A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

CHAPTER 6

QUALITY FUNCTION DEPLOYMENT AND PROCESS MODELS

1. INTRODUCTION
2. DESIGN PROCESS
3. RELATIONSHIP BETWEEN DESIGN PROCESS ATTRIBUTES
AND VARIABLES
4. OPTIMIZING THE QUALITY OF CRITICAL DESIGN ACTIVITIES
5. SUMMARY
REFERENCES
QUESTIONS
PROBLEMS

1. INTRODUCTION
Products should be designed to assure the highest possible quality and to facilitate manufacturing
and assembly in a short time and at a low cost. The engineering design process is to deliver high quality
products and systems, often subject to stringent time constraints.
Quality is an abstract term, often defined as the extent to which the customers (users) believe the
product meets their requirements and expectations. One of the techniques for improving products and
processes is the quality function deployment (QFD), known also as the house of quality (see Figure 1).
Toyota has used the QFD approach since 1977, following four years of training and preparation. Using
1977 as a base, Toyota reported a 20% reduction in a start-up cost on the launch of the new van in
1979, a 38% reduction in 1982, and a cumulative 61% reduction in 1984. During this period, the
product development cycle was reduced by one third with a corresponding improvement in quality,
largely due to the reduced number of engineering changes.
The basic idea of QFD is to transform customer requirements into specific design requirements.
The information in the various QFD matrices requires that different groups of individuals reach
consensus on the product, process, and design requirements necessary to effectively meet customer
requirements.
QFD begins by obtaining the customer requirements (CRi) with respect to the product being designed.
These requirements are not specifically related to the customer's product, but to the entire market
(Keats 1990). Next, the current or projected design features from the customer perspective are
compared with the list of technical descriptions of the product. The technical descriptions (TDj) might
be linked with other elements of the product life cycle process, including design activities (DAj). The
comparison of competitive products may be on the basis of ratings indicating how well the design
requirements are meet. The product features are then related to themselves in the roof of the diagram
using correlation symbols. As a result of this process, features with the conflicting design requirements
become apparent. This implies that some trade-off may be necessary and QFD has identified it early in
the design process. The next step is to benchmark the product as viewed by the customer with the
products of selected competitors with respect to meeting customer requirements using a scale of same,

1
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

better, or worse. Enhancements to the QFD process include adding importance measures to the
customer requirements, including target values for product design features and relating product design
features to part and mechanism characteristics. Figure 1 shows a typical QFD matrix that is the
foundation of QFD practices (Bossert 1991).

Technical descriptions
Relationship between ( TDj )
technical descriptions

Customer requirements Prioritized

( CRi ) customer needs
Importance
Competitive
analysis
Market potential

Relationship between Prioritized technical

CRi and TDj descriptions

Figure 1. The quality function deployment (QFD) matrix

The rows of the matrix in Figure 1 are the customer requirements, i.e., what the customer wants in the
product. The columns of the matrix show what the manufacturer does to ensure the quality of the
product. The right side of the matrix contains the planning information, i.e., the importance rating,
competitive analysis, target value, amount of scaling up necessary, and the sales points. The relationships
between customer requirements and design requirements are categorized in the body of the matrix. The
important goal of QFD is to turn the design requirements into the detailed design activities. A full
implementation of the QFD concept allows the customer requirements (CRi) to be cascaded down
through the technical descriptions (TDj) and functional descriptions (FDj) to design activities (DAj) as
illustrated in Figure 2.

2
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

Technical
description

requirement
Functional
Customer description
Design

description
Technical
activity

description
Functionl
Figure 2. QFD cascade

The following six steps summarize forming a QFD matrix.

