You are on page 1of 11

A Modified Analytic Hierarchy Process for Automated Manufacturing Decisions

Author(s): Stephen F. Weber


Source: Interfaces, Vol. 23, No. 4 (Jul. - Aug., 1993), pp. 75-84
Published by: INFORMS
Stable URL: https://www.jstor.org/stable/25061775
Accessed: 07-09-2022 10:53 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Interfaces

This content downloaded from 14.139.108.34 on Wed, 07 Sep 2022 10:53:24 UTC
All use subject to https://about.jstor.org/terms
A Modified Analytic Hierarchy Process for
Automated Manufacturing Decisions
STEPHEN F. WEBER United States Department of Commerce
National Institute of Standards and Technology
Office of Applied Economics
Computing and Applied Mathematics Laboratory
Gaithersburg, Maryland 20899

Decisions on adopting new manufacturing technologies are of


ten biased because they are based only on those impacts that
can be expressed in financial terms. Multi-criteria decision tools
that include nonfinancial impacts avoid this bias. One such
tool, the analytic hierarchy process (AHP), has been modified
for the manufacturing environment and incorporated into
AutoMan, decision support software for microcomputers. The
modified AHP is used here to decide on the best way to auto
mate a machine shop. The four steps are: (1) specify criteria
and alternatives; (2) weight criteria using pairwise comparison
judgments; (3) rate alternatives with respect to criteria; and (4)
compute the overall weighted score for each alternative. This
approach preserves existing reliable financial information and
integrates it with both nonfinancial quantitative performance
data and qualitative informed judgments. Judgments are made
explicit and checked for consistency. Over 800 industrial man
agers have purchased AutoMan to improve their automation
decisions.
when deliberations are limited to strictly
Decisions on adopting new manufac
turing technologies are often biased financial impacts. Managers faced with

Contribution of the National Institute of Standards and Technology, DECISION ANALYSIS?MULTIPLE CRITERIA
not subject to copyright in the United States ANALYTIC HIERARCHY PROCESS
0091-2102/93/2304/0075$01.25
This paper was refereed.

INTERFACES 23: 4 July-August 1993 (pp. 75-84)

This content downloaded from 14.139.108.34 on Wed, 07 Sep 2022 10:53:24 UTC
All use subject to https://about.jstor.org/terms
WEBER

manufacturing decisions can no longer rely criteria. For example, lexicographic order
solely on traditional measures of financial ing uses secondary criteria only as tie
worth because so many impacts cannot be breakers; the satisficing method imple
readily denominated in dollars [Berliner ments criteria only as lower or upper
and Brimson 1988; Canada and Sullivan bound hurdles; and dominance requires
1989; Choobineh and Suri 1986, pp. 269 that all criteria support the same ranking.
281; Drucker 1990; Jelinek and Goldhar Methods that allow direct trade-offs
1986; Owen 1990; Padmanabhan 1989; among all the criteria are preferable be
Sullivan 1986]. They must consider all the cause they integrate all criteria into a single
significant effects of their decisions. overall score for ranking alternatives. The
The impacts often ignored can be classi analytic hierarchy process (AHP) devel
fied as quantitative or qualitative. Nonfi oped by Saaty [1980] is one such scoring
nancial quantitative impacts can be mea method. It has recently gained popularity
sured in physical terms but cannot easily and been applied to a wide variety of deci
be converted to dollars. Such nonmonetary sion problems.
measures of performance include through The basic structure of the AHP is well
put, cycle time, and scrap. In contrast, suited to economic decisions regarding au
nonfinancial qualitative impacts cannot
even be measured in physical terms. They Multi-attribute decision
typically involve issues that affect the
methods account for both
long-term strategic position of the firm,
such as product quality, manufacturing
financial and nonfinancial
flexibility, and technology advancement. impacts.
Ignoring these often beneficial, yet intangi
ble, effects of automation may bias deci toma ted manufacturing investments. It can
sions against new manufacturing technolo incorporate into a unified analytic frame
gies. To evaluate automation technologies, work a wide variety of normally incom
managers need a method that includes all mensurable impact criteria, some of which
relevant impacts. may be financial, some nonfinancial yet
A Multi-Criteria Approach quantitative, and some purely qualitative.
Approaches that include more than one The AHP appeals to decision makers be
measure of performance in the evaluation cause they participate directly in the pro
process are called multi-attribute or multi cess and can generally understand how re
criteria decision methods [Canada and sults are derived.
Sullivan 1989; Gass 1985; Goicoechea, Researchers are beginning to apply the
Hansen, and Duckstein 1982; Sullivan AHP to automated manufacturing invest
and Liggett 1988; Zeleny 1982]. The ad ment decisions. Arbel and Seidmann
vantage of these methods is that they can [1984], Canada and Sullivan [1989], and
account for both financial and nonfinancial Varney, Sullivan, and Cochran [1985] di
impacts. Some multi-criteria methods do rectly applied the AHP to sample manu
not permit trade-offs among all decision facturing investment decisions. They all

