Professional Documents
Culture Documents
2004 CostIJAMT
2004 CostIJAMT
net/publication/225675889
CITATIONS READS
26 7,489
2 authors, including:
SEE PROFILE
All content following this page was uploaded by Yuo Tern Tsai on 29 January 2017.
Articles
Browse
> Home / Publication / Issue /
Article
Previous article
The International Journal of Advanced Manufacturing
Next article
Technology
Publisher: Springer-Verlag London Ltd Linking Options
ISSN: 0268-3768 (Paper) 1433-3015 (Online)
DOI: 10.1007/s00170-003-1564-7 Full Text Secured
The full text of this article is sec
Issue: Volume 23, Numbers 7-8 to subscribers. To gain access,
Date: April 2004 may:
Pages: 514 - 522 Add this item to your shopp
cart for purchase later.
Add to Shopping Car
Abstract When designing a new system, often cost is a critical factor in determining
whether the system will be carried out or not. For assessing the feasibility of a system
being developed in future, estimating the cost of system in the conceptual phase
becomes more and more important. This paper presents a cost estimation method
based on function characteristics and the technique of quality function deployment
(QFD). The cost structure of systems is first determined according to four cost items:
the direct material cost, the direct labour cost, the variable factory overhead and the
fixed factory overhead. Based on the cost structure, the total costs of some well-
known competitive products of the developed system are then investigated. The
function levels of these well-known products are identified one by one by means of
the QFD technique. By integrating the information about the competitor s costs and
the function levels, a set of cost factors which are used to evaluate the system costs is
derived by using the least-square-method. As a result, the total cost of the designed
system can be estimated quickly through simple steps of calculation. Furthermore, the
successful probability of a system being designed within someone budget can be
assessed in advance.
Yuo-Tern Tsai
Email: yttsai@dlit.edu.tw
file://D:\Cytt\Papers\Cost\Cost%20-%20Article.htm 2004/5/31
The International Journal of Advanced Manufacturing Technology
© Springer-Verlag London Limited 2003
DOI 10.1007/s00170-003-1564-7
Original Article
Y.-T. Tsai
Department of Mechanical Engineering, De-Lin Institute of Technology, 236 Tucheng, Taiwan
R.O.C.
Y.-M. Chang
Department of Civil Engineering, De-Lin Institute of Technology, 236 Tucheng, Taiwan R.O.C.
✉ Y.-T. Tsai
E-mail: yttsai@sitc.edu.tw
Received: 29 April 2002 / Accepted: 17 October 2002
Abstract When designing a new system, often cost is a critical factor in determining whether the
system will be supported or not. For assessing the feasibility of a system being developed and
assessing the progressive cost control, estimating the system cost in the conceptual phase becomes
more and more important. This paper presents a method of function-based cost estimation associated
with quality function deployment (QFD) to support system design. The cost structure of the systems
is first determined according to four cost items: the direct material cost, the direct labour cost, the
variable factory overhead and the fixed factory overhead. Based on the cost structure, the total
costs of some well-known competitive products of the developed system are then investigated.
The function levels of these well-known products are also identified one by one by means of the
QFD technique. By integrating the information about the competitor’s costs and the function levels,
a set of cost factors which are used to evaluate the system costs is derived by using the
least-square-method. As a result, the total cost of the designed system can be estimated quickly
by simple calculations. Furthermore, the probability of a system being designed within a budget
can be assessed in advance.
1
Keywords Cost estimation · System design · Quality function deployment (QFD)
1 Introduction
Huthwaite [1] indicated that 75% of the manufacturing cost and 80% of the product quality of a
product was determined by the end of the conceptual phase of the development of a product. It
emerged that a product’s quality and its cost are already decided early in the design phase. Since
product design is usually a techno-economic process, there is always a tradeoff between quality
goals and limited budgets. In the past, many design-related strategies have been proposed for
promoting a product’s quality and reducing its cost. Typically, concurrent engineering is one
example [2]; planning the design procedure to shorten the design time, as well as cutting down
the development cost of a products is another one [3]. However, product competition is usually
dependent on both the product functions and its costs. Thus, further controlling costs during the
design, and developing design systems for cost is increasingly important and is now emphasised
in system design now [4]. For success in target costing, good information is required. By using
quality function deployment (QFD) and spreadsheets, a variety of input information can be
effectively organised [5].
