This article was downloaded by: [Stanford University Libraries]

On: 23 September 2012, At: 20:09

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

IIE Transactions
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/uiie20

An Economic Order Quantity Model with Demand-

Dependent Unit Production Cost and Imperfect
Production Processes
a
T. C. E. CHENG
a
Senior Member IIE Department of Actuarial & Management Sciences, University of
Manitoba, Winnipeg, Manitoba, R3T 2N2, Canada

Version of record first published: 31 May 2007.

To cite this article: T. C. E. CHENG (1991): An Economic Order Quantity Model with Demand-Dependent Unit Production Cost
and Imperfect Production Processes, IIE Transactions, 23:1, 23-28

To link to this article: http://dx.doi.org/10.1080/07408179108963838

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,
proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.
 An Economic Order Quantity Model
with Demand-Dependent Unit
Production Cost and
Imperfect Production Processes
T. C. E. CHENG
Senior Member lIE
Department of Actuarial & Management Sciences
University of Manitoba
WiMipeg. Manitoba. Canada R3T 2N2
Downloaded by [Stanford University Libraries] at 20:09 23 September 2012

Abstract: The classical economic order quantity (EOQ) model assumes that items produced are of perfect quality
and that the unit cost of production is independent of demand. Product quality is not always perfect but directly
affected by the reliability of the production process used to produce the products. In addition, a relationship be-
tween unit production cost and demand may exist under certain circumstances. We propose an EOQ model with
demand-dependent unit production cost and imperfect production processes. We formulate this inventory decision
problem as a geometric program (GP) and solve it to obtain closed-form optimal solutions. An illustrative example
is provided to demonstrate the point that GP has potential as a valuable analytical tool for studying a certain class
of inventory control problems. We also discuss the aspect of sensitivity analysis based on the GP approach .

• Broadly defined, inventory is the stock of any item a
company keeps, be it a physical product or a service, to
be used in the company's output. Inventories are generally
regarded as necessary evils because the lack of synchro-
nization in the production system makes holding inventory
a necessity, yet keeping a high level of inventory is a costly
exercise. Representing a significant asset to a business op-
eration, inventories are used to serve a variety of functions,
chief of which are: (1) coordinating operations, (2) smooth-
ing production, (3) achieving economies of scale, and (4)
improving customer service.
Inventory planning and control is essentially concerned
with the design, operation and control of an inventory sys-
tem. While design involves the specification of inventory
system procedures, operation and control of an inventory
system are primarily concerned with answering two funda-
mental questions: (1) What should the order (production)
quantity be? (2) When and how frequent should an order
(production) be placed (initialized)?
The problem of determining the most desirable order quan-
tity under rather stable conditions is commonly known as
the classical. economic order quantity (EOQ) inventory prob-
Received August 1988; revised March 1989. Handled by the Department
of Inventory.
lem. Over the years an abundance of research on this topic
has been undertaken and many interesting results have been
reported in the literature: e.g. Clark [4], Nahmias [12), Sil-
ver [18], Urgeletti Tinarelli [20], Veinott [21], Wagner [22]
and Whitin [25].
Two major assumptions in the classical EOQ model are
that demand is constant and deterministic and that the unit
price (unit production cost) is independent of the order (pro-
duction) quantity (Silver and Peterson [l9}). Implicitly, the
model also assumes that items produced are of perfect qual-
ity (Hax and Candea [7] and Johnson and Montgomery [II)).
However, in reality, product quality is not always perfect
but a function of the reliability of the production process
employed to manufacture the products. Consequently, in-
vestment in improving the reliability of the production proc-
ess is the key to achieving consistently a high level of product
quality. However, this will increase the fixed cost of pro-
duction which will in turn push up the unit production cost
since the costing system, particularly an absorption costing
system commonly used in batch manufacturing, will inevi-
tably apportion the increased production overheads to each
individual product (Horngren [8]). Among researchers who
have considered the issues of process reliability and quality
improvement in EOQ problems are Cheng [2], Porteus [15],
Rosenblatt and Lee [16] and Silver [17].
In addition, under certain circumstances, the demand rate