Step 1. The primary customer requirements are normally expanded into secondary and tertiary
requirements using, for example, an AND/OT tree. The information usually comes from a variety of
sources such as marketing groups, dealers, sales departments, customer opinion surveys, and so on.
This step is critical and it is usually difficult, as it requires obtaining and expressing what the customer
truly wants.
Step 2. The technical descriptions must be related to the customer requirements and should be
selectively deployed throughout the manufacturing, assembly, and service process to manifest
themselves in the final product performance and customer acceptance.
Step 3. Developing a matrix describing the customer requirements and the technical descriptions is
accomplished. As they are correlated to a different degree, numbers are used to express the
significance of the relationships. A simple four-grade scale can be used to define strong, medium, weak,
and negative relationship. Of course, other scales for expressing the relationship are possible.
Step 4. Market evaluation that covers the customer-expressed importance ratings, requirements listed
and competitive products are evaluated. Market evaluations are compared with the design requirements
to determine areas of inconsistency between the customer requirements and strategy of the company.
Step 5. The target for each technical description is defined. These target values are based on the
customer importance index or strengths and weaknesses of the competing products. These values
should be measurable. The conformance with these targets is to be measured at each stage of the
product development and prove-out processes such as the evaluation of mechanical and functional
prototypes, engineering sign-offs, and units produced with production tooling just prior to production,
and on the first units at the beginning of production.
Step 6. Selection of the best solution should be deployed through the approach discussed later in this
chapter. Implementing QFD is a challenge for any company, but it requires significant effort in the early
design stages.

QFD brings several benefits to companies willing to undertake the study and training required
implementing the system:
. Product objectives based on customer requirements are not misinterpreted at the subsequent
stages.
. Particular marketing strategies or sales points do not become lost or blurred during the
translation process from marketing through planning to execution.

3
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

. Important production control points are not overlooked - everything necessary to achieve the
desired outcome is understood and in place.
. Efficiency is improved as the misinterpretation of design objectives, marketing perception, critical
control points, and the need for changes are minimized.
Integrated product and process design is a key aspect of concurrent engineering. One of the
objectives of concurrent engineering is to ensure that serious errors do not go undetected and that the
design intent is fully captured. Due to the simultaneous consideration of many factors, experts from
different functional groups need to contribute towards the design goal. In effect, teams perform the
design activities with members representing various disciplines. A design team can effectively utilize the
concurrent engineering approach only when it works cooperatively and consistently.
The design process consists of several phases ranging from the analysis of customer
requirements to manufacturing of the product. Each of these phases can in turn be decomposed into a
set of interrelated design activities. The main purpose of this decomposition is to gain control over total
duration of the design process and attain a proper utilization of resources during planning and execution
of design activities, so that the total cost is minimized (Pahl and Beitz 1988).
The quality function deployment (QFD) approach enables organizations to translate customer
requirements into relevant requirements throughout the design process. The house of quality is a tool for
recording and communicating product information to the groups that contribute to the realization of the
product (Hauser and Clausing, 1988). In the house of quality, certain design criteria are defined only,
which are important in evaluating the product. These criteria are related qualitatively to the engineering
domain and identify the relationship between the customer wants and what engineering must perform to
deliver them. The aim of this research is to use the house of quality to assess the performance of the
product development process.
The house of quality is modified to capture the relationships between design process attributes
and control variables (Locascio and Thurston 1994). These qualitative relationships are then
transformed into quantitative expressions. The expressions can either be used to examine the effect of
change of variable values on the design process attributes or they can be used as constraints to optimize
certain quality criteria in the design activity network.

2. DESIGN PROCESS
Consider the IDEF3 model of an electro-mechanical product shown in Figure 3 (Belhe and
Kusiak 1996). The first activity is to prepare system specification from the customer requirements.
Once this specification is ready, an activity to generate preliminary design begins which continuously
interacts with the activity to evaluate cost of different alternatives. This relationship is shown by the AND
logical junctions, J1 and EOR junctions J2 and J3. For the model clarity, only the fan-in logical junctions
have been incorporated. Once the preliminary design is ready, a prototype is built and then different
tests are performed on this prototype. The types of tests to be performed on this prototype depend on
the type of module to be designed. The test data is analyzed to confirm the suggested design. If the
suggested design is not acceptable, then it needs to be modified. Otherwise, the preliminary design data
is used in conjunction with cost estimates to finalize the design details. An EOR logical junction, J4,
indicates this relationship.

