ÷ 0.625 0.875 ÷ 0.75
·
ml
( fr) = max 0,min
(2.8.7)
fr ÷ 0.75 , 1÷ fr
(2.8.8)
0.875 ÷ 0.75 1÷ 0.875
·
el
( fr) = max 0,min 1, fr ÷ 0.875
1÷ 0.875
(2.8.9)
Table 2.1 Summary of LLs
Linguistic Likelihoods (LLs)
Items
Expected
value
Alpha-cut
at o = 0 . 5
Range of
fr
Cases
1
2
3
1
2
3
1
2
3
eu
-
-
0.042
-
-
[0,0.0625]
-
-
[0,0.0625)
mu
-
0.033
0.125
-
[0,0.05]
[0.0625,
0.1875]
-
[0,0.05)
[0.0625,
0.1875)
qu
0.1
0.133
0.25
[0,0.15]
[0.05,
0.2]
[0.1875,
0.3125]
[0,0.15)
[0.05,
0.2)
[0.1875,
0.3125)
su
0.267
0.3
0.375
[0.15,0.4]
[0.2,0.4]
[0.3125,
0.4375]
[0.15,0.4)
[0.2,0.4)
[0.3125,
0.4375)
ns
0.5
0.5
0.5
[0.4,0.6]
[0.4,0.6]
[0.4375,
0.5625]
[0.4,0.6)
[0.4,0.6)
[0.4375,
0.5625)
sl
0.733
0.7
0.625
[0.6,0.85]
[0.6,0.8]
[0.5625,
0.6875]
[0.6,0.85)
[0.6,0.8)
[0.5625,
0.6875)
gl
0.9
0.867
0.75
[0.85,1]
[0.8,0.95]
[0.6875,
0.8125]
[0.85,1]
[0.8,0.95)
[0.6875,
0.8125)
ml
-
0.967
0.875
-
[0.95,1]
[0.8125,
0.9375]
-
[0.95,1]
[0.8125,
0.9375)
el
-
-
0.958
-
-
[0.9375,
1]
-
-
[0.9375,
1]
Table 2.1 summarizes the expected values, alpha-cuts at o = 0.5, and ranges of fr
for three cases of LLs as defined above. In particular, the ranges of fr will be used to
find out the linguistic counterparts of a given fr. For example, if fr = 0.3 then its
??
linguistic counterpart is su according to Case 1, is su according to Case 2, is qu
according to Case 3. See the ranges of fr in Table 2.1 wherein fr = 0.3 belongs. Once
the linguistic counterpart of fr is determined, the corresponding linguistic likelihood
will be used to carry out the subsequent calculations. For example, if fr = 0.3 then its
linguistic counterpart is su according to Case 1 and the expected value of it (su),
0.267 according to Table 2.1, will be used to carry out the subsequent calculation.
Similarly, if fr = 0.3 then its linguistic counterpart is su according to Case 2 and the
expected value of it (su), 0.3 according to Table 2.1, will be used to carry out the
subsequent calculation. Again, if fr = 0.3 then its linguistic counterpart is qu
according to Case 3 and the expected value of it (qu), 0.25 according to Table 2.1,
will be used to carry out the subsequent calculation. This kind of scheme is helpful in
discrete event simulation using a relatively small set of data points.
2.4 Linguistic Truth Value
It is mentioned that the membership value can be considered the truth value of
proposition. For example, recall the proposition of Section 2.1: p = "20°C is a Cold
temperature." The truth value of p is equal to 0.2, TV(p) = 0.2, because ·
Comfortable
(x =
20°C) = 0.2 according to the membership function of the fuzzy number
"Comfortable" as shown in Fig. 2.1 and defined in equation (2.1).
However, there are cases wherein it would be difficult to explicitly construct a
membership function and calculate TV of a proposition from it. In such cases, one
can assign a preferential/judgmental TV using a phrase (mostly true, some false, etc.)
to a proposition. For example, consider the following proposition: p(Z, attractive) =
"Z is an attractive attribute for this product." One can assign a TV = "quite true" to p(Z,
attractive) based her/is judgment. To bring such preferential/judgmental TV, one can
use linguistic TV defined by a set of fuzzy numbers, as it is done for linguistic
likelihoods.
Figure 2.5 illustrates the linguistic TV defined by five fuzzy numbers labeled
mostly false (mf), perhaps false (pf), not sure (ns), perhaps true (pt), and mostly true
(mt). These TVs are used in Chapter 5. Note that the membership functions are the
same compared to those of Case 1 LLs.
??
mf pf ns pt mt
1
0.75
0.5
0.25
0
0 0.2 0.4 0.6 0.8 1
TV
Figure 2.5. Linguistic TV using five fuzzy numbers.
The membership functions are defined as follows:
·
mf
(TV ) = max 0,min 1,0.3 ÷ TV
0.3 ÷
0
(2.9.1.1)
·
pf
(TV ) = max 0,min TV ÷ 0 , 0.5 ÷ TV
(2.9.1.2)
0.3 ÷ 0 0.5 ÷ 0.3
·
ns
(TV ) = max 0,min TV ÷ 0.3 ,0.7 ÷ TV
(2.9.1.3)
0.5 ÷ 0.3 0.7 ÷ 0.5
·
pt
(TV ) = max 0,min TV ÷ 0.5 ,1 ÷ TV
(2.9.1.4)
0.7 ÷ 0.5 1 ÷ 0.7
·
mt
(TV ) = max 0,min 1,TV ÷ 0.7
(2.9.1.5)
1 ÷ 0.7
1
0
.
7
5
0
.
5
µ
(
.
)
(
T
V
)
µ
(
.
)
A
(
T
V
)
µ
0.25
0
mf
mu
qu qf
su
pf
ns
ns
psl
t
q
ql
t
mt
ml
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
TV
Figure 2.6. Linguistic TV using seven fuzzy numbers.
Similarly, Fig. 2.6 illustrates the linguistic TV defined by seven fuzzy numbers
??
labeled mostly false (mf), quite false (qf), perhaps false (pf), not sure (ns), perhaps
true (pt), quite true (qt), and mostly true (mt). The definitions of the membership
functions illustrated in Fig. 2.3 are as follows:
·
mf
(TV ) = max 0,min 1, 0.1 ÷ TV
0.1 ÷
0
(2.9.2.1)
·
qf
(TV ) = max 0,min TV ÷ 0 , 0.3 ÷ TV
(2.9.2.2)
0.1 ÷ 0 0.3 ÷ 0.1
·
pf
(TV ) = max 0,min TV ÷ 0.1 , 0.5 ÷ TV
(2.9.2.3)
0.3 ÷ 0.1 0.5 ÷ 0.3
·
ns
(TV ) = max 0,min TV ÷ 0.3 , 0.7 ÷ TV
(2.9.2.4)
0.5 ÷ 0.3 0.7 ÷ 0.5
·
pt
(TV ) = max 0,min TV ÷ 0.5 , 0.9 ÷ TV
(2.9.2.5)
0.7 ÷ 0.5 0.9 ÷ 0.7
·
qt
(TV ) = max 0,min TV ÷ 0.7 ,1 ÷ TV
(2.9.2.6)
0.9 ÷ 0.7 1 ÷ 0.9
·
mt
(TV ) = max 0,min 1,TV ÷ 0.9
1 ÷ 0.9
(
2
.
9
.
2
.
7
)
These TVs are used in Chapter 3. Note that the membership functions are the
same compared to those of Case 2 LLs.
Table 2.2 summarizes the expected values of the linguistic TVs defined above. If
one assigns the truth value of a proposition equal to perhaps true (pt), then the
subsequent computation will be carried out by using its expected value 0.733.
Table 2.2. Summary of linguistic TVs.
Linguistic TVs
Items
mf qf pf ns pt qt mt
Expected Five 0.1 - 0.267 0.5 0.733 - 0.9
Value Seven 0.033 0.133 0.3 0.3 0.7 0.867 0.967
2.5 Information Content
In 1940s, Shannon introduced the concept of information content as a part of his
information theory wherein an obvious event has low information content and less
likely event has high information content. Thus, if the probability of an event is Pr,
then the information content of the event is given by -log(1/Pr). In systems design,
??
Suh have utilized this concept introducing an axiom called the Information Axiom:
minimize the information content of a design (Suh 1990, 1998). According to the
information axiom, the information content of a functional requirement (FR) of a
system is defined as follows:
I (FR) = ÷log 1
S
(2.10)
In equation (2.10), S is the area under the probability density function of system
range (sr) (the performance of the system designed) for a given design range (dr) (the
requirement defined the designer). A schematic illustration of S, sr, and dr is shown
in Fig. 2.7.
15
.1
05
0
15
20
S
dr
25 FR 30
sr
35
40
Figure 2.7. Definition of information content for systems design.
I(FR) can be minimized by increasing the value of S, i.e., matching sr with respect
to dr. This means that minimization of information content means maximization of
requirement fulfillment. Therefore, information content defined in equation (2.10)
actually determines the degree of requirement fulfillment.
Note that in conceptual phase of product development (the focus of this thesis), it
would be difficult to clearly define the probability density function to represent sr
and the range called dr. Therefore, information content defined in equation (2.10)
(i.e., degree of fulfillment of requirement) may not be applied in conceptual phase of
product development. In addition, in conceptual phase of product development, not
only the degree of fulfillment requirement but also the degree of knowledge should
??
P
r
r
(
(
F
R
)
)
P
F
R
get proper attention (Ullah 2005a-b, Ullman 2006).
The above explanation implies that for conceptual phase of product development,
information content of a key solution has two facets, one is the degree of knowledge
and the other is the degree of fulfillment. In this thesis, a two-dimensional
information content (CE,RE) scheme is used. Here CE means Certainty Entropy that
measures the degree of knowledge in the interval [0,1] and RE means Requirement
Entropy (RE) that measures the degree of fulfillment of a linguistically defined
requirement (see Ullah 2005a, Ullah and Harib 2008). The calculation is done on the
truth values of a set of propositions, TV(P1),...,TV(Pn),TV(P
R
). Here, P1,...,Pn are
general propositions and P
R
is the requirement proposition. The truth values of
P1,...,Pn, TV(P1),...,TV(Pn), are assigned or calculated and the truth value of P
R
,
TV(P
R
), is calculated from TV(P1),...,TV(Pn).
In particular, CE is defined as follows:
n
CE =
¿I
i
c (TV (Pi))
n
(2.11.1)
This means that CE is the average epistemic information content, I
c
(.), of
TV(P1),...,TV(Pn). The epistemic information content I
c
(.) is determined as follows:
I
c
(TV (Pi)) = max 0,min TV ÷ 0 ,1 ÷ TV
0.5 ÷ 0 1 ÷
0.5
(2.11.2)
Figure 2.8 illustrates I
c
(.). As seen from Fig. 2.8, epistemic information content of the
truth value of a proposition is a tent function in the universe of discourse of [0,1].
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1
TV(.)
Figure 2.8. Epistemic information content.
This function ensures that a completely true or false proposition does not have
??
I
?
(
T
V
(
.
)
)
any information content (I
c
(TV = 1 or 0) = 0), whereas if the proposition is neither
true nor false, it has the highest information content (I
c
(TV = 0.5) = 1). For other
cases, the information content is between 0 and 1.
Recall the other coordinate of information content of a key solution in conceptual
phase of product development, RE, Requirement Entropy. RE measures the entropy
of the requirement given by P
R
, which is just the opposite of the degree of fulfillment
of requirement given by P
R
. Thus, if the requirement is fully fulfilled, RE should be
equal to zero (lowest entropy) and if the requirement is fully unfulfilled, RE should
be equal to unit (highest entropy). If the requirement is partially fulfilled, RE is
between 0 and 1. The following function can be used to measure RE (Ullah 2005a):
RE = max 0,min 1, a ÷ TV (P
R
)
a ÷ b
(2.11.3)
a = max(TV (Pi)| i = 1, ,n) b = min(TV (Pi)| i = 1, ,n)
The procedure to determine the TV(P
R
) from TV(P1),...,TV(Pn) shown in Ullah
2005a is used in this thesis. A typical nature of RE is illustrated in Fig. 2.9
corresponding to a = 0.9 and b = 0.05.
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1
TV(P
R
)
Figure 2.9. A function to determine requirement entropy.
The two-dimensional information content of key solutions can be plot on a RE
versus CE plot and a measure called coherency measure (ì) can be determined. The
??
R
E
coherency measure actually aggregates the variability in the information content of a
key solution using the following expression:
ì = e + f + g + h + ( f ÷ e)(h ÷ g) (2.11.4)
In ideal case, ì = 0 that means the solution fully fulfills the requirement the
knowledge of the solution is complete. In reality it does not happen. What is seen in
reality is schematically illustrated in Fig. 2.10. As seen from Fig. 2.10, both key
solutions suffer lack of knowledge and partial fulfillment of requirement, both CE,
RE> 0. However, the variability and magnitude in (CE,RE) points is less for key
solution 1 compared to those of key solution 2. This results a relatively low ì for key
solution 1 compared to that of key solution 2, i.e., ì
1
< ì
2
. Therefore, key solution 1
is a preferable key solution compared to the other. This way decisionmaking can be
carried out in conceptual phase of product development.
1
0.75
0.5
e
Solution 1
Solution 2
f
0.25 ì
1
< ì
2
g h
0
0 0.25 0.5 0.75 1
CE
Fig. 2.10. Decisionmaking using two-dimensional information content.
2.6 Monte Carlo Simulation of Discrete Events
Monte Carlo simulation of discrete events is a useful method to know the
unknown answers of external customers (Ullah and Tamaki 2011). In this thesis,
similar method is used to simulate the unknown answers of external customers. The
simulation steps are explained in Table 2.3. As explained in Table 2.3, the simulation
??
R
E
needs two inputs: a vector of events (ev(1) = Attractive (A), ev(2) = One-dimensional
(O), ev(3) = Must-be (M), ev(4) = Indifferent (I), ev(5) = Reverse (R), ev(6) =
Questionable (Q)) and another vector of their relative frequencies (fr(ev(i))) (e.g.,
(0.1,0.4,0.3,0.1,0.1,0). An example is shown in Table 2.3. Afterward, the user
chooses a set of linguistic likelihoods either from Case 1 or from Case 2 or from Case
3. The linguistic counterpart of each relative frequency fr(i) denoted by LL(i) is
determined by using the ranges listed in the "Range of fr" rows of Table 2.1. For
example, if fr(i) = 0.1, then its linguistic counterpart is qu. The reason is that fr(i) =
0.1 belongs to the range [0,0.15)--a range derived from the alpha-cut qu
corresponding to Case 1 (see Table 2.1). The expected values of the respective
linguistic likelihoods, E(i), are used in the subsequent calculations.
In the subsequent calculations, first, the probability of each event is calculated
by normalizing each expected value, as follows:
E(i)
Pr(i) =
6
¿E(i)
i=1
(2.12.1)
Afterward, the cumulative probability of an event is calculated, as follows:
i
CPr(i) =
¿Pr(i)
j=1
Table 2.3. An example of the settings of simulation
Vector of events (input)
(2.12.2)
Items
fr(i)
(input)
LL(i)
(based on LLs of Case 1)
E(i)
(calculated)
Pr(i)
(calculated)
CPr(i)
(calculated)
ev(1)
A
0.1
qu
0.1
0.079
0.079
ev(2)
O
0.4
ns
0.5
0.395
0.474
ev(3)
M
0.3
su
0.267
0.211
0.684
ev(4)
I
0.1
qu
0.1
0.079
0.763
ev(5)
R
0.1
qu
0.1
0.079
0.842
ev(6)
Q
0
qu
0.1
0.079
0.921
Using the following algorithm a set of simulated events, se(s), s = 1,...,N, are
generated.
??
Define the number iterations N (a relatively large integer)
For s = 1,...,N
Generate a random number RAND in the interval [0,1]
If RAND = [0,CPr(ev(1))) Then se(s) = ev(1)
For k = 2,...,6
If RAND = [CPr(ev(k-1)), CPr(ev(k))) Then se(s) = ev(k)
If RAND = [CPr(ev(5)), CPr(ev(6))] Then se(s) = ev(6)
Needless to say that se(s) e {A, O, M, I, R, Q} for all s = 1,...,N.
??
??
Chapter 3
Customer Needs Assessment
This section deals with the customer needs assessment based on the work of
Rashid et al. 2010 and Rashid et al. 2012. The following issues are emphasized:
How to deal with the unknown customer needs?
How to classify the key solutions based on customer responses?
The framework used in the customer needs assessment process is illustrated in
Fig. 3.1.
Step 1: Collection of customer needs
data using Kano model
Step 2: Selection of reliable answers
Step 3: Monte Carlo simulation of
unknown answers
Step 4: Determination of truth value of
product feature Kano evaluation
Step 8: Making final decision
Step 7: Determination of coherency
measures of product feature status
Step 6: Determination of information
content of product feature status
Step 5: Determination of truth value of
product feature status
Figure 3.1. Customer needs assessment framework.
As seen from Fig. 3.1, the framework consists of eight steps, as follows:
Step 1: Collection of customer needs data using Kano model
Step 2: Selection of reliable answers
Step 3: Monte Carlo simulation of unknown answers
Step 4: Determination of truth value of product feature Kano-evaluation
Step 5: Determination of truth value of product feature status
??
Step 6: Determination of information content of product feature status
Step 7: Determination of coherency measure of product feature status
Step 8: Making final decision
Step 1 deals with the customer needs data collection using Kano model from
Bangladesh on some selected features of small passenger vehicles (cars). This step is
described in details in Section 3.1. Step 2 deals with the determination of reliable
answers of the respondents. This step is described in Section 3.2. Step 3 deals with
the Monte Carlo simulation of unknown answers. This step is described in Section
3.3. Step 4 deals with the determination of truth value of Kano evaluation of product
feature. This step is described in Section 3.4. Step 5 deals with the determination of
truth value of product feature status. This step is described in Section 3.5. Step 6
deals with the determination of information content of product feature status. This
step is described in Section 3.6. Step 7 deals with the determination of coherency
measure of product feature status. This step is described in Section 3.7. The final step,
Step 8, deals with the final decisionmaking using the coherency measure. This step is
described in Section 8.
3.1. Customer needs data collection
Recall Fig. 1.1. As shown in Fig. 1.1, the external customers should get the
desired satisfaction using a product. Therefore, the internal customers (product
development team members) should be aware of the customer needs beforehand.
Otherwise, the product might not be a useful one. The internal customers first decide a
preliminary set of key solutions and prepare a set of questions. A selected segment of
potential external customers (real customers) then answer the questions. Using the
answers obtained, the internal customers try to identify the usefulness of a proposed
key solution or a set of key solutions. Kano model is one of the useful models by
which one can ask questions regarding a product feature so as to classify it (product
feature) into One-dimensional (O) feature, Attractive (A) feature, Must-be (M)
feature, Indifferent (I) feature, Reverse (R) feature, or Questionable (Q) feature
(Kano et al. 1984, Berger et al. 1993, Kahn 2004, Yang 2008, Xu et al. 2009, Ullah
and Tamaki 2011). Figure 3.2 schematically illustrates the implication of O, A, M, I,
??
R, and Q. According to Fig. 3.2, a feature is considered M, if its absence produces
absolute dissatisfaction and its presence does not increase the satisfaction. A feature
is considered O, if its fulfillment helps increase the satisfaction and vice versa. A
feature is considered A, if it leads to a greater satisfaction, whereas it is not expected
to be in the product. A feature is considered I, if its presence or absence does not
contribute to the customers' satisfaction. A feature is considered R if its presence
causes dissatisfaction and vice versa. If the customer answers inconsistently, the
feature is considered Q. However, to know whether or not a feature is O, A, M, I,R, Q
a two-dimensional questionnaire is prepared for each feature. Table 3.1 shows an
example of Kano questionnaire.
Figure 3.2. Kano model of customer needs (based on Ullah and Tamaki 2011).
Table 3.1. An example of Kano questionnaire
Functional answer Dysfunctional answer
Like Like
Must-be ? Must-be
My car is Sedan
Neutral My car is not Sedan Neutral ?
Live-with Live-with
Dislike Dislike
??
As seen from Table 3.2, the two-dimensional questionnaire has two questions,
one deals with customer opinion when function (or feature) is present (functional
answer) and other deals with customer opinion when function (or feature) is not
present (dysfunctional answer). The respondent needs to choose one of the options
(Like, Must-be, Neutral, Live-with, or Dislike) from the functional side. At the same
time, the respondent needs to choose one of the options (Like, Must-be, Neutral,
Live-with, or Dislike) from the dysfunctional side. The combination of the answers
provides the Kano evaluation of the feature of the product. For example, the case
shown in Table 3.1 represents a combination (Must-be, Neutral). As such, the product
feature, i.e., Sedan, is an Indifferent (I) feature--Sedan does not contribute to the
satisfaction or dissatisfaction of the respondent. All possible Kano evaluations with
are summarized in Table 3.2. Note that when a respondent answers Like for both
functional and dysfunctional sides or Dislike for both functional and dysfunctional
sides, the answer should not be trusted, i.e., the product feature is Questionable (Q)
feature.
Table 3.2. Kano evaluation of product feature and function.
Dysfunctional answer
Like Must-be Neutral Live-with Dislike
Functional
answer
Like
Must-be
Neutral
Live-with
Dislike
Q
R
R
R
R
A
I
I
I
R
A
I
I
I
R
A
I
I
I
R
O
M
M
M
Q
A total of 100 respondents are selected at random from Bangladesh and asked to
answer according to Kano questionnaire (e.g., Table 3.1) on 38 features of small
vehicles. The goal is to know the preferences of the respondents so that a key
solutions (or a key solutions) can be determined for product development (i.e., here
the product means a small passenger vehicle). The functional questions of these 38
features are listed in Table 3.3. The demographic and psychographic details of the
respondents are shown in Appendix A.
However, one of the important key solutions for developing a small passenger
vehicle is the type of vehicles. In Table 3.3, there are three types of vehicle, namely,
??
SUV, Sedan, and Van (No. 21-23). The Kano evaluation of these types of vehicles is
shown in Table 3.4 that has been determined using the answers of 100 respondents.
Table 3.3. Small passenger vehicle feature.
No
1
2
3
4
5
6
7
8
9
10
11
I bought a new vehicle
I bought a pre-owned vehicle
My vehicle runs 10-15 km/liter
My vehicle runs 15-20 km/liter
My vehicle runs 25-30 km/liter
My vehicle runs 30-35 km/liter
My vehicle runs 5-10 km/liter
My vehicle color is black
My vehicle color is metallic
My vehicle color is red
My vehicle color is white
Feature
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
My vehicle engine has a 2 year warranty My
vehicle engine has a 3 year warranty My
vehicle engine has a 5 year warranty
My vehicle engine is above 1300 cc My
vehicle engine is below 1000 cc
My vehicle engine is between 1000-1300 cc
My vehicle is a 2-door vehicle My
vehicle is a 4-door vehicle My
vehicle is a 5-door vehicle
My vehicle is a Sedan type vehicle My
vehicle is a SUV type vehicle
My vehicle is a Van (microbus) type vehicle
My vehicle is equipped with airbags
My vehicle is equipped with bumper guards
My vehicle is equipped with keyless-entry system
My vehicle is equipped with seatbelts
My vehicle is made in Germany
My vehicle is made in India My
vehicle is made in Japan My
vehicle is made in Korea
My vehicle needs regular maintenance every after 10,000 km
My vehicle needs regular maintenance every after 20,000 km
My vehicle needs regular maintenance every after 25,000 km
My vehicle needs regular maintenance every after 5,000 km
My vehicles runs on CNG My
vehicles runs on Diesel My
vehicles runs on Petrol
??
Table 3.4. Kano evaluation of different types of vehicle
Vehicle
Type
SUV
Sedan
Van
Kano evaluation
AOMI RQ 14 10 17 41
16 2
20 10 12 37 17 4 15 8
11 21 43 2
The evaluation listed in Table 3.4 exhibits a complex situation as far as formal
computation is concerned. Most of the respondents evaluated a vehicle either I or R.
In Bangladesh, Sedan is the most frequently used type of small passenger vehicle.
This type of vehicle is not that much suitable for the users in Bangladesh because of
the road condition, average size of a family, and life-style (travelling in a large
group). Thus, the Kano evaluation shown in Table 3.4 does not match the reality in
Bangladesh. This necessities the subsequent steps of customer needs assessment.
3.2. Selection of reliable answers
In Kano model, a respondent needs to choose an element drawn from {Like,
Must-be, Neutral, Live-with, Dislike} for both functional and dysfunctional sides.
Not the option called Neutral. Due to the lack of motivation and/or comprehensibility,
the respondent tends to answer this option for frequently. Since the goal is to get
anopinionative answer not an indecisive answer (Neutral), the answers equal to Neutral
should be avoided. For example, consider the case shown in Table 3.4. The
functional and dysfunctional answers are listed in Tables 3.5-6. As seen from Table
3.5, around 30% of the answers are "Neutral" for SUV and Sedan. This number is
however low for Van because it has been disliked by many respondents. On the other
hand, for all three types of vehicle more than 30% of the respondents have answered
Neutral, which is the highest percentage compared to those of other answers.
Table 3.5. Functional answers on vehicle type
Functional Answer
Question
Like Must-be Neutral Live-with Dislike
My car is SUV 26 19 34 9 12
My car is Sedan 31 16 30 8 15
My car is Van 24 7 17 11 41
??
Table 3.6. Dysfunctional answers on vehicle type
Dysfunctional Answer
Question
Like Must-be Neutral Live-with Dislike
My car is not SUV 11 9 37 16 27
My car is not Sedan 11 9 31 24 25
My car is not Van 26 12 31 11 20
As mentioned before, since the goal is to get an opinionative answer not an
indecisive answer (Neutral), the answers equal to Neutral either from functional side
or from dysfunctional side or from both sides should be not be considered for the
assessment. Applying this elimination strategy results the Kano evaluation shown in
Table 3.7. As seen from Table 3.7, the acceptable answers have reduced to 45, 53,
and 61 from 100 for SUV, Sedan, and Van, respectively. Compare Table 3.4 and
Table 3.7. Needless to say that an acceptable answer means here an answer that is not a
Neutral (i.e., not indecisive answer).
Table 3.7. Kano evaluation based on the answers of acceptable respondents
Vehicle
Type
SUV
Sedan
Van
A
6
10
6
Kano evaluation
OMIR 10 9 9
9
10 6 8 15
8 6 5 34
Q
2
4
2
Number of acceptable
respondents
45
53 61
3.3. Monte Carlo simulation of unknown answers
The above section describes that a limited number of answers are available for
the customer needs assessment. This means that a large number of answers are
unknown. Monte Carlo simulation can be used to know the unknown answers (Ullah
and Tamaki 2011, Rashid et al. 2012). The simulation process described in Section
2.6 is adopted here. The explanation is as follows: Table 3.8 shows the settings of
probability of events (A, O, M, I, R, Q) using the Case 1 LLs (see Chapter 2). The
relative frequencies listed in Table 3.8 correspond to results shown in Table 3.7.
??
Table 3.8 Settings of probability using Case 1 LLs.
Case 1
Feature
SUV
Sedan
Van
ev(i)
A
O
M
I
R
Q
A
O
M
I
R
Q
A
O
M
I
R
Q
fr(i)
0.133
0.222
0.200
0.200
0.200
0.044
0.189
0.189
0.113
0.151
0.283
0.075
0.098
0.131
0.098
0.082
0.557
0.033
LL(i)
qu
su
su
su
su
qu
su
su
qu
su
su
qu
qu
qu
qu
qu
ns
qu
E(i) Pr(i) CPr(i)
0.1 0.750 0.079
0.267 2.003 0.289
0.267 2.003 0.500
0.267 2.003 0.711
0.267 2.003 0.921
0.1 0.750 1.000
0.267 2.003 0.211
0.267 2.003 0.421
0.1 0.750 0.500
0.267 2.003 0.711
0.267 2.003 0.921
0.1 0.750 1.000 0.1
0.750 0.100 0.1
0.750 0.200 0.1
0.750 0.300 0.1
0.750 0.400 0.5
3.750 0.900 0.1
0.750 1.000
Table 3.9 Settings of probability using Case 2 LLs.
Case 2
Feature ev(i) fr(i)
LL(i) E(i) Pr(i) CPr(i)
A 0.133 qu 0.133 0.097 0.097
SUV
O
M
I
R
0.222
0.200
0.200
0.200
su
su
su
su
0.3 0.220 0.317 0.3
0.220 0.537 0.3
0.220 0.756 0.3
0.220 0.976
Q 0.044 mu 0.033 0.024 1.000 A
0.189 qu 0.133 0.138 0.138 O
0.189 qu 0.133 0.138 0.276 M
0.113 qu 0.133 0.138 0.413
Sedan
I 0.151 qu 0.133 0.138 0.551
R 0.283 su 0.3 0.311 0.862
Q 0.075 qu 0.133 0.138 1.000 A
0.098 qu 0.133 0.125 0.125 O
0.131 qu 0.133 0.125 0.250 M
0.098 qu 0.133 0.125 0.375
Van
I 0.082 qu 0.133 0.125 0.500
R 0.557 ns 0.5 0.469 0.969
Q 0.033 mu 0.033 0.031 1.000
??