March 1991, lIE Transactions, Volume 23, Number I 0740-817X/91/$3.00x.00 © 1991 "lIE" 23
 for a product and its unit production cost are not totally in-
dependent. Such a relationship between demand and unit
cost exists in two typical situations. First, in a manufactur-
ing environment; when demand is high, a company can pro-
duce more items so as to spread the fixed cost of production
more widely, which will result in lower unit production cost.
Second, in a distribution environment; if sufficient demand
exists, a firm can arrange for longer-term supply contracts
with the suppliers to take advantage of a quantity discount
on the unit price. Thus it is quite natural to expect that the
unit cost of production is an increasing function of process
reliability and a decreasing function of demand rate.
In this paper we propose to model the relationship between
unit production cost and process reliability and demand rate
as a general power function. The objective is to miniIilize
the long-term average annual cost, i.e. the sum of setup,
Downloaded by [Stanford University Libraries] at 20:09 23 September 2012
unit production and inventory carrying costs over a long
period of time, which is a function of process reliability,
demand rate, and production quantity and the relevant cost
parameters in the present case. Wagner and Whitin [23] have
shown that minimizing the long-term average annual cost,
in many cases, is mathematically equivalent to minimizing
the total cost per unit time which is generally a simpler per-
formance measure to deal with.
The differential calculus is a natural analytical tool for
solving this type of optimization problem involving contin-
uous decision variables. However, due to the general nature
of the power function used to model the relationship between
process reliability and demand and unit production cost, op-
timization of the resulting cost function by calculus will often
lead to a system of non-linear equations which, in general,
are hard to solve explicitly so that numerical methods are
needed to obtain approximate solutions. Consequently,
closed-form optimal solutions to this particular type of EOQ
problems are not easy to obtain from using the calculus-
based optimization technique.
A geometric programming (GP) approach to the solution
of the EOQ problem is presented in this paper. We show
that a closed-form optimal solution is readily obtained from
GP. This is due to the fact that the objective function of
the EOQ problem to be minimized consists of posynomiaLs
and GP provides an efficient solution method for optimiz-
ing posynomials. Another major advantage of using GP is
that it allows easy post-optimality sensitivity analysis to be
performed (Beightler et al. [1]). While the use of GP for
solving operations research problems is not new (Federo-
wicz and Mazumdar [6], Jefferson [9], Passy [13], Petro-
poulos [14] and Wall et al. [24]), it appears that there are
only a handful of papers concerned with the solution of in-
ventory problems by GP (Cheng [2,3] and Worrall and Hall
[26]).
This paper is organized as follows. The next section
presents the model with process reliability and demand de-
pendent unit production cost and the underlying assumptions.
The following section derives the optimal solution to the
EOQ problem using the geometric programming solution
24
technique. A numerical example is then solved to illustrate
how the closed-form optimal solution to a problem is de-
rived, followed by a section on discussing the aspect of post-
optimality sensitivity analysis. In the final section some con-
cluding remarks are presented.
Model and Assumptions
Consider the case of a company manufacturing a single
product for which demand is in excess of supply. The mar-
ket is able to absorb virtually any quantity of the product
rolled out from the production line. This situation is typical
of a successful technologically-advanced product entering
the growth phase of its life cycle. Since demand for the prod-
uct is high, the manufacturer will increase production to
meet the demand which will result in lower unit cost of pro-
duction for each product because production overheads are
spread among more items.
As for the manufacture of the product, the company uses
a production process with a certain level of reliability. The
process reliability depends on a number of factors such as
machine capability, use of on-line monitoring devices, the
skill of the operating personnel and maintenance and replace-
ment policies. Higher reliability means products with accept-
able quality are more consistently produced by the process,
thereby reducing the costs of scrap and rework of substand-
ard products, wasted materials and labour hours. However,
high reliability can only be achieved with additional costs,
both fixed and variable, which will eventually push up the
unit cost of production.
To construct a model for this inventory problem, we de-
fine the following variables.
S = setup cost per batch,
H = inventory carrying cost per item per unit time,
D = demand rate (a decision variable),
q = production quantity per batch (a decision var-
iable),
r = production process reliability (a decision var-
iable),
V(D,r) unit cost of production,
C(D,q,r) total cost of setup, production and inventory
holding per unit time.
We make the following basic assumptions about the model:
(1) Production is instantaneous,
(2) No back-order is allowed,
lIE Transactions, March 1991
 (3) Demand for the product exceeds supply, C(D,q,r) = (Total cost per cycle)/(qr/D)
(4) The unit cost of production is directly related to process subject to
reliability and inversely related to demand rate accord-
.ing to the following general power function: r:s1

