Вы находитесь на странице: 1из 8

EUROPEAN

JOURNAL
OF OPERATIONAL
RESEARCH
ELSEVIER Journal of Operational Research 88 (1996) 320-327
Theory and Methodology
Project management with time, cost, and quality considerations
A . J . G . B a b u a . , , Na l i n a S u r e s h b
a W. Paul Stillman School o f Business, Seton Hall University, South Orange, NJ 07079-2692, USA
t, University o f Wi sconson- Eau Claire, W! 54702, USA
Received December 1993; revised March 1994
Abstract
In the project management literature, quantitative models were developed for project crashing to determine the
appropriate activities for crashing at minimal cost. In this paper, we suggest that the project quality may be affected
by project crashing and develop linear programming models to study the tradeoffs among time, cost, and quality.
Each of the three models developed optimizes one of these entities by assigning desired bounds on the other two.
An illustrative example with a project network consisting of 13 nodes, 14 activities, and 2 dummy activities is
provided. The computational study includes tabulation of the interrelationships among time, cost, and quality.
Ke y wo r d s : Project management; Resource allocation; Optimization; Critical path programming; Linear program-
ming
1. Int roduct i on
A pr oj ect is a col l ect i on of activities t o ac-
compl i sh a specific obj ect i ve. Pr oj ect manage-
me nt i nvol ves pr oj ect pl anni ng, moni t or i ng, and
cont r ol . Pr oj ect pl anni ng i ncl udes defi ni t i on of
wor k speci fi cat i on, det er mi nat i on of quant i t y of
wor k, and est i mat i on of r esour ces r equi r ed. Suc-
cessful pr oj ect ma na ge me nt i nsures t he com-
pl et i on of pr oj ect in t i me, wi t hi n budget , and t o
t he pr oj ect speci fi cat i ons. It is faci l i t at ed by t he
i dent i fi cat i on and successful appl i cat i on of me-
t hodol ogy f or t r adeof f anal yses. Whi l e t i me - c os t
t r adeof f anal yses wer e devel oped in t he l i t era-
t ur e, t he aut hor s ar e not awar e of any publ i shed
* Corresponding author.
quant i t at i ve r esear ch t hat consi ders t i me - c o s t -
qual i t y t r adeof f s in pr oj ect ma na ge me nt simulta-
neousl y. Thi s paper devel ops opt i mi zat i on mod-
els t o consi der t hese t r adeof f s.
Eve n t hough pr oj ect ma na ge me nt t echni ques
have t r adi t i onal l y been ut i l i zed in bot h const ruc-
t i on and def ense r el at ed pr oj ect s, its appl i cabi l i t y
has been ext ended t o such di verse fields as
sof t war e devel opment , mai nt enance pl anni ng,
space vehi cl e depl oyment , and compl ex surgery.
A l arge pr oj ect may be br oke n down i nt o several
activities. The pr ecedence- r el at i onshi ps among
t hem may be vi sual i zed in a net wor k r epr esent a-
t i on of t he pr oj ect wher e arcs de not e activities.
The critical pat h me t hod ( CPM) is a f undament al
quant i t at i ve t echni que devel oped f or pr oj ect
management . Assumi ng det er mi ni st i c activity
0377-2217/96/$09.50 1996 Elsevier Science B.V. All rights reserved
S S DI 0377-2217(94)00202-9
A.J.G. Babu, N. Suresh / European Journal of Operational Research 88 (1996) 320-327 321
completion times, CPM determines the minimum
time needed to complete the project.
Generally, there is a due date for project com-
pletion and one needs to expedite the project.
This implies completing some of these activities
faster than normal. For each activity, normal time
of completion and crash time of completion are
determined and correspondingly normal and
crash costs for the activity are estimated. Crash
completion of an activity may involve overtime
charges, specialized resources, and faster wear
and tear of equipment. Thus crashing an activity
saves time but increases the cost. A decision
problem considered in the project management
literature is to determine the activities for crash-
ing and the extent of crashing [12]. By assuming
that the direct cost of an activity varies with time,
mathematical programming models were de-
veloped to minimize the project direct cost for a
given project completion time.