INTERFACES 23:4 76

This content downloaded from 14.139.108.34 on Wed, 07 Sep 2022 10:53:24 UTC
All use subject to https://about.jstor.org/terms
MODIFIED AHP

used benefit-to-cost ratios (BCR) and the applications is the failure to preserve valid
nine-point integer scale (1 to 9) of relative quantitative performance data. Many of
importance for evaluating pairwise com the criteria important to automation deci
parisons, both originally suggested by sions are quantified, though not always in
Saaty [1980, pp. 54 and 115-118]. Frazelle dollar terms. These data must be preserved
[1985] used the same integer scale to de and their informing power brought to bear
cide among material handling technolo on the decision. Most of these recent AHP
gies. Wabalickis [1988] applied the AHP to applications force decision makers to aban
evaluate a flexible manufacturing system don such crucial data. A better approach
(FMS). Troxler and Blank [1989] borrowed would be to combine and integrate reliable
the hierarchy concept from the AHP to il quantitative data with the informed quali
lustrate one decision model for manufac tative judgments of the decision maker.
turing investments, although they fail to Thus, the decision would use actual data
describe how to implement the AHP with directly without resorting to the pairwise
real data. comparison judgment method when rating
Applying the AHP to Manufacturing alternatives with respect to quantitative
Three problems arose in these recent criteria. Judgments would be used only
AHP applications to manufacturing auto when reliable data reflecting the perfor
mation decisions. First, the integer scale mance of the alternative were not available
originally proposed by Saaty [1980, p. 54] for a particular criterion. Wabalickis [1988]
is easily misused. For example, the scale does suggest using data on inventory costs,
calls for the value 3 to indicate "weak im although he offers no clear rationale for his
portance" and "slight" preference of A method of integrating the cost data with
over B. The eigenvalue method Saaty rec the judgment scores.
ommends computes this comparison into The third problem concerns the use of
weights of 0.75 for A and 0.25 for B, in the BCR approach for deciding among mu
strict proportion to the integer value en tually exclusive alternatives. Saaty [1980,
tered. Thus, the weakly more important A pp. 115-118] illustrates this approach for a
gets a score three times as large as that of
B. The verbal definitions and explanations
of the integer scale lead decision makers to
The integer scale proposed by
violate the critical principle of proportion Saaty is easily misused.
ality in pairwise judgments. Pairwise com
parison values expressing a slight prefer decision among alternative transportation
ence should more properly be less than systems for crossing a river. All the studies
1.5, a ratio resulting in weights of 0.6 for cited earlier except Frazelle [1985] and
A and 0.4 for B. Indeed, most comparison Wabalickis [1988] use this BCR approach,
values in manufacturing decisions will which has separate hierarchies for benefit
likely be expressed as nonintegers in the elements and for cost elements. The deci
1.0 to 3.0 range. sion maker evaluates each hierarchy to de
The second problem that arose in these rive benefit scores and cost scores and then