4
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

2 X 4 5 6
J2
J1
1 & X
7
J4
J3
3 X

1. Prepare system specifications

2. Generate preliminary design
3. Evaluate cost of different alternatives
4. Build prototype
5. Perform tests on prototype
6. Analyze test data
7. Finalize design details

Figure 3. IDEF3 design process model of an electro-mechanical product

Different types of logical relationships between design activities result in alternative paths through
the model. These alternate paths are associated with certain probabilities that may change over time.
One can not exactly predict the path followed through the model or the number of iterations in the
design process. The number of iterations depends, for example, on the quality of performing the design
tasks and quality of the results of these tasks, however, the quantitative dependence can not be
obtained.
In addition, the number of iterations taking place throughout the design process is not
deterministic. This number may largely vary from one project to the other and it may not be possible to
collect data and fit a distribution for the duration of different design phases. Also, the nature of the
design process changes over time due to new developments in technology, organizational improvements
and son on. The duration of some of the design activities may also depend upon the availability of
resources which itself may change over time. Due to such dynamic nature of the design process, each
design project is unique in terms of its time requirements. The quality of schedule depends on the
accuracy with which the activity duration and the set of successor activities is determined. The
predictability and standardization found in well planned manufacturing systems is not present to the same
degree in product development process. In design projects, uncertainty and diversity seem to be the
predominant features. Therefore, relationships between various design process attributes and variables
can not be captured quantitatively. In this paper, the house of quality is used to qualitatively express
these relationships.

3. RELATIONSHIP BETWEEN DESIGN PROCESS ATTRIBUTES AND

VARIABLES

5
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

An effort is required is to improve attributes associated with this design process. An attribute is
a function of design process variables and is normally an element of a performance measure or a simple
performance measure itself.
In the house of quality, design criteria are qualitatively related to the engineering domain,
identifying the relationships between what the customer wants and what engineering must do to provide
it. In this paper, it is proposed that a modified version of the house of quality be used to establish the
qualitative relationships between various design process attributes and the corresponding design process
variables. In this “modified” house of quality, design process attributes replace customer attributes and
design process variables replace engineering characteristics.
The design process attributes are defined so that they clearly represent the management's
objectives. Two attributes are combined when each is too specific and their total effect on the design
process can be represented by a single goal. An attribute is eliminated if it is not relevant to the goal. An
attribute is decomposed into two subattributes, when it can be better represented by two distinct goals.
It is required to redefine a design process attribute when it is possible to describe the corresponding
goal on a measurable scale.
A number of variables are involved in the design process. The level of these variables can be
directly controlled. The exact analytical relationship between these variables and the attributes identified
is not easy to establish. However, the qualitative impact of each of the design process attributes on the
variables can be established. It is assumed that the variables affect the attributes according to the degree
of dependency between them. The design process variables selected from the IDEF3 process model
may interact with each other.
For example, consider the design process model in Figure 3. The activities in this model are at a
high level of abstraction. One of the main advantages of using IDEF methodology is that it allows us to
decompose an activity at high level of abstraction into lower level activities. The activity “Prepare system
specifications” in Figure 3 can be decomposed into the activities in Figure 4, listed also in Table 1.

6
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

2 4

J1 J2
5 X 7
1 &

3 6

9
J3 J4
8 O 11 X 12

10

Critical activity

Figure 4. A process model with critical activities

A design project following the design process model shown in Figure 4 may undergo a number
of iterations. As a result, the last activity in the model may or may not be completed before the required
due date. The design activities have certain resource requirements, without which they can not be
performed. The total number of resources available is limited and that may cause resource conflict
between different activities in the design activity network. The duration of design activities is not fixed. It
is directly or indirectly influenced by the way design activities are performed. The stochastic duration of
design activities affects the length of the critical path. These characteristics of the process are important
from the point of view of the management of the design process, however, they can not be controlled
directly. In the approach presented in this chapter, the characteristics of the IDEF3 model are treated as
design process attributes.

Table 1. List of Activities of the Design Process in Figure 4

Activity
No. Activity Name
1 Perform project planning
2 Review and analyze customer requirements
3 Develop project coordination document
4 Define design requirements
5 Establish system design goals
6 Perform system tradeoffs
7 Finalize product requirements

7
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

8 Develop system requirements

9 Conduct internal requirements review
10 Review requirements with customer
11 Analyze modifications of system specifications
12 Finalize the system specifications

Unlike the design process attributes, the design process variables can be directly controlled. For
example, in the design model in Figure 4, the number of resources made available and number of design
projects undertaken simultaneously can be controlled. Similarly, the way the design effort is performed,
i.e., the level of interaction and involvement of different functional groups can also be directly controlled.
These control variables have an impact on one or more of the design process attributes. However, the
exact quantitative relationships between these design process attributes and the design process variables
are not known. Therefore, we use the house of quality to represent these relationships (see Figure 6)
using symbols shown in Figure 5. The house of quality in Figure 6 allows analyzing the relationship
between design process variables and designing process attributes.