In this paper, we suggest that the quality of
a completed project may be affected by project
crashing. For simplicity, we adapt the continuous
scale from Zero to One to specify quality attained
at each activity. The overall project quality is a
function of quality levels attained at the individ-
ual activities. For illustration, three simple func-
tions of defining this relationship, the arithmetic
mean, the geometric mean, and the minimum are
considered in this paper. The actual functional
form depends on both the problem scenario and
the definition of quality adapted. Optimization
models involving these project t i me-cost quality
tradeoffs are developed.
Section 2 briefly summarizes the existing litera-
ture on project crashing. Section 3 develops the
optimization models and an illustrative example
is presented in Section 4. Both the Sections 3 and
4 consider the average of quality levels attained
at the activities as the overall project quality ob-
jective. Section 5 extends these models to include
two other functional forms of aggregating the
quality levels attained at the activities, the geo-
metric mean and the minimum. Section 6 con-
cludes this paper with a summary and impli-
cations for further research.
2. Literature review
Kelly [8] presents a linear programming ap-
proach for project crashing assuming that cost
varies linearly with activity completion time. Re-
lated work is also presented in [10] and [6]. Sev-
eral researchers [3-5,9] have developed models
and solution procedures to incorporate a non-
linear relationship between cost and activity com-
pletion time. If the nonlinear relationship is con-
vex, a nonlinear programming model could be
formulated that can be solved easily since every
local optimal solution is a global optimal solution.
For an appreciation of related work in
PERT/ CPM, the reader is referred to Wiest and
Levy [12]. The authors are not aware of any work
in the literature that has studied time, cost, and
quality simultaneously as three objectives. Similar
models in the context of software development
were developed by the authors in an earlier paper
[11]. Portions of the results of this paper are pre-
sented in WDSI conference [1]. This paper illus-
trates the concept in a general project setting
using linearity assumptions. More complex re-
lationships could be modeled in a similar fashion
to capture reality more closely.
3. Mathematical programming models
In this section we develop linear programming
models that would assist in expediting a project,
weighing t i me-cost -qual i t y trio. Each of the
three models developed optimizes one of these
three entities by assigning desired levels (bounds)
on the other two. A project is divided into several
activities that are related by precedence relation-
ships. Each activity has a normal time of com-
pletion and a crash time of completion. Associ-
ated with the normal time are normal cost and
normal quality, and with crash time are crash
cost and crash quality. We develop the notation,
requirements, and models below.
K
I ndi ces:
for events (nodes in the project network),
K=l , 2 . . . . . M.
322 A.J.G. Babu, N. Suresh / European Journal of Operational Research 88 (1996) 320-327
L
LCD
for activities (arcs in the project network),
L =1 , 2 . . . . , N.
for dummy activities (dummy arcs in the
project network), LCD = N + 1,
N+2 . . . . . N +D.
Scalars:
M Number of events (nodes).
N Number of activities (arcs).
D Number of dummy activities (dummy arcs).
Data:
NTI ME( L)
CRTI ME( L)
NCOST (L)
CRCOST (L)
NQUAL( L)
CRQUAL( L)
Normal time for activity L.
Crash time for activity L.
Normal cost for activity L.
Crash cost for activity L.
Normal quality level for activity
L.
Crash quality level for activity L.
Decision variables:
X( L) activity completion times.
Auxiliary variables:
Y( K) Earliest time for event (node in the project
network) K.
Objectives:
OBJT Crashing time objective.
OBJC Crashing cost objective.
OAVQ Crashing (average) quality objective.
Additional notation:
S(L) Slope of activity cost curve for activity L:
S(L) = NCOST (L) - CRCOST (L)
NTI ME( L) - CRTI ME( L) "
I NTERCPT( L) Intercept of activity cost curve
for activity L:
I NTERCPT (L)
-- CRCOST (L) - S(L) CRTI ME( L) .
SQ(L) Slope of activity quality curve for activity
L:
SQ(L) -- NQUAL( L) - CRQUAL( L)
NTI ME( L) - CRTI ME( L)
I NTERCPTQ( L) Intercept of activity cost curve
for activity L:
I NTERCPTQ( L)
= CRQUAL( L) - SQ(L) CRTI ME( L) .