July-August 1993 77

This content downloaded from 14.139.108.34 on Wed, 07 Sep 2022 10:53:24 UTC
All use subject to https://about.jstor.org/terms
WEBER

forms BCRs for each alternative. The alter?Buy a new computer numerical control
native with the highest BCR is best. This (CNC) mill with an interactive graphic
BCR approach is unsuitable for the most controller and computer-aided design
common type of automation decision?se(CAD); and
?Replace the mill with a machining center
lecting the best alternative among mutually
exclusive options. The option with the and programmable tool changer.
highest BCR may have only a moderate The general approach of the AHP is to
benefit with a very low cost. Since decisionstructure such a complex decision into a
makers may select only one option, they hierarchy of its elements (criteria and alter
might easily forego a preferable option natives). The particular AHP hierarchy I
with greater net benefits. The proper ap used to evaluate the machine shop alterna
proach is to construct a single consolidated tives has three levels (Figure 1). Three cat
hierarchy including both advantageous egories seem sufficient for this decision:
(benefit) and disadvantageous (cost) ele monetary (financial) criteria, engineering
ments, instead of using the two separate performance measures, and strategic con
hierarchies to form a ratio. siderations. Monetary criteria include ini
A Machine Shop Decision tial cost, operation and maintenance costs,
I modified the AHP to address and solve and training costs. Engineering perfor
these problems using four basic steps. mance criteria are production rate, cycle
First, specify the evaluation criteria and thetime, setup time, and scrap. Strategic con
investment alternatives to be evaluated. siderations are manufacturing flexibility,
Second, make pairwise comparisons be product quality, business growth, and im
tween criteria to establish their weights. plementation risk. Table 1 shows the data
Third, rate the investment alternatives used in rating the alternatives.
with respect to the criteria, using the pair At each level in the hierarchy, the deci
wise comparison process to rate the quali sion maker establishes scores among ele
tative criteria. For quantitative criteria, ments by constructing a matrix of pairwise
however, use measured performance data,
such as cost, setup time, or throughput.
The AHP allows for and deals
Fourth, combine the ratings with criteria
weights into an overall rating for each in explicitly with inconsistencies
vestment alternative and then rank the al among pairwise comparison
ternatives accordingly. judgments.
I have applied the modified AHP to a
typical machine shop decision. A shop comparison judgments about relative dif
manager considering alternatives to an exferences between any two elements. The a?j
isting conventional milling machine faces value of the matrix represents the relative
several options: importance or preference of the fth ele
?Retrofit with a power feed, digital readment over the ;th element. Both reflexivity
out of positions, and numerical control (a? = 1 along the diagonal) and reciprocity
(NC) features; (Uij = 1/ttji) are assumed. Thus, only values

INTERFACES 23:4 78

This content downloaded from 14.139.108.34 on Wed, 07 Sep 2022 10:53:24 UTC
All use subject to https://about.jstor.org/terms
MODIFIED AHP

CATEGORIES CRITERIA ALTERNATIVES


(Level 1) (Level 2) (Level 3)
Initial Cost

O&M Cost

Training

Production Rate

Cycle Time

Setup Time

Scrap

Flexibility

Quality

Growth

Risk

Figure 1: The analytic hierarchy process hierarchy for the machine shop decision shows cate
gories of criteria on the left (Level 1), criteria themselves in the middle (Level 2), and alterna
tives on the right (Level 3).

above the diagonal must be entered, so sistencies among the pairwise comparison
that for n elements, one needs n(n?1)/2 judgments. For example, A need not be
distinct comparison judgments. four times as important as C just because
I used the pairwise comparison method A is twice as important as B and B is twice
to establish weights for the criteria given in as important as C. In technical terms, this
the top two levels of the machine shop hi means that the user is not forced to pre
erarchy. First I compared the categories serve cardinal transitivity among compari
(Level 1) with each other, then the criteria sons. The AHP even provides an explicit
(Level 2) within each category. Because measure of inconsistency that indicates
there are three categories, I needed only how far the pairwise judgments deviate
three [ = 3(3?1)/2] comparison judgments from perfect cardinal transitivity. The mea
on Level 1 (Table 2). sure can suggest whether to revisit pair
A major strength of the AHP is that it wise comparisons.
allows for and deals explicitly with incon In the modified AHP discussed here, the