For efficiently making use of resources and achieving the goals of cost control, accurately
estimating the total cost of the system in the early design phase is necessary. If the total cost of
the system can be evaluated correctly in advance, the appropriate expenditures can be allocated
so that the system targets can be achieved reasonably within the preset budget. As a result, the
target of designing for cost will be easily reached. Because the costs in the conceptual phase are
frequently more or less uncertain, the cost estimates need to be to reviewed frequently or the
estimated cost will have a very large deviation. A feasible method to estimate the cost of a system
in the conceptual stage was introduced in [6].
Normally, cost estimation during designing usually refers to a simple method based on a defined
cost entity which is at the designer’s disposal at the given design stage. Moreover, it is dependent
on the comprised components and the characteristics of these components. For instance, the cost
of a machined component can be estimated from 3 aspects: (1) the shape and cost of the stock
material (2) the amount and the shape of the material that needs to be removed, and (3) how
precisely it must be removed [7]. Based on these aspects, Ou and Lin [8] proposed a framework
in terms of a feature-based approach for estimating the manufacturing cost of a design. Similarly,
a feature-based product cost estimation method using a back-propagation neural network was
2
reported by Zhang [9] for estimating the product costs according to design information only. In
addition, Samid [10] proposed a conceptual framework for cost estimation. This framework is a
preliminary step towards what will hopefully develop into a full-fledged comprehensive general
theory of cost estimation. Smith [11] combined artificial neural networks with statistical models
for examining the performance, stability and ease of cost estimation modelling, to develop a cost
estimating relationship. Jha [12] developed a mathematical model in which the probable range of
cost was estimated by a stochastic geometric program. Aiming at a methods parametric estimation,
Duverlie [13] compared two manners of economic evaluation of a mechanical design: a parametric
method and a case-based reasoning method, to produce an efficient cost estimation model.
However, the noteworthy methods of cost estimation are those based on: weight, material,
throughput parameters, physical relationships, regression analysis and similar laws. Based on
quick or all-inclusive calculations, several methods for estimating the cost of components were
introduced by Pahl and Beitz [14] as follows:
(1) Comparing with a reference cost (a reference value)
(2) Estimating using a share of material cost; this manner is suitable in some particular applications
of which the rates of material to manufacturing costs is known.
(3) Estimating using regression analysis; the relation between the cost and the characteristic
parameters (output, weight, diameter, shaft length, etc.) are determined based on the statistical
analysis of data
(4) Extrapolating using a similarity relationship; the cost is evaluated from the geometric similarity
of known components.
Due to the diversity of products and the complexity in manufacturing, automatic technology is
extensively utilised in many fields for product manufacturing. The proportion of the manufacturing
cost of products in the production cost is obviously increased and the percentage of the direct
labour cost is gradually reduced due to the automatic technology adopted. The product cost is
easily distorted if the direct labour cost is shared by the manufacturing cost in a traditional
manufacturing process [15]. The calculated deviation of the cost results from mistakes made in
the decision making. Cooper and Kaplan [16] suggested activity-based costing (ABC) to overcome
the faults in the customary estimation methods for the costs of modern production technology.
The fundamental conceptual contribution of ABC is its recognition that cost elements are not
intrinsically fixed or valuable. It emphasises the distinction between value-adding and
non-value-adding activities to allow an easier identification of cost reduction opportunities [17].
3
The ABC approach, therefore, can supplement or replace the traditional cost systems, and provides
additional insights and information that are valuable for system planning, and enable a search to
be made for low-cost strategic alternatives in system design.
Although many cost estimation methods have been reported, one aspect that they have in
common is that their cost estimations are always related to a specific product for which the related
cost variables (labour, materials, manufacturing processes, etc.) are already established. For a
product in the conceptual phase, the cost estimations are few and the problems have not yet been
investigated completely due to insufficient cost information. The technique of QFD is extensively
used in early product design. To facilitate decision-making in product design and development,
integrating design costs into the QFD framework was proposed in [18]. A further study in integrating
the life-cycle-cost into QFD matrices and deploying quality, environmental and cost requirements
throughout the entire product development process was introduced in [19]. The facts imply that
combining some well-known cost information with QFD is a feasible method for accurately
estimating product costs at an early design stage.
In this paper, we integrate a statistical analysis technique with QFD for quickly estimating the
system development cost. After the function requirements are clarified, the total costs of the
possible competitive products are investigated according to the set cost structure. The magnitude
of the cost is expressed as a probability distribution in accordance with real conditions. The function
levels of the competitive products are analysed one by one through QFD matrices. After that, the
mathematical relationship between the function levels and the cost magnitudes is determined by
using the least square method. As soon as the function-cost relationship is decided, the total cost
of the designed system can be calculated when its function requirements are given. Furthermore,
the probability of the system achieved in the required performance can also be assessed.