(I, J) Arc L starts from node I and ends in
node J.
UBTME Upper bound for project completion
time.
UBCST upper bound for project direct cost
LBQAV Lower bound for project (average)
quality.
Equations and inequalities:
OBJFNT Project time objective function:
(OBJFNT)
Y( ' N' ) = OBJT.
OBJFNC Project cost objective function:
(OBJFNC)
SUM(L, INTERCPT (L) + S(L) X( L) )
= OBJC.
OBJAVQ Project quality objective function:
(OBJAVQ)
~4(SUM(L, I NTERCPTQ( L) + SQ(L)
X( L) ) ) = OAVQ.
ETIME Earliest time constraint for each arc (in-
cluding dummy arcs) joining from node I to node
J:
(ETIME)
Y( ' I ' ) + X( L) - Y( ' J' ) ~< O.
This constraint is written for each arc L: Sum of
the earliest time at event I and activity completion
t i me o f acti vi ty L i s l es s t han or equal t o t he
earliest time at event J,
T( ' I ' ) - Y( ' J' ) ~< 0.
This constraint is written for each dummy arc"
Since the activity completion time of a dummy
activity is zero, we write the corresponding con-
straint as: Earliest time at event I is less than or
equal to the earliest time at event J,
A .J.G. Babu, N. Suresh I European Journal of Operational Research 88 (1996) 320-327 323
Y(' I' ) = O,
since we can start on first node at time-zero.
LBTACT Lower bound constraint for each ac-
tivity completion time:
(LBTACT)
X( L) >! CRTI ME( L) .
UBTACT Upper bound constraint for each ac-
tivity completion time:
( UBTACT)
X( L) <~ NTI ME( L) .
UBTI ME Upper bound constraint for project
completion time:
(UBTIME)
OBJT ~< UBTME.
UBCOST Upper bound constraint for project di-
rect cost:
(UBCOST)
OBJC ~< UBCST.
LBQUAV Lower bound constraint for project
quality:
(LBQUAV)
OAVQ i> LBQAV.
Model s deoel oped
1. Model MINTME-AV:
Minimize OBJT subject to OBJFNT, ETI ME,
LBTACT, UBTACT, UBCOST, and LBQUAV.
2. Model MINCST-AV:
Minimize OBJC subject to OBJFNC, ETI ME,
LBTACT, UBTACT, UBTI ME, and LBQUAV.
3. Model MAXQAV-AV:
Maximize OAVQ subject to OBJAVQ, ETI ME,
LBTACT, UBTACT, UBTI ME, and UBCOST.
4. Illustrative example
In this section, we illustrate t he models de-
veloped in the previous section with an example.
This example is an extension of a time-cost trade-
offs example from Hillier and Li eberman [7]. The
data are given below. An analysis of the relation-
ships among the objective functions are presented
by summarizing several comput er runs of the
models.
K
L
LCD
Indi ces:
for events (nodes in the project network),
K= 1, 2 . . . . . 13.
for activities (arcs in the project network),
L =l , 2 . . . . . 14.
for dummy activities (dummy arcs in the
project network), LCD = 15, 16.
M= 13
N= 14
D=2
Scalars:
Number of events (nodes).
Number of activities (arcs). These are
(1, 2), (2, 3), (3, 4), (4, 5), (4, 6), (4, 7),
(5, , 7), (6, 8), (7, 9), (8, 10), (9, 11),
(9, 12), (10, 13), and (12, 13).
Number of dummy activities (dummy
arcs). These are (5, 8) and (11, 12).
Dat a:
NTI ME( L) Normal time in work days:
{2, 4, 10, 4, 6, 7, 5, 7, 8, 9, 4, 5, 2, 6}.
CRTI ME( L) Crash time in work days:
{ 1 , 2 , 7 , 3 , 4 , 5 , 3 , 4 , 6 , 6 , 3 , 3 , 1,3}.