Relationship Matrix
Strong positive
Interaction Matrix
Medium positive Positive

Weak positive Negative

Negative
Figure 5. Symbols used in the house of quality for design process attributes and variables

8
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

Average resource level savings

Design Process Variables

No. of supporting functional

Frequency of interaction
No. of design projects

among different design

functional groups

groups involved
undertaken
Design Process Attributes
No. of design activities delayed
due to the resources being not available
Risk of violating the due date

Average lateness of design activities

Completion time of the last design

activity in the network

Figure 6. House of quality for a design process

Let n be the total number of attributes. The value of the design process attributes are given by
the vector:
y = (y1, y2, ..., yn)

For attribute value yi, the lower limit and upper limit on its values are denoted by yiL and yiU,
respectively. Therefore, the feasible range for yi is:

yiL = yi = yiU

The design process attributes and their feasible ranges for this example are given in Table 2. These
ranges can be obtained from the past data.

Table 2. Design Process Attributes and Their Feasible Ranges

Attribute Range

9
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

y1 No. of design activities delayed due to 0 = y1 = 8

shortage of resources

y2 Risk of violating the due date 0 = y2 = 1

y3 Average lateness of design activities 2.5 = y3 = 9.4

y4 Completion time of the last design 18 = y4 = 35

activity of the process

Now, let m be the total number of design process variables. The vector gives values of the
design process variables:
x = (x1, x2, ..., xm)
For variable value xi, the lower limit and upper limit on its values are indicated by xiL and xiU,
respectively. Therefore, the feasible range for xi is:
xiL = xi = xiU
The design process variables and their feasible ranges for this example are given in Table 3.
The design process variables are not required to be independent. The interaction among these variables
is shown in the roof of the house of quality in Figure 7.

Table 3. Design Process Variables and their Feasible Ranges

Variable Range

x1 Average savings in resource level 3.2 = x1 = 8.4

x2 Number of design projects pursued 4 = x2 = 10

x3 Frequency of interaction between different 2 = x3 = 6

functional design groups

x4 Number of supporting functional groups 5 = x4 = 10

involved

10
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

The lower and upper limits on the design attributes as well as on the design process variables
represent their possible minimum and maximum values, respectively. For example, the risk of violating
the due date is the probability with which the due date is violated and hence it ranges between 0 and 1.
Similarly, in this specific example, there are at least 5 functional groups involved but it is not feasible to
include more than 10 functional groups.
Once the ranges on the values of the design process attributes and the design process variables
are established, one may want to quantify the relationships between the attributes and variables of this
process to examine the effect of increasing or reducing the values of certain design variables on the
attribute of interest. The commonly used conversion values for symbols in the house of quality are
shown in Table 4.

Table 4. Conversion of Symbols

Symbol Coefficient Value

+9

+3

+1

-3

The value of yi, the design process attribute i, is represented as a function of the design process
variables.
yi = f(x) (1)

Following the scheme presented in Locascio and Thurston (1994), this functional relationship
includes first order effects from the relationship matrix and second order effects from interaction matrix.
The design process variables are scaled on to the range of 0 to 1 to obtain the vector of relative design
process variable values:
x′ = (x1′, x2′, ..., xm′)

When the relative values of the design process variables, x′, are used in place of the original values of
process variables, x, in equation (1), the relative values of design process attribute are obtained:
yi′ = f(x′) (2)

This transformation from original values to the relative values is done to capture the relationship
between variables and attributes captured in the house of quality. The use of this relationship is
discussed next.