Table 3.10 Settings of probability using Case 3 LLs.
Case 3
Feature
SUV
Sedan
Van
ev(i)
A
O
M
I
R
Q
A
O
M
I
R
Q
A
O
M
I
R
Q
fr(i)
0.133
0.222
0.200
0.200
0.200
0.044
0.189
0.189
0.113
0.151
0.283
0.075
0.098
0.131
0.098
0.082
0.557
0.033
LL(i)
mu
qu
qu
qu
qu
eu
qu
qu
mu
mu
qu
mu
mu
mu
mu
mu
ns
eu
E(i)
0.125
0.25
0.25
0.25
0.25
0.042
0.25
0.25
0.125
0.125
0.25
0.125
0.125
0.125
0.125
0.125
0.5
0.042
Pr(i) CPr(i)
0.107 0.107
0.214 0.321
0.214 0.536
0.214 0.750
0.214 0.964
0.036 1.000
0.222 0.222
0.222 0.444
0.111 0.556
0.111 0.667
0.222 0.889
0.111 1.000
0.120 0.120
0.120 0.240
0.120 0.360
0.120 0.480
0.480 0.960
0.040 1.000
0.5
0.4
0.3
0.2
0.1
0
A O M I R Q
events
Figure 3.2. Variability in fr for SUV and Case 3 LLs due to Monte Carlo simulation.
Using the settings listed in Tables 3.8-10 and the simulation process shown in
Section 2.6, the Monte Carlo simulation of the events (A, O, M, I, R, Q) has been
performed for all SUV, Sedan, and Van. In the simulation, N = 100 (number of
??
f r
iteration) because the original number of respondents was 100 (see Table 3.4). The
variability in the relative frequencies of the simulated events (A, O, M, I, R, Q) has
been determined by repeating the simulation process for all features SUV, Sedan, and
Van. As an example, the variability in the relative frequencies of events for SUV and
Case 3 LLs is shown in Fig. 3.2. The dark rectangular point on each vertical bar in
Fig. 3.2 is the original relative frequency of the event (see Table 3.7).
3.4. Determination of truth values of product feature Kano-evaluation
This section deals with the truth value (TV) determination process of Kano-
evaluation of product feature. Needless to say that Kano evaluation of a product
feature is either A, or O, or M, or I, or R, or Q. The truth value determination process
uses the linguistic TVs (LTk), k = 1,2,..., defined by the seven fuzzy numbers namely,
mostly false (mf), quite false (qf), perhaps false (pf), not sure (ns), perhaps true (pt),
quite true (qt), and mostly true (mt), as defined by the membership functions in
equations (2.9.2.1-7) (see Chapter 2).
Let p(Fi,Xj) be a proposition of the following form: Fi is Xj. Here, Fi e {SUV,
Sedan, Van} and Xj e {A, O, M, I, R, Q}. The problem is to assign a TV e [0,1] to
p(Fi,Xj) from the relative frequency of Xj, fr(Xj), obtained after performing Monte
Carlo simulation as explained in the previous section. An example of how to
determine TV is illustrated in Fig. 3.3. In this particular case, the linguistic
counterpart of fr(Xj) = 0.25 is perhaps false (pf) (linguistic TV shown in bold in Fig.
3.3) because fr(Xj) = 0.25 belongs to pf more it belongs to any other linguistic TV.
1
0.75
0.5
0.25
0
mf
mu
qu qf
fr(Xj)
su
pf ns
ns
psl
t
q
ql
t
mt
ml
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
TV
Figure. 3.3. Converting a relative frequency to a linguistic TV.
??
µ
(
.
)
A
(
T
V
)
µ
In general, the membership value of a linguistic TV denoted asµLTk(TV = fr(Xj))
represents the degree of belongingness of fr(Xj) to a linguistic TV, LTk e {mf, qf, pf,
ns, pt, qt, mt}. IfµLTk(fr(Xj)) >µLTl(fr(Xj)), l e {1, ., 7} - {k} and -k e{1, .,7},
then fr(Xj) belongs to LTk more than it belongs to any other linguistic TV. This way,
LTk is the linguistic counterpart of fr(Xj) and the expected value of LTk, E(LTk), is
the truth value of p(.,Xj), i.e., TV(p(.,Xj)) = E(LTk).
Note that fr(Xj) is not a constant value in the interval [0,1]. It depends on the
simulation instance. For example, consider the variability in the relative frequencies
shown in Fig. 3.2. In particular, consider the variability of relative frequency of A,
fr(Xj = A) = [0.04,0.19]. This means that the truth value (TV(p)) of the proposition
p(SUV,A) = "SUV is A" (i.e., SUV is an attractive car) is around [0.04,0.19], i.e.,
TV(p) ~ fr(A) = [0.04,0.19]. If TV(p) ~ 0.04 in some simulations, then its linguistic
counterpart is mostly false (mf) and the expected value of mf, E(mf) = 0.033,
becomes the truth value of p, i.e., the truth value of the proposition "SUV is an
attractive car is equal to 0.033." Similarly, if TV(p) ~ 0.19 in some simulations, then
its linguistic counterpart is quite false (qf) and the expected value of qf, E(qf) = 0.133,
becomes the truth value of this p, i.e., the truth value of the proposition "SUV is an
attractive car is equal to 0.133." This means that in some other cases TV(p) = 0.133.
Therefore, the TV of p(Fi,Xj) may vary based on the result of Monte Carlo
simulation. As a result, variability in the information content of product feature
might be observed.
3.5. Determination of truth value of product feature status
This section deals with the determination process of truth value of the status of a
product feature. Product feature status is defined by using the Kano evaluations, O, A,
M, I, R, and Q. Therefore, the truth value of the product feature status is calculated
from the truth value of its Kano evaluation. Here a product feature status means one
of the following: must be included, should be included, and could be included. Let Fi
be a product feature and Yj be an element of {must be included, should be included,
and could be included}. Therefore, the problem is to determine the truth value of a
proposition of the following form: Fi Yj in the product.
Recall the schematic illustration of Kano evaluation shown in Fig. 3.2.
??
According to Fig. 3.2, if a product feature is classified as One-dimensional (O) or
Must-be (M) and it is not included in the product, the customers are not satisfied.
Therefore, a product feature "must be included" in the product means it is "either O
or M." This leads to the following formulation:
Fi must be included ÷ (F
i
is O) v (F
i
is M )
TV (Fi must be included ) = max(TV (Fi,O),TV (Fi,M ))
(3.1)
Recall the definition of A illustrated in Fig. 3.2. If the feature is classified as
Attractive (A), then it is an unexpected but customer satisfaction-enriching feature.
Thus, a product feature classified as A, it "should be included" in the product for
enhancing the level of customer satisfaction. This yields the following formulation:
Fi should be included ÷ F
i
is A
TV (Fi should be included ) = TV (Fi, A)
(3.2)
On the other hand, if a feature is classified as Indifferent (I), it is not that much
helpful in increasing or decreasing the level of customer satisfaction even though it is
included or not included in the product, respectively. In addition, if a feature is
classified as Reverse (R), its inclusion in the product creates a great deal of
dissatisfaction. Moreover, if a feature is classified as Questionable (Q), then it is an
unreliable feature. This means that if a feature is "I or not R or not Q," it "could be
included" in the product. This yields the following formulation:
Fi could be included ÷ Fi is I v Fi is ÷R v Fi is ÷Q
( )( )( )
(3.3)
TV (Fi could be included ) = max(TV (Fi, I ),1 ÷ TV (Fi, R),1 ÷ TV (Fi, Q))
Table 3.11 shows an example of truth value determination process of the status of
product features defined in equations (3.1-3). The relative frequencies of Kano
evaluation found after performing Monte Carlo simulation (Section 3.3) are used to
determine the TV of Kano evaluation, A, O, M, I, R, and Q. Afterward, the TV of the
status of the product features, namely, must be include, should be included, and could
be included, are calculated using equations (3.1-3), respectively. It is observed that
the relative frequency of Kano evaluation depends on both the simulation instance
and the case of linguistic likelihoods. The example shown in Table 3.11 refers to a
simulation instance. The simulated relative frequencies are quite different (compare
the relative frequencies of Cases 1-3) but the truth values (calculated from the
linguistic counterparts as explained in Section 3.4) are quite similar (not exactly the
same (compare the truth values of status of Cases 1-3)). This is not only true for the
??
case shown in Table 3.11 but also for other simulation instances.
Table 3.11. An example of product feature status truth value determination.
Case 1
Kano
evaluation
fr
TV
A
0.11
0.133
O
0.2
0.3
M
0.23
0.3
I
0.2
0.3
R
0.17
0.133
Q
0.09
0.133
Remarks
After
simulation
Status
TV
Kano
evaluation
A
must be
included
0.3
O
M
should be
included
0.133
Case 2
I
R
could be
included
0.867
Q
fr
TV
0.15
0.133
0.21
0.3
0.22
0.3
0.2
0.3
0.19
0.133
0.03
0.033
After
simulation
Status
TV
Kano
evaluation
A
must be
included
0.3
O
M
should be
included
0.133
Case 3
I
R
could be
included
0.967
Q
fr
TV
0.16
0.133
0.2
0.3
0.22
0.3
0.2
0.3
0.17
0.133
0.05
0.133
After
Simulation
Status
TV
must be
included
0.3
should be
included
0.133
could be
included
0.867
However, consider the relative frequencies of Kano evaluation corresponding to
Case 1 in Table 3.11: fr(A) = 0.11, fr(O) = 0.2, fr(M) = 0.23, fr(I) = 0.2, fr(R) = 0.17,
fr(Q) = 0.09. The linguistic counterpart of fr(A) = 0.11 is quite false (qf), fr(O) = 0.2
is perhaps false (pf), fr(M) = 0.23 is perhaps false (pf), fr(I) = 0.2 is perhaps false (pf),
fr(R) = 0.17 is quite false (qf), and fr(Q) = 0.09 is quite false (qf). This is in
accordance with the procedure explained in Section 3.3 and with the seven-fuzzy-
number-based linguistic truth values defined in Chapter 2. Thus, the truth value of
the proposition "Sedan is A" is equal to the expected value of qf (0.133), i.e., TV(A)
= 0.133. Similarly, the propositions "Sedan is R," "Sedan is Q" also have the truth
value 0.133, i.e., TV(R) = 0.133 and TV(Q) = 0.133. On the other
??
hand, the propositions "Sedan is O," "Sedan is M," and Sedan is I" have the truth
value 0.3 because 0.3 is the expected value of the linguistic truth values of these
propositions, i.e., perhaps true (pf). This means that TV(O), TV(M), and TV(I) = 0.3.
Thus, the truth value of the proposition "Sedan must be included in the product" is
equal to max(TV(O),TV(M)) = max(0.3,0.3) = 0.3. The truth value of the proposition
"Sedan should be included in the product" is equal to TV(A) = 0.13. The truth value of
the proposition "Sedan could be included in the product" is equal to max(TV(I),1-
TV(R),1-TV(Q)) = max(0.3,1-0.133,1-0.133) = 0.867.
3.6. Determination of information content of product feature status
This section describes the information content determination process of product
feature status. The information content means here the two-dimensional information
content wherein one of the dimensions is the Certainty Entropy (CE) and the other
dimension is the Requirement Entropy (RE). Note that CE and RE have already been
defined in Section 2.5 (equations 2.11.1-3). CE measures the variability in the truth
values of a feature and RE measures the degree of fulfillment of requirement given
by P
R
. For this particular case, P
R
= "The feature is a must be included feature," or
"The feature is a should be included feature," or "The feature is a could be included
feature." For an example, consider the truth values of the status of the product feature
shown in Table 3.11. For the sake of better understanding these truth values are
organized in Table 3.12. Table 3.12 also lists the calculated CE and RE based on
these truth values and also on the P
R
. The P
R
in Table 3.12 is The feature is a must be
included feature for all three cases. Case 1 and Case 3 underlie the same amount of
information content (CE,RE) = (0.377,0.722), whereas the information content
underlying the Case 2 is (CE,RE) = (0.311,0.8). The value of CE = 0.377 means that
there is great deal of consensus among the respondents. The degree of consensus is
comparatively much higher for Case 2 because for Case 2 CE has reduced to 0.311
(less than that of Case 1 and Case 3). One the other hand RE is quite high for all three
cases. This means that the requirement "the feature is a must be feature" has hardly
been fulfilled. This means that if one considers this feature a must be feature, it
might create problem in fulfilling this expectation. However, if one resets the
requirement to "the feature is a could be feature," then RE = 0. This also means that
??
the feature is a could be feature rather than a must be or should be feature.
Table 3.12. An example of information content determination process
Case 1
Status
TV
Ic
must be
included
0.3
0.6
should be
included
0.133
0.266
could be
included
0.867
0.266
CE 0.377
P
R
The feature is a must be included feature
RE 0.722
Case 2
Status
TV
Ic
must be
included
0.3
0.6
should be
included
0.133
0.266
could be
included
0.967
0.067
CE 0.311
P
R
The feature is a must be included feature
RE
Status
TV
Ic
0.8
must be
included
0.3
0.6
Case 3
should be
included
0.133
0.266
could be
included
0.867
0.266
CE 0.377
P
R
The feature is a must be included feature
RE 0.722
3.7. Determination of coherency measures of product feature status
This section describes the coherency measure of product feature status. As
explained in the previous sections, the values of truth values product feature status
might change due to the result of simulation. As such, the information contents
(CE,RE) might also vary with the simulation instance. Therefore, the variability in
the information content (CE,RE) should play a role in the customer needs assessment
process. As explained in Section 2.5 (see equation (2.11.4) and Fig. 2.10) a quantity
called coherency measure (ì) measures the variability in (CE,RE) for a given feature
and requirement.
Figure 3.4 shows the variability in (CE,RE) for the product feature called SUV
for all cases, Cases 1-3. The information content is high for should be and must be
included and low for could be included.
??
1.0
0.8
0.6
0.4
0.2
0.0
must be included
should be included
could be included
0.0 0.2 0.4 0.6 0.8 1.0
CE
Figure 3.4. Variability in the information content of SUV.
1.0
0.8
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
CE
Figure 3.5. Determining the coherency measure of SUV for the status called "must be
included."
Based on the data shown in Fig. 3.4, the coherency measure of SUV for three
requirements can be calculated separately. For example, consider the requirement
??
R
E
R
E
"SUV is a must be included feature." The variability in the information content for
this requirement is shown in Fig. 3.5, which is the segment of data points already
shown in Fig. 3.4 corresponding to must be included. The value of coherency
measure ì is equal to 2.313 because e = 0.199, f = 0.377, g = 0.68, h = 1 (see
equation (2.11.4) and Fig. 2.10). Similarly, the values of coherency measure of SUV
for should be included and could be included have been found to be 2.577 and 0.577,
respectively.
However, Figs. 3.6-7 show the variability in (CE,RE) for the other two product
features called Sedan and Van for all cases, Cases 1-3. Similar to the case shown in
Fig. 3.4, the information content is high for should be and must be included and low
for could be included for both cases in Figs. 3.6-7. Based on the data points shown in
Figs. 3.4,6-7, the value of the coherency measure has been determined using the
procedure illustrated in Fig. 3.5. The values are listed in Table 3.13. Note that Sedan
exhibits high values of coherency measure compared to those of SUV and Van. This
means that SUVs and Vans might be good options to replace Sedans.
1.0
0.8
must be included
0.6
should be included
0.4
could be included
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
CE
Figure 3.6. Variability in the information content of Sedan.
??
R
E
1.0
0.8
0.6
0.4
0.2
0.0
must be included
should be included
could be included
0.0 0.2 0.4 0.6 0.8 1.0
CE
Figure 3.7. Variability in the information content of Van.
Table 3.13. Coherency measure of product features.
Status
Features
SUV
Sedan
Van
must be
included
2.314
2.578
2.332
should be
included
2.577
2.578
2.332
could be
included
0.577
0.755
0.577
3.8. Making final decision
This section describes process of how to making a final decision (customer
needs assessment). In this case, the customer needs assessment means to identify the
level of satisfaction of SUV, Sedan, and Van based on the value of the respective
coherency measure (overall information content). Note that a low value of coherency
measure is desirable. Section 2.5 describes the details of the implication of coherency
measure.
However, recall the situation in Bangladesh. Sedan is the most frequently used
vehicles in Bangladesh. This type of vehicle is not that much suitable for the users in
Bangladesh because of the road condition, average size of a family, and life-style
(travelling in a large group). Thus, all Sedan, SUV, and Vans could be suitable for but
??
R
E
it would be difficult to conclude precisely that they are must/should be the vehicle for
users in Bangladesh. At least it can be said that increase in the number of SUV and
Van compared to that of Sedan might lead to an enhancement in the customer
satisfaction.
Whether or not the above conclusion holds if a decision is made based on the
values of coherency measure is an important issue to investigate. As such, the values
of the coherency measure are plotted separately for each requirement. Figure 3.8
shows the value of coherency measure when the requirement refers to "could be
included" for Sedan, SUV, and Van. Needless to say that the values correspond to the
values listed in Table 3.13. As seen from Fig. 3.8, if SUV and Van are introduced side
by side Sedan in a large volume in Bangladesh, the level of satisfaction of vehicle
users "could" increase. In this case, SUV and Van are indifferent.
1.0
0.8
0.6
0.4
0.2
0.0
Sedan SUV Van
Features
Figure 3.8. Reduction in overall information content for could be included.
Figure 3.9 shows the value of coherency measure when the requirement refers
to "should be included" for Sedan, SUV, and Van. Needless to say that the values
correspond to the values listed in Table 3.13. As seen from Fig. 3.9, if Van is
introduced side by side Sedan in a large volume in Bangladesh, the level of
satisfaction of vehicle users "should" increase. This time, SUV does not increase the
level of satisfaction compared to that of Sedan. This decision however, underlies a
great deal of uncertainty (a large value of coherency measure).
??
ì
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Sedan SUV Van
Features
Figure 3.9. Reduction in overall information content for should be included.
Figure 3.10 shows the value of coherency measure when the requirement refers
to "must be included" for Sedan, SUV, and Van. Needless to say that the values
correspond to the values listed in Table 3.13. As seen from Fig. 3.10, if SUV and Van
are introduced side by side Sedan in a large volume in Bangladesh, the level of
satisfaction of vehicle users "must" increase. The trend seen here similar to that of
could be included (Fig. 3.8). This time a great deal of uncertainty (a large value of
coherency measure) is associated with the decision.
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Sedan SUV Van
Features
Figure 3.10. Reduction in overall information content for must be included.
??
ì
ì
In synopsis, the following statements can be made:
level of satisfaction of car users in Bangladesh (Fig. 3.8).
Bangladesh (Fig. 3.9)
Bangladesh but this conclusion possesses a great deal of uncertainty (Fig.
3.10).
Similar to the case of car types (Sedan, SUV, Van), other features in Table 3.3
can be studied and similar conclusions can be made. This way, customer needs
assessment can be carried out and the key solutions to develop the product can be
determined in the conceptual phase of product development.
??
??
Chapter 4
Sustainability Assessment
This chapter deals with the sustainability assessment of key solutions in
conceptual phase of product development. This chapter is based on work of Rashid et
al. 2011. In this chapter, a product means a grinding wheel, i.e., a cutting tool used to
remove hard materials and also to finish surfaces of precision parts.
For the sake of better understanding, recall Fig. 1.1 as repeated in Fig. 4.1.
External Customers
(Real Customers)
Customer needs
Use
(Satisfaction)
Internal Customers
(Product Development Team)
Creativity
Conceptual Phase
(Key Solutions)
Detailed Design
Disposal
(Recycle, Downcycle, Landfill)
Manufacturing
Sustainability
Primary Materials
Production
Figure 4.1. A product development scenario (Fig. 1.1 repeated).
As mentioned in Chapter 1, sustainability has become an important issue of
product development (Fiksel 2009) that refer to the fact that the product is
environmentally benign on top of other desired performances. One should
incorporate so-called Life-Cycle Assessment (LCA) into the product development
processes to ensure the sustainability (Donnelly et al. 2004, Kobayashi, 2006). In
??
addition to conventional sustainability assessment (i.e., LCA), it is important to do
scenario analysis (Umeda 2009, Fukushige et al. 2012) taking a broader perspective
into the consideration. However, one of the remarkable finding underlying
scenario-analysis-based sustainability assessment is that the primary production of
materials used in the product plays a critical role to ensure the sustainability (Higuchi
et al. 2012). This implies the following question:
How to deal with the sustainability of materials (used in the product) in the key
solution determination process in the conceptual phase of product development?
This section provides an answer to this question by taking the example of a
grinding wheel. This means that the grinding wheel is the product here and
sustainability factors of the primary material production of the materials used in
manufacturing a grinding wheel are the important decision-relevant information. This
also means that the possible type of materials used to manufacture a grinding wheel
is the key solutions. The remainder of this chapter is as follows:
Section 4.1 describes the information relevant to the sustainability of primary
material production of the materials used to manufacture a grinding wheel. Section
4.2 describes the fuzzy numbers used to formally compute the
uncertainty/imprecision associated with decision-relevant information. Section 4.3
describes the results and discusses the implication of the results.
4.1. Decision-relevant information
In general, very hard materials based on technical ceramics are used to produce
the abrasive grains of a grinding wheel. The abrasive grains actually provide the
main cutting action, i.e., they are the main ingredients of a grinding wheel. There are
many hard materials (Ullah et al. 2011), e.g., price, annual production, density,
energy consumption for primary production of bulk material, energy for processing
(powder formation, bulk deformation, etc.) materials, CO
2
footprint for primary
production and processing, NO
X
emission of primary production, SO
X
emission of
primary production, water usage of primary production, recycle fraction, CO
2
emission and energy for recycling, and alike. Obtaining reliable data on these
sustainability attributes is not an easy task. In most cases, an estimation is given in
the form of a numerical range on an eco-attribute compiling data/information from
??
many sources. Some of the sustainability attributes do not have any information (e.g.,
recycle fraction, gel formation, etc.).
However, more than 320 types of technical ceramics based hard materials available
in the database of a material evaluation system called CES Selector (version 5.1.0)
developed by the Granta Design Limited have been studied (reference [1]). The
maximum and minimum ranges of each sustainability attribute (in particular, CO
2
footprint (or emission), NO
X
emission, SO
X
emission, and water usage of primary
production) for five different classes of hard material, namely, Alumina (AN), Silica
(SC), Boron Nitride (BN), Boron Carbide (BC), and Zirconia (ZN) are identified.
The variants of AN, SC, BN, BC, and ZN not used for producing abrasive grains are
excluded from this study. Figure 4.2 shows the variability in CO
2
footprint
(kg-CO
2
/kg-material) and water usage (l-water/kg-material) of AN, SC, BN, BC, and
ZN. As seen from Fig. 4.2, the information of CO
2
footprint underlies low
uncertainty/impression whereas water usage exhibit a relatively high
uncertainty/imprecision.
300
250
200
150
100
50
0
Alumina
Silicon Carbide
Zirconia
Boron Nitride
Boron Carbide
0 3 6 9 12 15
CO
2
Footprint (kg/kg)
Figure 4.2. CO
2
footprint and water usage of some selected hard materials.
??
W
a
t
e
r
U
s
a
g
e
(
l
/
k
g
)
120
90
60
Alumina
30 Silicon Carbide
Zirconia
Boron Nitride
Boron Carbide
0
0 20 40 60 80
NO
X
(g/kg)
Figure 4.3. NO
X
/SO
X
emissions of some selected hard materials.
Figure 4.3 shows the variability in NO
X
and SO
X
emissions (g-NO
X
or
SO
X
/kg-material) of the primary production of AN, SC, BN, BC, and ZN. As seen
from Fig. 4.3, the variability in the information increases with the increase in NO
X
or
SO
X
emission. This means that underlying uncertainty/impression increases with the
increase in the emission of NO
X
or SO
X
.
4.2. Computational framework
To deal with the uncertainty/imprecision associated with the parameters
described above computational framework based on "range compliance" is proposed
here. Range compliance has already been explained in Section 2.2 (equation (2.5).
Range compliance is an operation on a fuzzy number using a numerical range. In
particular, range compliance R(L,A) of a numerical range L is its average
membership value with respect to a fuzzy number A. The expression of R(L,A) is as
follows:
}·
A
(x)dx
R(L, A) =
L
L?
(L? _ L). (L? _ Supp(A))
(4.1)
??
S
O
X
(
g
/
k
g
)
In equation (4.1), L? is the segment of L that belongs to Supp(A). Note that the
equation (4.1) is the repetition of equation (2.5).
However, to be more specific consider the following objects. Let G be a
member of the set of materials {AN, SC, BN, BC, ZN}, i.e., G e {AN, SC, BN, BC,
ZN}. Let S be a member of the set of sustainability parameters {CO
2
footprint, water
usage, NO
X
emission, SO
X
emission}, i.e., S e {CO
2
footprint, water usage, NO
X
emission, SO
X
emission}. Let F
S
be a member of the set of fuzzy numbers {VL
S
, L
S
,
M
S
, H
S
, VH
S
}. Here, VL refers to very low, L refers to low, M refers to moderate, H
refers to high, and VH refers to very high. The subscript "S" means that the
sustainability parameter is S. Let X
S
be a point on the real line, X
S
e ?. The interval
[0, X
S
] is the universe of discourse of a fuzzy set F
S
. The membership functions of
the fuzzy numbers VL
S
, L
S
, M
S
, H
S
, and VH
S
(in general F
S
) can thus be defined as
follows:
·
VL
S
(x
S
) = max 0,min 1, 0.3X
S
÷ x
S
0.3X
S
÷
0
(4.2)
·
L
S
(x
S
) = max 0,min x
S
÷ 0.1X
S
, 0.5X
S
÷ x
S
0.3X
S
÷ 0.1X
S
0.5X
S
÷ 0.3X
S
(4.3)
·
M
S
(x
S
) = max 0,min x
S
÷ 0.3X
S
, 0.7 X
S
÷ x
S
0.5X
S
÷ 0.3X
S
0.7 X
S
÷ 0.5X
S
(4.4)
·
H
S
(x
S
) = max 0,min x
S
÷ 0.5X
S
, 0.7 X
S
÷ x
S
0.7 X
S
÷ 0.5X
S
0.9 X
S
÷ 0.9 X
S
(4.5)
·
VH
S
(x
S
) = max 0,min 1, x
S
÷ 0.7X
S
X
S
÷ 0.7 X
S
(4.6)
Figures 4.2-3 provide an estimation of X
S
. As seen from Figs. 4.2-3, X
S
could be
a point in the interval [12,15] if S = CO
2
footprint, X
S
could be a point in the interval
[280,300] if S = water usage, X
S
could be a point in the interval [70,80] if S = NO
X
emission, and X
S
could be a point in the interval [110,120] if S = SO
X
emission.
Figure 4.4 illustrates the membership functions for S = CO
2
footprint and X
S
= 15. At
the core L
S
, M
S
, or H
S
(a point corresponding to unit membership value) the
membership values of other fuzzy numbers are equal to zero. This nature of remains
the same for all fuzzy numbers irrespective of the state of S and the value of X
S
because of the definitions in equations (4.2-6).
??
·
M
S
(x
S
)
1
0.75
0.5
0.25
0
0
VL
S
3
L
S
6
M
S
9
H
S
12
VH
S
15
x
S
Figure 4.4. Membership functions when S = CO
2
footprint and X
S
= 15.
Let L
S
(G) is the range of sustainability parameter S for a material G. One can
calculate the range compliance R(L
S
(G), F
S
) with respect to F
S
, as follows:
}
(
·)
F
S
(x
S
)dx
S
R(L
S
(G),F
S
) =
L
S
G
L
S
? (G)
(L?
S
(G) _ L
S
(G)). (L?
S
(G) _ Supp(F
S
))
(4.7)
In equation (4.7), L?
S
(G) is the largest segment of L
S
(G) that belongs to the support of
F
S
, Supp(F
S
).
For example, if S = CO
2
footprint, G = SC, the L
S
(G) = [6,7.8] (see Fig. 4.2).
This yields the following range compliances: R(L
S
(G),F
S
) = 0 for F
S
= VL
S
,
R(L
S
(G),F
S
) = 0.25 for F
S
= L
S
, R(L
S
(G),F
S
) = 0.783 for F
S
= M
S
, R(L
S
(G),F
S
) = 0.05
for F
S
= H
S
, R(L
S
(G),F
S
) = 0 for F
S
= VH
S
.
To achieve a better sustainability all CO
2
footprint, NO
X
emission, SO
X
emission, and water usage should be minimized. This means that the material that
complies more with VL
S
or L
S
is good material and the material that complies more
with M
S
, H
S
, or VH
S
is not-so-good material. Based on this contemplation two indices
can be derived called Desirable Impact (DI
S
(G)) and Undesirable Impact (UI
S
(G)),
as follows:
DI
S
(G) = R(L
S
(G),VL
S
)+ R(L
S
(G),L
S
) (4.8)
UI
S
(G) = R(L
S
(G),M
S
) + R(L
S
(G),H
S
) + R(L
S
(G),VH
S
) (4.9)
??