4
the material and the labour cost for producing a specific component. Indirect costs are those costs
that cannot be allocated directly, for example the costs of running the stores and illuminating the
workshop [20].
According to the properties the of system costs, the cost structure of a system can be expressed
as shown in Fig 1. The detailed contents of each item is explained in the following sections.
Fig. 1 The cost structure of the system
5
system, the direct labour cost can be computed according to the conversion time of the product,
and the needed labour and the degree of factory automation. Naturally, more labour costs are
involved if the product is more complex (more conversion steps) or the degree of factory automation
is low.
the influence of these production-related factors, the cost variable, Xij, may not be a fixed value.
Alternately, it may be a range which has upper and lower limits or is a function. The cost variable
is set as a probability distribution for appropriately modelling the real condition of the cost analysis.
Typically, if the cost variable has a Gaussian distribution and the probability density function is
f(x), then the mean and the variance of the cost can be calculated by the following equation:
[Table 1. will appear here. See end of document.]
6
(1a)
(1b)
While calculating the system cost according to the cost distribution, we must consider the
coupled relationship between the cost variables. Normally, the four cost items (j direction) can be
regarded as independent since their cost attributes are obviously different. On the other hand, the
subsystems (i direction) may have more or less of a correlation due to the functions coupling with
each other. Thus, the mean of the total cost of a system can be computed by:
(2)
where n denotes the number of the subsystems. Meanwhile, the variance of the total cost can be
calculated by:
(3)
standard deviation of the cost variable, X.
7
(4)
Fj the function score which is a product of the function importance (or weight) and the function
level (describing the quality). The function score can be measured by QFD which is discussed in
the next section.
As Fj is determined, the cost factor, kj, becomes a critical factor to estimate the system cost. To
evaluate kj, we suppose that a system has m competitive products and n function attributes. The
costs of these competitive products (μ(c), Var(c)) are obtained by a system cost analysis.
Considering the function attributes and the costs of all the competitive products, Eq. 4 is expressed
in a matrix form. It is:
(5)
Simplifying the equation, we have the following equation:
(6)
evaluated by using the least square method under the condition of m≥n. It is:
(7)
As soon as the cost factors are obtained, the total mean cost of a new product can be quickly
estimated by Eq. 4 in cooperation with the given function scores Fj.
The variance of the cost factors can also be acquired by a similar method. According to the
format of Eq. 4, the variance of the total cost of the product can be expressed as:
(8)
Considering m number of products, the matrix form of the cost variance can also be written as:
8
(9)
Similarly, Eq. 9 is expressed as:
(10)
Then the variances of the cost factors can be calculated by:
(11)
where σk is the standard deviation of the cost factor. Accordingly, substituting the results into
Eq. 4, the total cost variance can also be obtained.
4 QFD
QFD is a structured framework for ensuring that a customer’s needs are folded into every phase
of the design and the manufacturing [2]. The basic goals of QFD are to increase the customer
satisfaction while reducing the cycle time of the product development. A series of matrices are
used to deploy the customer’s demands to assist the company in moving from product planning
through manufacturing planning. The most frequently applied QFD matrix, which is also called
the “house of quality”, includes the six basic steps shown in Fig. 2. An important issue in the QFD
process is that these steps are all completed by a cross-functional team. The cross-functional nature
of the group helps shorten the product’s total development time and increases its marketability.
The detail contents of the six steps are summarised in the following section.
9
Fig. 2 The six steps of a “house of quality”
10
to you? How well are the competitors doing?” For each customer need, the existing design is
assessed on a scale of 1 to 5, where 1 means the design doesn’t meet the need at all, and 5 means
the design fulfils the need completely. Though these are not refined ratings, they do give an
indication of how the competition is perceived by the customer. Furthermore, the competition will
be reviewed again in step 5 of the QFD technique for technical comparisons.
11
4.5 Technical comparisons
This step compares the technical specifications of the competing products for determining the
engineering values of the function characteristics. This ensures that all products are examined
under the same environmental conditions, with the same testing equipment; this may be the only
way of collecting data concerning the technical measures as identified in step 4. For convenience,
we can translate the technical specifications of the competitors into function levels which are
expressed in a scale of poor to excellent. According to the investigated results of the function
levels, a designer can judge the merits and faults of well-known competitive products so that a set
of target levels of the function characteristics can be appropriately chosen. A target level is the
degree of improvement that the company would like to have after the new product is presented.