11
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

The functional relationship between design process attributes and design process variables can be
further expressed by considering the conversion of symbols in Table 4. For example, the expression for
“Risk of violating the due date” y2′ as a function of scaled process variables is expressed as:
y2′ = 3x1′ + 9x2′ + 3x3′ + 3x1′x2′ + 3x2′x3′ (3)

Similarly, the expressions for the remaining relative attributes are given by :

y1′ = 9x1′ + 3x2′ + 9x4′ + 3x1′x2′ − 3x1′x4′ (4)

y3′ = 3x1′ + 9x2′ + x4′ + 3x1′x2′ − 3x1′x4′ (5)

y4′ = 3x1′ + 3x2′ + 9x3′ + x4′+ 3x1′x2′ + 3x2′x3′

− 3x1′x4′ + 3x2′x4′ − 3x3′x4′ (6)

The allowable range for each relative design process variable is [0, 1]. The allowable range for
the relative design process attribute is computed by minimizing and maximizing the expressions for the
relative design process attributes.
When one wants to determine the attribute value corresponding to a specific combination of
values assigned to the design process variables, the values of the relative design process variables are
determined first. Then the value of the relative design process attribute is determined. This value is then
mapped back on to the original range of the design process attribute to obtain the actual value of the
design process attribute.
For example, the minimum (maximum) values 0 (21) of y2′ is obtained by minimizing (maximizing)
expression (3).

Therefore, the expression for y2 is:

y2 = 0 + ((1 - 0) (3x1′ + 9x2′ + 3x3′ + 3x1′x2′ + 3x2′x3′) / 21) (7)

Suppose that one is interested in assessing the change in “Risk of violating the due date” when
values of design process variables changed to a new set of values. The old values of the design process
variables and the corresponding new values are provided in Table 5.

Table 5. The Values of Design Process Variables

Design Variable Old Value New Value

xi xi′ xi xi′

Average savings in resource level 5.6 0.46 6.8 0.69

Number of design projects pursued 6 0.33 6 0.33

12
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

Frequency of interaction between 3 0.25 5 0.75

different functional design groups

Number of supporting functional 8 0.6 8 0.6

design groups involved

The value of y2 is determined from equation (7). Initially, the value of y2 was 0.28 and after
increasing the value of variables “Average resource level savings” and “Frequency of interaction
between different functional groups” (see Table 5), the new value of y2 is 0.42.
Similar to expression (7), the equations for other design process attributes are obtained:

y1 = (8 - 0)(9x1′ + 3x2′ + 9x4′ + 3x1′x2′ − 3x1′x4′) / 21 (8)

y3 = 2.5 + (9.4 - 2.5)(3x1′ + 9x2′ + x4′ + 3x1′x2′ − 3x1′x4′) / 13 (9)

y4 = 18 + ((35 - 18)(3x1′ + 3x2′ + 9x3′ + x4′+ 3x1′x2′ + 3x2′x3′

− 3x1′x4′ + 3x2′x4′ − 3x3′x4′) / 19 (10)

The approach presented above allows to analyze the impact of a change of any variable value on
the values of design process attributes by deriving quantitative relationships from the house of quality for
the design process. Based on these relationships, any desired changes in the values of design process
attributes can be implemented by making appropriate changes in the design process variables.

4. OPTIMIZING THE QUALITY OF CRITICAL DESIGN ACTIVITIES

The quality of results of design activities depends upon the quality of performing these activities.
A design engineer may select activities that are critical to the quality of the design. For example, in the
design activity network, shown in Figure 5, activities 4, 7, 8, and 11 are critical.
In this section, we analyze the relationship between the quality of performance of critical design
activities and the design process control variables. In the house of quality, the critical design activities
replace the design process attributes and the corresponding design process variables are determined. In
order to measure the quality of the critical design activities, quality indices are used. The quality index for
each design activity is in the range of 0 to 1, where 1 corresponds to the best quality and 0 corresponds
to the lowest quality. The design process attributes y1, y2, y3, and y4 and the corresponding ranges are
shown in Table 6.

Table 6. Critical Design Activities and Their Limits

Activity No. Critical Activity Name Limits

13
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

y1 4 Define design requirements 0 = y1 = 1

y2 7 Finalize product requirements 0 = y2 = 1

y3 8 Develop system requirements 0 = y3 = 1

y4 11 Analyze modifications of system 0 = y4 = 1

specifications

The set of design process variables includes two of the previously defined variables, “Frequency
of interaction between different functional groups” and “Average resource level”. In addition, “Level of
expertise” is used as a design process variable. This variable refers to the average experience of a
designer in terms of number of similar projects handled in the past. Since the designers are also assigned
other responsibilities and these may affect the quality of input from the team member, “Number of
resource preemptions” is used as a variable. The information required by the designer may become
available from a variety of sources. The designer may try to minimize the risk of using wrong information
by referring to more than one source of information so that he / she can verify the information obtained
from other sources. Therefore, “Number of information sources” is treated as a variable affecting quality
of performance of certain design activities. The variables and the corresponding limits are listed in Table
7.