July-August 1993 79

This content downloaded from 14.139.108.34 on Wed, 07 Sep 2022 10:53:24 UTC
All use subject to https://about.jstor.org/terms
WEBER

Categories Alternatives to be Evaluated


Criteria Units Add NC CNC + CAD Machining Center

Monetary Criteria
Initial Cost 12,000 25,000 120,000
O & M Cost 2,000 4,000 15,000
Training 3,000 8,000 20,000
Performance Measures
Production Rate units/day 14 40
Cycle Time days 10 6 2
Setup Time minutes 30 20 3
Scrap kilograms/day 200 75 20
Strategic Considerations
Flexibility [Not quantified]
Quality [Not quantified]
Growth [Not quantified]
Risk [Not quantified]

Table 1: For the three monetary criteria and the four performance measures, data are available
to rate the three alternatives of the machine shop decision. The four strategic criteria require
the pairwise comparison method, since they are not quantified.

category weights and inconsistency ratio inconsistency index is the difference be


are both computed by methods recom tween the largest eigenvalue and the num
mended by Saaty [1980]. Thus, I found ber of elements, divided by the number of
weights by normalizing the eigenvector of elements minus one. It turns out that the
the largest eigenvalue of the complete ma largest eigenvalue of a matrix of perfectly
trix of pairwise comparison values above consistent pairwise comparisons (that is, a
the diagonal, their reciprocals below the matrix that preserves cardinal transitivity
diagonal, and ones along the diagonal. The
throughout) equals the number of ele

Monetary Performance Strategic

Monetary 1.500 0.500


Performance 0.250
Strategic

Weights taken from eigenvector of largest eigenvalue:


Monetary Performance Strategic
0.263 0.159 0.578
Inconsistency Ratio: 0.008

Query posed to solicit comparison values


"For the overall decision, the row catego
category?"
Table 2: Pairwise comparison judgments, such as "The monetary criteria are 1.5 times as im
portant as the performance criteria/' establish the category weights. Inconsistency among
comparisons less than 0.10 is not significant.

INTERFACES 23:4 80

This content downloaded from 14.139.108.34 on Wed, 07 Sep 2022 10:53:24 UTC
All use subject to https://about.jstor.org/terms
MODIFIED AHP

ments. Thus, the numerator of the incon comparison method is similar to the
sistency index measures how far from per weighting procedure except the query
fect consistency the matrix is. This index is posed for positive criteria becomes: "With
then divided by a similar index based on respect to this criterion, the row alternative
randomly selected pairwise comparisons to is how many times as desirable as the col
form the inconsistency ratio. Saaty sug umn alternative?" When rating alternatives
gests that acceptable levels of consistency with respect to negative criteria (such as
are assured as long as this ratio does not costs), the query becomes: "A is how
exceed 0.10. Table 2 shows the weights many times as desirable (in terms of being
and inconsistency ratio computed for the more available or easier to obtain) as B?"
categories (Level 1). Thus if A costs twice as much as B, A
To compare criteria within the perfor should be rated as one half as desirable as
mance category (Level 2), which in this B. The eigenvector of the largest eigen
case has four elements, I had to make six value then represents the ratings of each
comparisons (Table 3). The procedure is alternative for the criterion in question.
similar to that illustrated in Table 2. I re One must use this pairwise comparison
peated the process for the other two cate method to rate alternatives whenever data
gories. Then I multiplied the criterion that are directly or inversely proportional
weight within each category by the cate to the desirability of alternatives are un
gory weight to get a combined or global available.
weight for each criterion. If proportional data (or other data, such
After I established all the weights, I as exponential or logarithmic, that can eas
needed to rate the alternatives (Level 3) ily be transformed into proportional scales)
with respect to each criterion using either are available, they may be used directly.
the pairwise comparison method or the For example, production rate data are di
proportional data method. The pairwise rectly proportional if 200 units per day are

Production Rate Cycle Time Setup Time Scrap


Production Rate 0.500 2.000 3.000
Cycle Time 2.000 6.000
Setup Time 4.000
Scrap
Weights taken from eigenvector of largest eigenvalue:
Production Rate Cycle Time Setup Time Scrap
0.274 0.448 0.208 0.070
Inconsistency Ratio: 0.036

Query posed to solicit comparison values:


' 'Within performance criteria, the row criterion
criterion?"