NCOST (L) Normal cost in dollars:
{800, 3200, 6200, 4100, 2600, 2100, 1800, 9000,
4300, 2000, 1600, 2500, 1000, 3300}
CRCOST (L) Crash cost in dollars:
{2300, 3600, 7300, 4900, 3000, 2400, 2200, 9600,
4600, 2500, 1800, 3000, 1500, 4000}.
NQUAL( L) Normal quality:
{ 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 , 1 } .
CRQUAL( L) Crash quality:
{0.9, 0.7, 0.8, 0.8, 0.7, 0.7, 0.3, 0.75, 0.4, 0.5,
0.9, 0.95, 0.9, 0.95}.
Comput at i onal experi ence
1. Model MINTIME-AV: This model is solved
with a combination of values for UBCOST
($45 500 through $52 500 in increments of $500;
and a last value of $52 700) and LBQUAV (0.73,
0.8, 0.85, 0.9, 0.95 and 0.98). The optimal project
completion times are present ed in Table 1 and
Fig. 1. For low values of budget, the project com-
324 A. J. G. Babu, N. Suresh / European Journal of Operational Research 88 (1996) 320-327
Table 1
Optimal project completion time when cost and OAVQ are
bounded
Upper bound
on project
c o s t
Lower bound on project quality (OAVQ)
0.73 0.8 0.85 0.9 0.95 0.98
45500
46000
46500
47000
47500
48000
48500
49000
49500
50000
50500
51000
51500
52000
52500
52700
44 44 44 44 44 44
41 41 41 41 41.05 41.56
38.5 38. 5 38. 5 38.51 38.96 39.49
36.91 36.91 36.91 36.98 37.31 38.08
35.55 35.55 35.55 35.64 36.03 37.08
34.25 34.25 34.25 34.38 34. 8 36.95
33 33 33.02 33.25 33.91 36.95
31.78 31.78 31.87 32.31 33.88 36.95
30.7 30. 7 30. 93 31.72 33.88 36.95
29.77 29.77 30.26 31. 5 33. 88 36.95
29 29 29.96 31.39 33.88 36.95
28.57 28.69 29.76 31.39 33.88 36.95
28.13 28.53 29. 7 31. 39 33.88 36.95
28 28.46 29. 7 31. 39 33.88 36.95
28 28.46 29. 7 31. 39 33.88 36.95
28 28.46 29. 7 31. 39 33.88 36.95
pl et i on t i me is hi gher and is affect ed onl y mar-
ginally by the quality r equi r ement . As the budget
is i ncreased, t he proj ect compl et i on time is re-
duced and depends heavi l y on the quality requi re-
ment ; hi gher qual i t y r equi r ement s demand l onger
proj ect compl et i on times.
2. Model MI NCST- AV: This model is solved
with a combi nat i on of values for UBTI ME (28
t hr ough 44 in i ncr ement s of 2 wor k days) and
LBQUAV (0.73, 0.8, 0.85, 0.9, 0.95, and 0.98).
The opt i mal proj ect di rect costs are pr esent ed in
Tabl e 2 and Fig. 2. When t he proj ect time al l owed
is l ower, t he cost is enor mous and coul d onl y
accommodat e modest qual i t y requi rement s. As
t he pr oj ect t i me al l owed is i ncreased, t he proj ect ' s
direct cost is r educed and is able t o accommodat e
for mor e st ri ngent r equi r ement s on quality.
3. Model MAXQAV- AV: This model is sol ved
with a combi nat i on of values for UBTI ME (28
t hr ough 44 in i ncr ement s of 2 wor k days) and
UBCOST ($45 500 t hr ough $52 500 in i ncrement s
of $500; and a last val ue of $52 700). The opt i mal
proj ect (average) qual i t y levels are pr esent ed in
4 5
4 4
4 3
4 2
41
P
~ 4o
o 39
~j
f 99
P- 3 7
g 66
- 34
8 33
32
~" ~ o
2g
28
27
V
x
V
i v
i
V V V V V V V V V V V
X X
i
i
X X X X X X X X
& & & A A & & A
g
o
g 0 0 0 0 0 0
[]
4 6 ~ o o 1 4 6 ~ o o I 4 v ~ o o 1 4 . ~ o o 1 4 9 ~ o o I s o ~ o o I s , e o o I s = ~ o o I
4 6 0 0 0 4 7 0 0 0 4 8 0 0 0 4 9 0 0 0 5 0 0 0 0 5"1000 5 2 0 0 0 52" 700
P r o J e c t C o e t C u p p e r boundS)
Q OAVQ. GE. 0 . ' 7 3 + 0AVQ. GE. 0 . 6 0 0 OAVQ. GE. 0 . 8 5 ~. 0AVQ. GE. 0 . 9 0
x 0~, VG. GE. 0 . 9 5 v OAVO. GE. 0 . 9 8
Fi g. 1. Op t i ma l pr oj ec t c o mp l e t i o n t i me when cost and O A V Q ar e bounded.