Table 7. Variables Impacting Quality of a Design Process and Their Limits

Variable Limits

x1 Average level of expertise 0 = x1 = 8

x2 Average resource level 4.6 = x2 = 12.34

x3 Frequency of interaction between 1 = x3 = 5

different functional design groups

x4 Number of resource preemptions 8 = x4 = 25

14
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

x5 Number of information sources 3 = x5 = 11

The next step is to construct the relationship matrix and the interaction matrix for the
relationships between the quality indices of the critical design activities and the corresponding design
process variables. The house of quality is shown in Figure 7.

Number of resource preemptions

Frequency of interaction among

Number of information sources

Design Process Variables

different functional groups

Average level of expertise

Average resource level

Critical Design Activities

Define derived requirements

Finalize product requirements

Develop system
requirements specifications
Analyze modifications suggested
in system specifications

Figure 7. The house of quality for relationships between critical activities and
process variables

As previously discussed, expressions for each of the attributes in terms of the process variables
and their interaction are developed. The values of the design process variables are scaled to [0, 1]. The
quality indices for each critical design activity in terms of the relative design process variables are
obtained. These expressions are as follows :

y1 = 0 + (1 - 0)(3x1′ + x2′ - x4′ + 9x5′ - x1′x2′ - x1′x5′ + 3x2′x4′) / 13 (11)

y2 = 0 + (1 - 0)(9x1′ + 3x3′ + x5′ - x1′x5′ + 3x3′x5′) / 15 (12)

15
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

y3 = 0 + (1 - 0)(9x1′ + x2′ + 9x3′ + 3x5′ - x1′x2′ - x1′x5′ + 3x3′x5′) / 23 (13)

y4 = 0 + (1 - 0)(3x1′ + 9x2′ + 3x3′ - x4′ - x1′x2′ + 3x2′x4′) / 16 (14)

Assume that wi, i = 1, 2, 3, 4 are the weights assigned to four critical design activities listed in
Table 6. These weights are assigned to the critical design activities. Our goal is to determine the values
of design process variables to maximize the combination of the quality indices of the critical activities.
The optimization problem can be modeled as follows :

Maximize w1y1 + w2y2 + w3y3 + w4y4 (15)

s.t. (11) - (14)
0 ≤ xi′ ≤ 1 for i = 1, 2, 3, 4

The objective function (15) is the weighted sum of the quality indices of the critical design
activities. The expressions (11) - (14) form the constraints for the problem, where the relative design
process variables used in these expressions have a range [0, 1]. The constraints (11) - (14) can be
substituted to the objective function. The resulting constrained geometric programming problem can then
be solved using standard techniques. The optimal solution to this problem are the values for the relative
design process variables xi′, i = 1, 2, 3, 4. These values can be mapped back on to their original ranges.

Example
Consider w1 = 10, w2 = 35, w3 = 20, w4 = 15

The resulting problem is:

Maximize 10y1 + 35y2 + 20y3 + 15y4
s.t. (11) - (14)
0 ≤ xi′ ≤ 1 for i = 1, 2, 3, 4

Substituting (11) - (14) to the objective function the following program is obtained:

Maximize 39.94x1′ + 10.38x2′ + 21.47x3′ − 1.71x4′ + 13.74x5′

−2.88x1′x2′ − 4.63x1′x5′ + 5.12x2′x4′ + 11.6x2′x5′ (16)

s.t. 0 ≤ xi′ ≤ 1 for i = 1, 2, 3, 4 (17)

The problem (16) - (17) is a signomial geometric programming problem (see Beightler, 1976 for further
details).

The optimal solution for this problem is:

x1′ = 0.75, x2′ = 0.93 , x3′ = 0.832, x4′ = 0.1, x5′ = 0.95

16
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

These values of the relative design process variables are then mapped on to their original ranges to
obtain the best values for control variables as shown in Table 8.