The above formulation provides an two-dimensional decision space, wherein a
sustainable material G means its DI
S
(G) is high and UI
S
(G) is low with respect to all
Ss. This decision space is schematically illustrated in Fig. 4.5.
Less
sustainable
material
Highly
sustainable
material
0
0 DI
S
(G)
Figure 4.5. Decision space for sustainability assessment.
4.3. Results
This section describes the results obtained by using the framework described in
the previous section. The range compliance R(L
S
(G),F
S
) has been calculated for all
materials and sustainability parameters for two different cases Case 1 and Case 2.
These cases are listed in Table 4.1. Needless to say that the values of X
S
in Table 4.1
define the universe of discourses of the fuzzy numbers defined in equations (4.2-6).
Table 4.1 Cases of setting the universe of discourse
X
S
Case
1
2
CO
2
footprint
15
12
water usage
300
280
NO
X
emission
80
70
SO
X
emission
120
110
The values of X
S
listed in Table 4.1 underlies the observation mentioned before,
i.e., X
S
could be a point in the interval [12,15] if S = CO
2
footprint, X
S
could be a
??
U
I
S
(
G
)
point in the interval [280,300] if S = water usage, X
S
could be a point in the interval
[70,80] if S = NO
X
emission, and X
S
could be a point in the interval [110,120] if S =
SO
X
emission. In particular the maximum and minimum values are considered to see
the sensitivity.
Table 4.2 shows the range compliances of AN, R(L
S
(AN),F
S
). Figure 4.6 shows
the position of AN in the two-dimensional decision space, wherein the data points
(DI
S
(AN),UI
S
(AN)) have been calculated using the data points listed in Table 4.2.
Table 4.2. Range compliances of AN.
F
S
S Case
VL
S
L
S
M
S
H
S
VH
S
CO
2
1 0.344 0.483 0 0 0
footprint 2 0.181 0.729 0 0 0
water 1 0.333 0.5 0 0 0
usage 2 0.321 0.736 0.05 0 0
NO
X
1 0.315 0.531 0 0 0
emission 2 0.214 0.678 0 0 0
SO
X
1 0.152 0.87 0.083 0 0
emission 2 0.121 0.835 0.159 0 0
1.5
1.25
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1 1.25 1.5
DI
S
(AN)
Figure 4.6. Sustainability of AN.
As seen from Fig. 4.6, the undesirable impact is very low and the desirable
impact is very high for AN for all four sustainability parameters. Therefore, AN is a
good material from the view point of sustainability and one should use this material
??
U
I
S
(
A
N
)
to manufacture a grinding wheel unless there are other problems.
Table 4.3 shows the range compliances of SC, R(L
S
(SC),F
S
). Figure 4.7 shows
the position of SC in the two-dimensional decision space, wherein the data points
(DI
S
(SC),UI
S
(SC)) have been calculated using the data points listed in Table 4.3.
Table 4.3. Range compliances of SC.
F
S
S Case
VL
S
L
S
M
S
H
S
VH
S
CO
2
1 0 0.25 0.783 0.05 0
footprint 2 0 0 0.625 0.375 0
water 1 0 0.25 0.75 0 0
usage 2 0 0.178 0.851 0.08 0
NO
X
1 0.104 0.938 0.031 0 0
emission 2 0.048 0.866 0.143 0 0
SO
X
1 0 0.625 0.375 0 0
emission 2 0 0.625 0.375 0 0
1.5
1.25
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1 1.25 1.5
DI
S
(SC)
Figure 4.7. Sustainability of SC.
As seen from Fig. 4.7, the undesirable impact is high and the desirable impact is
low for SC for all most of the sustainability parameters. For some parameters, the
scenario is the opposite one. Therefore, SC is perhaps a not-so-good material from
the view point of sustainability and one should avoid using this material for
manufacturing grinding wheel unless there are other problems.
??
U
I
S
(
S
C
)
Table 4.4 shows the range compliances of SC, R(L
S
(BN),F
S
). Figure 4.8 shows
the position of BN in the two-dimensional decision space, wherein the data points
(DI
S
(BN),UI
S
(BN)) have been calculated using the data points listed in Table 4.4.
Table 4.4. Range compliances of BN.
F
S
S Case
VL
S
L
S
M
S
H
S
VH
S
CO
2
1 0 0.25 0.789 0.06 0
footprint 2 0 0 0.604 0.396 0
water 1 0 0.208 0.791 0 0
usage 2 0 0.134 0.883 0.08 0
NO
X
1 0 0.186 0.84 0.046 0
emission 2 0 0.036 0.794 0.232 0
SO
X
1 0 0.042 0.535 0.559 0.05
emission 2 0 0 0.432 0.635 0.167
1.5
1.25
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1 1.25 1.5
DI
S
(BN)
Figure 4.8. Sustainability of BN.
As seen from Fig. 4.8, the undesirable impact is very high and the desirable
impact is very low for BN for all sustainability parameters. Therefore, BN is a not-
so-good material from the view point of sustainability and one should avoid using
this material for manufacturing grinding wheel unless there are other problems.
Table 4.5 shows the range compliances of BC, R(L
S
(BC),F
S
). Figure 4.9 shows
the position of BC in the two-dimensional decision space, wherein the data points
??
U
I
S
(
B
N
)
(DI
S
(BC),UI
S
(BC)) have been calculated using the data points listed in Table 4.5.
Table 4.5. Range compliances of BC.
F
S
S Case
VL
S
L
S
M
S
H
S
VH
S
CO
2
1 0 0 0.533 0.467 0
footprint 2 0 0 0.104 0.739 0.208
water 1 0.027 0.535 0.5 0.537 0.277
usage 2 0 0.491 0.504 0.5 0.381
NO
X
1 0 0 0.468 0.531 0
emission 2 0 0 0.25 0.766 0.143
SO
X
1 0 0 0 0.375 0.444
emission 2 0 0 0 0.206 0.697
1.5
1.25
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1 1.25 1.5
DI
S
(BC)
Figure 4.9. Sustainability of BC.
For BC, all sustainability parameters, except water usage, provide zero desirable
impact, as seen from Fig. 4.9. The undesirable impact for all parameters are also very
high. Therefore, similar to BN, BC is also an less preferable material to manufacture a
grinding wheel. Thus, BC should also be avoided for manufacturing a grinding wheel
unless there are other problems.
Table 4.6 shows the range compliances of ZN, R(L
S
(ZN),F
S
). Figure 4.10 shows
the position of ZN in the two-dimensional decision space, wherein the data points
(DI
S
(ZN),UI
S
(ZN)) have been calculated using the data points listed in Table 4.6. The
??
U
I
S
(
B
C
)
sustainability scenario of ZN is similar to that of SC (compare Fig. 4.10 with Fig.
4.7). Therefore, ZN is perhaps a not-so-good material from the view point of
sustainability and one should avoid using this material for manufacturing grinding
wheel unless there are other problems.
Table 4.5. Range compliances of ZN.
F
S
S Case
VL
S
L
S
M
S
H
S
VH
S
CO
2
1 0 0.7 0.3 0 0
footprint 2 0 0.25 0.75 0 0
water 1 0.027 0.919 0.083 0 0
usage 2 0 0.848 0.152 0 0
NO
X
1 0 0.75 0.25 0 0
emission 2 0 0.5 0.5 0 0
SO
X
1 0 0.27 0.756 0.041 0
emission 2 0 0.182 0.828 0.159 0
1.5
1.25
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1 1.25 1.5
DI
S
(ZN)
Figure 4.10. Sustainability of ZN.
Based on the findings described in the above, the following decision can be
made, as shown in Table 4.7. As listed in Table 4.7, Alumina based hard materials are
highly sustainable materials for manufacturing abrasive grains of a grinding wheel.
Boron Nitride/Carbide based hard materials are less sustainable materials. Whereas,
Zirconia/Silicon Carbide based materials are moderately sustainable materials for
??
U
I
S
(
Z
N
)
manufacturing abrasive grains of a grinding wheel. One may use this finding while
developing more sustainable material removal tools (products) for precision
engineering.
Table 4.7. Sustainability assessment of selected hard materials for abrasive grains of
grinding wheel.
Categories Materials
Highly sustainable materials Alumina based hard materials
Moderately sustainable materials Zirconia/Silicon Carbide based hard materials
Less sustainable material Boron Nitride/Carbide based hard materials
??
??
Chapter 5
Creativity Assessment
This chapter deals with the assessment of creativity in key solutions
determination process in the conceptual phase of product development. This chapter is
based on the work of Ullah et al. 2012. For the sake of better understanding, recall
Fig. 1.1 as repeated in Fig. 5.1.
External Customers
(Real Customers)
Customer needs
Use
(Satisfaction)
Internal Customers
(Product Development Team)
Creativity
Conceptual Phase
(Key Solutions)
Detailed Design
Disposal
(Recycle, Downcycle, Landfill)
Manufacturing
Sustainability
Primary Materials
Production
Fig. 5.1. A product development scenario (Fig. 1.1 repeated).
In the conceptual phase of product development the internal customers (product
development team members) need to be creative so that a great deal of useful The
internal customers need to be creative to suggest many potential key solutions for
satisfying the needs of external customers (real customers). Therefore, the following
question arise in the conceptual phase of product development:
How to differentiate a creative key solution from a non-creative key solution?
??
Creativity is a complex and multifaceted phenomenon (Puccio et al. 2010). In
the industry in particular, lateral thinking (de Bono 1970) has been practiced to be
creative. In addition, TRIZ (theory of innovative problem solving) (Altshuller 2001)
has also been practiced to be creative (Puccio et al. 2010). However, to describe
reasoning and processes of human creativity in product development, mappings of
objects from one domain (or space) to another have been found effective. For
example, consider the mappings i) among Functions (F), Behaviors (B), and
Structures (S) introduced by Gero (Gero, 2000), ii) between Functional
Requirements (FR) and Design Parameters (DP) introduced by Suh (Suh, 1998), and
iii) between Concept (C) and Knowledge (K) introduced by Hatchuel and Weil
(Hatchuel and Weil 2003, 2009). In particular, C-K theory (Hatchuel and Weil 2003,
2009) provides an clear definition of creative concept-- a creative concept is
undecided entity with respect to the existing knowledge at the point of time when it
(the concept) is conceived. Therefore, creation of new knowledge is associated with
acceptance/rejection of a creative concept. Ullah et al. 2012 have shown that the
processes involved in adopting a creative concept may not necessarily be an outcome
of such logical processes as deduction, induction, and abduction (Yoshikawa 1981,
Tomiyama et al. 2009, Ullah 2008, Zeng and Cheng 1991, Kazakci et al. 2005). Two
different kinds of motivation called epistemic challenge and compelling reason are
involved in adopting a creative concept and if one uses epistemic information content
(CE,RE), as defined in Chapter 2 and used in Chapter 3, then one can easily
differentiate a creative key solution from a non-creative key solution in the
conceptual phase of product development. Based on this contemplation, this chapter is
written. The remainder of this chapter is organized as follows: Section 5.1
describes the main elements of C-K theory. Section 5.2 explains the epistemic
information contents for differentiating a creative key solution from a non-creative
key solution. Section 5.3 describes the results and discusses the implication of the
findings.
5.1. C-K theory
This section describes the main elements of C-K theory (Hatchuel and Weil
2003, 2009, Braha and Reich 2003, Kazakci et al. 2005, Ullah et al. 2012).
??
A schematic illustration of C-K theory is shown in Fig. 5.2. As seen from Fig.
5.2, there are two interdependent domains called Concept Domain and Knowledge
Domain in C-K theory.
Figure 5.2. An illustration of C-K theory (Ullah et al. 2012).
In addition, there are mappings between C and K, i.e., C÷K, K÷C, C÷C, and
K÷K. This mapping is somewhat different compared to those in other theories. For
example, in Axiomatic Design (Suh, 1998) the mapping is allowed in a hierarchical
manner: FR÷DP÷FR (new)÷DP (new). The mapping FR to FR or DP to DP is not
allowed in Axiomatic Design. However, one of the most remarkable features of C-K
mapping is that it is able to deal with a creative concept—a concept that is undecided
??
with respect to the existing knowledge at the point of time when it (the concept) is
conceived. If such an undecided concept is pursued further, new knowledge might
evolve in favor of the concept. As a result, both the knowledge evolved and the
concept conceived become a part K Domain and C Domain, respectively. Thus, C-K
mapping expands enriching both domains by the addition of undecided concepts and
co-creation of new knowledge.
5.2. Differentiating creative and non-creative concepts
Let C1 be an existing concept (ordinary key solution) and C2 be an creative
concept (creative key solution). Let K1 be the knowledge of suitableness of C1 and
K2 be the knowledge of performance of C1. In addition, let K3 be the knowledge of
suitableness of C2 and K4 be the knowledge of performance of C2.
One considers C2 because C1 is perhaps not suitable for the perceived need.
This means that C1 should be replaced by C2 for the better fulfillment of the
perceived need. This is called compelling reason. Thus, compelling reason acts as
one of the motivations behind perusing C2 instead of C1 for a given need. One the
other hand, at the beginning (when C2 is being conceived), K4 is empty (K4 = {?}),
i.e., there is a lack of knowledge regarding the performance of C2. The performance
of C2 is somewhat unknown when C2 is being conceived. This is called epistemic
challenge. Thus, a challenge of seeking new knowledge emerges. Overcoming this
challenge acts as the other motivation for pursing C2 instead of C1.
The motivations, compelling reason and epistemic challenge, can quantitatively
be measured by the Certainty Entropy and Requirement Entropy (CE,RE). To do this,
consider a set of propositions for C1 and C2.
First, consider the propositions regarding C1 (an ordinary or existing concept).
The propositions regarding C1 (P11,...,P14) and their truth values are shown in Table
5.1. First, the linguistic truth values defined in Chapter 2 (i.e., five fuzzy numbers
(mostly false (mf), perhaps false (pf), not sure (ns), perhaps true (pt), and mostly true
(mt))) are used to determine the TV of the proposition P11,...,P14. The expected
values (listed in Table 2.2) of the linguistic TV are used as the numerical TV for
calculating CE and RE. Note that the TVs of P11,...,P14 underlie the knowledge K1
and K2. Based on the settings shown in Table 5.1, the two-dimensional information
??
content (i.e., (CE,RE)) of C1 is calculated using the functions described in Chapter 2.
The results are shown in Fig. 5.3. As seen from Fig. 5.3, the epistemic challenge
exhibit low information content whereas compelling reason exhibits high information
content. They are placed opposite to each other. High information content of
compelling reason implies that it is not serving as a compelling reason as such. Low
information content of epistemic challenge implies that it is not a challenge as such.
Table 5.1. State of ordinary concept (C1).
Propositions
Linguistic
TV
Numerical
TV
Requirement
(P
R
)
TV
P11: Cerceisuitanble for the
1 is
mostly
0.1
C1 should be
p ved eed false (mf) suitable for 0.1
C1 is not suitable for
P12: the perceived need
P13: C1 performs well
perhaps
true (pt)
mostly
true (mt)
0.733
0.9
perceived need
C1 should
0.9
P14: C1 ldoes not perform mossetl(ymf) perform well
0.1
wel
1.2
1
0.8
0.6
0.4
0.2
0
fal
Compelling
reason
(suitableness)
Epistemic
challenge
(performance)
0 0.2 0.4 0.6 0.8 1 1.2
Certainty Entropy
Figure 5.3. Information content of C1.
??
R
e
q
u
i
r
e
m
e
n
t
E
n
t
r
o
p
y
Similar to C1, consider the propositions regarding C2 (a creative concept). The
propositions regarding C2 (P21,...,P24) and their truth values are shown in Table 5.2
and the information content is shown in Fig. 5.4. Note the opposite positions of
epistemic challenge and compelling reason in Fig. 5.4. This time, the epistemic
challenge has a very high information content (i.e., it is really a challenge), whereas
compelling reason has a low information content (i.e., it is indeed a compelling
reason).
Table 5.2. State of creative concept (C2).
Propositions
Linguistic
TV
Numerical
TV
Requirement
(P
R
)
TV
P21: Cerceisuitanbeedfor the
p 2 is ved le perhaps
true (pt)
0.733 C2 should be
suitable for
0.733
C2 is not suitable for
P22: the perceived need
P23: C2 performs well
perhaps
false (pf)
not sure
(ns)
0.267
0.5
perceived need
C2 should
0.5
P24: C2 ldoes not perform not)sure perform well
0.5
wel
1.2
1
0.8
0.6
0.4
0.2
0
(ns
Epistemic
challenge
(performance)
Compelling
reason
(suitableness)
0 0.2 0.4 0.6 0.8 1 1.2
Certainty Entropy
Figure 5.4. Information content of C2.
The overall information content of C2 is high compared to that of C1. Needless
??
R
e
q
u
i
r
e
m
e
n
t
E
n
t
r
o
p
y
to say that the overall information content means here the value of coherency
measure (ì) as defined in Chapter 2. This means a creative concept possesses high
epistemic information content while being conceived. This information content
however should reduce when new knowledge is available, i.e., K4 transforms to K?4
so that the propositions like P23 and P24 exhibit truth value similar to that of P21
and P22.
Lack of knowledge Rejection of concept
adopted
Motivation
Concept
conceived
Continuation of concept
ì
Gain of knowledge
adopted
Concept adopted
Pursing a creative concept
Figure 5.5. The states of a creative concept.
Figure 5.5 illustrates the states of a creative concept in terms of overall
information content ì. As seen from Fig. 5.5, the Information Axiom (minimize the
information content of design (Suh 1998)) does not hold as such for creative concept.
Sometimes the information content should be minimized, sometimes it should be
maximized. In addition, due to the lack of knowledge the information content of a
creative concept jumps to its peak. At the same time, if motivating factors called
compelling reason and epistemic challenge prevail, then a creative concept is
??
conceived. When a substantial amount of knowledge becomes available, the
information content of the conceived concept should go down significantly. In this
case, the conceived concept becomes a part of C-K mapping, i.e., the concept is
adopted as a key solution. Otherwise, the creative concept should be abundant and a
new course of direction should be explored.
5.3. Results and discussions
This section describes the results of how a creative concept (an engine for Mars
exploration) has been differentiated from an ordinary engine (an existing fossil-fuel
based engine) using the method described in the previous section. At the beginning
the C-K mapping takes the form of the map shown in Fig. 5.6.
Figure 5.6. C-K mapping of a creative key solution for Mars exploration.
As seen from Fig. 5.6, two solutions C1 = Fossil-fuel based propulsion engine
??
and C2 = Mg-CO
2
based propulsion engine have been considered. C1 is suitable for
earth whereas C2 is suitable for Mars. The performance of C1 is known whereas the
performance of C2 is quite unknown. This implies the propositions and their truth
values as shown in Table 5.3. The information contents of C1 (fossil-fuel based
propulsion engine) and C2 (Mg-CO
2
based propulsion engine) can be expressed by
the information contents shown in Fig. 5.3 an Fig. 5.4, respectively because of the
settings in Table 5.1 is similar to that of Table 5.3 for C1 and in Table 5.2 is similar to
that of Table 5.3 for C2. Therefore, Mg-CO
2
based propulsion engine is a creative
concept and it can be pursued further.
Table 5.3. The states of ordinary and creative concepts.
C1 = Fossil-fuel based propulsion engine
Propositions
Linguistic
TV
Numerical
TV
Requirement
(P
R
)
TV
P11: C1arsis psluirtatblen
for a
M ex o io
mostly
false (mf)
0.1
C1 should be
suitable for
0.1
P12: C1ariss explsuitaone for
M not orati bl perhaps
true (pt)
mostly
0.733 Mars
exploration
P13: C1 performs well
true (mt)
0.9
C1 should
0.9
P14: C1 ldoes not perform mossetl(ymf) perform well
0.1
wel fal
C2 = Mg-CO
2
based propulsion engine
Propositions
Linguistic
TV
Numerical
TV
Requirement
(P
R
)
TV
P11: C2arsisxpluitaatble
for s
M e or ion
perhaps
true (pt)
0.733
C2 should be
suitable for
0.733
P12: C2ariss explsuitaone for
M not orati bl perhaps
false (pf)
not sure
0.267 Mars
exploration
P13: C2 performs well
(ns)
0.5
C2 should
0.5
P14: C3 ldoes not perform not)sure perform well
0.5
wel (ns
If C2 (Mg-CO2 based propulsion engine) is pursued further new knowledge can
be gained (K4 transforms to K?4). Figure 5.7 shows the state of K?4.
??
Figure 5.7. When K4 transforms to K?4.
As seen from Fig. 5.7, for Mars exploration, a propulsion engine is needed that
should use in-situ fuel and oxidizer. Given the fact that Mars atmosphere consists of
more than 95% CO
2
, it (CO
2
) can be used as an oxidizer, even if it is an unusual
choice. This necessitates a particular type of fuel either metals (Be, Mg, Al, Li, Ca,
??
etc.) or their hydrates (e.g., BeH
2
, MgH
2
, etc.). The fundamental studies conducted
by Shafirovich et al. 1992, 1993 have revealed that the fuels, namely, Mg, Al, Be,
BeH2 are probably the most useful fuels when CO
2
acts as the oxidizer. It has also
been found that Mg-CO
2
combination produces almost the same amount of Specific
Impulse (an important performance measure of propulsion devices) compared to that
of other combinations (i.e., Al-CO
2
, Be-CO
2
, and BeH
2
-CO
2
). In terms of other
important performance measures (i.e., combustion characteristics, such as toxicity,
ignitability, combustion rate, slag formation, etc.) Mg-CO
2
combination produces
relatively better result.
Table 5.4. States of C2 and C2 based on C-K mapping in Fig. 5.7.
Requirement
P1
Propositions
C2 is acceptable in terms of Specific pt
Impulse
Truth Values
0.733
An
(P
R
)
engine
C2 is not acceptable in terms of Specific mf should be
P2 Impulse 0.1 acceptable in
C3 is acceptable in terms of Specific mt 0.9 terms of
P3 Impulse Specific
Impulse
C3 is not acceptable in terms of Specific
P4 Impulse mf 0.1
P5
C2 is acceptable in terms of toxicity, mt
ignitability, combustion rate, slag
formation, etc.
0.9
An engine
C2 is not acceptable in terms of toxicity, should be
ignitability, combustion rate, slag acceptable in
P6
formation, etc.
mf 0.1
terms
toxicity,
of
P7 C3 is acceptable in terms of toxicity, mf
ignitability, combustion rate, slag
formation, etc.
0.1
ignitability,
combustion
rate,
sl a g
formation, etc.
C3 is not acceptable in terms of toxicity,
P8 ignmabiility,etc. combustion rate, slag pt ita
for t on,
0.733
C2 = Mg-CO
2
Propulsion Engine, C3 = Y-CO
2
Propulsion Engine, Y e{Be, BeH
2
,
Al}
??
Given the C-K mapping in Fig. 5.7, is it possible to show that the information
content of concept C2 (Mg-CO2-based propulsion engine) has come down
significantly? An answer to this question is needed to make sure the effectiveness of
the transformation of knowledge from K4 to K?4. Otherwise, new knowledge (K?4)
does not add any value to key solution determination process.
To answer the question, as set of propositions P1,.,P8 and two alternatives C2
(same as before) and C3 (=Y-CO
2
-based propulsion engine, Y is either Be or BeH
2
or
Al) are considered. The propositions and their truth values are listed in Table 5.4.
Needless to say that the truth values of the propositions listed in Table 3 reflect the
facts in Fig. 5.7.
1
0.8
0.6
0.4
0.2
0
(a) C2
1
0.8
0.6
0.4
0.2
0
(b) C3
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Certainty Entropy Certainty Entropy
Figure 5.8. Information content of C2 and C3 based the settings in Table 5.4.
The information content in terms of Certainty and Requirement Entropies
(CE,RE) are determined by using the same methods used in the previous section. The
results are shown in Fig. 5.8. C2 has information contents (0.37,0) and (0.2,0) for
{P1,P2} and {P5,P6}, respectively. The overall information content of C2 is now
equal to 0.57. On the other hand, C3 has information content (0.2,0) and (0.37,1) for
{P3,P4} and {P7,P8}, respectively. The overall information content of C3 is equal to
1.74. Thus, C2 is preferred over C3, as the key solution to develop a propulsion
engine for Mars exploration, the decisionmaking now underlies "minimization of
information content," i.e., the process holds the Information Axiom (Suh 1998).
??
R
e
q
u
i
r
e
m
e
n
t
E
n
t
r
o
p
y
R
e
q
u
i
r
e
m
e
n
t
E
n
t
r
o
p
y
Therefore, the results described in this section hold the scenario described in Fig.
5.5. In synopsis, creativity is first controlled by the maximization of information
content in presence of such motivating factors as compelling reason and epistemic
challenge and then by the minimization of information content in presence of new
knowledge.
??
??
Chapter 6
Concluding Remarks
Making decisions, i.e., identifying a key solution (or a set of key solution), in
conceptual phase of product development is not only critical but also difficult. It is
critical in a sense that around 80% cost of a product is decided by the key solution
determination process in the conceptual phase of product development and it cannot
be rectified by making adjustments in the downstream of a product lifecycle. It is
difficult in a sense that in conceptual phase of product development, the knowledge
is very limited and there is an abundance of choice. To shed some light on this issue
(decisionmaking in conceptual phase of product development) this thesis poses and
answers the following questions:
How to differentiate a creative key solution from a non-creative key solution?
What is the appropriate customer need model?
How to deal with the unknown customer needs?
How to classify the key solutions based on customer responses?
How to deal with the sustainability of materials (used in the product) in key
solution determination process?
Nevertheless, the following remarks can be made on the findings:
On the customer needs assessment:
1.
2.
3.
One of the ways to identify a key solution to develop a product is to take
opinions of customers regarding a set of key solutions.
To deal with the intrinsic complexity of customer responses, logical
aggregation of customer opinions is a better choice compared to frequency
based analysis. This faculty of thought is demonstrated to be true by
logically aggregating the field data of customer needs collected from
Bangladesh on small passenger vehicles using Kano model.
It has been found that a product feature needs to be classified either into a
??
must be included feature, or into a should be included feature, and or into a
could be included feature. The link among these classifiers and Kano
evaluations (Must-be, Attractive, One-Dimensional, Indifferent, Reverse,
and Questionable) has been established.
4.
5.
The multi-valued logic plays an important role in the customer needs
assessment. In particular, a two-dimensional information content (in
epistemic sense) scheme has been found effective in logically computing
the degree of customer satisfaction of a given product feature in terms of
must be included, should be included, and could be included.
To increase the degree of satisfaction of vehicle users in Bangladesh, it is
important to develop SUV- and Van-type passenger vehicles replacing
some of the Sedan-type vehicles.
On the sustainability assessment:
1.
2.
3.
4.
Sustainability of a product largely depends on the materials used to
manufacturing it. Therefore, the material used to manufacture the product
become one of the key solutions.
To deal with the imprecision associated with the material related
sustainability parameters in the conceptual phase of product development,
an entity called range compliance has been found effective. The compliance
of an sustainability parameter given by a numerical range is determined by
calculating its compliance with five fuzzy numbers of the parameters
labeled very low, low, moderate, high, and very high.
As an example, the imprecision associated with four sustainability
parameters namely, CO
2
footprint, NO
X
emission, SO
X
emission and water
usage (i.e., resource depletion) of five classes of hard materials (the
materials used to produce abrasive grains of grinding wheel or other
material removal tools) based on Alumina, Zirconia, Silicon Carbide, Boron
Nitride, and Boron Carbide have been quantified by using the range
compliance.
The sustainability parameter complying more with very low or low less
negative impact on the sustainability, whereas the sustainability parameter
??
5.
complying more with moderate, high, or very high has high negative impact
on sustainability.
It is found that Alumina based hard materials have low negative impact
followed by Zirconia and Silicon Carbide based hard materials. Boron
Nitride/Carbide based materials have the highest negative impact.
On the creativity assessment:
1.
2.
3.
4.
5.
To identify a useful key solution in conceptual phase of product
development, the product development team members needs to be creative.
To differentiate a creative concept from a non-creative concept, Concept-
Knowledge mapping as prescribed in C-K theory can be employed.
Creative concept means a concept which is undecided when it is being
conceived.
Conceiving a creative concept is rather a motivation driven process.
Information content of a creative concept is high compared to that of a
non-creative concept. The information content means here the
6.
7.
8.
two-dimensional information content in epistemic sense.
When a creative concept is pursued and new knowledge becomes available,
the information content should go down significantly. Otherwise, the new
knowledge does not add any value to product development process.
A non-creative key solution does not exhibit the abovementioned behavior
of information content.
The effectiveness of the abovementioned approach has been demonstrated
by calculating the information contents of two concepts Mg-CO
2
based
propulsion engine (a creative concept) and fossil-fuel base propulsion
engine (an non-creative concept). It has been found that the Mg-CO
2
based
propulsion engine exhibits high information content compared to that of
fossil-fuel base propulsion engine for Mars exploration. The information
content of Mg-CO
2
based propulsion engine have gone down significantly
under the presence of new knowledge.
??
??
??
Bibliography
1.
2.
3.
4.
5.
6.
7.
8.
9.
—.CES Selector¹ (Version 5.1.0). 2009. Granta Design Limited, UK,
http://www.grantadesign.com
Altshuller, G. (2001). And Suddenly the Inventor Appeared: TRIZ, the
Theory of Inventive Problem Solving (Shulyak, L., Trans.). Worcester, MA:
Technical Innovation Center.