V(D,r) = aD-hrc (I) which, after substituting (1) and (2), becomes

(here a, band c are non-negative real numbers to be

chosen to provide the best fit of the estimated cost func-
tion.) subject to

The first two assumptions are the basic assumptions used r :s I . (3)
in the classical EOQ model. Because of the third assump-
tion, that demand is in excess of supply, the manufacturer Geometric Programming Solution
is free to choose a preferred level of production, since the
Downloaded by [Stanford University Libraries] at 20:09 23 September 2012

market guarantees to snap up any quantity of the product To solve the inventory optimization problem (3) by GP,
produced. Thus production volume equals demand rate which we first let
becomes a decision variable in this problem. The fourth as-
sumption follows partly from the third assumption, for high U. SDq-'r' , (4)
demand justifies a higher volume of production which re-
aD-(b-')r~t
duces the fixed cost per unit, thus resulting in lower unit U2 (5)
production cost. In addition, since qr is the number of usa-
ble products per batch, higher r means more "good" UJ Hqr/2 , (6)
products will be available therefore more capable of absorb-
ing the fixed charge within the batch. However, to reduce U. r. (7)
the cost of scrap and rework on shoddy products, invest-
ment in improving the reliability of the production process Using the fact that for a given set of positive real numbers
is necessary which gives rise to higher unit production cost. the arithmetic mean is greater than or equal to the geomet-
It is easy to see from (1) that iW(D,r)/iJD < 0 and ric mean, we write the following:
iJV(D,r)/iJr > 0 which are in accordance with the assumed
relationships between unit cost of production and demand
and process reliability. In addition, the use of power func-
tions to estimate production costs is common practice in both (8)
manufacturing and chemical industries (Jelen and Black [10]).
In particular, the exponential forms of power functions are where
widely used because they offer the advantage of being plot-
ted as straight lines on log-log graph papers which greatly
simplifies the estimation process.
As with the classical EOQ model, the production cycle To take account of the process reliability constraint, we
is implicitly assumed to repeat indefinitely throughout an write
infinite planning horizon and so we can focus our analysis
on one typical production cycle. A process reliability level (10)
of r means of all the items produced in a production run
only r percent are of the acceptable quality that can be used where
to meet demand. It is thus evident that the length of a pro-
duction cycle is qr/D. The total cost incurred in a produc- if u. < I

I
(1,00)
tion cycle is the sum of setup, production and inventory Q'.E
carrying costs. That is ( - 00 , 00) if u. = I .

Total cost per cycle ':" S + V(D,r)q + Htfr/2D. (2) Multiplying (8) and (10) yields

Our objective is to minimize the total cost per unit time C(D,q,r) = u. + U2 + UJ

subject to the constraint that process reliability cannot ex-

ceed 100%. So we wish to minimize . (11)

March 1991, TIE Transactions 25

 In OP terminology C(D,q,r) and ¢(a) are known as the ities, all (u/a;)(i = 1,2,3) in (8) must be equal and u"a4 =
primal and dual functions respectively and a = (a. a1 a3 I in (10). Hence, from (21), for the minimal C(D,q,r), all
(4) is a vector of dual variables (Duffin et al. [5]). (u'/a,*)(i = 1,2,3) are equal. Let
Substituting (4), (5), (6) and (7) into (11), we get
(U/a,*) = k for i = 1,2,3 (22)
¢(a) =
and

U4
a4' = 1
(12)
or
Since a is arbitrary, it can be chosen according to con-
venience subject to (9) and (10). We choose a such that the (23)
exponents of D, q and r are zero thus making the right hand
side of (12) independent of the decision variables. To do Substituting (22) and (23) into (11), we obtain
this we require
= k = ¢(a*)
Downloaded by [Stanford University Libraries] at 20:09 23 September 2012

k"1 +,,2+,,3 . (24)

(13)
Hence, combining (24) and (22) we get
(14)
u, = a,*¢(a*) for i = 1,2,3 . (25)