17
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

Table 8. The Values of Variables Affecting the Quality of Critical Design Activities

Variable Best Value Alternative

Value

x1 Average level of expertise 6 6

x2 Average resource level 11 9

x3 Frequency of interaction between 4 4

functional design groups

x4 Number of resource preemptions 10 10

x5 Number of information sources 9 7

Value of the Objective Function 68.1 59.33

The best set of values for the design process variables shown in Figure 8 indicate that for a fairly
high quality performance of design activities, we need high level of expertise, high resource level,
reasonable frequency of interaction among functional groups, small resource preemptions, and a high
number of information sources. The value of the objective function in (16) is indicative of the overall
quality of performance of critical activities. As shown in Table 8, if we slightly reduce the values of
variables “Average resource level” (from 11 to 9) and “Number of information sources” (from 9 to 7),
then the value of the objective function reduces from 68.1 to 59.33 (by almost 13 %). In this way,
optimal values of the design process variables corresponding to a function of design process attributes,
with a given set of qualitative relationships between them represented by the house of quality, are
determined. This approach also further helps to determine the impact of change in values of design
process variables involved on the overall quality of the performance of critical design activities.

5. SUMMARY
In this chapter, the relationships among design process attributes and variables were studied.
The design process was represented with an IDEF3 model. Due to the randomness involved in the
design process, qualitative relationships between design process attributes and the corresponding
variables were established. These relationships were represented by the house of quality, and then
transformed into quantitative relationships. It was shown that such quantitative relationships could be
used to study the impact of design process variable on the design process attributes. The use of these

18
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

quantitative relationships as constraints in maximization of the desired function of the process attributes
was discussed.

REFERENCES
1. Ang, C., Luo, M., and Gay, R. 1994, “Development of a Knowledge-based Manufacturing
Modeling System Based on IDEF0 for the Metal-cutting Industry,” International Journal of
Production Economics, 34, 267 - 281.
2. Beightler, C., 1976, Applied Geometric Programming, Wiley, New York.
3. Belhe, U. and Kusiak, A., 1996, The Quality Function Deployment in a Design Process,
International Journal of Production Research, 34, 2119-2131.
4. Bossert, J. L., 1991, Quality Function Deployment: A Practitioner's Approach (Quality and
Reliability Series, Vol. 21), Marcel Dekker, New York.
5. Colquhoun, G., Baines, R., and Crossly, R., 1993, “A State of the Art Review of IDEF0,”
International Journal of Computer Integrated Manufacturing, 6, 252 - 264.
6. Cook, H., 1991, “On Competitive Manufacturing Enterprises II: S-Model Paradigm,”
Manufacturing Review, 4, 106 - 114.
7. Cook, H, 1992, “New Avenues to Total Quality Management,” Manufacturing Review, Vol. 5,
No. 4, pp. 284 - 292.
8. Hauser, J. and Clausing, D., 1988, “The House of Quality,” Harvard Business Review, May -
June ‘88, 63 - 73.
9. Hsu, C., 1994, “Manufacturing Information Systems,” in R. Dorf and A. Kusiak, ed., Handbook of
Design, Manufacturing and Automation, Wiley, New York, 737 - 766.
10. Keats, L. P., 1990, Quality Control Issues in Automated Manufacturing. The Automated Factory
Handbook: Technology and Management (Cleland, D. I., and Bidanda, B.), TAB Books,
Blue Ridge Summit, PA, pp. 680-704.
11. Keeny, R. and Raiffa, H., 1976, Decisions with Multiple Objectives: Preferences and Value
Tradeoffs, Wiley, New York.
12. Kusiak, A., (Ed.), 1993, Concurrent Engineering: Automation, Tools, and Techniques, John
Wiley, New York.
13. Kusiak, A., Larson, T.N., and Wang, J., 1994, “Reengineering of Design and Manufacturing
Process,” Computers and Industrial Engineering, 26, 521 - 536.
14. Locascio, A. and Thurston, D., 1994, “Quantifying the House of Quality for Optimal Product
Design,” DE-Vol. 68, Design Theory and Methodology - DTM ‘94, The American Society of
Mechanical Engineers, 43 - 54.
15. Menzel, C., Mayer, R., and Edwards, D., 1994, “IDEF3 Process Descriptions and Their
Semantics,” in C. Dagli and A. Kusiak, ed., Intelligent Systems in Design and Manufacturing,
ASME Press, New York, pp. 171 - 212.
16. Mayer, R., Cullinane, T., deWitte, P., Knappenberger, W., Perakath, B., and Wells, M., 1992,
“Information Integration for Concurrent Engineering (IICE),” Report # AL-TR-1992-0057, Air
Force Systems Command, Wright-Patterson Air Force Base, Ohio.
17. Menon, H., 1992, TQM in New Product Manufacturing, McGraw-Hill, New York.
18. Nevins, J. and Whitney, D., 1989, Concurrent Design of Products and Processes, McGraw-Hill,
New York.
19. Nichols, K., 1992, “Better, Cheaper, Faster Products - By Design,” Journal of Engineering
Design, 3, 217 - 228.