Table 3: Within performance criteria, pairwise comparison judgments, such as "The produc
tion rate is half (0.5) as important as cycle time," establish the criteria weights. Inconsistency
among comparisons less than 0.10 is not significant.

July-August 1993 81

This content downloaded from 14.139.108.34 on Wed, 07 Sep 2022 10:53:24 UTC
All use subject to https://about.jstor.org/terms
WEBER

judged to be twice as desirable as 100 Alternative Overall Weighted Ratin


units. Setup time is inversely proportional
Machining Center 0.385
if a 40-minute setup is judged half as de AddNC 0.318
sirable (twice as bad) as a 20-minute setup. CNC with CAD 0.297
Since performance data must be propor
Table 5: The machining center alter
tional to desirability, the pairwise comparithe highest overall weighted rating.
son method must be used for negative and
zero values. For directly proportional data,tion with AutoMan 2.0, microcomp
the ratings of the alternatives are given bysoftware that incorporates my mo
the normalized vector of the data. For in AHP approach and is designed spe
versely proportional data, the ratings are to support investment decisions ab
given by the normalized vector of the re tomated manufacturing equipment
ciprocals of the data. I can best explain theLippiatt, and Johnson 1991]. AutoM
rationale for using reciprocals through an users combine up to 49 quantitati
example. A process that can accomplish aqualitative criteria in evaluating in
task in half the time is twice as desirable, alternatives. Quantitative criteria c
since it can complete twice as many tasks clude such traditional financial me
in the same time. Table 4 shows the case life-cycle cost and net benefits as
of inversely proportional data for the ini such engineering performance mea
tial cost criterion. throughput and setup time. Qualit
Once I established the weights and rat teria could include flexibility and p
ings, I combined them to compute an overquality.
all weighted rating for each alternative I developed AutoMan at the National
(Table 5). This overall rating equals the Institute of Standards and Technology
sum of the products of the criteria global with support from the US Navy Manufac
weights times the criteria ratings. The ma turing Technology Program. It has been
chine shop manager should select the al purchased by over 800 industrial decision
ternative with the highest overall rating, makers since its first release in August
the machining center. 1989. AutoMan includes sample decision
AutoMan Software models and a user manual with tutorial,
I formulated the decision model and carglossary of evaluation criteria, bibliogra
ried out all computations for this illustra phy, and index. Bennett and Neises [1989],
Owen [1991], and Stout, Libera tore, and
Alternative Initial Cost ($) Rating
Monahan [1991] describe the software and
illustrate how to apply it.
AddNC 12,000 0.633 Conclusion
CNC with CAD 25,000 0.304 The modified AHP and the AutoMan
Machining Center 120,000 0.063
software are appropriate for team decisions
Table 4: When rating alternatives with re
requiring input from senior managers, ac
spect to cost criteria, take reciprocals of the
data before normalizing the vector so the countants, financial analysts, and engi
least costly alternative is rated highest. neers. Senior managers can structure the