Substituting (4), (5) and (6) into (25) and taking natural
These are called the orthogonality conditions, which, together logarithms of the resulting equations, we obtain three linear
with the normality condition (9), are sufficient to determine equations in three variables as follows
a uniquely. Solving (9), (13), (14) and (15) yields
In(H/2) + Inq + lnr =In¢(a*) + In[(b-I)/(2b-I»), (26)
a.* = (b-I)/(2b-l), (16)
InS + InD -lnq -Inr =Inq,(a*) + In[(b-I)/(2b-I») , (27)
(17)
Ina - (b-I)lnD + (c-I)lnr = Inq,(a*) + In[l/(2b-l)). (28)
a3* = (b-I)/(2b-l) , (18)
Solving equations (26)-(28) for D, q and r yields the opti-
a4* = (l-c)/(2b-l) . (19) mal values of the decision variables:

In view of constraint (9) which requires a,*(i = 1,2,3) > (29)

0, we see from (16)-(18) that either 0 :$ b < 1/2 or b >
I must hold for the unit production cost function (1). With q* = 2(b-I)¢(a*)[(2b-I)Hr*r· , (30)
a* determined, (11) is reduced to

= {a(b-I)[SHI2)b-·[(2b-I)I(b-I)]U-.p/(1b-.). (20)
Since equality is possible for (20),
min C(D,q,r) = max q,(a*) = q,(a*) . (21)
This is so because ¢(a*) is a constant in the present case.
Thus the original constrained optimization problem (3) has
been reduced to a mere evaluation of the quantity q,(a*) as
determined from (20) where all a/* (i = 1,2,3,4) are de-
termined from equations (16)-(19).
To find the optimal values of D, q and r, we proceed as
follows. Since the optimal solution requires equality in (II)
which is possible if and only if (8) and (10) are also equal-
An Illustrative Example
For a particular EOQ problem, the relevant cost param-
eters are given as S = $IOO/batch and H = $l!itemlday.
In addition, it is estimated that a = 5000, b = 2 and c =
1/2 in (1). Thus the unit production cost takes the following
form:
(32)
The three decision variables are D. q and r whose optimal
values are to be determined to minimize the total cost per
unit time, that is:

26 lIE Transactions, March 1991

 subject to (41)

r S 1. (33) The new optimal values of demand, production quantity

and process reliability are then easily determined using
To find the optimal demand D*, production quantity q* (29)-(31), with ¢",(a*) replacing ¢(a*).
and process reliability r*, we use (16)-( 19) to determine a, * While changes in cost coefficients are more frequent, oc-
(i = 1,2,3,4) as follows: casionally changes in the exponents b and c of the unit pro-
duction cost function (1) may be required. These changes
(34) give rise to a new set of orthogonality conditions (13)-(15),
which, together with the normality condition (9), are suf-
(35) ficient to determine the new dual variables a,*, Substituting
these a,* values into the modified dual function ¢",(a)
Thus the corresponding cP(a*) as determined from (20) has yields the new minimum cost C",(D,q,r) = cP",(a*). It is
the value: then easy to determine the optimal demand D*, the produc-
tion quantity q* and the process reliability r* with the aid
cP(a*) ,= 3(aSH/l)1/3 = $189/day . (36) of equations (29)-(31). These equations should also have
Downloaded by [Stanford University Libraries] at 20:09 23 September 2012