19
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

20. Pahl, G., and Beitz, W., 1988, Engineering Design: A Systematic Approach, Springer-Verlag,
Berlin, Germany.
21. Ross, D., 1977, “Structured Analysis (SA): A Language for Communicating Ideas,” IEEE
Transactions on Software Engineering, SE-3, 16 - 24.

QUESTIONS
1. What is the house of quality?
2. How the quality function deployment model differs from the process model?
3. What is common between the quality function deployment model and the process model?
4. What is the quality function deployment cascade? Define it for a product development
process.
5. Why a four grade rather than a ten grade scale is used in QFD?

PROBLEMS
1. Transform number x from the interval [0, 9] into number y from the interval [6, 18].

2. Transform number x from the interval [3, 9] into number y from the interval [6, 18].

3. Figure A1 shows the relationship between customer requirements and subsystems of an

electrical product in the form of a house of quality. The customer evaluations as well as the
current (product model A) and desired (product model B) technical evaluations are
provided in the house of quality.
The ranges for each requirement and each technical subsystem are identical integer numbers
in the interval [1, 5]. The numerical scale for the relationships in the main QFD matrix is
(strong positive = 9, medium positive = 3, weak positive = 1).

20
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

Strong positive +1

Meduim positive -1 +1

Weak positive

perception (1-5)
1 - Low preference (evaluation)

Panel design
5 - High preference (evaluation)

mechanism

Product A
Electrical

customer
Casing
-1 - Negative relationship
+1 - Positive relationship

Easy to read 5

Long life 3

Easy to adjust 4

Easy to carry 1

Product A
4 2 3
evaluation (1-5)

Product B
evaluation (1-5) 5 5 4

Figure A1. The house of quality for an electrical product

(a) Determine the values of customer requirements that would meet the new levels of technical
product evaluations
(b) For the following values of customer requirements [easy to read = 5, long life = 4, easy to
adjust = 4, and easy to carry = 4], determine the level of technical product evaluations
4. Figure A2 shows the relationship between customer requirements and technical specifications
of a consumer product. The range of values of the customer relationships is (0, 10). The
range of values for technical specifications is as follow: Spec 1 (2, 12), Spec 2 (1, 10),
Spec 3 (5, 15), and Spec 4 (10, 15).
The weights associated with the customer requirements are as follows: Low weight (1),
High reliability (3), High durability (2), and High functionality (5).

21
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

Technical Specifications

Specification 2

Specification 3
Specification 1

Specification 4
Customer Requirements

Low weight

High reliability

High durability

High functionality

Figure A2. The house of quality of a consumer product

Assuming the usual values of the entries of QFD, formulate a mathematical programming
model that maximizes the weighted total of customer requirements while meeting the ranges
of values for the technical specifications.

5. For the QFD table in Figure A3 for the door of an aircraft set up a model maximizing
the total satisfaction of customer requirements. Hint: (1) assume the range for each technical
characteristic, (2) assume that each customer requirement is equally important.

22
A. Kusiak, Engineering Design: Products, Processes, and Systems, Academic Press, San Diego, CA, 1999.

Relationship matrix

Strong positive
X
Medium positive
Weak positive X X

aluminum alloy (low) or steel (high)

Lower door seal thickness
Technical characteristics
Interaction matrix

Lower peak closing force

Tighter fit on closure

Lower force to hold

Positive

Material
X Negative

for assembly
Customer requirements

Cost (mfg and material)

Ease of maintenance

Reliability

Maintenance cost

Figure A3. The QFD table for the door of an aircraft

23