According to the given target level, the engineering value of the function characteristic, i.e., the
design specification, can be derived for establishing the design goal. Normally, the best target is
set for a specific value. Less precise, but still usable, are those targets set within a specific range.
Some extremes such as “larger” or “as small as possible” are not good targets, since they give no
information that indicates when the performance of the new product is acceptable.
In addition, the total costs of the competitive products must also be analysed one by one according
to the approach described in an earlier section. The results are combined with function levels for
finding out the cost factors of the function characteristics, to estimate the total cost of the designed
product.
5 A numerical example
To show the cost estimation method, a mechatronic system which consists of five subsystems
(E1~E5) (control, power, mechanisms, tools and sensors, as shown in Fig. 3) is used as an example.
12
Focusing on the system, the needs of customer are given by a representative group of customers.
They are: (1) very stable (2) very rapid (3) high accessibility (4) consuming a small amount of
electricity (5) sufficient horse power (6) high precision; (7) operating quietly (8) activity agility
(9) good maintainability and (10) failure wearout. Next, the customer needs are expressed with
measurable engineering indexes which are defined as follows:
Fig. 3 A typical mechatronic system
– Weight (kg)
6
– Failure rate (times/10 hr)
– Response time of the controller (second)
– Power consuming (kw)
– Maximum horsepower (Hp)
– Positioning error (mm)
– Maximum moving velocity (m/s)
– Maintenance time (MTTR) (hr)
– Expected life of the system (year)
– Operating noise (db)
The QFD matrix representing the correlations between the customer needs and the function
characteristics are recorded in Fig. 4. Integrating the weight information and the correlated values
(rij), the importance-score of each engineering index is calculated by:
13
Fig. 4 The correlations of the customer needs and the function characteristics in the example
(12)
On the other hand, the function characteristics of the competitive products can be investigated.
The function characteristics are recorded from level-1 (poor) to level-10 (excellent), here for
convenience in a technical comparison. As a result, the function score (Fj) is:
(13)
coefficients, ρ, are all set to 0 since these subsystems possess independent tendencies. Similarly,
the total costs of the other competitive products can also be obtained by the analysing procedure.
[Table 2. will appear here. See end of document.]
[Table 3. will appear here. See end of document.]
14
In this example, we suppose that the designed system has 10 competitive products. The analysed
results in the function level and total cost of these products are recorded in Table 4. Combining
the function weights in Fig. 4 and the data in Table 4, the mean and the standard deviation of cost
factors (μk,σk) of the example can be derived by Eqs. 7 and 11, respectively. The evaluated results
are recorded in Fig. 5. Next, the appropriate design targets of the system can be set reasonably
and the tradeoffs among the function characteristics can also be given according to the existing
correlations. The corresponding data about this example is listed in Fig. 5. The total cost of the
system will be μ(c)=$571, Var(c)=$425 according to the cost estimating method.
[Table 4. will appear here. See end of document.]
Fig. 5 The corresponding data of the system in design targets, cost factors and the tradeoff of function
characteristics
Based on the estimated cost information, the reliability index of cost can be expressed as:
(14)
where Y is the cost budget. With the abovementioned equation, we can compute the probability
of the system accomplishment for different expenditures. Several typical results are listed in
Table 5. For example, the achieved probability is only 5% if we spend $537.1; however, the
probability would reach 95% if the budget is set to $605. By means of the reliability information,
15
if the estimated cost is too high or too low related to the budget, the designer can adjust the function
levels of the system according to the technical tradeoffs in Fig. 5 to satisfy customer requirements,
meanwhile, quickly estimating the needed cost. This provides a rapid manner for assessing the
feasibility of a system developed under some cost constraints. As a result, a decision-maker can
avoid the problems that not only is the budget set too high (and the product competition declining)
but also that the preset costs are too low (with the budget increasing in the future) so that the goals
of cost control can be easily achieved.
[Table 5. will appear here. See end of document.]
6 Conclusion
This paper provides a method of a quick cost estimation based on system function characteristics
to support system design. The cost estimating is integrated with the technique of QFD. The function
levels and the cost distributions of competitive products are investigated in advance to find out
the relationship between the function levels and the costs. The relationship is formulated by a set
of cost factors which are obtained by the least square method. After the cost factors are determined,
the cost of the new system can be quickly estimated as soon as its function levels are given. To
summarise:
(1) System cost is expressed as a distribution so that the successful probability of the system
being achieved for different costs can be assessed. Based on the analysed results, a designer
can quickly adjust the development strategy of system and avoid bad decision making.