A.J.G. Babu, N. Suresh / European Journal of Operational Research 88 (1996) 320-327
Tabl e 2
Op t i ma l pr oj ect cost w h e n t i me a n d O A V Q a r e b o u n d e d
325
Up p e r b o u n d Lower b o u n d on pr oj ect qual i t y ( OAVQ)
on pr oj ect 0. 73 0. 8 0. 85 0. 9 0. 95 0. 98
compl et i on t i me
28 51650 I NFEA a I NFEA I NFEA I NFEA I NFEA
30 49850 49850 50400 I NFEA I NFEA I NFEA
32 48900 48900 48938 49221 I NFEA I NFEA
34 48100 48100 48100 48158 48437 I NFEA
36 47333 47333 47333 47367 47514 I NFEA
38 46600 46600 46600 46620 46735 47040
40 46200 46200 46200 46200 46250 46373
42 45800 45800 45800 45800 45800 45894
44 45500 45500 45500 45500 45500 45500
a I n f e a s i b i l i t y .
6 '
, , j o
Wf-
o
D
~ 3
5 2
51
5 0
4 9
4 E
4 7
4 6
4 5
A
D

/
I I I I I I
2 9 3 0 3 2 3 4 3 6 3 9
i
i i
I I I
4 0 4 2 4 . 4
0 A V Q. GE . 0 . ' 7 3
P r o j e c t C o m p l e t l o n T l m e C u ~ a e r b o u n d )
+ 0 A V Q. GE . 0 . 8 0 O OAVQ. GE. 0 . 8 5
X 0 A V O. GE . 0 . 9 5 V OAVO. GE. 0 . 9 8
,'~ 0 AVQ. GE. 0 . 9 0
Fig. 2. Opt i ma l pr oj ect cost w h e n t i me a n d O A V Q a r e bounde d.
Table 3. Allocating higher budgets and allowing
for longer project times are congenial for at-
taining higher (average) quality levels.
5. Additional models and computational
experience
In Section 1 we have noted that the overall
project quality is a function of quality levels at-
tained at the individual activities and have ex-
plored the arithmetic mean form in Sections 3-
4. In [2], we report additional computational ex-
perience with two other functional forms, the ge-
ometric mean and the minimum. We summarize
the computational experience below.
The trends observed for both minimum quality
and geometric mean quality are similar to those
for average quality. In addition we note that for
a given budget, the optimal time is in the nonin-
creasing order corresponding to the same bound
326 A.J.G. Babu, N. Suresh I European Journal of Operational Research 88 (1996) 320-327
Table 3
Optimal project quality (OAVQ) when cost and time are bounded
Upper bound Lower bound on project time
on
project 28 30 32 34 36 38 40 42 44
cost
45500 INF a INF INF INF INF INF INF INF 1.0
46000 INF INF INF INF INF INF INF 0.998 1.0
46500 INF INF INF INF INF INF 0.995 0.998 1.0
47000 INF INF INF INF INF 0.979 0.995 0.998 1.0
47500 INF INF INF INF 0.947 0.979 0.995 0.998 1.0
48000 INF INF INF INF 0.973 0.985 0.995 0.998 1.0
48500 INF INF INF 0.952 0.973 0.985 0.995 0.998 1.0
49000 INF INF 0.867 0.952 0.973 0.985 0.995 0.998 1.0
49500 INF INF 0.910 0.952 0.973 0.985 0.995 0.998 1.0
50000 INF 0.835 0.916 0.952 0.973 0.985 0.995 0.998 1.0
50500 INF 0.852 0.0917 0.952 0.973 0.985 0.995 0.998 1.0
51000 INF 0.860 0.917 0.952 0.973 0.985 0.995 0.998 1.0
51500 INF 0.861 0.917 0.952 0.973 0.985 0.995 0.998 1.0
52000 0.778 0.861 0.917 0.952 0.973 0.985 0.995 0.998 1.0
52500 0.779 0.861 0.917 0.952 0.973 0.985 0.995 0.998 1.0
52700 0.779 0.861 0.917 0.952 0.973 0.985 0.995 0.998 1.0
a Infeasibility.