Bellman, R.E. and Zadeh, L.A. (1970). Decision-Making in a Fuzzy
Environment, Management Science, 17(4), B-141-B-164.
Berger, C., Blauth, R., Boger, D., Bolster, C., Burchill, G., DuMouchel, W.,
Pouliot, F., Richter, R., Rubinoff, A., Shen, D., Timko, M. and Walden, D.
(1993). Kano's method for understanding customer-defined quality, Center for
Quality of Management Journal, 2(4), 3-35.
Braha, D. and Reich, Y. (2003). Topological structures for modelling
engineering design processes, Research in Engineering Design, 14(4), 185-
199.
Briggs, C. and Little, P. (2008). Impact of Organizational Culture and
Personality Traits on Decision-making in Technical Organization, Systems
Engineering, 11(1), 15-26.
Chen, Z. and Ben-Arieh, D. (2006). On the fusion of multi-granularity
linguistic label sets in group decision making, Computers & Industrial
Engineering, 51(3), 526-541.
de Bono, E. (1970). Lateral Thinking: Creativity Step by Step, Harper
&Row, Publisher: NY.
Dieter, G.E. and Schmidt, L.C. (2009). Engineering Design, Fourth Edition,
McGraw-Hill: New York.
10. Donnelly, K., Olds, R., Blechinger, F., Reynolds, D. and Beckett-Furnell, Z.
(2004). ISO 14001 - effective management of sustainable design, The
Journal of Sustainable Product Design, 4(1-4), 43-54.
??
11. Dubois, D. and Prade, H. (1978). Operations on fuzzy numbers,
International Journal of Systems Science, 9(6), 613-626.
12. Fang, L., Hipel, K.W. and Kilgour, D.M. (1993). Interactive Decision
Making: The graph Model for Conflict Resolution, Wiley & Sons: New
York.
13. Fiksel, J. (2009). Design for Environment: A Guide to Sustainable Product
Development, Second Edition, McGraw-Hill: New York.
14. Foote, J.P. and Litchford, R.J. (2007). Powdered Magnesium—Carbon
Dioxide Rocket Combustion Technology for In Situ Mar Propulsion, NASA
Technical Publication, NASA/TP-2007-215077, available from
https://www2.sti.nasa.gov.
15. Fukushige, S. Yamamoto, K. and Umeda, Y. (2012). Lifecycle scenario
design for product end-of-life strategy, Journal of Remanufacturing, 2012,
2:1, doi:10.1186/2210-4690-2-1.
16. Gero, J.S. (2000). Computational Models of Innovative and Creative Design
Processes, Technological Forecasting and Social Change, 64(2-3), 183-196.
17. Hatchuel, A. and Weil, B. (2009). C-K design theory: an advanced
formulation, Research in Engineering Design, 19(4), 181-192.
18. Hatchuel. A. and Weil., B. (2003). A new approach of innovative design: an
introduction to C-K theory, Proceedings of the 14th International
Conference on Engineering Design Research for Practice (ICED'03), 19-21
August 2003, Stockholm, Sweden, pp 109-124.
19. Herrera, F. and Herrera-Viedma, E. (2000). Linguistic decision analysis:
steps for solving decision problems under linguistic information, Fuzzy Sets
and Systems, 115(1), 67-82.
20. Higuchi, Y., Wada, N., Nakakubo, T. and Tokai, A. (2012). Scenario
Analysis on the Impact of Diffusion of Next Generation Vehicles on Material
Consumption and GHG Emissions. In: Design for Innovative Value Towards
a Sustainable Society (Editors: Matsumoto, M., Umeda, Y., Masui, K. and
Fukushige, S.), pp. 599-604, Springer: The Netherland.
21. Hutton, R.J.B. and Klein, G. (1999). Expert Decision Making, Systems
Engineering, 2(1), 32-45.
??
22. Inohara, T. and Hipel, K.W. (2008). Coalition Analysis in the Graph Model
for Conflict Resolution, Systems Engineering, 11(4), 343-359.
23. Kahn, K.B. (Ed.) (2004). The PDMA Handbook of New Product
Development (Second Edition), Wiley: New York.
24. Kano, N., Seraku, N., Takahashi, F. and Tsuji, S. (1984). Attractive quality
and must-be quality, Hinshitsu, 14(2), 39-48. (In Japanese).
25. Kazakci, A.O. and Tsoukias, A. (2005). Extending the C-K design theory: A
theoretical background for personal design assistants, Journal of
Engineering Design, 16(4), 399-411.
26. Keller, L.R., Kirkwood, C.W. and Jones, N.S. (2008). Assessing Stakeholder
Evaluation Concerns: An Application to the Central Arizona Water
Resources Systems, Systems Engineering, 13(1), 58-71.
27. Khozaimy, O., Al-Dhaheri, A. and Sharif Ullah, A.M.M. (2011). A
decision-making approach using point-cloud-based granular information,
Applied Soft Computing Journal, 11(2), 2576-2586.
28. Klien, G.A. (2008). Naturalistic Decision Making, Human Factors, 50(3),
456-460.
29. Kobayashi, H. (2006). A systematic approach to eco-innovative product
design based on life cycle planning, Advanced Engineering Informatics,
20(2), 113-125.
30. Kujawski, E. (2005). A Reference-Dependent Regret Model for
Deterministic Tradeoff Studies, Systems Engineering, 8(2), 119-137.
31. Noor-E-Alam, M., Lipi, T.F., Hasin, M.A.A. and Sharif Ullah, A.M.M.
(2011). Algorithms for fuzzy multi expert multi criteria decision making
(ME-MCDM), Knowledge-Based Systems, 24(3), 367-377.
32. O'Hagan, A. and Oakley, J.E. (2004). Probability is perfect, but we can't
elicit it perfectly, Reliability Engineering and System Safety, 85, 239-248.
33. Puccio, G.J., Cabra, J.F., Fox, J.M., and Cahen, H. (2010). Creativity on
demand: Historical approaches and future trends, Artificial Intelligence for
Engineering Design, Analysis and Manufacturing, 24(2), 153-159.
34. Rashid, M.M., Tamaki, J., Ullah, A.M.M.S. and Kubo, A. (2010). A Virtual
Customer Needs System for Product Development, Proceedings of the 2010
??
Annual Meeting of Japan Society for Precision Engineering, September 04,
2010, Sapporo, Japan, Pages 53-54.
35. Rashid, M.M., Ullah, A.M.M.S., Sarker, M.A.R., Tamaki, J. and Kubo, A.
(2012). Logical Aggregation of Customer Needs Assessment, Proceedings
of the Fifth International Symposium on Computational Intelligence and
Industrial Applications (ISCIIA2012), Sapporo, Japan, August 20-26, 2012.
36. Rashid, M.M., Ullah, A.M.M.S., Tamaki, J., and Kubo, A. (2011).
Evaluation of Hard Materials using Eco-Attribute, Advanced Materials
Research, 325, 693-698
37. Rasmussen, J. (1993). Deciding and doing: Decision making in natural
contexts. In: Decision Making in Action, Klein, G., Orasallu, J., Calderwood,
R., and Zsambok, C.E. (Eds), Ablex, Norwood, New Jersey, pages 158 - 171.
38. Saaty, T.L. (1980). The Analytic Hierarchy Process, McGraw-Hill: New
York.
39. Saaty, T.L. (1990). Multicriteria Decision Making: The Analytic Hierarchy
Process, RWS Publications, Pittsburgh: Pennsylvania.
40. Shafirovich, E.Y., and Goldshleger, U.I. (1992). Combustion of Magnesium
Particles in CO2/CO Mixtures, Combustion Science and Technology, 84(1),
33-43.
41. Shafirovich, E.Y., Shiryaev, A.A., and Goldshleger, U.I. (1993). Magnesium
and Carbon Dioxide: A Rocket Propellant for Mars Missions, Journal of
Propulsion and Power, 9(2), 197-203.
42. Shamsuzzaman, M., Ullah, A.M.M.S. and Bohez, E.L.J. (2003). Applying
linguistic criteria in FMS selection: fuzzy-set-AHP approach, Integrated
Manufacturing Systems, 14(3), 247-254.
43. Smith, E.D., Son, Y.J., Palmarini-Piattelli, M. and Bahill, T. (2005).
Ameliorating Mental Mistakes in Tradeoff Studies, Systems Engineering,
10(3), 222-240.
44. Suh, N.P. (1990). The principles of design. Oxford University Press: New
York.
??
45. Suh, N.P. (1998). Axiomatic Design Theory for Systems, Research in
Engineering Design, 10(4), 189-209.
46. Tomiyama, T., Gu, P. Jin, Y., Lutters, D., Kind, C., and Kimura, F. (2009).
Design methodologies: Industrial and educational applications, CIRP
Annals-Manufacturing Technology, 58(2), 543-565.
47. Ullah, A.M.M.S. (2005a). A Fuzzy Decision Model for Conceptual Design,
Systems Engineering, 8(4), 296-308.
48. Ullah, A.M.M.S. (2005b). Handling design perceptions: an axiomatic design
perspective, Research in Engineering Design, 16(3), 109-117.
49. Ullah, A.M.M.S. (2008). Logical interaction between domain knowledge
and human cognition in design, International Journal of Manufacturing
Technology and Management, 14(1-2), 215-227.
50. Ullah, A.M.M.S. and Harib, K.H. (2006). Zadehian Paradigms for
Knowledge Extraction in Intelligent Manufacturing. In: Manufacturing the
Future: Concepts-Technologies-Visions. Editors: Kordic, V., Lazinica, A.
and Merdan., M. Pro Literatur Verlag Robert Mayer-Scholz, Mammendorf,
Germany, pages 291-308.
51. Ullah, A.M.M.S. and Harib, K.H. (2008). An Intelligent Method for
Selecting Optimal Materials and its Application, Advanced Engineering
Informatics, 22(4), 473-483.
52. Ullah, A.M.M.S. and Tamaki, J. (2011). Analysis of Kano-Model-Based
Customer Needs for Product Development, Systems Engineering, 14(2),
154-172.
53. Ullah, A.M.M.S., Harib, K.H., Al-Awar, A. (2007). Minimizing Information
Content of a Design using Compliance Analysis, SAE Technical Paper
2007-01-1209, SAE International: Pennsylvania.
54. Ullah, A.M.M.S., Kitajima, K., Akamatsu, T., Furuno, M., Tamaki, J., and
Kubo, A. On Some Eco-indicators of Cutting Tools, Proceedings of the
ASME 2011 International Manufacturing Science and Engineering
Conference, June 13-17, 2011, Corvallis, Oregon, USA.
??
55. Ullah, A.M.M.S., Rashid, M.M. and Tamaki, J. (2012). On Some Unique
Features of C-K Theory of Design, CIRP Journal of Manufacturing Science
and Technology, 5(1), 55-66.
56. Ullman, D.G. (2006). Making Robust Decisions: Decision Management For
Technical, Business, & Service Teams, Trafford Publishing: British
Colombia.
57. Ullman, D.G. (2009). Mechanical Design Process, Fourth Edition,
McGraw-Hill: New York.
58. Ulrich, K.T. and Eppinger, S.D. (2004). Product Design and Development,
Third Edition, McGraw-Hill: New York.
59. Umeda, Y. Nishiyama, T., Yamasaki, Y., Kishita, Y. and Fukushige, S.
(2009). Proposal of sustainable society scenario simulator, CIRP Journal of
Manufacturing Science and Technology, 1(4), 272-278.
60. Walley, P. (1991). Statistical Reasoning with Imprecise Probabilities.
Chapman and Hall: London.
61. Wood, W.H.III. and Agogino, A.M. (1996). Case-based conceptual design
information server for concurrent engineering, Computer-Aided Design,
28(5), 361-369.
62. Xu, Q., Jiao, R.J., Yang, X., Helander, M., Khalid, H.M. and Opperud, A.
(2009). An analytical Kano model for customer need analysis, Design
Studies, 30(1), 87-110.
63. Yager, R.R. (1978). Fuzzy decision making including unequal objectives,
Fuzzy Sets and Systems, 1(2), 87-95.
64. Yang, K. (2008). Voice of the Customer: Capture and Analysis.
McGraw-Hill: New York.
65. Yoshikawa, H. (1981). General Theory of Design Process, Journal of the
Japan Society of Precision Engineering, 47(4), 405-410. (In Japanese)
66. Zadeh, L.A. (1965). Fuzzy sets, Information and Control, 8, 338-353.
67. Zadeh, L.A. (1975). Fuzzy logic and approximate reasoning, Synthese,
30(3-4), 407-428.
??
68. Zadeh, L.A. (1997). Toward a theory of fuzzy information granulation and
its centrality in human reasoning and fuzzy logic, Fuzzy Sets and Systems,
90(2), 111-127.
69. Zadeh, L.A. (2005). From imprecise to granular probabilities, Fuzzy Sets
and Systems, 154(3), 370-374.
70. Zeng, Y. and Cheng, G.D. (1991). On the logic of design, Design Studies,
12(3), 137-141.
??
Appendix A: Customer Needs Data
Collection
Customer needs data has been collected on the some features of small passenger
vehicles. The data collection period was January 2012 to April 2012. The data on the
38 features has been collected using the Kano model. The features are listed in Table
3.3. In addition to the data on 38 features, the physiographic and demographic data of
the respondents has also been collected, as follows:
Table A.1 shows the demographic questions that have been asked to the
respondents in Bangladesh. Needless to say that demographic questions means the
questions related the profession, income level, gender, and alike of an individual. The
questions related to income of an individual are from the context of Bangladesh.
Table A.1. Demographic questions and answers
Statement
private service holder
government service holder
housewife
businessperson
Choose one Frequency
14
12 7
11
I am a/n
My income is
I am a
engineer
doctor
lawyer
student
others
very high
high
moderate
low
Female
Male
11 2
12
26 5
48
60
28
82
18
The frequencies of the answers are also shown in the last column in Table A.1.
On the other hand, Table A.2 shows the psychographic questions that have been
??
asked to the respondents in Bangladesh. Needless to say that psychographic
questions means the questions related the life-style and values of an individual. A
respondents can choose multiple answers from given options. The frequencies of the
answers are also shown in the last column in Table A.2.
Table A.1. Psychographic questions and answers
Check as many as
Statements
I prefer to drive my vehicle by
myself
I prefer to hire a driver to drive
my vehicle
I use my personal vehicle for long
trips
I always use my own vehicle for
commuting to office/school
A vehicle is an essential means of
transportation for me
A vehicle is a luxurious means of
transportation for me
I prefer environmentally friendly
vehicles
you like
•
•
•
•
•
•
•
Frequency
52
39
35
45
53
21
60
??
??
??
List of Research Achievements
The following research publications have been submitted to defend the thesis.
1.
2.
3.
4.
Md. Mamunur Rashid, A.M.M. Sharif Ullah, Junichi Tamaki, and Akihiko
Kubo. (2011). Evaluation of Hard Materials using Eco-Attribute, Advanced
Materials Research, Volume 325, Pages 693-698 [Trans Tech Publications,
Switzerland] [http://dx.doi.org/10.4028/www.scientific.net/AMR.325.693]
A.M.M. Sharif Ullah, Md. Mamunur Rashid and Junichi Tamaki. (2012). On
Some Unique Features of C-K Theory of Design, CIRP Journal of
Manufacturing Science and Technology, Volume 5, Number 1, Pages 55-66.
[Elsevier, The Netherlands] [http://dx.doi.org/10.1016/j.cirpj.2011.09.001].
Md. Mamunur Rashid, A.M.M. Sharif Ullah, M.A. Rashid Sarker, Junichi
Tamaki, and Akihiko Kubo. (2012). Logical Aggregation of Customer Needs
Assessment, Proceedings of the Fifth International Symposium on
Computational Intelligence and Industrial Applications (ISCIIA2012),
Sapporo, Japan, August 20-26, 2012.
Md. Mamunur Rashid, Junichi Tamaki, A.M.M. Sharif Ullah and Akihiko
Kubo. (2010). A Virtual Customer Needs System for Product Development,
Proceedings of the 2010 Annual Meeting of Japan Society for Precision
Engineering, September 04, 2010, Sapporo, Japan, Pages 53-54.
??
??
??
??
doc_481834431.docx
Innovative new products are the fuel for the most powerful growth engine you can connect to. You can grow without new products--AT&T sold essentially the same telephones for decades while becoming the world's largest telecommunications concern--but most small companies will find it difficult to grow at all, much less rapidly, without a constant stream of new products that meet customer needs.
Study on Decision making in Conceptual Phase
of Product Development
Table of Contents
Heading Page
Abstract 1
Chapter 1: Introduction 5
Section 1.1: Product Development
Section 1.2: Decisionmaking in General
Section 1.3: Decisionmaking in Conceptual Phase of
Product Development
Section 1.4: Thesis Structure
Chapter 2: Mathematical Settings
Section 2.1: Fuzzy Number
Sections 2.2: Operations on Fuzzy Number
Section 2.3: Linguistic Likelihood
Section 2.4: Linguistic Truth Value
Section 2.5: Information Content
Section 2.6: Monte Carlo Simulation of Discrete
Events
Chapter 3: Customer Needs Assessment
Section 3.1: Customer needs data collection
Section 3.2: Selection of reliable answers
Section 3.3: Monte Carlo simulation of unknown
answers
Section 3.4: Determination of truth values of product
feature Kano-evaluation
Section 3.5: Determination of truth value of product
feature status
Section 3.6: Determination of information content of
product feature status
Section 3.7: Determination of coherency measures of
product feature status
Section 3.8: Making final decision
Chapter 4: Sustainability Assessment
Section 4.1: Decision-relevant information
Section 4.2: Computational framework
i
5
7
9
9
13
13
14
15
19
21
25
29
30
34
35
38
39
42
43
47
51
52
54
Table of Contents continues
Section 4.3: Results
Chapter 5: Creativity Assessment
Section 5.1: C-K theory
Section 5.2: Differentiating creative and non-creative
concepts
Section 5.3: Results and discussions
Chapter 6: Concluding Remarks
Bibliography
Appendix A: Customer Needs Data Collection
List of Research Achievements
Acknowledgments
57
65
66
68
72
79
83
91
95
97
Pages 3-4, 12, 28, 50, 64, 78, 82, 90, 93-4, 96, and 98-9 are intentionally left blank.
ii
Abstract
Product Development studies the activities underlying a product life cycle in a
concurrent manner. In the conceptual phase of product development, a set of key
solutions are determined by using an appropriate decisionmaking approach. However,
decisionmaking in this phase is a difficult task to perform due to incomplete
information, lack of knowledge, and abundance of choice. This thesis describes
logical approaches for making decisions in conceptual phase of product development.
In particular, the emphasis is given on such issues as customer needs, sustainability,
and creativity. Multi-valued logic and information content from the context of
epistemic uncertainty have been used for the sake of computation. The thesis is
structured, as follows:
Chapter 2 describes the key mathematical entities used in this study such as
fuzzy number, imprecise probability, information content, and discrete event
simulation.
Chapter 3 describes customer needs assessment in conceptual phase of product
development. In this study, a set of field data has been collected from Bangladesh by
using Kano-model-based questionnaires on some features of small passenger
vehicles. The opinions obtained exhibit a high variability and controversy and simple
relative-frequency-based approaches have been found less effective in assessing the
product features, correctly. To solve this problem, a logical customer needs
assessment approach has been developed. The approach has been found effective in
classifying a product feature into the following categories: the feature "must,"
"should," and/or "could" be included in the product. The findings are useful for
developing more customer-focused passenger vehicles.
Chapter 4 describes a decisionmaking approach for assessing the sustainability
in conceptual phase of product development. One of the key sustainability issues of a
product is whether or not the product is made of environmentally friendly materials.
To assess the environmental burden of a material in conceptual phase of product
development, the information of greenhouse gas emissions (CO
2
, NO
X
, SO
X
) and
?
resource consumptions (e.g., water usage) of primary material production is needed.
This kind of information exhibits a high variability and incompleteness. To deal with
this issue, an entity called range compliance is used that logically defines the degree
of belongingness of a given numerical range to a linguistically defined class. Using
range compliance, the environmental burdens of hard materials used to produce a
grinding wheel (i.e., Alumina, Zirconia, Silicon Carbide, and Boron Carbide/Nitride)
have been identified. The findings help develop more environmentally friendly
abrasive tools (products) used in precision engineering.
Chapter 5 describes a decisionmaking approach for assessing a creative concept
in conceptual phase of product development. Creative concept is one of key
ingredients of developing a product. According to a theory called C-K theory, a
creative concept means a concept that is undecided with respect to the existing
knowledge at the point of time when it (the concept) is conceived. In this study, a
logical decisionmaking approach has been developed to assess the degree of
creativeness of a set of concepts. In particular, it has been found that for conceiving a
creative concept one should maximize the information content (in the epistemic
sense) of conceptual phase of product development. The findings help manage the
cognitive processes of conceptual phase of product development.
The last chapter, Chapter 6, provides the concluding remarks and discusses the
scope of further research opportunities.
?
?
?
Chapter 1
Introduction
This chapter describes the general background, scope, and limitation of this
thesis. For the sake of better understanding this chapter is structured into the
following sections: Product Development, Decisionmaking in General,
Decisionmaking in Conceptual Phase of Product Development, and Thesis Structure.
1.1. Product Development
Product Development is a field of study wherein the activities underlying
product lifecycle are studied in a concurrent manner (Ulrich and Eppinger 2004,
Dieter and Schmidt 2009). The lifecycle of a product can be represented in many
ways. One of the ways, which is relevant to this thesis, is schematically illustrated in
Fig. 1.1.
External Customers
(Real Customers)
Customer needs
Use
(Satisfaction)
Internal Customers
(Product Development Team)
Creativity
Conceptual Phase
(Key Solutions)
Detailed Design
Disposal
(Recycle, Downcycle, Landfill)
Manufacturing
Sustainability
Primary Materials
Production
Figure 1.1. A product development scenario.
?
A product development process first decides a set of key solutions (conceptual
phase). The key solutions are further pursued and a detailed design of the product is
produced. Based on the detailed design, manufacturing of the product is conducted.
The manufactured product is then made available to the external customers (real
customers) for use. Using the product, an external customer gets satisfied. If the
product can no longer be used, the product is disposed off. In this phase, the product
might experience recycling, down-cycling, remanufacturing, and/or land-filling.
However, the activities of product development are carried out by a team
(hereinafter referred to as internal customers) wherein a group of individuals from
different departments of an organization (or different organizations) work together to
materialize a product or a family of products. The internal customers first try to make
sure the liking-disliking of external customers (the potential real-customers who will
use the product to get satisfied). The internal customers need to be creative to suggest
many potential key solutions for satisfying the needs of external customers.
Therefore, the following questions might arise in the conceptual phase of product
development:
How to differentiate a creative key solution from a non-creative key solution?
What is the appropriate customer need model?
How to deal with the unknown customer needs?
How to classify the key solutions based on customer responses?
Nowadays, sustainability has become an important issue (Fiksel 2009).
Sustainability often means that the product is environmentally benign on top of other
desired performances. One should incorporate so-called Life-Cycle Assessment
(LCA) into the product development processes to ensure the sustainability (Donnelly
et al. 2004, Kobayashi, 2006). In addition to conventional sustainability assessment
(i.e., LCA), it is important to do scenario analysis (Umeda 2009, Fukushige et al.
2012) taking a broader perspective into the consideration. One of the remarkable
finding underlying scenario-analysis-based sustainability assessment is that the
primary production of materials used in the product plays a critical role to ensure the
sustainability (Higuchi et al. 2012). This implies the following question:
How to deal with the sustainability of materials (used in the product) in key
solution determination process?
?
However, around 80% cost of a product is decided by the key solution
determination process (in the conceptual phase) and it cannot be rectified by making
adjustments in the downstream of product lifecycle (Wood and Agonigo 1996). This
means that the decisionmaking in conceptual phase of product development is a
critical task. In addition, in conceptual phase the knowledge is very limited and there
is an abundance of choice (Dieter and Schmidt 2009). This means that the
decisionmaking in conceptual phase is a very difficult task to perform on top of its
criticalness, as mentioned before.
1.2. Decisionmaking in General
In late 1940s, Neumann and Morgenstern introduced a theory called game
theory. This theory has been accepted as a means to develop methods and tools for
rational decisionmaking. Two approaches have emerged from the game theoretic
practices. One of the approaches uses traditional settings of game theory (e.g.,
conflict/coalition analysis method using graph theory (e.g., see Fang et al. 1993,
Inohara and Hipel 2008 and the references therein)). The other approach has taken
the form of multiple-attribute utility analysis, wherein a set of attributes and their
relative weights are used to simultaneously evaluate (tradeoff) a set of given
alternatives, and, thereby, to select the optimal alternative corresponding to the
maximal utility (Saaty 1980, 1990). However, many authors have studied the
applicability of the multi-attribute utility analysis from the context of real-life
decisionmaking. Some of the salient points are briefly described below. In real-life
decisionmaking, a decision-maker often seeks a balanced alternative rather than an
optimal alternative and it is important to visualize the state of an alternative rather
than to automate the decisionmaking process (Kujawski 2005). Sometimes mental
biases of decision-makers affect the utility-based tradeoff and it is important to take
measures for reducing the biases in terms of problem statement, weights of
importance, alternative solution, evaluation data, scoring function, and combining
function (Smith et al. 2007). Sometimes the sequence of acts (i.e., bring the required
parties together, determine the needs, analyze the data, make a decision and
implement it) is important than the calculation process of tradeoff (Briggs and Little
2008). Sometimes determining the relevant set of criteria and their weights for
?
tradeoffs is a cumbersome task that involves the opinions of stakeholders (Keller et
al. 2008). Thus, in real-life settings it is not an easy task to utilize the utility based
decisionmaking approaches (i.e., rational decisionmaking approaches).
Opposed to rational decisionmaking, there is a faculty of thought of
decisionmaking called naturalistic decisionmaking (Klein 1989, Rasmussen 1993,
Hutton and Klein 1999, Klein 2008). In particular, human experts perform
naturalistic decisionmaking under the following context: time pressure,
incomplete/unreliable information, ill-defined goal, organizational constraints,
multiple decision-makers, and alike. Humans make decision under the
abovementioned context using a decisionmaking approach called recognition-primed
process (Klein 2008) that consists of the following steps: plausible goals, cues to
monitor, expectancies, and sequential action evaluation (Klein 1989, Hutton and
Klein 1999). There are three types of cognitive controls in recognition-primed
process, namely, 1) skill-based spontaneous act, 2) ruled-based conscious attention
and selection of relatively familiar action, and 3) knowledge-based conscious
attention and selection of relatively unfamiliar action (Rasmussen 1993).
Either it is a rational decisionmaking process or it is a naturalistic
decisionmaking process, the decision-relevant information may not necessarily be
crisp in nature. It might be granular in nature (Bellman and Zadeh 1970, Zadeh 1965,
1975, 1997). Zadeh and his colleagues have argued that the manifestation of human
cognitive is a set of "granular information"—imprecisely defined linguistic classes or
clusters of points—and multi-valued logic (known as fuzzy logic) is needed to
formally compute the linguistically expressed imprecise arguments (i.e., granular
information). Multi-attribute utility analysis community (i.e., rational
decisionmaking community) has integrated this idea to make the rational
decisionmaking more realistic (Yager 1978, Herrera and Herrera-Viedma 2000).
There are different models available to deal with the computational complexity of
stakeholder-driven heterogeneous formulation of decision problem and imprecisely
defined decision-relevant information (e.g., Herrera and Herrera-Viedma 2000,
Shamsuzzaman et al. 2003, Chen and Ben-Arieh 2006, Ullah 2005, Noor-E-Alam et
al. 2011). This kind of decisionmaking approach is suitable when the decision-
relevant information is dominated by personal preferences, judgments, and
?
vaguely defined alternatives, weights, and requirements.
1.3. Decisionmaking in Conceptual Phase of Product Development
As mentioned before, decisionmaking in conceptual phase of product
development decides around 80% cost of the product and the decisionmaking process
suffers lack of knowledge and abundance of choice (Wood and Agonigo 1996, Dieter
and Schmidt 2009, Ullman 2009, Ulrich and Eppinger 2004, Ullah 2005). Therefore,
the decision-relevant information in conceptual phase is predominated by personal
preferences, judgments, and vaguely defined alternatives, weights, and requirements.
As a result, granular information based decisionmaking approach is suitable for
making decisions in conceptual phase of product development (Ullah 2005).
However, decisionmaking in conceptual phase of product development requires
an explicit measure that quantifies the lack/abundance of knowledge. For example,
consider the measures called degree of certainty of knowledge in robust
decisionmaking (Ullman 2006) and certainty compliance (entropy) in
general-pinion-desire based decisionmaking (Ullah 2005). In addition, a measure is
needed to quantify the degree of fulfillment of requirement, though the requirement
might be vaguely defined or vary across the external customers. For example,
consider the measure called criteria satisfaction in robust decisionmaking (Ullman
2006) and requirement compliance (entropy) in general-opinion-desire based
decisionmaking (Ullah 2005).