(2) The function characteristics of a system will be increased following the technological evolution.
Thus, using more cost factors (kj) to estimate the system cost will insure the accuracy of the
predicted cost and the achievement of the required performance.
(3) If the cost information of the competitive products is less than the number of the function
characteristics, i.e., m<n, the cost estimating method becomes impractical. In this situation, we
can choose some key characteristics, i.e., less cost factors, so that the method can be implemented
successfully.
Acknowledgements The work was supported by a grant from the National Science Council under
contract no. NSC 90-2212-E-237-004. The authors would also like to express their appreciation
to the reviewers for their valuable suggestions.
16
References
1. Huthwaite B (1988) Designing in quality. Quality 27(11):26–35
2. Parsaei HR, Sullivan WG (1993) Concurrent engineering: contemporary issues and modern design
tools. Chapman & Hall, London
3. Tsai YT, Wang KS (1999) The development of modular-based design in considering technology
complexity. Europ J Oper Res 119(3):138–149
4. Michaels JV, Wood WP (1989) Design to cost. Wiley, New York
5. Clifton BT (1998) Inputs for successful target costing. In: Proceedings of the 1998 ASME International
Mechanical Engineering Congress and Exposition, Anaheim, CA, 15–20 November 1988, pp 21–27
6. Tsai YT, Wang KS (2000) A method of cost estimation of system in conceptual design stage. Comm
Manage Quart 1(2):139–152
7. Ulman DG (1992) The mechanical design process. McGraw-Hill, New York
8. Ou-Yang C, Lin TS (1997) Developing an integrated framework for feature-based early manufacturing
cost estimation. Int J Adv Manufact Technol 13(9):618–629
9. Zhang YF, Fuh YJH (1998) Neural network approach for early cost estimation of packing products.
Comput Indust Engin 34(2):433–450
10. Samid G (1996) Estimation knowledge: a conceptual framework for cost estimation. Cost Engin
38(5):25–31
11. Smith AE, Mason AK (1997) Cost estimation prediction modeling: regression versus neural networks.
Engin Econ 42(2):137–162
12. Jha NK (1996) Probabilistic cost estimation in advance of production in a computerized manufacturing
system through stochastic geometric programming. Comput Indust Engin 30(4):809–822
13. Duverlie P, Castelain JM (1999) Cost estimation during the design step: the parametric method versus
the case-based reasoning method. Int J Adv Manufact Technol 15(12):895–906
14. Pahl G, Beitz W (1996) Engineering design: a systematic approach. Springer, London
15. Cooper R, Kaplan RS (1998) Measure costs right: make the right decision. Harv Bus Rev 66:96–103
16. Cooper R (1992) Activity-based costing for improved costing. Handbook of Cost Management, B1.
Warren, Gorham & Lamont, Boston
17. Dhavale DG (1993) Activity-based costing in cellular manufacturing systems. J Cost Manage 8(1):13–27
18. Bode JF, Richard YK (1998) Cost engineering with quality function deployment. Comput Indust Engin
35(3–4):587–590
19. Zhang Y, Wang HP, Zhang C (1999) Green QFD-II: a life cycle approach for environmentally conscious
manufacturing by integrating LCA and LCC into QFD matrices. Int J Prod Res 37(5):1075–1091
20. Horngren CT, Foster G (1991) Cost accounting: a managerial emphasis, 7th edn. Prentice-Hall,
Englewood Cliffs, NJ
17
Table 1. The cost variables of a uncertainty system
Subsystems I1 I2 I3 I4
E1 X11 X12 X13 X14
E2 X21 X22 X14 X24
⋮ ⋮ ⋮ ⋮ ⋮
En Xn1 Xn2 Xn3 Xn4
18
Table 2. The supposed costs of some well-known products
Subsystems Cost items
U1(a,b) U2(a,b) U3(a,b) U4(a,b)
E1 70 90 25 35 33 42 14 22
E2 50 70 23 35 19 30 21 32
E3 30 50 15 25 22 30 19 29
E4 20 35 12 22 14 23 12 21
E5 10 20 15 20 12 18 10 16
19
Table 3. The mean and the variance of the cost elements of the well-known product
20
Table 4. The function levels and the total costs of some competitive products in the example
21
View publication stats
Table 5. The successful probabilities of the system accomplished under different cost budgets
Cost budgets Probability
$537.1 5%
$544.6 10%
$571.0 50%
$597.5 90%
$605.0 95%
22