INTERFACES 23:4 82

This content downloaded from 14.139.108.34 on Wed, 07 Sep 2022 10:53:24 UTC
All use subject to https://about.jstor.org/terms
MODIFIED AHP

decision into a hierarchy of its elements Christoph Witzgall, and two anonymous
and establish priorities among the ele reviewers.
ments; accountants and financial analysts References
can develop financial data for the invest Arbel, A. and Seidmann, A. 1984, "Perfor
mance evaluation of flexible manufacturing
ment alternatives; and engineers can pro
systems," IEEE Transactions on Systems, Man
vide the nonfinancial data and judgments and Cybernetics, Vol. SMC-14, No. 4 (July
on the performance of alternatives. Users August), pp. 606-617.
must be aware that adding or deleting an Bennett, Robert E. and Neises, Steven J. 1989,
"AutoMan?Decision support software,"
alternative from an AHP hierarchy may
Management Accounting, Vol. 71, No. 5 (No
change the ranking of the original set of vember), pp. 58-60.
alternatives. This phenomenon, called rank Berliner, Callie and Brimson, James A., eds.
reversal, is considered a weakness by util 1988, Cost Management for Today's Advanced
Manufacturing: The CAM-I Conceptual Design,
ity theorists, such as Dyer [1990], but is Harvard Business School Press, Boston,
well defended by such AHP theorists as Massachusetts.
Harker and Vargas [1987; 1990] and Saaty Canada, John R. and Sullivan, William G. 1989,
Economic and Multiattribute Evaluation of Ad
[1987; 1990]. I support the AHP view that
vanced Manufacturing Systems, Prentice Hall,
rank reversal is possible in rational deci Englewood Cliffs, New Jersey.
sions. Choobineh, Fred and Suri, Rajan, eds. 1986,
Several benefits result from using Flexible Manufacturing Systems: Current Issues
AutoMan to make decisions on automation and Models, Industrial Engineering and Man
agement Press, Norcross, Georgia.
investments. The software facilitates deci
Drucker, Peter F. 1990, "The emerging theory
sion making with its sample decision mod of manufacturing," Harvard Business Review,
els, glossary of criteria, facility for building Vol. 68, No. 3 (May-June), pp. 94-102.
Dyer, James S. 1990, "Remarks on the analytic
custom models, and automatic computa
hierarchy process," Management Science, Vol.
tion of results. Reliable financial analyses 36, No. 3 (March), pp. 249-258.
and performance data are fully preserved Frazelle, Ed 1985, "Suggested techniques enable
multi-criteria evaluation of material handling
and have direct impact on the final deci
alternatives," Industrial Engineering, Vol. 17,
sion. Decision makers can include impor No. 2 (February), pp. 42-48.
tant nonfinancial quantitative performance Gass, Saul I. 1985, Decision Making, Models, and
data as well as qualitative informed judg Algorithms, John Wiley and Sons, New York.
Goicoechea, Ambrose; Hansen, Don R.; and
ments. All judgments are automatically
Duckstein, Lucien 1982, Multiobjective Deci
checked for internal consistency. Resulting sion Analysis with Engineering and Business
decisions are based on more complete in Applications, John Wiley and Sons, New
York.
formation than permitted by traditional
Harker, Patrick T. and Vargas, Luis G. 1987,
financial analysis.
"The theory of ratio scale estimation: Saaty's
Acknowledgments analytic hierarchy process," Management Sci
The research was supported by the US ence, Vol. 33, No. 11 (November), pp. 1383
1403.
Navy Manufacturing Technology Program.
Harker, Patrick T. and Vargas, Luis G. 1990,
For their helpful comments, I thank Saul
"Reply to 'Remarks on the analytic hierarchy
Gass, Mary Haight, Harold Marshall, process' by J. S. Dyer," Management Science,
Frederic Murphy, Mary Ellen Weber, Vol. 36, No. 3 (March), pp. 269-273.

July-August 1993 83

This content downloaded from 14.139.108.34 on Wed, 07 Sep 2022 10:53:24 UTC
All use subject to https://about.jstor.org/terms
WEBER

Jelinek, Mariann and Goldhar, Joel D. 1986, ceedings, pp. 181-190.