The explanation refers to the fact that a two-dimensional decision measure is
needed for making decisions in conceptual phase of product development. One of the
coordinates of the measure should measure the degree of certainty of knowledge and
the other should measure the degree of fulfillment. However, it would be convenient
if the decision measure is directly related to some of the important principles of
systems design. In this case, general-opinion-desire based decisionmaking is a
desirable one because the certainty entropy and requirement entropy (Ullah 2005) are
directly related to general systems design principles (i.e., information axiom of
axiomatic design of systems) (Suh 1990, 1998, Ullah 2005b).
1.4. Thesis Structure
?
The remainder of this thesis is organized as follows:
Chapter 2 describes the key mathematical entities used in this study namely,
fuzzy number, range compliance, linguistic likelihood, information content, and
discrete event simulation.
Chapter 3 describes customer needs assessment in conceptual phase of product
development. In this study, a set of field data has been collected from Bangladesh by
using Kano-model-based questionnaires on some features of small passenger
vehicles. The opinions obtained exhibit a high variability and controversy and simple
relative-frequency-based approaches have been found less effective in assessing the
product features, correctly. To solve this problem, a logical customer needs
assessment approach has been developed. The approach has been found effective in
classifying a product feature into the following categories: the feature "must,"
"should," and/or "could" be included in the product. The findings are useful for
developing more customer-focused passenger vehicles.
Chapter 4 describes a decisionmaking approach for assessing the sustainability in
conceptual phase of product development. One of the key sustainability issues of a
product is whether or not the product is made of environmentally friendly materials.
To assess the environmental burden of a material in conceptual phase of product
development, the information of greenhouse gas emissions (CO
2
, NO
X
, SO
X
) and
resource consumptions (e.g., water usage) of primary material production is needed.
This kind of information exhibits a high variability and incompleteness. To deal with
this issue, an entity called range compliance is used that logically defines the degree
of belongingness of a given numerical range to a linguistically defined class. Using
range compliance, the environmental burdens of hard materials used to produce
agrinding wheel (i.e., Alumina, Zirconia, Silicon Carbide, and Boron Carbide/Nitride)
have been identified. The findings help develop more environmentally friendly
abrasive tools (products) used in precision engineering.
Chapter 5 describes a decisionmaking approach for assessing a creative concept
in conceptual phase of product development. Creative concept is one of key
ingredients of developing a product. According to a theory called C-K theory, a
creative concept means a concept that is undecided with respect to the existing
knowledge at the point of time when it (the concept) is conceived. In this study, a
??
logical decisionmaking approach has been developed to assess the degree of
creativeness of a set of concepts. In particular, it has been found that for conceiving a
creative concept one should maximize the information content (in the epistemic
sense) of conceptual phase of product development. The findings help manage the
cognitive processes of conceptual phase of product development.
The last chapter, Chapter 6, provides the concluding remarks and discusses the
scope of further research opportunities.
??
??
Chapter 2
Mathematical Settings
Chapter 2 describes the mathematical entities used in this thesis, namely, fuzzy
number, range compliance, linguistic likelihood, information content, and discrete
event simulation. For the sake of better understanding, this chapter is organized into
the following five sections: Fuzzy Number, Operations on Fuzzy Numbers,
Linguistic Likelihood, Information Content, and Monte Carlo Simulation of Discrete
Events.
2.1 Fuzzy Number
A fuzzy number A is a special type of fuzzy subset (Zadeh 1965) wherein the
universe of discourse X is a segment of real-line ? and the maximum membership
value of the membership function ·
A
(x) is equal to 1 (condition of normality). In
addition, ·
A
(x) fulfills the conditions of convexity, continuity, and compactness
(Dubois and Prade 1978, Ullah and Harib 2006). Fuzzy numbers are suitable for
bringing the linguistically defined qualitative entities into a formal computation.
1
Comfortable
0 .7 5
0.5
Cold
0 .2 5
0
0
5
Hot
10 15 20 25 30 35 40 45 50 55 60
Temperature x, (°C)
Figure 2.1. Example of fuzzy number.
For the sake of understanding, consider the three fuzzy numbers labeled Hot,
??
(
.
)
(
x
)
·
Cold, and Comfortable in the universe of discourse X = [0°C,60°C]. The membership
functions are as follows:
·
A
: X = |0,60| ÷ |01| ,
x ·
A
(x) = ÷ x
25 ÷
0
A = Cold
x ·
A
(x) = x ÷ 15 , 35 - x (2.1)
25 ÷ 15 35 ÷
25
A = Comfortable
x ·
A
(x) = x ÷
60 ÷
25
A = Hot
Note that ·
A
(x) is called the membership value of x with respect to A. It, ·
A
(x), is
also called the degree of belief of x in terms of A. The Truth-Value (TV) of the
proposition "x is A" is also equal to ·
A
(x).
For example, let x be 20°C. According to equation (2.1), the membership value of
x = 20°C is equal to 0.2 for A = Cold, 0.5 for A = Comfortable, and 0 for A = Hot.
This means that the TV of a proposition "20°C is a Cold temperature" is equal to 0.2, of
a proposition "20°C is a Comfortable temperature" is equal to 0.5, and of a
proposition "20°C is a Hot temperature" is equal to 0.
In this thesis, fuzzy numbers are used to define such entities as linguistic
likelihoods, linguistic truth values, and classes like high, low, moderate, and alike for
the sake of formal computation.
2.2 Operations on Fuzzy Number
There are many operations applied on a fuzzy number or on a set of fuzzy
numbers. The operations relevant to this thesis are Alpha-cut (Dubois and Prade
1978), Expected Value (Dubois and Prade 1978), and Range Compliance (Ullah
2008).
An Alpha-cut of a fuzzy number A denoted by A
o
is the crisp range for all ·
A
(x) >
o, o e (0,1]. Thus,
A
o
= {x | ·
A
(x) > o e (0 1|} , (2.2)
For example, A
0.5
= [20°C,30°C], if A = Comfortable (see equation (2.1)). The largest
alpha-cut is called Support (Supp(.)). Thus,
Supp(A) = (a,b) {a,b}= |a,b| max(A
o
| ¬o e (0,1|) = (a,b) (2.3)
For example, Supp(A) = [15°C,35°C], if A = Comfortable (see equation (2.1)).
??
Expected value of a fuzzy number A is the first moment of the shape of its
membership function ·
A
, denoted as E(A). Thus,
E ( A) =
} · (x)xdx A
(2.4)
} · (x)dx A
For example, E(A) = 25°C, if A = Comfortable (see equation (2.1)).
Range compliance of a numerical range L is its average membership value with
respect to a fuzzy number A, denoted as R(L,A).
}·
A
(x)dx
R(L, A) =
L
L?
(L? _ L). (L? _ Supp(A))
(2.5)
Here, L? is the segment of L that belongs to Supp(A).
For example, R(L,A) = 0.583, if L = [10°C,30°C] and A = Comfortable.
Range compliance as defined in equation (2.5) will be used in Chapter 4 for
sustainability assessment. Other operations will be used in Chapters 3 and 5.
2.3 Linguistic Likelihood
In real-life cases, probability of an event is difficult to known accurately
(O'Hagan and Oakley 2004). One of the useful representations of real-life probability
is called imprecise probability (Pr
imp
) wherein both upper and lower limits of
probability are used to define the probability of an event (Walley 1991). Alternatively,
the imprecision associate with the probability of an event can be represented by a set
of fuzzy numbers defined in the universe of discourse [0,1] (i.e., possible values of
probability). The labels of the fuzzy numbers are called Linguistic Likelihoods (LLs)
(most likely, less unlikely, and alike). It is useful in determining the probability of an
event given the relative frequency (fr) of the event (Ullah and Harib 2006, Ullah and
Tamaki 2011). It is needless to say that the membership functions of LLs denoted as
·
LLi
(.), i = 1,...,n, are defined in the universe of discourse [0,1].
However, the linguistic counterpart of fr corresponds to an LLj, j e{1,...,n} that
corresponds to the maximum membership value for fr. This means that if
max(·
LLi
(fr) i = 1,...,n) = ·
LLj
(fr) then Pr
imp
(fr) = LLj, j e {1,...,n}. Later the fuzzy
number corresponding to LLj can be used for having a crisp probability for the sake
of calculation. In this thesis, three cases of LL are considered.
??
qu su ns sl ql 1
0.75
0.5
0.25
0
0 0.2 0.4 0.6 0.8 1
fr
Figure 2.2. Membership functions of five LL (Case 1).
1
0.75
0.5
0.25
0
mu
mu
qu qu
su su ns ns
sls
l
ql
ql
ml ml
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
fr
Figure 2.3. Membership functions of seven LL (Case2).
eu
1
eu
mu
u m
qu
qu
su
su
ns
ns
sl
sl
ql
ql
ml
ml e
el
l
0.75
0.5
0.25
0
0 0.125 0.25 0.375 0.5 0.625 0.75 0.875 1
fr
Figure 2.4. Membership functions of seven LL (Case 3).
Case 1 corresponds to five LLs, namely, quite unlikely (qu), some unlikely (su),
not sure (ns), some likely (sl), and quite likely (ql). Case 2 corresponds to seven LLs,
namely, most unlikely (mu), quite unlikely (qu), some unlikely (su), not sure (ns),
some likely (sl), quite likely (ql), and most likely (ml). Case 3 corresponds to nine
??
µ
(
.
)
(
f
r
)
µ
µ
(
.
A
)
(
f
r
)
µ
(
.
)
(
f
r
)
LLs, namely, extremely unlikely (eu), most unlikely (mu), quite unlikely (qu), some
unlikely (su), not sure (ns), some likely (sl), quite likely (ql), most likely (ml), and
extremely likely (el).
Figure 2.2 illustrates the membership functions corresponding to Case 1. The
definitions of the membership functions illustrated in Fig. 2.2 are as follows:
·
qu
( fr) = max 0,min 1, 0.3 ÷ fr
0.3 ÷ 0
(2.6.1)
·
su
( fr) = max 0,min fr ÷ 0 , 0.5 ÷ fr
0.3 ÷ 0 0.5 ÷
0.3
(2.6.2)
·
ns
( fr) = max 0,min fr ÷ 0.3 , 0.7 ÷ fr
(2.6.3)
0.5 ÷ 0.3 0.7 ÷ 0.5
·
sl
( fr) = max 0,min fr ÷ 0.5 , 1 ÷ fr
(2.6.4)
0.7 ÷ 0.5 1 ÷ 0.7
·
ql
( fr) = max 0,min 1, fr ÷ 0.7
1 ÷
0.7
(2.6.5)
Figure 2.3 illustrates the membership functions corresponding to Case 2. The
definitions of the membership functions illustrated in Fig. 2.3 are as follows:
·
mu
( fr) = max 0,min 1, 0.1÷ fr
0.1÷
0
(2.7.1)
·
qu
( fr) = max 0,min fr ÷ 0 , 0.3 ÷ fr
(2.7.2)
0.1÷ 0 0.3 ÷ 0.1
·
su
( fr) = max 0,min fr ÷ 0.1 , 0.5 ÷ fr
(2.7.3)
0.3 ÷ 0.1 0.5 ÷ 0.3
·
ns
( fr) = max 0,min fr ÷ 0.3 , 0.7 ÷ fr
0.5 ÷ 0.3 0.7 ÷
0.5
(2.7.4)
·
sl
( fr) = max 0,min fr ÷ 0.5 , 0.9 ÷ fr
0.7 ÷ 0.5 0.9 ÷ 0.7
(2.7.5)
·
ql
( fr) = max 0,min fr ÷ 0.7 , 1÷ fr
0.9 ÷ 0.7 1÷
0.9
(2.7.6)
·
ml
( fr) = max 0,min 1, fr ÷ 0.9
1÷
0.9
(2.7.7)
Figure 2.4 illustrates the membership functions corresponding to Case 3. In the
conceptual phase of product development, a set of key solutions are determined by
using an appropriate decisionmaking approach.
??
·
eu
( fr) = max 0,min 1, 0.125 ÷ fr
0.125 ÷
0
(2.8.1)
·
mu
( fr) = max 0,min
fr ÷ 0 , 0.25 ÷ fr
0.125 ÷ 0 0.25 ÷
0.125
(2.8.2)
·
qu
( fr) = max 0,min fr ÷ 0.125 , 0.375 ÷ fr
(2.8.3)
0.25 ÷ 0.125 0.375 ÷ 0.25
·
su
( fr) = max 0,min
fr ÷ 0.25 , 0.5 ÷ fr
(2.8.4)
0.375 ÷ 0.25 0.5 ÷ 0.375
·
ns
( fr) = max 0,min fr ÷ 0.375 , 0.625 ÷ fr
(2.8.5)
0.5 ÷ 0.375 0.625 ÷ 0.5
·
sl
( fr) = max 0,min
fr ÷ 0.5 , 0.75 ÷ fr
(2.8.6)
0.625 ÷ 0.5 0.75 ÷ 0.625
·
ql
( fr) = max 0,min fr ÷ 0.625 , 0.875 ÷ fr
0.75 ÷ 0.625 0.875 ÷ 0.75
·
ml
( fr) = max 0,min
(2.8.7)
fr ÷ 0.75 , 1÷ fr
(2.8.8)
0.875 ÷ 0.75 1÷ 0.875
·
el
( fr) = max 0,min 1, fr ÷ 0.875
1÷ 0.875
(2.8.9)
Table 2.1 Summary of LLs
Linguistic Likelihoods (LLs)
Items
Expected
value
Alpha-cut
at o = 0 . 5
Range of
fr
Cases
1
2
3
1
2
3
1
2
3
eu
-
-
0.042
-
-
[0,0.0625]
-
-
[0,0.0625)
mu
-
0.033
0.125
-
[0,0.05]
[0.0625,
0.1875]
-
[0,0.05)
[0.0625,
0.1875)
qu
0.1
0.133
0.25
[0,0.15]
[0.05,
0.2]
[0.1875,
0.3125]
[0,0.15)
[0.05,
0.2)
[0.1875,
0.3125)
su
0.267
0.3
0.375
[0.15,0.4]
[0.2,0.4]
[0.3125,
0.4375]
[0.15,0.4)
[0.2,0.4)
[0.3125,
0.4375)
ns
0.5
0.5
0.5
[0.4,0.6]
[0.4,0.6]
[0.4375,
0.5625]
[0.4,0.6)
[0.4,0.6)
[0.4375,
0.5625)
sl
0.733
0.7
0.625
[0.6,0.85]
[0.6,0.8]
[0.5625,
0.6875]
[0.6,0.85)
[0.6,0.8)
[0.5625,
0.6875)
gl
0.9
0.867
0.75
[0.85,1]
[0.8,0.95]
[0.6875,
0.8125]
[0.85,1]
[0.8,0.95)
[0.6875,
0.8125)
ml
-
0.967
0.875
-
[0.95,1]
[0.8125,
0.9375]
-
[0.95,1]
[0.8125,
0.9375)
el
-
-
0.958
-
-
[0.9375,
1]
-
-
[0.9375,
1]
Table 2.1 summarizes the expected values, alpha-cuts at o = 0.5, and ranges of fr
for three cases of LLs as defined above. In particular, the ranges of fr will be used to
find out the linguistic counterparts of a given fr. For example, if fr = 0.3 then its
??
linguistic counterpart is su according to Case 1, is su according to Case 2, is qu
according to Case 3. See the ranges of fr in Table 2.1 wherein fr = 0.3 belongs. Once
the linguistic counterpart of fr is determined, the corresponding linguistic likelihood
will be used to carry out the subsequent calculations. For example, if fr = 0.3 then its
linguistic counterpart is su according to Case 1 and the expected value of it (su),
0.267 according to Table 2.1, will be used to carry out the subsequent calculation.
Similarly, if fr = 0.3 then its linguistic counterpart is su according to Case 2 and the
expected value of it (su), 0.3 according to Table 2.1, will be used to carry out the
subsequent calculation. Again, if fr = 0.3 then its linguistic counterpart is qu
according to Case 3 and the expected value of it (qu), 0.25 according to Table 2.1,
will be used to carry out the subsequent calculation. This kind of scheme is helpful in
discrete event simulation using a relatively small set of data points.
2.4 Linguistic Truth Value
It is mentioned that the membership value can be considered the truth value of
proposition. For example, recall the proposition of Section 2.1: p = "20°C is a Cold
temperature." The truth value of p is equal to 0.2, TV(p) = 0.2, because ·
Comfortable
(x =
20°C) = 0.2 according to the membership function of the fuzzy number
"Comfortable" as shown in Fig. 2.1 and defined in equation (2.1).
However, there are cases wherein it would be difficult to explicitly construct a
membership function and calculate TV of a proposition from it. In such cases, one
can assign a preferential/judgmental TV using a phrase (mostly true, some false, etc.)
to a proposition. For example, consider the following proposition: p(Z, attractive) =
"Z is an attractive attribute for this product." One can assign a TV = "quite true" to p(Z,
attractive) based her/is judgment. To bring such preferential/judgmental TV, one can
use linguistic TV defined by a set of fuzzy numbers, as it is done for linguistic
likelihoods.
Figure 2.5 illustrates the linguistic TV defined by five fuzzy numbers labeled
mostly false (mf), perhaps false (pf), not sure (ns), perhaps true (pt), and mostly true
(mt). These TVs are used in Chapter 5. Note that the membership functions are the
same compared to those of Case 1 LLs.
??
mf pf ns pt mt
1
0.75
0.5
0.25
0
0 0.2 0.4 0.6 0.8 1
TV
Figure 2.5. Linguistic TV using five fuzzy numbers.
The membership functions are defined as follows:
·
mf
(TV ) = max 0,min 1,0.3 ÷ TV
0.3 ÷
0
(2.9.1.1)
·
pf
(TV ) = max 0,min TV ÷ 0 , 0.5 ÷ TV
(2.9.1.2)
0.3 ÷ 0 0.5 ÷ 0.3
·
ns
(TV ) = max 0,min TV ÷ 0.3 ,0.7 ÷ TV
(2.9.1.3)
0.5 ÷ 0.3 0.7 ÷ 0.5
·
pt
(TV ) = max 0,min TV ÷ 0.5 ,1 ÷ TV
(2.9.1.4)
0.7 ÷ 0.5 1 ÷ 0.7
·
mt
(TV ) = max 0,min 1,TV ÷ 0.7
(2.9.1.5)
1 ÷ 0.7
1
0
.
7
5
0
.
5
µ
(
.
)
(
T
V
)
µ
(
.
)
A
(
T
V
)
µ
0.25
0
mf
mu
qu qf
su
pf
ns
ns
psl
t
q
ql
t
mt
ml
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
TV
Figure 2.6. Linguistic TV using seven fuzzy numbers.
Similarly, Fig. 2.6 illustrates the linguistic TV defined by seven fuzzy numbers
??
labeled mostly false (mf), quite false (qf), perhaps false (pf), not sure (ns), perhaps
true (pt), quite true (qt), and mostly true (mt). The definitions of the membership
functions illustrated in Fig. 2.3 are as follows:
·
mf
(TV ) = max 0,min 1, 0.1 ÷ TV
0.1 ÷
0
(2.9.2.1)
·
qf
(TV ) = max 0,min TV ÷ 0 , 0.3 ÷ TV
(2.9.2.2)
0.1 ÷ 0 0.3 ÷ 0.1
·
pf
(TV ) = max 0,min TV ÷ 0.1 , 0.5 ÷ TV
(2.9.2.3)
0.3 ÷ 0.1 0.5 ÷ 0.3
·
ns
(TV ) = max 0,min TV ÷ 0.3 , 0.7 ÷ TV
(2.9.2.4)
0.5 ÷ 0.3 0.7 ÷ 0.5
·
pt
(TV ) = max 0,min TV ÷ 0.5 , 0.9 ÷ TV
(2.9.2.5)
0.7 ÷ 0.5 0.9 ÷ 0.7
·
qt
(TV ) = max 0,min TV ÷ 0.7 ,1 ÷ TV
(2.9.2.6)
0.9 ÷ 0.7 1 ÷ 0.9
·
mt
(TV ) = max 0,min 1,TV ÷ 0.9
1 ÷ 0.9
(
2
.
9
.
2
.
7
)
These TVs are used in Chapter 3. Note that the membership functions are the
same compared to those of Case 2 LLs.
Table 2.2 summarizes the expected values of the linguistic TVs defined above. If
one assigns the truth value of a proposition equal to perhaps true (pt), then the
subsequent computation will be carried out by using its expected value 0.733.
Table 2.2. Summary of linguistic TVs.
Linguistic TVs
Items
mf qf pf ns pt qt mt
Expected Five 0.1 - 0.267 0.5 0.733 - 0.9
Value Seven 0.033 0.133 0.3 0.3 0.7 0.867 0.967
2.5 Information Content
In 1940s, Shannon introduced the concept of information content as a part of his
information theory wherein an obvious event has low information content and less
likely event has high information content. Thus, if the probability of an event is Pr,
then the information content of the event is given by -log(1/Pr). In systems design,
??
Suh have utilized this concept introducing an axiom called the Information Axiom:
minimize the information content of a design (Suh 1990, 1998). According to the
information axiom, the information content of a functional requirement (FR) of a
system is defined as follows:
I (FR) = ÷log 1
S
(2.10)
In equation (2.10), S is the area under the probability density function of system
range (sr) (the performance of the system designed) for a given design range (dr) (the
requirement defined the designer). A schematic illustration of S, sr, and dr is shown
in Fig. 2.7.
15
.1
05
0
15
20
S
dr
25 FR 30
sr
35
40
Figure 2.7. Definition of information content for systems design.
I(FR) can be minimized by increasing the value of S, i.e., matching sr with respect
to dr. This means that minimization of information content means maximization of
requirement fulfillment. Therefore, information content defined in equation (2.10)
actually determines the degree of requirement fulfillment.
Note that in conceptual phase of product development (the focus of this thesis), it
would be difficult to clearly define the probability density function to represent sr
and the range called dr. Therefore, information content defined in equation (2.10)
(i.e., degree of fulfillment of requirement) may not be applied in conceptual phase of
product development. In addition, in conceptual phase of product development, not
only the degree of fulfillment requirement but also the degree of knowledge should
??
P
r
r
(
(
F
R
)
)
P
F
R
get proper attention (Ullah 2005a-b, Ullman 2006).
The above explanation implies that for conceptual phase of product development,
information content of a key solution has two facets, one is the degree of knowledge
and the other is the degree of fulfillment. In this thesis, a two-dimensional
information content (CE,RE) scheme is used. Here CE means Certainty Entropy that
measures the degree of knowledge in the interval [0,1] and RE means Requirement
Entropy (RE) that measures the degree of fulfillment of a linguistically defined
requirement (see Ullah 2005a, Ullah and Harib 2008). The calculation is done on the
truth values of a set of propositions, TV(P1),...,TV(Pn),TV(P
R
). Here, P1,...,Pn are
general propositions and P
R
is the requirement proposition. The truth values of
P1,...,Pn, TV(P1),...,TV(Pn), are assigned or calculated and the truth value of P
R
,
TV(P
R
), is calculated from TV(P1),...,TV(Pn).
In particular, CE is defined as follows:
n
CE =
¿I
i
c (TV (Pi))
n
(2.11.1)
This means that CE is the average epistemic information content, I
c
(.), of
TV(P1),...,TV(Pn). The epistemic information content I
c
(.) is determined as follows:
I
c
(TV (Pi)) = max 0,min TV ÷ 0 ,1 ÷ TV
0.5 ÷ 0 1 ÷
0.5
(2.11.2)
Figure 2.8 illustrates I
c
(.). As seen from Fig. 2.8, epistemic information content of the
truth value of a proposition is a tent function in the universe of discourse of [0,1].
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1
TV(.)
Figure 2.8. Epistemic information content.
This function ensures that a completely true or false proposition does not have
??
I
?
(
T
V
(
.
)
)
any information content (I
c
(TV = 1 or 0) = 0), whereas if the proposition is neither
true nor false, it has the highest information content (I
c
(TV = 0.5) = 1). For other
cases, the information content is between 0 and 1.
Recall the other coordinate of information content of a key solution in conceptual
phase of product development, RE, Requirement Entropy. RE measures the entropy
of the requirement given by P
R
, which is just the opposite of the degree of fulfillment
of requirement given by P
R
. Thus, if the requirement is fully fulfilled, RE should be
equal to zero (lowest entropy) and if the requirement is fully unfulfilled, RE should
be equal to unit (highest entropy). If the requirement is partially fulfilled, RE is
between 0 and 1. The following function can be used to measure RE (Ullah 2005a):
RE = max 0,min 1, a ÷ TV (P
R
)
a ÷ b
(2.11.3)
a = max(TV (Pi)| i = 1, ,n) b = min(TV (Pi)| i = 1, ,n)
The procedure to determine the TV(P
R
) from TV(P1),...,TV(Pn) shown in Ullah
2005a is used in this thesis. A typical nature of RE is illustrated in Fig. 2.9
corresponding to a = 0.9 and b = 0.05.
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1
TV(P
R
)
Figure 2.9. A function to determine requirement entropy.
The two-dimensional information content of key solutions can be plot on a RE
versus CE plot and a measure called coherency measure (ì) can be determined. The
??
R
E
coherency measure actually aggregates the variability in the information content of a
key solution using the following expression:
ì = e + f + g + h + ( f ÷ e)(h ÷ g) (2.11.4)
In ideal case, ì = 0 that means the solution fully fulfills the requirement the
knowledge of the solution is complete. In reality it does not happen. What is seen in
reality is schematically illustrated in Fig. 2.10. As seen from Fig. 2.10, both key
solutions suffer lack of knowledge and partial fulfillment of requirement, both CE,
RE> 0. However, the variability and magnitude in (CE,RE) points is less for key
solution 1 compared to those of key solution 2. This results a relatively low ì for key
solution 1 compared to that of key solution 2, i.e., ì
1
< ì
2
. Therefore, key solution 1
is a preferable key solution compared to the other. This way decisionmaking can be
carried out in conceptual phase of product development.
1
0.75
0.5
e
Solution 1
Solution 2
f
0.25 ì
1
< ì
2
g h
0
0 0.25 0.5 0.75 1
CE
Fig. 2.10. Decisionmaking using two-dimensional information content.
2.6 Monte Carlo Simulation of Discrete Events
Monte Carlo simulation of discrete events is a useful method to know the
unknown answers of external customers (Ullah and Tamaki 2011). In this thesis,
similar method is used to simulate the unknown answers of external customers. The
simulation steps are explained in Table 2.3. As explained in Table 2.3, the simulation
??
R
E
needs two inputs: a vector of events (ev(1) = Attractive (A), ev(2) = One-dimensional
(O), ev(3) = Must-be (M), ev(4) = Indifferent (I), ev(5) = Reverse (R), ev(6) =
Questionable (Q)) and another vector of their relative frequencies (fr(ev(i))) (e.g.,
(0.1,0.4,0.3,0.1,0.1,0). An example is shown in Table 2.3. Afterward, the user
chooses a set of linguistic likelihoods either from Case 1 or from Case 2 or from Case
3. The linguistic counterpart of each relative frequency fr(i) denoted by LL(i) is
determined by using the ranges listed in the "Range of fr" rows of Table 2.1. For
example, if fr(i) = 0.1, then its linguistic counterpart is qu. The reason is that fr(i) =
0.1 belongs to the range [0,0.15)--a range derived from the alpha-cut qu
corresponding to Case 1 (see Table 2.1). The expected values of the respective
linguistic likelihoods, E(i), are used in the subsequent calculations.
In the subsequent calculations, first, the probability of each event is calculated
by normalizing each expected value, as follows:
E(i)
Pr(i) =
6
¿E(i)
i=1
(2.12.1)
Afterward, the cumulative probability of an event is calculated, as follows:
i
CPr(i) =
¿Pr(i)
j=1
Table 2.3. An example of the settings of simulation
Vector of events (input)
(2.12.2)
Items
fr(i)
(input)
LL(i)
(based on LLs of Case 1)
E(i)
(calculated)
Pr(i)
(calculated)
CPr(i)
(calculated)
ev(1)
A
0.1
qu
0.1
0.079
0.079
ev(2)
O
0.4
ns
0.5
0.395
0.474
ev(3)
M
0.3
su
0.267
0.211
0.684
ev(4)
I
0.1
qu
0.1
0.079
0.763
ev(5)
R
0.1
qu
0.1
0.079
0.842
ev(6)
Q
0
qu
0.1
0.079
0.921
Using the following algorithm a set of simulated events, se(s), s = 1,...,N, are
generated.
??
Define the number iterations N (a relatively large integer)
For s = 1,...,N
Generate a random number RAND in the interval [0,1]
If RAND = [0,CPr(ev(1))) Then se(s) = ev(1)
For k = 2,...,6
If RAND = [CPr(ev(k-1)), CPr(ev(k))) Then se(s) = ev(k)
If RAND = [CPr(ev(5)), CPr(ev(6))] Then se(s) = ev(6)
Needless to say that se(s) e {A, O, M, I, R, Q} for all s = 1,...,N.
??
??
Chapter 3
Customer Needs Assessment
This section deals with the customer needs assessment based on the work of
Rashid et al. 2010 and Rashid et al. 2012. The following issues are emphasized:
How to deal with the unknown customer needs?