on minimum quality, geometric mean quality,
and average quality respectively. For a given due
date, the optimal cost is in the nonincreasing
order corresponding to the same bound on mini-
mum quality, geometric mean quality, and aver-
age quality respectively. The optimal values of
average quality, minimum quality, and geometric
mean quality are in the nonincreasing order cor-
responding to each pair of bounds on cost and
time.
6. Summary and concl usi ons
In this paper we suggest ed that the project
quality shoul d al so be taken i nt o consi derat i on
al ong with t he traditional practice o f consi deri ng
t he t i me - c o s t tradeoffs in project expedi t i on de-
ci si ons. We present ed a me t hodol ogy t o offer de-
ci si on support in crashing a project. Wi t h si mpl e
linearity assumpt i ons, we devel oped easi l y sol v-
abl e mat hemat i cal programs. The project quality
is as s umed by aggregati ng quality l evel s attai ned
at t he activities in three different ways usi ng arith-
met i c mean, geomet ri c mean, and t he mi ni mum
functi ons. The sol ut i on o f t hese mode l s not onl y
support our i ntui ti ons regarding the i ntertwi ned
effects of time, cost, and quality in project man-
agement, but also offer quantitative results. The
models devel oped in this paper could be easily
ext ended to accommodat e nonlinear relation-
ships.
References
[1] Babu, A. J. G. , and Suresh, N., "Project crashing: Mod-
eling and optimization", in: WDSI Conference Proceed-
ings, March 1994, 751-755.
[2] Babu, A. J. G. , and Suresh, N., "Quality in project man-
agement ", Working Paper, W. Paul Stillman School of
Business, Seton Hall University, South Orange, NJ,
1994.
[3] Berman, E. B. , "Resource allocation in a PERT network
under continuous activity t i me-cost functions", Manage-
ment Science 10/4 (1964) 734-745.
[4] Falk, J. E. , and Horowitz, J. L. , "Critical path problems
with concave cost -t i me curves", Management Science
19/4 (1974) 446-455.
[5] Fulkerson, D. , "A network flow computation for project
cost curves", Management Science 7/2 (1961) 167-178.
[6] Goyal, S. K. , "A note on a simple CPM time-cost trade-
off algorithm", Management Science 21/6 (1975) 718-
722.
[7] Hillier, F. S. , and Lieberman, G. J. , Introduction to
Mathematical Programming, McGraw-Hill, New York,
1990.
A.J.G. Babu, N. Suresh / European Journal of Operational Research 88 (1996) 320-327 327
[8] Kelly, J., "Critical-path planning and scheduling: Mathe-
matical basis", Operations Research 9/3 (1961) 296-320.
[9] Meyer, W.L., and Shaffer, L.R., "Extensions of the
critical path method through the application of integer
programming", Department of Civil Engineering, Uni-
versity of Illinois, 1963.
10] Siemens, N., "A simple CPM time-cost trade-off algo-
rithm", Management Science 17/6 (1971) B354-B363.
[11] Suresh, N., and Babu, A. J. G. , "Software development:
Enhancing quality and cost-effectiveness", in: Proceed-
ings of the International Conference on Software Quality,
ASQC, 1993.
[12] Wiest, J.D., and Levy, F.K., A Management Guide to
PERT/CPM, 2nd ed., Prentice-Hall, Englewood Cliffs,
NJ, 1977.

Вам также может понравиться