"Economics in the factory of the future," CI M Wabalickis, Roger N. 1988, "Justification of
Review, Vol. 2, No. 2 (Winter), pp. 21-27. FMS with the analytic hierarchy process,"
Owen, Jean V. 1990, "Flexible justification for Journal of Manufacturing Systems, Vol. 7, No.
flexible cells," Manufacturing Engineering, Vol. 3, pp. 175-182.
105, No. 9 (September), pp. 39-45. Weber, Stephen F.; Lippiatt, Barbara C; and
Owen, Jean V. 1991, "Justifying manufacturing Johnson, Katherine S. 1991, AutoMan 2.0: De
flexibility," Manufacturing Engineering, Vol. cision Support Software for Automated Manu
106, No. 3 (March), pp. 65-71. facturing Investments?User Manual, NISTIR
Padmanabhan, Srikanth 1989, "A tandem ex 4543, National Institute of Standards and
pert support system as justification for a flexi Technology, Washington, DC. [The AutoMan
ble manufacturing system," Journal of Manu 2.0 software and User Manual can be ob
facturing Systems, Vol. 8, No. 3, pp. 195-205. tained for $55 plus $3 handling from the Na
Saaty, Thomas L. 1980, The Analytic Hierarchy tional Technical Information Service, Spring
Process: Planning, Priority Setting, Resource Al field Virginia 22161, (703) 487-4650. Request
location, McGraw-Hill, New York. PB91-506568 (5.25" diskette) or PB91
Saaty, Thomas L. 1987, "Rank generation, pres 507194 (3.5" diskette).]
ervation, and reversal in the analytic hierar Zeleny, Milan 1982, Multiple Criteria Decision
chy process," Decision Sciences, Vol. 18, pp. Making, McGraw-Hill, New York.
157-177.
Saaty, Thomas L. 1988, Multicriteria Decision
Making: The Analytic Hierarchy Process, Uni
Barry Davis, Program Manager, Indus
versity of Pittsburgh, Pittsburgh, Pennsylva trial Modernization, Austin Operations,
nia. Tracor Inc., 6500 Tracor Lane, Austin,
Saaty, Thomas L. 1990, "An exposition of the
Texas, 78721, writes, "\ have recently used
AHP in reply to the paper 'Remarks on the
analytic hierarchy process/ " Management Sci your AutoMan software program version
ence, Vol. 36, No. 3 (March), pp. 259-268. 2.0 to aid in selection of a prototype ma
Stout, David E.; Liberatore, Matthew J.; and chine from three divergent concepts. The
Monahan, Thomas F. 1991, "Decision sup
program provided a method of placing a
port software for capital budgeting," Manage
ment Accounting, Vol. 73, No. 1 (July), pp. more objective view on the results of an
50-53. evaluation team's grading of the different
Sullivan, William G. 1986, "Models IEs can use
concepts. The team's original scoring
to include strategic, non-monetary factors in
automation decisions," Industrial Engineering,
methods provided a linear weighted evalu
Vol. 18, No. 3 (March), pp. 42-50. ation with results very heavily selecting
Sullivan, William G. and Liggett, Hampton R. one machine. AutoMan's analytical hierar
1988, "A decision support system for evaluat
chy process showed the design concepts to
ing investments in manufacturing local area
networks," Manufacturing Review, Vol. 1, No. be more closely competitive. It did how
3 (October), pp. 151-157. ever, confirm the original recommendation.
Troxler, Joel W. and Blank, Leland 1989, "A "The program was easy to use. This ease
comprehensive methodology for manufactur
of use allowed me to recompute results af
ing system evaluation and comparison," Jour
nal of Manufacturing Systems, Vol. 8, No. 3, ter applying various alternative conditions.
pp. 175-183. The sensitivity analyses plots were very
Varney, Mark S.; Sullivan, William G; and helpful with analysis of areas that possibly
Cochran, Jeffrey K. 1985, "Justification of have more risk."
flexible manufacturing systems with the ana
lytical hierarchy process," Annual Interna
tional Industrial Engineering Conference Pro

INTERFACES 23:4 84

This content downloaded from 14.139.108.34 on Wed, 07 Sep 2022 10:53:24 UTC
All use subject to https://about.jstor.org/terms

You might also like