How to classify the key solutions based on customer responses?
The framework used in the customer needs assessment process is illustrated in
Fig. 3.1.
Step 1: Collection of customer needs
data using Kano model
Step 2: Selection of reliable answers
Step 3: Monte Carlo simulation of
unknown answers
Step 4: Determination of truth value of
product feature Kano evaluation
Step 8: Making final decision
Step 7: Determination of coherency
measures of product feature status
Step 6: Determination of information
content of product feature status
Step 5: Determination of truth value of
product feature status
Figure 3.1. Customer needs assessment framework.
As seen from Fig. 3.1, the framework consists of eight steps, as follows:
Step 1: Collection of customer needs data using Kano model
Step 2: Selection of reliable answers
Step 3: Monte Carlo simulation of unknown answers
Step 4: Determination of truth value of product feature Kano-evaluation
Step 5: Determination of truth value of product feature status
??
Step 6: Determination of information content of product feature status
Step 7: Determination of coherency measure of product feature status
Step 8: Making final decision
Step 1 deals with the customer needs data collection using Kano model from
Bangladesh on some selected features of small passenger vehicles (cars). This step is
described in details in Section 3.1. Step 2 deals with the determination of reliable
answers of the respondents. This step is described in Section 3.2. Step 3 deals with
the Monte Carlo simulation of unknown answers. This step is described in Section
3.3. Step 4 deals with the determination of truth value of Kano evaluation of product
feature. This step is described in Section 3.4. Step 5 deals with the determination of
truth value of product feature status. This step is described in Section 3.5. Step 6
deals with the determination of information content of product feature status. This
step is described in Section 3.6. Step 7 deals with the determination of coherency
measure of product feature status. This step is described in Section 3.7. The final step,
Step 8, deals with the final decisionmaking using the coherency measure. This step is
described in Section 8.
3.1. Customer needs data collection
Recall Fig. 1.1. As shown in Fig. 1.1, the external customers should get the
desired satisfaction using a product. Therefore, the internal customers (product
development team members) should be aware of the customer needs beforehand.
Otherwise, the product might not be a useful one. The internal customers first decide a
preliminary set of key solutions and prepare a set of questions. A selected segment of
potential external customers (real customers) then answer the questions. Using the
answers obtained, the internal customers try to identify the usefulness of a proposed
key solution or a set of key solutions. Kano model is one of the useful models by
which one can ask questions regarding a product feature so as to classify it (product
feature) into One-dimensional (O) feature, Attractive (A) feature, Must-be (M)
feature, Indifferent (I) feature, Reverse (R) feature, or Questionable (Q) feature
(Kano et al. 1984, Berger et al. 1993, Kahn 2004, Yang 2008, Xu et al. 2009, Ullah
and Tamaki 2011). Figure 3.2 schematically illustrates the implication of O, A, M, I,
??
R, and Q. According to Fig. 3.2, a feature is considered M, if its absence produces
absolute dissatisfaction and its presence does not increase the satisfaction. A feature
is considered O, if its fulfillment helps increase the satisfaction and vice versa. A
feature is considered A, if it leads to a greater satisfaction, whereas it is not expected
to be in the product. A feature is considered I, if its presence or absence does not
contribute to the customers' satisfaction. A feature is considered R if its presence
causes dissatisfaction and vice versa. If the customer answers inconsistently, the
feature is considered Q. However, to know whether or not a feature is O, A, M, I,R, Q
a two-dimensional questionnaire is prepared for each feature. Table 3.1 shows an
example of Kano questionnaire.
Figure 3.2. Kano model of customer needs (based on Ullah and Tamaki 2011).
Table 3.1. An example of Kano questionnaire
Functional answer Dysfunctional answer
Like Like
Must-be ? Must-be
My car is Sedan
Neutral My car is not Sedan Neutral ?
Live-with Live-with
Dislike Dislike
??
As seen from Table 3.2, the two-dimensional questionnaire has two questions,
one deals with customer opinion when function (or feature) is present (functional
answer) and other deals with customer opinion when function (or feature) is not
present (dysfunctional answer). The respondent needs to choose one of the options
(Like, Must-be, Neutral, Live-with, or Dislike) from the functional side. At the same
time, the respondent needs to choose one of the options (Like, Must-be, Neutral,
Live-with, or Dislike) from the dysfunctional side. The combination of the answers
provides the Kano evaluation of the feature of the product. For example, the case
shown in Table 3.1 represents a combination (Must-be, Neutral). As such, the product
feature, i.e., Sedan, is an Indifferent (I) feature--Sedan does not contribute to the
satisfaction or dissatisfaction of the respondent. All possible Kano evaluations with
are summarized in Table 3.2. Note that when a respondent answers Like for both
functional and dysfunctional sides or Dislike for both functional and dysfunctional
sides, the answer should not be trusted, i.e., the product feature is Questionable (Q)
feature.
Table 3.2. Kano evaluation of product feature and function.
Dysfunctional answer
Like Must-be Neutral Live-with Dislike
Functional
answer
Like
Must-be
Neutral
Live-with
Dislike
Q
R
R
R
R
A
I
I
I
R
A
I
I
I
R
A
I
I
I
R
O
M
M
M
Q
A total of 100 respondents are selected at random from Bangladesh and asked to
answer according to Kano questionnaire (e.g., Table 3.1) on 38 features of small
vehicles. The goal is to know the preferences of the respondents so that a key
solutions (or a key solutions) can be determined for product development (i.e., here
the product means a small passenger vehicle). The functional questions of these 38
features are listed in Table 3.3. The demographic and psychographic details of the
respondents are shown in Appendix A.
However, one of the important key solutions for developing a small passenger
vehicle is the type of vehicles. In Table 3.3, there are three types of vehicle, namely,
??
SUV, Sedan, and Van (No. 21-23). The Kano evaluation of these types of vehicles is
shown in Table 3.4 that has been determined using the answers of 100 respondents.
Table 3.3. Small passenger vehicle feature.
No
1
2
3
4
5
6
7
8
9
10
11
I bought a new vehicle
I bought a pre-owned vehicle
My vehicle runs 10-15 km/liter
My vehicle runs 15-20 km/liter
My vehicle runs 25-30 km/liter
My vehicle runs 30-35 km/liter
My vehicle runs 5-10 km/liter
My vehicle color is black
My vehicle color is metallic
My vehicle color is red
My vehicle color is white
Feature
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
My vehicle engine has a 2 year warranty My
vehicle engine has a 3 year warranty My
vehicle engine has a 5 year warranty
My vehicle engine is above 1300 cc My
vehicle engine is below 1000 cc
My vehicle engine is between 1000-1300 cc
My vehicle is a 2-door vehicle My
vehicle is a 4-door vehicle My
vehicle is a 5-door vehicle
My vehicle is a Sedan type vehicle My
vehicle is a SUV type vehicle
My vehicle is a Van (microbus) type vehicle
My vehicle is equipped with airbags
My vehicle is equipped with bumper guards
My vehicle is equipped with keyless-entry system
My vehicle is equipped with seatbelts
My vehicle is made in Germany
My vehicle is made in India My
vehicle is made in Japan My
vehicle is made in Korea
My vehicle needs regular maintenance every after 10,000 km
My vehicle needs regular maintenance every after 20,000 km
My vehicle needs regular maintenance every after 25,000 km
My vehicle needs regular maintenance every after 5,000 km
My vehicles runs on CNG My
vehicles runs on Diesel My
vehicles runs on Petrol
??
Table 3.4. Kano evaluation of different types of vehicle
Vehicle
Type
SUV
Sedan
Van
Kano evaluation
AOMI RQ 14 10 17 41
16 2
20 10 12 37 17 4 15 8
11 21 43 2
The evaluation listed in Table 3.4 exhibits a complex situation as far as formal
computation is concerned. Most of the respondents evaluated a vehicle either I or R.
In Bangladesh, Sedan is the most frequently used type of small passenger vehicle.
This type of vehicle is not that much suitable for the users in Bangladesh because of
the road condition, average size of a family, and life-style (travelling in a large
group). Thus, the Kano evaluation shown in Table 3.4 does not match the reality in
Bangladesh. This necessities the subsequent steps of customer needs assessment.
3.2. Selection of reliable answers
In Kano model, a respondent needs to choose an element drawn from {Like,
Must-be, Neutral, Live-with, Dislike} for both functional and dysfunctional sides.
Not the option called Neutral. Due to the lack of motivation and/or comprehensibility,
the respondent tends to answer this option for frequently. Since the goal is to get
anopinionative answer not an indecisive answer (Neutral), the answers equal to Neutral
should be avoided. For example, consider the case shown in Table 3.4. The
functional and dysfunctional answers are listed in Tables 3.5-6. As seen from Table
3.5, around 30% of the answers are "Neutral" for SUV and Sedan. This number is
however low for Van because it has been disliked by many respondents. On the other
hand, for all three types of vehicle more than 30% of the respondents have answered
Neutral, which is the highest percentage compared to those of other answers.
Table 3.5. Functional answers on vehicle type
Functional Answer
Question
Like Must-be Neutral Live-with Dislike
My car is SUV 26 19 34 9 12
My car is Sedan 31 16 30 8 15
My car is Van 24 7 17 11 41
??
Table 3.6. Dysfunctional answers on vehicle type
Dysfunctional Answer
Question
Like Must-be Neutral Live-with Dislike
My car is not SUV 11 9 37 16 27
My car is not Sedan 11 9 31 24 25
My car is not Van 26 12 31 11 20
As mentioned before, since the goal is to get an opinionative answer not an
indecisive answer (Neutral), the answers equal to Neutral either from functional side
or from dysfunctional side or from both sides should be not be considered for the
assessment. Applying this elimination strategy results the Kano evaluation shown in
Table 3.7. As seen from Table 3.7, the acceptable answers have reduced to 45, 53,
and 61 from 100 for SUV, Sedan, and Van, respectively. Compare Table 3.4 and
Table 3.7. Needless to say that an acceptable answer means here an answer that is not a
Neutral (i.e., not indecisive answer).
Table 3.7. Kano evaluation based on the answers of acceptable respondents
Vehicle
Type
SUV
Sedan
Van
A
6
10
6
Kano evaluation
OMIR 10 9 9
9
10 6 8 15
8 6 5 34
Q
2
4
2
Number of acceptable
respondents
45
53 61
3.3. Monte Carlo simulation of unknown answers
The above section describes that a limited number of answers are available for
the customer needs assessment. This means that a large number of answers are
unknown. Monte Carlo simulation can be used to know the unknown answers (Ullah
and Tamaki 2011, Rashid et al. 2012). The simulation process described in Section
2.6 is adopted here. The explanation is as follows: Table 3.8 shows the settings of
probability of events (A, O, M, I, R, Q) using the Case 1 LLs (see Chapter 2). The
relative frequencies listed in Table 3.8 correspond to results shown in Table 3.7.
??
Table 3.8 Settings of probability using Case 1 LLs.
Case 1
Feature
SUV
Sedan
Van
ev(i)
A
O
M
I
R
Q
A
O
M
I
R
Q
A
O
M
I
R
Q
fr(i)
0.133
0.222
0.200
0.200
0.200
0.044
0.189
0.189
0.113
0.151
0.283
0.075
0.098
0.131
0.098
0.082
0.557
0.033
LL(i)
qu
su
su
su
su
qu
su
su
qu
su
su
qu
qu
qu
qu
qu
ns
qu
E(i) Pr(i) CPr(i)
0.1 0.750 0.079
0.267 2.003 0.289
0.267 2.003 0.500
0.267 2.003 0.711
0.267 2.003 0.921
0.1 0.750 1.000
0.267 2.003 0.211
0.267 2.003 0.421
0.1 0.750 0.500
0.267 2.003 0.711
0.267 2.003 0.921
0.1 0.750 1.000 0.1
0.750 0.100 0.1
0.750 0.200 0.1
0.750 0.300 0.1
0.750 0.400 0.5
3.750 0.900 0.1
0.750 1.000
Table 3.9 Settings of probability using Case 2 LLs.
Case 2
Feature ev(i) fr(i)
LL(i) E(i) Pr(i) CPr(i)
A 0.133 qu 0.133 0.097 0.097
SUV
O
M
I
R
0.222
0.200
0.200
0.200
su
su
su
su
0.3 0.220 0.317 0.3
0.220 0.537 0.3
0.220 0.756 0.3
0.220 0.976
Q 0.044 mu 0.033 0.024 1.000 A
0.189 qu 0.133 0.138 0.138 O
0.189 qu 0.133 0.138 0.276 M
0.113 qu 0.133 0.138 0.413
Sedan
I 0.151 qu 0.133 0.138 0.551
R 0.283 su 0.3 0.311 0.862
Q 0.075 qu 0.133 0.138 1.000 A
0.098 qu 0.133 0.125 0.125 O
0.131 qu 0.133 0.125 0.250 M
0.098 qu 0.133 0.125 0.375
Van
I 0.082 qu 0.133 0.125 0.500
R 0.557 ns 0.5 0.469 0.969
Q 0.033 mu 0.033 0.031 1.000
??
Table 3.10 Settings of probability using Case 3 LLs.
Case 3
Feature
SUV
Sedan
Van
ev(i)
A
O
M
I
R
Q
A
O
M
I
R
Q
A
O
M
I
R
Q
fr(i)
0.133
0.222
0.200
0.200
0.200
0.044
0.189
0.189
0.113
0.151
0.283
0.075
0.098
0.131
0.098
0.082
0.557
0.033
LL(i)
mu
qu
qu
qu
qu
eu
qu
qu
mu
mu
qu
mu
mu
mu
mu
mu
ns
eu
E(i)
0.125
0.25
0.25
0.25
0.25
0.042
0.25
0.25
0.125
0.125
0.25
0.125
0.125
0.125
0.125
0.125
0.5
0.042
Pr(i) CPr(i)
0.107 0.107
0.214 0.321
0.214 0.536
0.214 0.750
0.214 0.964
0.036 1.000
0.222 0.222
0.222 0.444
0.111 0.556
0.111 0.667
0.222 0.889
0.111 1.000
0.120 0.120
0.120 0.240
0.120 0.360
0.120 0.480
0.480 0.960
0.040 1.000
0.5
0.4
0.3
0.2
0.1
0
A O M I R Q
events
Figure 3.2. Variability in fr for SUV and Case 3 LLs due to Monte Carlo simulation.
Using the settings listed in Tables 3.8-10 and the simulation process shown in
Section 2.6, the Monte Carlo simulation of the events (A, O, M, I, R, Q) has been
performed for all SUV, Sedan, and Van. In the simulation, N = 100 (number of
??
f r
iteration) because the original number of respondents was 100 (see Table 3.4). The
variability in the relative frequencies of the simulated events (A, O, M, I, R, Q) has
been determined by repeating the simulation process for all features SUV, Sedan, and
Van. As an example, the variability in the relative frequencies of events for SUV and
Case 3 LLs is shown in Fig. 3.2. The dark rectangular point on each vertical bar in
Fig. 3.2 is the original relative frequency of the event (see Table 3.7).
3.4. Determination of truth values of product feature Kano-evaluation
This section deals with the truth value (TV) determination process of Kano-
evaluation of product feature. Needless to say that Kano evaluation of a product
feature is either A, or O, or M, or I, or R, or Q. The truth value determination process
uses the linguistic TVs (LTk), k = 1,2,..., defined by the seven fuzzy numbers namely,
mostly false (mf), quite false (qf), perhaps false (pf), not sure (ns), perhaps true (pt),
quite true (qt), and mostly true (mt), as defined by the membership functions in
equations (2.9.2.1-7) (see Chapter 2).
Let p(Fi,Xj) be a proposition of the following form: Fi is Xj. Here, Fi e {SUV,
Sedan, Van} and Xj e {A, O, M, I, R, Q}. The problem is to assign a TV e [0,1] to
p(Fi,Xj) from the relative frequency of Xj, fr(Xj), obtained after performing Monte
Carlo simulation as explained in the previous section. An example of how to
determine TV is illustrated in Fig. 3.3. In this particular case, the linguistic
counterpart of fr(Xj) = 0.25 is perhaps false (pf) (linguistic TV shown in bold in Fig.
3.3) because fr(Xj) = 0.25 belongs to pf more it belongs to any other linguistic TV.
1
0.75
0.5
0.25
0
mf
mu
qu qf
fr(Xj)
su
pf ns
ns
psl
t
q
ql
t
mt
ml
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
TV
Figure. 3.3. Converting a relative frequency to a linguistic TV.
??
µ
(
.
)
A
(
T
V
)
µ
In general, the membership value of a linguistic TV denoted asµLTk(TV = fr(Xj))
represents the degree of belongingness of fr(Xj) to a linguistic TV, LTk e {mf, qf, pf,
ns, pt, qt, mt}. IfµLTk(fr(Xj)) >µLTl(fr(Xj)), l e {1, ., 7} - {k} and -k e{1, .,7},
then fr(Xj) belongs to LTk more than it belongs to any other linguistic TV. This way,
LTk is the linguistic counterpart of fr(Xj) and the expected value of LTk, E(LTk), is
the truth value of p(.,Xj), i.e., TV(p(.,Xj)) = E(LTk).
Note that fr(Xj) is not a constant value in the interval [0,1]. It depends on the
simulation instance. For example, consider the variability in the relative frequencies
shown in Fig. 3.2. In particular, consider the variability of relative frequency of A,
fr(Xj = A) = [0.04,0.19]. This means that the truth value (TV(p)) of the proposition
p(SUV,A) = "SUV is A" (i.e., SUV is an attractive car) is around [0.04,0.19], i.e.,
TV(p) ~ fr(A) = [0.04,0.19]. If TV(p) ~ 0.04 in some simulations, then its linguistic
counterpart is mostly false (mf) and the expected value of mf, E(mf) = 0.033,
becomes the truth value of p, i.e., the truth value of the proposition "SUV is an
attractive car is equal to 0.033." Similarly, if TV(p) ~ 0.19 in some simulations, then
its linguistic counterpart is quite false (qf) and the expected value of qf, E(qf) = 0.133,
becomes the truth value of this p, i.e., the truth value of the proposition "SUV is an
attractive car is equal to 0.133." This means that in some other cases TV(p) = 0.133.
Therefore, the TV of p(Fi,Xj) may vary based on the result of Monte Carlo
simulation. As a result, variability in the information content of product feature
might be observed.
3.5. Determination of truth value of product feature status
This section deals with the determination process of truth value of the status of a
product feature. Product feature status is defined by using the Kano evaluations, O, A,
M, I, R, and Q. Therefore, the truth value of the product feature status is calculated
from the truth value of its Kano evaluation. Here a product feature status means one
of the following: must be included, should be included, and could be included. Let Fi
be a product feature and Yj be an element of {must be included, should be included,
and could be included}. Therefore, the problem is to determine the truth value of a
proposition of the following form: Fi Yj in the product.
Recall the schematic illustration of Kano evaluation shown in Fig. 3.2.
??
According to Fig. 3.2, if a product feature is classified as One-dimensional (O) or
Must-be (M) and it is not included in the product, the customers are not satisfied.
Therefore, a product feature "must be included" in the product means it is "either O
or M." This leads to the following formulation:
Fi must be included ÷ (F
i
is O) v (F
i
is M )
TV (Fi must be included ) = max(TV (Fi,O),TV (Fi,M ))
(3.1)
Recall the definition of A illustrated in Fig. 3.2. If the feature is classified as
Attractive (A), then it is an unexpected but customer satisfaction-enriching feature.
Thus, a product feature classified as A, it "should be included" in the product for
enhancing the level of customer satisfaction. This yields the following formulation:
Fi should be included ÷ F
i
is A
TV (Fi should be included ) = TV (Fi, A)
(3.2)
On the other hand, if a feature is classified as Indifferent (I), it is not that much
helpful in increasing or decreasing the level of customer satisfaction even though it is
included or not included in the product, respectively. In addition, if a feature is
classified as Reverse (R), its inclusion in the product creates a great deal of
dissatisfaction. Moreover, if a feature is classified as Questionable (Q), then it is an
unreliable feature. This means that if a feature is "I or not R or not Q," it "could be
included" in the product. This yields the following formulation:
Fi could be included ÷ Fi is I v Fi is ÷R v Fi is ÷Q
( )( )( )
(3.3)
TV (Fi could be included ) = max(TV (Fi, I ),1 ÷ TV (Fi, R),1 ÷ TV (Fi, Q))
Table 3.11 shows an example of truth value determination process of the status of
product features defined in equations (3.1-3). The relative frequencies of Kano
evaluation found after performing Monte Carlo simulation (Section 3.3) are used to
determine the TV of Kano evaluation, A, O, M, I, R, and Q. Afterward, the TV of the
status of the product features, namely, must be include, should be included, and could
be included, are calculated using equations (3.1-3), respectively. It is observed that
the relative frequency of Kano evaluation depends on both the simulation instance
and the case of linguistic likelihoods. The example shown in Table 3.11 refers to a
simulation instance. The simulated relative frequencies are quite different (compare
the relative frequencies of Cases 1-3) but the truth values (calculated from the
linguistic counterparts as explained in Section 3.4) are quite similar (not exactly the
same (compare the truth values of status of Cases 1-3)). This is not only true for the
??
case shown in Table 3.11 but also for other simulation instances.
Table 3.11. An example of product feature status truth value determination.
Case 1
Kano
evaluation
fr
TV
A
0.11
0.133
O
0.2
0.3
M
0.23
0.3
I
0.2
0.3
R
0.17
0.133
Q
0.09
0.133
Remarks
After
simulation
Status
TV
Kano
evaluation
A
must be
included
0.3
O
M
should be
included
0.133
Case 2
I
R
could be
included
0.867
Q
fr
TV
0.15
0.133
0.21
0.3
0.22
0.3
0.2
0.3
0.19
0.133
0.03
0.033
After
simulation
Status
TV
Kano
evaluation
A
must be
included
0.3
O
M
should be
included
0.133
Case 3
I
R
could be
included
0.967
Q
fr
TV
0.16
0.133
0.2
0.3
0.22
0.3
0.2
0.3
0.17
0.133
0.05
0.133
After
Simulation
Status
TV
must be
included
0.3
should be
included
0.133
could be
included
0.867
However, consider the relative frequencies of Kano evaluation corresponding to
Case 1 in Table 3.11: fr(A) = 0.11, fr(O) = 0.2, fr(M) = 0.23, fr(I) = 0.2, fr(R) = 0.17,
fr(Q) = 0.09. The linguistic counterpart of fr(A) = 0.11 is quite false (qf), fr(O) = 0.2
is perhaps false (pf), fr(M) = 0.23 is perhaps false (pf), fr(I) = 0.2 is perhaps false (pf),
fr(R) = 0.17 is quite false (qf), and fr(Q) = 0.09 is quite false (qf). This is in
accordance with the procedure explained in Section 3.3 and with the seven-fuzzy-
number-based linguistic truth values defined in Chapter 2. Thus, the truth value of
the proposition "Sedan is A" is equal to the expected value of qf (0.133), i.e., TV(A)
= 0.133. Similarly, the propositions "Sedan is R," "Sedan is Q" also have the truth
value 0.133, i.e., TV(R) = 0.133 and TV(Q) = 0.133. On the other
??
hand, the propositions "Sedan is O," "Sedan is M," and Sedan is I" have the truth
value 0.3 because 0.3 is the expected value of the linguistic truth values of these
propositions, i.e., perhaps true (pf). This means that TV(O), TV(M), and TV(I) = 0.3.
Thus, the truth value of the proposition "Sedan must be included in the product" is
equal to max(TV(O),TV(M)) = max(0.3,0.3) = 0.3. The truth value of the proposition
"Sedan should be included in the product" is equal to TV(A) = 0.13. The truth value of
the proposition "Sedan could be included in the product" is equal to max(TV(I),1-
TV(R),1-TV(Q)) = max(0.3,1-0.133,1-0.133) = 0.867.
3.6. Determination of information content of product feature status
This section describes the information content determination process of product
feature status. The information content means here the two-dimensional information
content wherein one of the dimensions is the Certainty Entropy (CE) and the other
dimension is the Requirement Entropy (RE). Note that CE and RE have already been
defined in Section 2.5 (equations 2.11.1-3). CE measures the variability in the truth
values of a feature and RE measures the degree of fulfillment of requirement given
by P
R
. For this particular case, P
R
= "The feature is a must be included feature," or
"The feature is a should be included feature," or "The feature is a could be included
feature." For an example, consider the truth values of the status of the product feature
shown in Table 3.11. For the sake of better understanding these truth values are
organized in Table 3.12. Table 3.12 also lists the calculated CE and RE based on
these truth values and also on the P
R
. The P
R
in Table 3.12 is The feature is a must be
included feature for all three cases. Case 1 and Case 3 underlie the same amount of
information content (CE,RE) = (0.377,0.722), whereas the information content
underlying the Case 2 is (CE,RE) = (0.311,0.8). The value of CE = 0.377 means that
there is great deal of consensus among the respondents. The degree of consensus is
comparatively much higher for Case 2 because for Case 2 CE has reduced to 0.311
(less than that of Case 1 and Case 3). One the other hand RE is quite high for all three
cases. This means that the requirement "the feature is a must be feature" has hardly
been fulfilled. This means that if one considers this feature a must be feature, it
might create problem in fulfilling this expectation. However, if one resets the
requirement to "the feature is a could be feature," then RE = 0. This also means that
??
the feature is a could be feature rather than a must be or should be feature.
Table 3.12. An example of information content determination process
Case 1
Status
TV
Ic
must be
included
0.3
0.6
should be
included
0.133
0.266
could be
included
0.867
0.266
CE 0.377
P
R
The feature is a must be included feature
RE 0.722
Case 2
Status
TV
Ic
must be
included
0.3
0.6
should be
included
0.133
0.266
could be
included
0.967
0.067
CE 0.311
P
R
The feature is a must be included feature
RE
Status
TV
Ic
0.8
must be
included
0.3
0.6
Case 3
should be
included
0.133
0.266
could be
included
0.867
0.266
CE 0.377
P
R
The feature is a must be included feature
RE 0.722
3.7. Determination of coherency measures of product feature status
This section describes the coherency measure of product feature status. As
explained in the previous sections, the values of truth values product feature status
might change due to the result of simulation. As such, the information contents
(CE,RE) might also vary with the simulation instance. Therefore, the variability in
the information content (CE,RE) should play a role in the customer needs assessment
process. As explained in Section 2.5 (see equation (2.11.4) and Fig. 2.10) a quantity
called coherency measure (ì) measures the variability in (CE,RE) for a given feature
and requirement.
Figure 3.4 shows the variability in (CE,RE) for the product feature called SUV
for all cases, Cases 1-3. The information content is high for should be and must be
included and low for could be included.
??
1.0
0.8
0.6
0.4
0.2
0.0
must be included
should be included
could be included
0.0 0.2 0.4 0.6 0.8 1.0
CE
Figure 3.4. Variability in the information content of SUV.
1.0
0.8
0.6
0.4
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
CE
Figure 3.5. Determining the coherency measure of SUV for the status called "must be
included."
Based on the data shown in Fig. 3.4, the coherency measure of SUV for three
requirements can be calculated separately. For example, consider the requirement
??
R
E
R
E
"SUV is a must be included feature." The variability in the information content for
this requirement is shown in Fig. 3.5, which is the segment of data points already
shown in Fig. 3.4 corresponding to must be included. The value of coherency
measure ì is equal to 2.313 because e = 0.199, f = 0.377, g = 0.68, h = 1 (see
equation (2.11.4) and Fig. 2.10). Similarly, the values of coherency measure of SUV
for should be included and could be included have been found to be 2.577 and 0.577,
respectively.
However, Figs. 3.6-7 show the variability in (CE,RE) for the other two product
features called Sedan and Van for all cases, Cases 1-3. Similar to the case shown in
Fig. 3.4, the information content is high for should be and must be included and low
for could be included for both cases in Figs. 3.6-7. Based on the data points shown in
Figs. 3.4,6-7, the value of the coherency measure has been determined using the
procedure illustrated in Fig. 3.5. The values are listed in Table 3.13. Note that Sedan
exhibits high values of coherency measure compared to those of SUV and Van. This
means that SUVs and Vans might be good options to replace Sedans.
1.0
0.8
must be included
0.6
should be included
0.4
could be included
0.2
0.0
0.0 0.2 0.4 0.6 0.8 1.0
CE
Figure 3.6. Variability in the information content of Sedan.
??
R
E
1.0
0.8
0.6
0.4
0.2
0.0
must be included
should be included
could be included
0.0 0.2 0.4 0.6 0.8 1.0
CE
Figure 3.7. Variability in the information content of Van.
Table 3.13. Coherency measure of product features.
Status
Features
SUV
Sedan
Van
must be
included
2.314
2.578
2.332
should be
included
2.577
2.578
2.332
could be
included
0.577
0.755
0.577
3.8. Making final decision
This section describes process of how to making a final decision (customer
needs assessment). In this case, the customer needs assessment means to identify the
level of satisfaction of SUV, Sedan, and Van based on the value of the respective
coherency measure (overall information content). Note that a low value of coherency
measure is desirable. Section 2.5 describes the details of the implication of coherency
measure.
However, recall the situation in Bangladesh. Sedan is the most frequently used
vehicles in Bangladesh. This type of vehicle is not that much suitable for the users in
Bangladesh because of the road condition, average size of a family, and life-style
(travelling in a large group). Thus, all Sedan, SUV, and Vans could be suitable for but
??
R
E
it would be difficult to conclude precisely that they are must/should be the vehicle for
users in Bangladesh. At least it can be said that increase in the number of SUV and
Van compared to that of Sedan might lead to an enhancement in the customer
satisfaction.
Whether or not the above conclusion holds if a decision is made based on the
values of coherency measure is an important issue to investigate. As such, the values
of the coherency measure are plotted separately for each requirement. Figure 3.8
shows the value of coherency measure when the requirement refers to "could be
included" for Sedan, SUV, and Van. Needless to say that the values correspond to the
values listed in Table 3.13. As seen from Fig. 3.8, if SUV and Van are introduced side
by side Sedan in a large volume in Bangladesh, the level of satisfaction of vehicle
users "could" increase. In this case, SUV and Van are indifferent.
1.0
0.8
0.6
0.4
0.2
0.0
Sedan SUV Van
Features
Figure 3.8. Reduction in overall information content for could be included.
Figure 3.9 shows the value of coherency measure when the requirement refers
to "should be included" for Sedan, SUV, and Van. Needless to say that the values
correspond to the values listed in Table 3.13. As seen from Fig. 3.9, if Van is
introduced side by side Sedan in a large volume in Bangladesh, the level of
satisfaction of vehicle users "should" increase. This time, SUV does not increase the
level of satisfaction compared to that of Sedan. This decision however, underlies a
great deal of uncertainty (a large value of coherency measure).
??
ì
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Sedan SUV Van
Features
Figure 3.9. Reduction in overall information content for should be included.
Figure 3.10 shows the value of coherency measure when the requirement refers
to "must be included" for Sedan, SUV, and Van. Needless to say that the values
correspond to the values listed in Table 3.13. As seen from Fig. 3.10, if SUV and Van
are introduced side by side Sedan in a large volume in Bangladesh, the level of
satisfaction of vehicle users "must" increase. The trend seen here similar to that of
could be included (Fig. 3.8). This time a great deal of uncertainty (a large value of
coherency measure) is associated with the decision.
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Sedan SUV Van
Features
Figure 3.10. Reduction in overall information content for must be included.
??
ì
ì
In synopsis, the following statements can be made:
level of satisfaction of car users in Bangladesh (Fig. 3.8).
Bangladesh (Fig. 3.9)
Bangladesh but this conclusion possesses a great deal of uncertainty (Fig.
3.10).
Similar to the case of car types (Sedan, SUV, Van), other features in Table 3.3
can be studied and similar conclusions can be made. This way, customer needs
assessment can be carried out and the key solutions to develop the product can be
determined in the conceptual phase of product development.
??
??
Chapter 4
Sustainability Assessment
This chapter deals with the sustainability assessment of key solutions in
conceptual phase of product development. This chapter is based on work of Rashid et
al. 2011. In this chapter, a product means a grinding wheel, i.e., a cutting tool used to
remove hard materials and also to finish surfaces of precision parts.
For the sake of better understanding, recall Fig. 1.1 as repeated in Fig. 4.1.
External Customers
(Real Customers)
Customer needs
Use
(Satisfaction)
Internal Customers
(Product Development Team)
Creativity
Conceptual Phase
(Key Solutions)
Detailed Design
Disposal
(Recycle, Downcycle, Landfill)
Manufacturing
Sustainability
Primary Materials
Production
Figure 4.1. A product development scenario (Fig. 1.1 repeated).
As mentioned in Chapter 1, sustainability has become an important issue of
product development (Fiksel 2009) that refer to the fact that the product is
environmentally benign on top of other desired performances. One should
incorporate so-called Life-Cycle Assessment (LCA) into the product development
processes to ensure the sustainability (Donnelly et al. 2004, Kobayashi, 2006). In
??
addition to conventional sustainability assessment (i.e., LCA), it is important to do
scenario analysis (Umeda 2009, Fukushige et al. 2012) taking a broader perspective
into the consideration. However, one of the remarkable finding underlying
scenario-analysis-based sustainability assessment is that the primary production of
materials used in the product plays a critical role to ensure the sustainability (Higuchi
et al. 2012). This implies the following question:
How to deal with the sustainability of materials (used in the product) in the key
solution determination process in the conceptual phase of product development?
This section provides an answer to this question by taking the example of a
grinding wheel. This means that the grinding wheel is the product here and
sustainability factors of the primary material production of the materials used in
manufacturing a grinding wheel are the important decision-relevant information. This
also means that the possible type of materials used to manufacture a grinding wheel
is the key solutions. The remainder of this chapter is as follows:
Section 4.1 describes the information relevant to the sustainability of primary
material production of the materials used to manufacture a grinding wheel. Section
4.2 describes the fuzzy numbers used to formally compute the
uncertainty/imprecision associated with decision-relevant information. Section 4.3
describes the results and discusses the implication of the results.
4.1. Decision-relevant information
In general, very hard materials based on technical ceramics are used to produce
the abrasive grains of a grinding wheel. The abrasive grains actually provide the
main cutting action, i.e., they are the main ingredients of a grinding wheel. There are
many hard materials (Ullah et al. 2011), e.g., price, annual production, density,
energy consumption for primary production of bulk material, energy for processing
(powder formation, bulk deformation, etc.) materials, CO
2
footprint for primary
production and processing, NO
X
emission of primary production, SO
X
emission of
primary production, water usage of primary production, recycle fraction, CO
2
emission and energy for recycling, and alike. Obtaining reliable data on these
sustainability attributes is not an easy task. In most cases, an estimation is given in
the form of a numerical range on an eco-attribute compiling data/information from
??
many sources. Some of the sustainability attributes do not have any information (e.g.,
recycle fraction, gel formation, etc.).
However, more than 320 types of technical ceramics based hard materials available
in the database of a material evaluation system called CES Selector (version 5.1.0)
developed by the Granta Design Limited have been studied (reference [1]). The
maximum and minimum ranges of each sustainability attribute (in particular, CO
2
footprint (or emission), NO
X
emission, SO
X
emission, and water usage of primary
production) for five different classes of hard material, namely, Alumina (AN), Silica
(SC), Boron Nitride (BN), Boron Carbide (BC), and Zirconia (ZN) are identified.
The variants of AN, SC, BN, BC, and ZN not used for producing abrasive grains are
excluded from this study. Figure 4.2 shows the variability in CO
2
footprint
(kg-CO
2
/kg-material) and water usage (l-water/kg-material) of AN, SC, BN, BC, and
ZN. As seen from Fig. 4.2, the information of CO
2
footprint underlies low
uncertainty/impression whereas water usage exhibit a relatively high
uncertainty/imprecision.
300
250
200
150
100
50
0
Alumina
Silicon Carbide
Zirconia
Boron Nitride
Boron Carbide
0 3 6 9 12 15
CO
2
Footprint (kg/kg)
Figure 4.2. CO
2
footprint and water usage of some selected hard materials.
??
W
a
t
e
r
U
s
a
g
e
(
l
/
k
g
)
120
90
60
Alumina
30 Silicon Carbide
Zirconia
Boron Nitride
Boron Carbide
0
0 20 40 60 80
NO
X
(g/kg)
Figure 4.3. NO
X
/SO
X
emissions of some selected hard materials.
Figure 4.3 shows the variability in NO
X
and SO
X
emissions (g-NO
X
or
SO
X
/kg-material) of the primary production of AN, SC, BN, BC, and ZN. As seen
from Fig. 4.3, the variability in the information increases with the increase in NO
X
or
SO
X
emission. This means that underlying uncertainty/impression increases with the
increase in the emission of NO
X
or SO
X
.
4.2. Computational framework
To deal with the uncertainty/imprecision associated with the parameters
described above computational framework based on "range compliance" is proposed
here. Range compliance has already been explained in Section 2.2 (equation (2.5).
Range compliance is an operation on a fuzzy number using a numerical range. In
particular, range compliance R(L,A) of a numerical range L is its average
membership value with respect to a fuzzy number A. The expression of R(L,A) is as
follows:
}·
A
(x)dx
R(L, A) =
L
L?
(L? _ L). (L? _ Supp(A))
(4.1)
??
S
O
X
(
g
/
k
g
)
In equation (4.1), L? is the segment of L that belongs to Supp(A). Note that the
equation (4.1) is the repetition of equation (2.5).
However, to be more specific consider the following objects. Let G be a
member of the set of materials {AN, SC, BN, BC, ZN}, i.e., G e {AN, SC, BN, BC,
ZN}. Let S be a member of the set of sustainability parameters {CO
2
footprint, water
usage, NO
X
emission, SO
X
emission}, i.e., S e {CO
2
footprint, water usage, NO
X
emission, SO
X
emission}. Let F
S
be a member of the set of fuzzy numbers {VL
S
, L
S
,
M
S
, H
S
, VH
S
}. Here, VL refers to very low, L refers to low, M refers to moderate, H
refers to high, and VH refers to very high. The subscript "S" means that the
sustainability parameter is S. Let X
S
be a point on the real line, X
S
e ?. The interval
[0, X
S
] is the universe of discourse of a fuzzy set F
S
. The membership functions of
the fuzzy numbers VL
S
, L
S
, M
S
, H
S
, and VH
S
(in general F
S
) can thus be defined as
follows:
·
VL
S
(x
S
) = max 0,min 1, 0.3X
S
÷ x
S
0.3X
S
÷
0
(4.2)
·
L
S
(x
S
) = max 0,min x
S
÷ 0.1X
S
, 0.5X
S
÷ x
S
0.3X
S
÷ 0.1X
S
0.5X
S
÷ 0.3X
S
(4.3)
·
M
S
(x
S
) = max 0,min x
S
÷ 0.3X
S
, 0.7 X
S
÷ x
S
0.5X
S
÷ 0.3X
S
0.7 X
S
÷ 0.5X
S
(4.4)
·
H
S
(x
S
) = max 0,min x
S
÷ 0.5X
S
, 0.7 X
S
÷ x
S
0.7 X
S
÷ 0.5X
S
0.9 X
S
÷ 0.9 X
S
(4.5)
·
VH
S
(x
S
) = max 0,min 1, x
S
÷ 0.7X
S
X
S
÷ 0.7 X
S
(4.6)
Figures 4.2-3 provide an estimation of X
S
. As seen from Figs. 4.2-3, X
S
could be
a point in the interval [12,15] if S = CO
2
footprint, X
S
could be a point in the interval
[280,300] if S = water usage, X
S
could be a point in the interval [70,80] if S = NO
X
emission, and X
S
could be a point in the interval [110,120] if S = SO
X
emission.
Figure 4.4 illustrates the membership functions for S = CO
2
footprint and X
S
= 15. At
the core L
S
, M
S
, or H
S
(a point corresponding to unit membership value) the
membership values of other fuzzy numbers are equal to zero. This nature of remains
the same for all fuzzy numbers irrespective of the state of S and the value of X
S
because of the definitions in equations (4.2-6).
??
·
M
S
(x
S
)
1
0.75
0.5
0.25
0
0
VL
S
3
L
S
6
M
S
9
H
S
12
VH
S
15
x
S
Figure 4.4. Membership functions when S = CO
2
footprint and X
S
= 15.
Let L
S
(G) is the range of sustainability parameter S for a material G. One can
calculate the range compliance R(L
S
(G), F
S
) with respect to F
S
, as follows:
}
(
·)
F
S
(x
S
)dx
S
R(L
S
(G),F
S
) =
L
S
G
L
S
? (G)
(L?
S
(G) _ L
S
(G)). (L?
S
(G) _ Supp(F
S
))
(4.7)
In equation (4.7), L?
S
(G) is the largest segment of L
S
(G) that belongs to the support of
F
S
, Supp(F
S
).
For example, if S = CO
2
footprint, G = SC, the L
S
(G) = [6,7.8] (see Fig. 4.2).
This yields the following range compliances: R(L
S
(G),F
S
) = 0 for F
S
= VL
S
,
R(L
S
(G),F
S
) = 0.25 for F
S
= L
S
, R(L
S
(G),F
S
) = 0.783 for F
S
= M
S
, R(L
S
(G),F
S
) = 0.05
for F
S
= H
S
, R(L
S
(G),F
S
) = 0 for F
S
= VH
S
.
To achieve a better sustainability all CO
2
footprint, NO
X
emission, SO
X
emission, and water usage should be minimized. This means that the material that
complies more with VL
S
or L
S
is good material and the material that complies more
with M
S
, H
S
, or VH
S
is not-so-good material. Based on this contemplation two indices
can be derived called Desirable Impact (DI
S
(G)) and Undesirable Impact (UI
S
(G)),
as follows:
DI
S
(G) = R(L
S
(G),VL
S
)+ R(L
S
(G),L
S
) (4.8)
UI
S
(G) = R(L
S
(G),M
S
) + R(L
S
(G),H
S
) + R(L
S
(G),VH
S
) (4.9)
??
The above formulation provides an two-dimensional decision space, wherein a
sustainable material G means its DI
S
(G) is high and UI
S
(G) is low with respect to all
Ss. This decision space is schematically illustrated in Fig. 4.5.
Less
sustainable
material
Highly
sustainable
material
0
0 DI
S
(G)
Figure 4.5. Decision space for sustainability assessment.
4.3. Results
This section describes the results obtained by using the framework described in
the previous section. The range compliance R(L
S
(G),F
S
) has been calculated for all
materials and sustainability parameters for two different cases Case 1 and Case 2.
These cases are listed in Table 4.1. Needless to say that the values of X
S
in Table 4.1
define the universe of discourses of the fuzzy numbers defined in equations (4.2-6).
Table 4.1 Cases of setting the universe of discourse
X
S
Case
1
2
CO
2
footprint
15
12
water usage
300
280
NO
X
emission
80
70
SO
X
emission
120
110
The values of X
S
listed in Table 4.1 underlies the observation mentioned before,
i.e., X
S
could be a point in the interval [12,15] if S = CO
2
footprint, X
S
could be a
??
U
I
S
(
G
)
point in the interval [280,300] if S = water usage, X
S
could be a point in the interval
[70,80] if S = NO
X
emission, and X
S
could be a point in the interval [110,120] if S =
SO
X
emission. In particular the maximum and minimum values are considered to see
the sensitivity.
Table 4.2 shows the range compliances of AN, R(L
S
(AN),F
S
). Figure 4.6 shows
the position of AN in the two-dimensional decision space, wherein the data points
(DI
S
(AN),UI
S
(AN)) have been calculated using the data points listed in Table 4.2.
Table 4.2. Range compliances of AN.
F
S
S Case
VL
S
L
S
M
S
H
S
VH
S
CO
2
1 0.344 0.483 0 0 0
footprint 2 0.181 0.729 0 0 0
water 1 0.333 0.5 0 0 0
usage 2 0.321 0.736 0.05 0 0
NO
X
1 0.315 0.531 0 0 0
emission 2 0.214 0.678 0 0 0
SO
X
1 0.152 0.87 0.083 0 0
emission 2 0.121 0.835 0.159 0 0
1.5
1.25
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1 1.25 1.5
DI
S
(AN)
Figure 4.6. Sustainability of AN.
As seen from Fig. 4.6, the undesirable impact is very low and the desirable
impact is very high for AN for all four sustainability parameters. Therefore, AN is a
good material from the view point of sustainability and one should use this material
??
U
I
S
(
A
N
)
to manufacture a grinding wheel unless there are other problems.
Table 4.3 shows the range compliances of SC, R(L
S
(SC),F
S
). Figure 4.7 shows
the position of SC in the two-dimensional decision space, wherein the data points
(DI
S
(SC),UI
S
(SC)) have been calculated using the data points listed in Table 4.3.
Table 4.3. Range compliances of SC.
F
S
S Case
VL
S
L
S
M
S
H
S
VH
S
CO
2
1 0 0.25 0.783 0.05 0
footprint 2 0 0 0.625 0.375 0
water 1 0 0.25 0.75 0 0
usage 2 0 0.178 0.851 0.08 0
NO
X
1 0.104 0.938 0.031 0 0
emission 2 0.048 0.866 0.143 0 0
SO
X
1 0 0.625 0.375 0 0
emission 2 0 0.625 0.375 0 0
1.5
1.25
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1 1.25 1.5
DI
S
(SC)
Figure 4.7. Sustainability of SC.
As seen from Fig. 4.7, the undesirable impact is high and the desirable impact is
low for SC for all most of the sustainability parameters. For some parameters, the
scenario is the opposite one. Therefore, SC is perhaps a not-so-good material from
the view point of sustainability and one should avoid using this material for
manufacturing grinding wheel unless there are other problems.
??
U
I
S
(
S
C
)
Table 4.4 shows the range compliances of SC, R(L
S
(BN),F
S
). Figure 4.8 shows
the position of BN in the two-dimensional decision space, wherein the data points
(DI
S
(BN),UI
S
(BN)) have been calculated using the data points listed in Table 4.4.
Table 4.4. Range compliances of BN.
F
S
S Case
VL
S
L
S
M
S
H
S
VH
S
CO
2
1 0 0.25 0.789 0.06 0
footprint 2 0 0 0.604 0.396 0
water 1 0 0.208 0.791 0 0
usage 2 0 0.134 0.883 0.08 0
NO
X
1 0 0.186 0.84 0.046 0
emission 2 0 0.036 0.794 0.232 0
SO
X
1 0 0.042 0.535 0.559 0.05
emission 2 0 0 0.432 0.635 0.167
1.5
1.25
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1 1.25 1.5
DI
S
(BN)
Figure 4.8. Sustainability of BN.
As seen from Fig. 4.8, the undesirable impact is very high and the desirable
impact is very low for BN for all sustainability parameters. Therefore, BN is a not-
so-good material from the view point of sustainability and one should avoid using
this material for manufacturing grinding wheel unless there are other problems.
Table 4.5 shows the range compliances of BC, R(L
S
(BC),F
S
). Figure 4.9 shows
the position of BC in the two-dimensional decision space, wherein the data points
??
U
I
S
(
B
N
)
(DI
S
(BC),UI
S
(BC)) have been calculated using the data points listed in Table 4.5.
Table 4.5. Range compliances of BC.
F
S
S Case
VL
S
L
S
M
S
H
S
VH
S
CO
2
1 0 0 0.533 0.467 0
footprint 2 0 0 0.104 0.739 0.208
water 1 0.027 0.535 0.5 0.537 0.277
usage 2 0 0.491 0.504 0.5 0.381
NO
X
1 0 0 0.468 0.531 0
emission 2 0 0 0.25 0.766 0.143
SO
X
1 0 0 0 0.375 0.444
emission 2 0 0 0 0.206 0.697
1.5
1.25
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1 1.25 1.5
DI
S
(BC)
Figure 4.9. Sustainability of BC.
For BC, all sustainability parameters, except water usage, provide zero desirable
impact, as seen from Fig. 4.9. The undesirable impact for all parameters are also very
high. Therefore, similar to BN, BC is also an less preferable material to manufacture a
grinding wheel. Thus, BC should also be avoided for manufacturing a grinding wheel
unless there are other problems.
Table 4.6 shows the range compliances of ZN, R(L
S
(ZN),F
S
). Figure 4.10 shows
the position of ZN in the two-dimensional decision space, wherein the data points
(DI
S
(ZN),UI
S
(ZN)) have been calculated using the data points listed in Table 4.6. The
??
U
I
S
(
B
C
)
sustainability scenario of ZN is similar to that of SC (compare Fig. 4.10 with Fig.
4.7). Therefore, ZN is perhaps a not-so-good material from the view point of
sustainability and one should avoid using this material for manufacturing grinding
wheel unless there are other problems.
Table 4.5. Range compliances of ZN.
F
S
S Case
VL
S
L
S
M
S
H
S
VH
S
CO
2
1 0 0.7 0.3 0 0
footprint 2 0 0.25 0.75 0 0
water 1 0.027 0.919 0.083 0 0
usage 2 0 0.848 0.152 0 0
NO
X
1 0 0.75 0.25 0 0
emission 2 0 0.5 0.5 0 0
SO
X
1 0 0.27 0.756 0.041 0
emission 2 0 0.182 0.828 0.159 0
1.5
1.25
1
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1 1.25 1.5
DI
S
(ZN)
Figure 4.10. Sustainability of ZN.
Based on the findings described in the above, the following decision can be
made, as shown in Table 4.7. As listed in Table 4.7, Alumina based hard materials are
highly sustainable materials for manufacturing abrasive grains of a grinding wheel.
Boron Nitride/Carbide based hard materials are less sustainable materials. Whereas,
Zirconia/Silicon Carbide based materials are moderately sustainable materials for
??
U
I
S
(
Z
N
)
manufacturing abrasive grains of a grinding wheel. One may use this finding while
developing more sustainable material removal tools (products) for precision
engineering.
Table 4.7. Sustainability assessment of selected hard materials for abrasive grains of
grinding wheel.
Categories Materials
Highly sustainable materials Alumina based hard materials
Moderately sustainable materials Zirconia/Silicon Carbide based hard materials
Less sustainable material Boron Nitride/Carbide based hard materials
??
??
Chapter 5
Creativity Assessment
This chapter deals with the assessment of creativity in key solutions
determination process in the conceptual phase of product development. This chapter is
based on the work of Ullah et al. 2012. For the sake of better understanding, recall
Fig. 1.1 as repeated in Fig. 5.1.
External Customers
(Real Customers)
Customer needs
Use
(Satisfaction)
Internal Customers
(Product Development Team)
Creativity
Conceptual Phase
(Key Solutions)
Detailed Design
Disposal
(Recycle, Downcycle, Landfill)
Manufacturing
Sustainability
Primary Materials
Production
Fig. 5.1. A product development scenario (Fig. 1.1 repeated).
In the conceptual phase of product development the internal customers (product
development team members) need to be creative so that a great deal of useful The
internal customers need to be creative to suggest many potential key solutions for
satisfying the needs of external customers (real customers). Therefore, the following
question arise in the conceptual phase of product development:
How to differentiate a creative key solution from a non-creative key solution?
??
Creativity is a complex and multifaceted phenomenon (Puccio et al. 2010). In
the industry in particular, lateral thinking (de Bono 1970) has been practiced to be
creative. In addition, TRIZ (theory of innovative problem solving) (Altshuller 2001)
has also been practiced to be creative (Puccio et al. 2010). However, to describe
reasoning and processes of human creativity in product development, mappings of
objects from one domain (or space) to another have been found effective. For
example, consider the mappings i) among Functions (F), Behaviors (B), and
Structures (S) introduced by Gero (Gero, 2000), ii) between Functional
Requirements (FR) and Design Parameters (DP) introduced by Suh (Suh, 1998), and
iii) between Concept (C) and Knowledge (K) introduced by Hatchuel and Weil
(Hatchuel and Weil 2003, 2009). In particular, C-K theory (Hatchuel and Weil 2003,
2009) provides an clear definition of creative concept-- a creative concept is
undecided entity with respect to the existing knowledge at the point of time when it
(the concept) is conceived. Therefore, creation of new knowledge is associated with
acceptance/rejection of a creative concept. Ullah et al. 2012 have shown that the
processes involved in adopting a creative concept may not necessarily be an outcome
of such logical processes as deduction, induction, and abduction (Yoshikawa 1981,
Tomiyama et al. 2009, Ullah 2008, Zeng and Cheng 1991, Kazakci et al. 2005). Two
different kinds of motivation called epistemic challenge and compelling reason are
involved in adopting a creative concept and if one uses epistemic information content
(CE,RE), as defined in Chapter 2 and used in Chapter 3, then one can easily
differentiate a creative key solution from a non-creative key solution in the
conceptual phase of product development. Based on this contemplation, this chapter is
written. The remainder of this chapter is organized as follows: Section 5.1
describes the main elements of C-K theory. Section 5.2 explains the epistemic
information contents for differentiating a creative key solution from a non-creative
key solution. Section 5.3 describes the results and discusses the implication of the
findings.
5.1. C-K theory
This section describes the main elements of C-K theory (Hatchuel and Weil
2003, 2009, Braha and Reich 2003, Kazakci et al. 2005, Ullah et al. 2012).
??
A schematic illustration of C-K theory is shown in Fig. 5.2. As seen from Fig.
5.2, there are two interdependent domains called Concept Domain and Knowledge
Domain in C-K theory.
Figure 5.2. An illustration of C-K theory (Ullah et al. 2012).
In addition, there are mappings between C and K, i.e., C÷K, K÷C, C÷C, and
K÷K. This mapping is somewhat different compared to those in other theories. For
example, in Axiomatic Design (Suh, 1998) the mapping is allowed in a hierarchical
manner: FR÷DP÷FR (new)÷DP (new). The mapping FR to FR or DP to DP is not
allowed in Axiomatic Design. However, one of the most remarkable features of C-K
mapping is that it is able to deal with a creative concept—a concept that is undecided
??
with respect to the existing knowledge at the point of time when it (the concept) is
conceived. If such an undecided concept is pursued further, new knowledge might
evolve in favor of the concept. As a result, both the knowledge evolved and the
concept conceived become a part K Domain and C Domain, respectively. Thus, C-K
mapping expands enriching both domains by the addition of undecided concepts and
co-creation of new knowledge.
5.2. Differentiating creative and non-creative concepts
Let C1 be an existing concept (ordinary key solution) and C2 be an creative
concept (creative key solution). Let K1 be the knowledge of suitableness of C1 and
K2 be the knowledge of performance of C1. In addition, let K3 be the knowledge of
suitableness of C2 and K4 be the knowledge of performance of C2.
One considers C2 because C1 is perhaps not suitable for the perceived need.
This means that C1 should be replaced by C2 for the better fulfillment of the
perceived need. This is called compelling reason. Thus, compelling reason acts as
one of the motivations behind perusing C2 instead of C1 for a given need. One the
other hand, at the beginning (when C2 is being conceived), K4 is empty (K4 = {?}),
i.e., there is a lack of knowledge regarding the performance of C2. The performance
of C2 is somewhat unknown when C2 is being conceived. This is called epistemic
challenge. Thus, a challenge of seeking new knowledge emerges. Overcoming this
challenge acts as the other motivation for pursing C2 instead of C1.
The motivations, compelling reason and epistemic challenge, can quantitatively
be measured by the Certainty Entropy and Requirement Entropy (CE,RE). To do this,
consider a set of propositions for C1 and C2.
First, consider the propositions regarding C1 (an ordinary or existing concept).
The propositions regarding C1 (P11,...,P14) and their truth values are shown in Table
5.1. First, the linguistic truth values defined in Chapter 2 (i.e., five fuzzy numbers
(mostly false (mf), perhaps false (pf), not sure (ns), perhaps true (pt), and mostly true
(mt))) are used to determine the TV of the proposition P11,...,P14. The expected
values (listed in Table 2.2) of the linguistic TV are used as the numerical TV for
calculating CE and RE. Note that the TVs of P11,...,P14 underlie the knowledge K1
and K2. Based on the settings shown in Table 5.1, the two-dimensional information
??
content (i.e., (CE,RE)) of C1 is calculated using the functions described in Chapter 2.
The results are shown in Fig. 5.3. As seen from Fig. 5.3, the epistemic challenge
exhibit low information content whereas compelling reason exhibits high information
content. They are placed opposite to each other. High information content of
compelling reason implies that it is not serving as a compelling reason as such. Low
information content of epistemic challenge implies that it is not a challenge as such.
Table 5.1. State of ordinary concept (C1).
Propositions
Linguistic
TV
Numerical
TV
Requirement
(P
R
)
TV
P11: Cerceisuitanble for the
1 is
mostly
0.1
C1 should be
p ved eed false (mf) suitable for 0.1
C1 is not suitable for
P12: the perceived need
P13: C1 performs well
perhaps
true (pt)
mostly
true (mt)
0.733
0.9
perceived need
C1 should
0.9
P14: C1 ldoes not perform mossetl(ymf) perform well
0.1
wel
1.2
1
0.8
0.6
0.4
0.2
0
fal
Compelling
reason
(suitableness)
Epistemic
challenge
(performance)
0 0.2 0.4 0.6 0.8 1 1.2
Certainty Entropy
Figure 5.3. Information content of C1.
??
R
e
q
u
i
r
e
m
e
n
t
E
n
t
r
o
p
y
Similar to C1, consider the propositions regarding C2 (a creative concept). The
propositions regarding C2 (P21,...,P24) and their truth values are shown in Table 5.2
and the information content is shown in Fig. 5.4. Note the opposite positions of
epistemic challenge and compelling reason in Fig. 5.4. This time, the epistemic
challenge has a very high information content (i.e., it is really a challenge), whereas
compelling reason has a low information content (i.e., it is indeed a compelling
reason).
Table 5.2. State of creative concept (C2).
Propositions
Linguistic
TV
Numerical
TV
Requirement
(P
R
)
TV
P21: Cerceisuitanbeedfor the
p 2 is ved le perhaps
true (pt)
0.733 C2 should be
suitable for
0.733
C2 is not suitable for
P22: the perceived need
P23: C2 performs well
perhaps
false (pf)
not sure
(ns)
0.267
0.5
perceived need
C2 should
0.5
P24: C2 ldoes not perform not)sure perform well
0.5
wel
1.2
1
0.8
0.6
0.4
0.2
0
(ns
Epistemic
challenge
(performance)
Compelling
reason
(suitableness)
0 0.2 0.4 0.6 0.8 1 1.2
Certainty Entropy
Figure 5.4. Information content of C2.
The overall information content of C2 is high compared to that of C1. Needless
??
R
e
q
u
i
r
e
m
e
n
t
E
n
t
r
o
p
y
to say that the overall information content means here the value of coherency
measure (ì) as defined in Chapter 2. This means a creative concept possesses high
epistemic information content while being conceived. This information content
however should reduce when new knowledge is available, i.e., K4 transforms to K?4
so that the propositions like P23 and P24 exhibit truth value similar to that of P21
and P22.
Lack of knowledge Rejection of concept
adopted
Motivation
Concept
conceived
Continuation of concept
ì
Gain of knowledge
adopted
Concept adopted
Pursing a creative concept
Figure 5.5. The states of a creative concept.
Figure 5.5 illustrates the states of a creative concept in terms of overall
information content ì. As seen from Fig. 5.5, the Information Axiom (minimize the
information content of design (Suh 1998)) does not hold as such for creative concept.
Sometimes the information content should be minimized, sometimes it should be
maximized. In addition, due to the lack of knowledge the information content of a
creative concept jumps to its peak. At the same time, if motivating factors called
compelling reason and epistemic challenge prevail, then a creative concept is
??
conceived. When a substantial amount of knowledge becomes available, the
information content of the conceived concept should go down significantly. In this
case, the conceived concept becomes a part of C-K mapping, i.e., the concept is
adopted as a key solution. Otherwise, the creative concept should be abundant and a
new course of direction should be explored.
5.3. Results and discussions
This section describes the results of how a creative concept (an engine for Mars
exploration) has been differentiated from an ordinary engine (an existing fossil-fuel
based engine) using the method described in the previous section. At the beginning
the C-K mapping takes the form of the map shown in Fig. 5.6.
Figure 5.6. C-K mapping of a creative key solution for Mars exploration.
As seen from Fig. 5.6, two solutions C1 = Fossil-fuel based propulsion engine
??
and C2 = Mg-CO
2
based propulsion engine have been considered. C1 is suitable for
earth whereas C2 is suitable for Mars. The performance of C1 is known whereas the
performance of C2 is quite unknown. This implies the propositions and their truth
values as shown in Table 5.3. The information contents of C1 (fossil-fuel based
propulsion engine) and C2 (Mg-CO
2
based propulsion engine) can be expressed by
the information contents shown in Fig. 5.3 an Fig. 5.4, respectively because of the
settings in Table 5.1 is similar to that of Table 5.3 for C1 and in Table 5.2 is similar to
that of Table 5.3 for C2. Therefore, Mg-CO
2
based propulsion engine is a creative
concept and it can be pursued further.
Table 5.3. The states of ordinary and creative concepts.
C1 = Fossil-fuel based propulsion engine
Propositions
Linguistic
TV
Numerical
TV
Requirement
(P
R
)
TV
P11: C1arsis psluirtatblen
for a
M ex o io
mostly
false (mf)
0.1
C1 should be
suitable for
0.1
P12: C1ariss explsuitaone for
M not orati bl perhaps
true (pt)
mostly
0.733 Mars
exploration
P13: C1 performs well
true (mt)
0.9
C1 should
0.9
P14: C1 ldoes not perform mossetl(ymf) perform well
0.1
wel fal
C2 = Mg-CO
2
based propulsion engine
Propositions
Linguistic
TV
Numerical
TV
Requirement
(P
R
)
TV
P11: C2arsisxpluitaatble
for s
M e or ion
perhaps
true (pt)
0.733
C2 should be
suitable for
0.733
P12: C2ariss explsuitaone for
M not orati bl perhaps
false (pf)
not sure
0.267 Mars
exploration
P13: C2 performs well
(ns)
0.5
C2 should
0.5
P14: C3 ldoes not perform not)sure perform well
0.5
wel (ns
If C2 (Mg-CO2 based propulsion engine) is pursued further new knowledge can
be gained (K4 transforms to K?4). Figure 5.7 shows the state of K?4.
??
Figure 5.7. When K4 transforms to K?4.
As seen from Fig. 5.7, for Mars exploration, a propulsion engine is needed that
should use in-situ fuel and oxidizer. Given the fact that Mars atmosphere consists of
more than 95% CO
2
, it (CO
2
) can be used as an oxidizer, even if it is an unusual
choice. This necessitates a particular type of fuel either metals (Be, Mg, Al, Li, Ca,
??
etc.) or their hydrates (e.g., BeH
2
, MgH
2
, etc.). The fundamental studies conducted
by Shafirovich et al. 1992, 1993 have revealed that the fuels, namely, Mg, Al, Be,
BeH2 are probably the most useful fuels when CO
2
acts as the oxidizer. It has also
been found that Mg-CO
2
combination produces almost the same amount of Specific
Impulse (an important performance measure of propulsion devices) compared to that
of other combinations (i.e., Al-CO
2
, Be-CO
2
, and BeH
2
-CO
2
). In terms of other
important performance measures (i.e., combustion characteristics, such as toxicity,
ignitability, combustion rate, slag formation, etc.) Mg-CO
2
combination produces
relatively better result.
Table 5.4. States of C2 and C2 based on C-K mapping in Fig. 5.7.
Requirement
P1
Propositions
C2 is acceptable in terms of Specific pt
Impulse
Truth Values
0.733
An
(P
R
)
engine
C2 is not acceptable in terms of Specific mf should be
P2 Impulse 0.1 acceptable in
C3 is acceptable in terms of Specific mt 0.9 terms of
P3 Impulse Specific
Impulse
C3 is not acceptable in terms of Specific
P4 Impulse mf 0.1
P5
C2 is acceptable in terms of toxicity, mt
ignitability, combustion rate, slag
formation, etc.
0.9
An engine
C2 is not acceptable in terms of toxicity, should be
ignitability, combustion rate, slag acceptable in
P6
formation, etc.
mf 0.1
terms
toxicity,
of
P7 C3 is acceptable in terms of toxicity, mf
ignitability, combustion rate, slag
formation, etc.
0.1
ignitability,
combustion
rate,
sl a g
formation, etc.
C3 is not acceptable in terms of toxicity,
P8 ignmabiility,etc. combustion rate, slag pt ita
for t on,
0.733
C2 = Mg-CO
2
Propulsion Engine, C3 = Y-CO
2
Propulsion Engine, Y e{Be, BeH
2
,
Al}
??
Given the C-K mapping in Fig. 5.7, is it possible to show that the information
content of concept C2 (Mg-CO2-based propulsion engine) has come down
significantly? An answer to this question is needed to make sure the effectiveness of
the transformation of knowledge from K4 to K?4. Otherwise, new knowledge (K?4)
does not add any value to key solution determination process.
To answer the question, as set of propositions P1,.,P8 and two alternatives C2
(same as before) and C3 (=Y-CO
2
-based propulsion engine, Y is either Be or BeH
2
or
Al) are considered. The propositions and their truth values are listed in Table 5.4.
Needless to say that the truth values of the propositions listed in Table 3 reflect the
facts in Fig. 5.7.
1
0.8
0.6
0.4
0.2
0
(a) C2
1
0.8
0.6
0.4
0.2
0
(b) C3
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Certainty Entropy Certainty Entropy
Figure 5.8. Information content of C2 and C3 based the settings in Table 5.4.
The information content in terms of Certainty and Requirement Entropies
(CE,RE) are determined by using the same methods used in the previous section. The
results are shown in Fig. 5.8. C2 has information contents (0.37,0) and (0.2,0) for
{P1,P2} and {P5,P6}, respectively. The overall information content of C2 is now
equal to 0.57. On the other hand, C3 has information content (0.2,0) and (0.37,1) for
{P3,P4} and {P7,P8}, respectively. The overall information content of C3 is equal to
1.74. Thus, C2 is preferred over C3, as the key solution to develop a propulsion
engine for Mars exploration, the decisionmaking now underlies "minimization of
information content," i.e., the process holds the Information Axiom (Suh 1998).
??
R
e
q
u
i
r
e
m
e
n
t
E
n
t
r
o
p
y
R
e
q
u
i
r
e
m
e
n
t
E
n
t
r
o
p
y
Therefore, the results described in this section hold the scenario described in Fig.
5.5. In synopsis, creativity is first controlled by the maximization of information
content in presence of such motivating factors as compelling reason and epistemic
challenge and then by the minimization of information content in presence of new
knowledge.
??
??
Chapter 6
Concluding Remarks
Making decisions, i.e., identifying a key solution (or a set of key solution), in
conceptual phase of product development is not only critical but also difficult. It is
critical in a sense that around 80% cost of a product is decided by the key solution
determination process in the conceptual phase of product development and it cannot
be rectified by making adjustments in the downstream of a product lifecycle. It is
difficult in a sense that in conceptual phase of product development, the knowledge
is very limited and there is an abundance of choice. To shed some light on this issue
(decisionmaking in conceptual phase of product development) this thesis poses and
answers the following questions:
How to differentiate a creative key solution from a non-creative key solution?
What is the appropriate customer need model?
How to deal with the unknown customer needs?
How to classify the key solutions based on customer responses?
How to deal with the sustainability of materials (used in the product) in key
solution determination process?
Nevertheless, the following remarks can be made on the findings:
On the customer needs assessment:
1.
2.
3.
One of the ways to identify a key solution to develop a product is to take
opinions of customers regarding a set of key solutions.
To deal with the intrinsic complexity of customer responses, logical
aggregation of customer opinions is a better choice compared to frequency
based analysis. This faculty of thought is demonstrated to be true by
logically aggregating the field data of customer needs collected from
Bangladesh on small passenger vehicles using Kano model.
It has been found that a product feature needs to be classified either into a
??
must be included feature, or into a should be included feature, and or into a
could be included feature. The link among these classifiers and Kano
evaluations (Must-be, Attractive, One-Dimensional, Indifferent, Reverse,
and Questionable) has been established.
4.
5.
The multi-valued logic plays an important role in the customer needs
assessment. In particular, a two-dimensional information content (in
epistemic sense) scheme has been found effective in logically computing
the degree of customer satisfaction of a given product feature in terms of
must be included, should be included, and could be included.
To increase the degree of satisfaction of vehicle users in Bangladesh, it is
important to develop SUV- and Van-type passenger vehicles replacing
some of the Sedan-type vehicles.
On the sustainability assessment:
1.
2.
3.
4.
Sustainability of a product largely depends on the materials used to
manufacturing it. Therefore, the material used to manufacture the product
become one of the key solutions.
To deal with the imprecision associated with the material related
sustainability parameters in the conceptual phase of product development,
an entity called range compliance has been found effective. The compliance
of an sustainability parameter given by a numerical range is determined by
calculating its compliance with five fuzzy numbers of the parameters
labeled very low, low, moderate, high, and very high.
As an example, the imprecision associated with four sustainability
parameters namely, CO
2
footprint, NO
X
emission, SO
X
emission and water
usage (i.e., resource depletion) of five classes of hard materials (the
materials used to produce abrasive grains of grinding wheel or other
material removal tools) based on Alumina, Zirconia, Silicon Carbide, Boron
Nitride, and Boron Carbide have been quantified by using the range
compliance.
The sustainability parameter complying more with very low or low less
negative impact on the sustainability, whereas the sustainability parameter
??
5.
complying more with moderate, high, or very high has high negative impact
on sustainability.
It is found that Alumina based hard materials have low negative impact
followed by Zirconia and Silicon Carbide based hard materials. Boron
Nitride/Carbide based materials have the highest negative impact.
On the creativity assessment:
1.
2.
3.
4.
5.
To identify a useful key solution in conceptual phase of product
development, the product development team members needs to be creative.
To differentiate a creative concept from a non-creative concept, Concept-
Knowledge mapping as prescribed in C-K theory can be employed.
Creative concept means a concept which is undecided when it is being
conceived.
Conceiving a creative concept is rather a motivation driven process.
Information content of a creative concept is high compared to that of a
non-creative concept. The information content means here the
6.
7.
8.
two-dimensional information content in epistemic sense.
When a creative concept is pursued and new knowledge becomes available,
the information content should go down significantly. Otherwise, the new
knowledge does not add any value to product development process.
A non-creative key solution does not exhibit the abovementioned behavior
of information content.
The effectiveness of the abovementioned approach has been demonstrated
by calculating the information contents of two concepts Mg-CO
2
based
propulsion engine (a creative concept) and fossil-fuel base propulsion
engine (an non-creative concept). It has been found that the Mg-CO
2
based
propulsion engine exhibits high information content compared to that of
fossil-fuel base propulsion engine for Mars exploration. The information
content of Mg-CO
2
based propulsion engine have gone down significantly
under the presence of new knowledge.
??
??
??
Bibliography
1.
2.
3.
4.
5.
6.
7.
8.
9.
—.CES Selector¹ (Version 5.1.0). 2009. Granta Design Limited, UK,
http://www.grantadesign.com
Altshuller, G. (2001). And Suddenly the Inventor Appeared: TRIZ, the
Theory of Inventive Problem Solving (Shulyak, L., Trans.). Worcester, MA:
Technical Innovation Center.
Bellman, R.E. and Zadeh, L.A. (1970). Decision-Making in a Fuzzy
Environment, Management Science, 17(4), B-141-B-164.
Berger, C., Blauth, R., Boger, D., Bolster, C., Burchill, G., DuMouchel, W.,
Pouliot, F., Richter, R., Rubinoff, A., Shen, D., Timko, M. and Walden, D.
(1993). Kano's method for understanding customer-defined quality, Center for
Quality of Management Journal, 2(4), 3-35.
Braha, D. and Reich, Y. (2003). Topological structures for modelling
engineering design processes, Research in Engineering Design, 14(4), 185-
199.
Briggs, C. and Little, P. (2008). Impact of Organizational Culture and
Personality Traits on Decision-making in Technical Organization, Systems
Engineering, 11(1), 15-26.
Chen, Z. and Ben-Arieh, D. (2006). On the fusion of multi-granularity
linguistic label sets in group decision making, Computers & Industrial
Engineering, 51(3), 526-541.
de Bono, E. (1970). Lateral Thinking: Creativity Step by Step, Harper
&Row, Publisher: NY.
Dieter, G.E. and Schmidt, L.C. (2009). Engineering Design, Fourth Edition,
McGraw-Hill: New York.
10. Donnelly, K., Olds, R., Blechinger, F., Reynolds, D. and Beckett-Furnell, Z.
(2004). ISO 14001 - effective management of sustainable design, The
Journal of Sustainable Product Design, 4(1-4), 43-54.
??
11. Dubois, D. and Prade, H. (1978). Operations on fuzzy numbers,
International Journal of Systems Science, 9(6), 613-626.
12. Fang, L., Hipel, K.W. and Kilgour, D.M. (1993). Interactive Decision
Making: The graph Model for Conflict Resolution, Wiley & Sons: New
York.
13. Fiksel, J. (2009). Design for Environment: A Guide to Sustainable Product
Development, Second Edition, McGraw-Hill: New York.
14. Foote, J.P. and Litchford, R.J. (2007). Powdered Magnesium—Carbon
Dioxide Rocket Combustion Technology for In Situ Mar Propulsion, NASA
Technical Publication, NASA/TP-2007-215077, available from
https://www2.sti.nasa.gov.
15. Fukushige, S. Yamamoto, K. and Umeda, Y. (2012). Lifecycle scenario
design for product end-of-life strategy, Journal of Remanufacturing, 2012,
2:1, doi:10.1186/2210-4690-2-1.
16. Gero, J.S. (2000). Computational Models of Innovative and Creative Design
Processes, Technological Forecasting and Social Change, 64(2-3), 183-196.
17. Hatchuel, A. and Weil, B. (2009). C-K design theory: an advanced
formulation, Research in Engineering Design, 19(4), 181-192.
18. Hatchuel. A. and Weil., B. (2003). A new approach of innovative design: an
introduction to C-K theory, Proceedings of the 14th International
Conference on Engineering Design Research for Practice (ICED'03), 19-21
August 2003, Stockholm, Sweden, pp 109-124.
19. Herrera, F. and Herrera-Viedma, E. (2000). Linguistic decision analysis:
steps for solving decision problems under linguistic information, Fuzzy Sets
and Systems, 115(1), 67-82.
20. Higuchi, Y., Wada, N., Nakakubo, T. and Tokai, A. (2012). Scenario
Analysis on the Impact of Diffusion of Next Generation Vehicles on Material
Consumption and GHG Emissions. In: Design for Innovative Value Towards
a Sustainable Society (Editors: Matsumoto, M., Umeda, Y., Masui, K. and
Fukushige, S.), pp. 599-604, Springer: The Netherland.
21. Hutton, R.J.B. and Klein, G. (1999). Expert Decision Making, Systems
Engineering, 2(1), 32-45.
??
22. Inohara, T. and Hipel, K.W. (2008). Coalition Analysis in the Graph Model
for Conflict Resolution, Systems Engineering, 11(4), 343-359.
23. Kahn, K.B. (Ed.) (2004). The PDMA Handbook of New Product
Development (Second Edition), Wiley: New York.
24. Kano, N., Seraku, N., Takahashi, F. and Tsuji, S. (1984). Attractive quality
and must-be quality, Hinshitsu, 14(2), 39-48. (In Japanese).
25. Kazakci, A.O. and Tsoukias, A. (2005). Extending the C-K design theory: A
theoretical background for personal design assistants, Journal of
Engineering Design, 16(4), 399-411.
26. Keller, L.R., Kirkwood, C.W. and Jones, N.S. (2008). Assessing Stakeholder
Evaluation Concerns: An Application to the Central Arizona Water
Resources Systems, Systems Engineering, 13(1), 58-71.
27. Khozaimy, O., Al-Dhaheri, A. and Sharif Ullah, A.M.M. (2011). A
decision-making approach using point-cloud-based granular information,
Applied Soft Computing Journal, 11(2), 2576-2586.
28. Klien, G.A. (2008). Naturalistic Decision Making, Human Factors, 50(3),
456-460.
29. Kobayashi, H. (2006). A systematic approach to eco-innovative product
design based on life cycle planning, Advanced Engineering Informatics,
20(2), 113-125.
30. Kujawski, E. (2005). A Reference-Dependent Regret Model for
Deterministic Tradeoff Studies, Systems Engineering, 8(2), 119-137.
31. Noor-E-Alam, M., Lipi, T.F., Hasin, M.A.A. and Sharif Ullah, A.M.M.
(2011). Algorithms for fuzzy multi expert multi criteria decision making
(ME-MCDM), Knowledge-Based Systems, 24(3), 367-377.
32. O'Hagan, A. and Oakley, J.E. (2004). Probability is perfect, but we can't
elicit it perfectly, Reliability Engineering and System Safety, 85, 239-248.
33. Puccio, G.J., Cabra, J.F., Fox, J.M., and Cahen, H. (2010). Creativity on
demand: Historical approaches and future trends, Artificial Intelligence for
Engineering Design, Analysis and Manufacturing, 24(2), 153-159.
34. Rashid, M.M., Tamaki, J., Ullah, A.M.M.S. and Kubo, A. (2010). A Virtual
Customer Needs System for Product Development, Proceedings of the 2010
??
Annual Meeting of Japan Society for Precision Engineering, September 04,
2010, Sapporo, Japan, Pages 53-54.
35. Rashid, M.M., Ullah, A.M.M.S., Sarker, M.A.R., Tamaki, J. and Kubo, A.
(2012). Logical Aggregation of Customer Needs Assessment, Proceedings
of the Fifth International Symposium on Computational Intelligence and
Industrial Applications (ISCIIA2012), Sapporo, Japan, August 20-26, 2012.
36. Rashid, M.M., Ullah, A.M.M.S., Tamaki, J., and Kubo, A. (2011).
Evaluation of Hard Materials using Eco-Attribute, Advanced Materials
Research, 325, 693-698
37. Rasmussen, J. (1993). Deciding and doing: Decision making in natural
contexts. In: Decision Making in Action, Klein, G., Orasallu, J., Calderwood,
R., and Zsambok, C.E. (Eds), Ablex, Norwood, New Jersey, pages 158 - 171.
38. Saaty, T.L. (1980). The Analytic Hierarchy Process, McGraw-Hill: New
York.
39. Saaty, T.L. (1990). Multicriteria Decision Making: The Analytic Hierarchy
Process, RWS Publications, Pittsburgh: Pennsylvania.
40. Shafirovich, E.Y., and Goldshleger, U.I. (1992). Combustion of Magnesium
Particles in CO2/CO Mixtures, Combustion Science and Technology, 84(1),
33-43.
41. Shafirovich, E.Y., Shiryaev, A.A., and Goldshleger, U.I. (1993). Magnesium
and Carbon Dioxide: A Rocket Propellant for Mars Missions, Journal of
Propulsion and Power, 9(2), 197-203.
42. Shamsuzzaman, M., Ullah, A.M.M.S. and Bohez, E.L.J. (2003). Applying
linguistic criteria in FMS selection: fuzzy-set-AHP approach, Integrated
Manufacturing Systems, 14(3), 247-254.
43. Smith, E.D., Son, Y.J., Palmarini-Piattelli, M. and Bahill, T. (2005).
Ameliorating Mental Mistakes in Tradeoff Studies, Systems Engineering,
10(3), 222-240.
44. Suh, N.P. (1990). The principles of design. Oxford University Press: New
York.
??
45. Suh, N.P. (1998). Axiomatic Design Theory for Systems, Research in
Engineering Design, 10(4), 189-209.
46. Tomiyama, T., Gu, P. Jin, Y., Lutters, D., Kind, C., and Kimura, F. (2009).
Design methodologies: Industrial and educational applications, CIRP
Annals-Manufacturing Technology, 58(2), 543-565.
47. Ullah, A.M.M.S. (2005a). A Fuzzy Decision Model for Conceptual Design,
Systems Engineering, 8(4), 296-308.
48. Ullah, A.M.M.S. (2005b). Handling design perceptions: an axiomatic design
perspective, Research in Engineering Design, 16(3), 109-117.
49. Ullah, A.M.M.S. (2008). Logical interaction between domain knowledge
and human cognition in design, International Journal of Manufacturing
Technology and Management, 14(1-2), 215-227.
50. Ullah, A.M.M.S. and Harib, K.H. (2006). Zadehian Paradigms for
Knowledge Extraction in Intelligent Manufacturing. In: Manufacturing the
Future: Concepts-Technologies-Visions. Editors: Kordic, V., Lazinica, A.
and Merdan., M. Pro Literatur Verlag Robert Mayer-Scholz, Mammendorf,
Germany, pages 291-308.
51. Ullah, A.M.M.S. and Harib, K.H. (2008). An Intelligent Method for
Selecting Optimal Materials and its Application, Advanced Engineering
Informatics, 22(4), 473-483.
52. Ullah, A.M.M.S. and Tamaki, J. (2011). Analysis of Kano-Model-Based
Customer Needs for Product Development, Systems Engineering, 14(2),
154-172.
53. Ullah, A.M.M.S., Harib, K.H., Al-Awar, A. (2007). Minimizing Information
Content of a Design using Compliance Analysis, SAE Technical Paper
2007-01-1209, SAE International: Pennsylvania.
54. Ullah, A.M.M.S., Kitajima, K., Akamatsu, T., Furuno, M., Tamaki, J., and
Kubo, A. On Some Eco-indicators of Cutting Tools, Proceedings of the
ASME 2011 International Manufacturing Science and Engineering
Conference, June 13-17, 2011, Corvallis, Oregon, USA.
??
55. Ullah, A.M.M.S., Rashid, M.M. and Tamaki, J. (2012). On Some Unique
Features of C-K Theory of Design, CIRP Journal of Manufacturing Science
and Technology, 5(1), 55-66.
56. Ullman, D.G. (2006). Making Robust Decisions: Decision Management For
Technical, Business, & Service Teams, Trafford Publishing: British
Colombia.
57. Ullman, D.G. (2009). Mechanical Design Process, Fourth Edition,
McGraw-Hill: New York.
58. Ulrich, K.T. and Eppinger, S.D. (2004). Product Design and Development,
Third Edition, McGraw-Hill: New York.
59. Umeda, Y. Nishiyama, T., Yamasaki, Y., Kishita, Y. and Fukushige, S.
(2009). Proposal of sustainable society scenario simulator, CIRP Journal of
Manufacturing Science and Technology, 1(4), 272-278.
60. Walley, P. (1991). Statistical Reasoning with Imprecise Probabilities.
Chapman and Hall: London.
61. Wood, W.H.III. and Agogino, A.M. (1996). Case-based conceptual design
information server for concurrent engineering, Computer-Aided Design,
28(5), 361-369.
62. Xu, Q., Jiao, R.J., Yang, X., Helander, M., Khalid, H.M. and Opperud, A.
(2009). An analytical Kano model for customer need analysis, Design
Studies, 30(1), 87-110.
63. Yager, R.R. (1978). Fuzzy decision making including unequal objectives,
Fuzzy Sets and Systems, 1(2), 87-95.
64. Yang, K. (2008). Voice of the Customer: Capture and Analysis.
McGraw-Hill: New York.
65. Yoshikawa, H. (1981). General Theory of Design Process, Journal of the
Japan Society of Precision Engineering, 47(4), 405-410. (In Japanese)
66. Zadeh, L.A. (1965). Fuzzy sets, Information and Control, 8, 338-353.
67. Zadeh, L.A. (1975). Fuzzy logic and approximate reasoning, Synthese,
30(3-4), 407-428.
??
68. Zadeh, L.A. (1997). Toward a theory of fuzzy information granulation and
its centrality in human reasoning and fuzzy logic, Fuzzy Sets and Systems,
90(2), 111-127.
69. Zadeh, L.A. (2005). From imprecise to granular probabilities, Fuzzy Sets
and Systems, 154(3), 370-374.
70. Zeng, Y. and Cheng, G.D. (1991). On the logic of design, Design Studies,
12(3), 137-141.
??
Appendix A: Customer Needs Data
Collection
Customer needs data has been collected on the some features of small passenger
vehicles. The data collection period was January 2012 to April 2012. The data on the
38 features has been collected using the Kano model. The features are listed in Table
3.3. In addition to the data on 38 features, the physiographic and demographic data of
the respondents has also been collected, as follows:
Table A.1 shows the demographic questions that have been asked to the
respondents in Bangladesh. Needless to say that demographic questions means the
questions related the profession, income level, gender, and alike of an individual. The
questions related to income of an individual are from the context of Bangladesh.
Table A.1. Demographic questions and answers
Statement
private service holder
government service holder
housewife
businessperson
Choose one Frequency
14
12 7
11
I am a/n
My income is
I am a
engineer
doctor
lawyer
student
others
very high
high
moderate
low
Female
Male
11 2
12
26 5
48
60
28
82
18
The frequencies of the answers are also shown in the last column in Table A.1.
On the other hand, Table A.2 shows the psychographic questions that have been
??
asked to the respondents in Bangladesh. Needless to say that psychographic
questions means the questions related the life-style and values of an individual. A
respondents can choose multiple answers from given options. The frequencies of the
answers are also shown in the last column in Table A.2.
Table A.1. Psychographic questions and answers
Check as many as
Statements
I prefer to drive my vehicle by
myself
I prefer to hire a driver to drive
my vehicle
I use my personal vehicle for long
trips
I always use my own vehicle for
commuting to office/school
A vehicle is an essential means of
transportation for me
A vehicle is a luxurious means of
transportation for me
I prefer environmentally friendly
vehicles
you like
•
•
•
•
•
•
•
Frequency
52
39
35
45
53
21
60
??
??
??
List of Research Achievements
The following research publications have been submitted to defend the thesis.
1.
2.
3.
4.
Md. Mamunur Rashid, A.M.M. Sharif Ullah, Junichi Tamaki, and Akihiko
Kubo. (2011). Evaluation of Hard Materials using Eco-Attribute, Advanced
Materials Research, Volume 325, Pages 693-698 [Trans Tech Publications,
Switzerland] [http://dx.doi.org/10.4028/www.scientific.net/AMR.325.693]
A.M.M. Sharif Ullah, Md. Mamunur Rashid and Junichi Tamaki. (2012). On
Some Unique Features of C-K Theory of Design, CIRP Journal of
Manufacturing Science and Technology, Volume 5, Number 1, Pages 55-66.
[Elsevier, The Netherlands] [http://dx.doi.org/10.1016/j.cirpj.2011.09.001].
Md. Mamunur Rashid, A.M.M. Sharif Ullah, M.A. Rashid Sarker, Junichi
Tamaki, and Akihiko Kubo. (2012). Logical Aggregation of Customer Needs
Assessment, Proceedings of the Fifth International Symposium on
Computational Intelligence and Industrial Applications (ISCIIA2012),
Sapporo, Japan, August 20-26, 2012.
Md. Mamunur Rashid, Junichi Tamaki, A.M.M. Sharif Ullah and Akihiko
Kubo. (2010). A Virtual Customer Needs System for Product Development,
Proceedings of the 2010 Annual Meeting of Japan Society for Precision
Engineering, September 04, 2010, Sapporo, Japan, Pages 53-54.
??
??
??
??
doc_481834431.docx