been modified to take account of the new values of the ex-

It follows from (29)-(31) that the optimal values of demand,
production quantity and process reliability are respectively:
D* = (2/SH)[cP(a*)/3]1 = 80/day , (37)
q* = 2¢(a*)/[3Hr*] = 129/batch , (38)
r* = {3aIlD*¢(a*)]}1 = 98% , (39)
and the minimum total cost is
min C(D,q,r) = cP(a*) = $189/day , (40)
, problems. We also illustrate that the GP approach permits
It is noted that this optimal solution can also be obtained
using the calculus-based constrained optimization technique
(see Appendix).
Post-Optimality Sensitivity Analysis
According to the theory of GP, the values of the dual var-:
iables a,* represent the fractions of the minimum cost
C(D,q,r) accounted for by the respective terms in the ob-
jective function of the minimization problem (3). The der-
ivation of these a,* involves solution of the simultaneous
linear equations (9) and (13)-(15). These equations are in-
dependent of the coefficients of the total cost function S,
H and a. Thus the fractions are unaltered for changes in
these coefficients. The a,* only change if at least one ex-
ponent in the cost function (1) is changed. This property
makes finding a new optimal solution for a change in S,
H and a very easy. For instance, let the setup cost S be
changed by an amount of .:1 percent. The new minimum cost
C",(D,q,r) = ¢",(a*) can be calculated as follows:
or
ponent factors b and c,,
Conclusion
In this paper we first propose to use a general power func-
tion to model the relationship between unit production cost
and demand rate and process reliability. We then formulate
the EOQ problem as a GP and apply the theories of GP to
help derive the optimal solution in closed-form. We dem-
onstrate through an illustrative example the power and ef-
ficiency of GP compared to calculus as an analytical tool
for the analysis of this particular type of inventory decision
simple and easy sensitivity analysis to be performed on the
optimal solutions.
Acknowledgements
The author is thankful to Professor H. J. Boom and an
anonymous referee for many helpful comments. This re-
search was supported in part by a grant from the Faculty
of Management Associates Fund of the University of Man-
itoba.
REFERENCES
[I) Beightler, C. 5., Phillips, D, T. and Wilde, D. J" Foundations of
Optimization, 31e, Prentice-Hall, New Jersey (1985).
[21 Cheng, T. C. E., "An Economic Production Quantity Model with
Flexibility and Reliability Considerations," European Journal ofOp-
erational Research, 39, 174-179 (1989).
[31 Cheng, T. C. E., "An Economic Order Quantity Model with Demand-
Dependent Unit Cost," European Journal of Operational Research,
40, 252-256 (1989).
[41 Clark, A. J., "An Informal Survey of Multi-Echelon Inventory The-
ory," Naval Research Logistics Quarterly, 19,621-650 (1972).
lSI Duffin, R. J., Peterson, E. L. and Zener, C., Geometric Program-
ming, John Wiley and Sons, New York (1967).
[61 Federowicz, A. J. and Mazumdar, M., "Usc of Geometric Program-
ming to Maximize Reliability Achieved by Redundancy, " Operations
Research, 16, 948-954 (1968).

March 1991, lIE Transactions 27

 [71 Hax, A. C. and Candea, D., Production and Inventory Management.
Prentice-Hall, New Jersey (19&4).
181 Homgren, C. T .• Cost Accounting, a Managerial Emphasis, Prentice-
Hall, New Jersey (1977).
[91 Jefferson, T.• Geometric Programming with an Application to Trans-
portation Planning. Unpublished Ph.D. Thesis, Northwestern Uni-
versity, U1inois (1972).
IIOJ Jelen, F. C. and Black, J. H.• Cost and Optimization Engineering,
McGraw-Hili. New York (1983). q[ -SDq-2r - 1 + Hrl2] = 0 (AS)
IIIJ Johnson, L. A. and Montgomery, D. c., Operations in Production
Planning Scheduling and Inventory Control, John Wiley and Sons,
New York (1974).
1121 Nahmias, S., "Perishable Inventory Theory: A Review," Operations
Research, 30, 680-708 (1982).
[131 Passy, U., "Nonlinear Assignment Problems Treated by Geometric
Programming." Operations Research, 19. 1675-1690 (1971). which, together with the process reliability constraint and
1141 Petropoulos, P. G., "Optimal Selection of Machine Rate Variables its associated complementary slackness condition with the
by Geometric Programming," lnternational Journal of Production
Lagrangian multiplier )..,
Research, II, 305-312 (1973).
Downloaded by [Stanford University Libraries] at 20:09 23 September 2012

115] Porteus, E. L., "Optimal Lot-Sizing, Process Quality Improvement

WId Setup Cost Reduction," Operations Research, 34, 137-144 (1986).
[16] Rosenblatt, M. J. and Lee, H. L., "Economic Production Cycles
with Imperfect Production Processes." lIE Transactions, 14, 48-55
(1986).
[17] Silver, E. A., "Establishing the Order Quantity when the Amount
Received is Uncertain." INFOR, 14, 32-39 (1976).
[18] Silver, E. A., "Operations Research in Inventory Management: a
Review and Critique," Operations Research, 29, 628-645 (1981).
[19J Silver, E. A. and Peterson, R., Decision Systems for Inventory Man-
agement and Production Planning, John Wiley and Sons, New York
(1985).
[20J Urgeletti Tinarelli, G., "Inventory Control: Models and Problems,"
European Journal of Operational Research, 14, 1-12 (1983).
[21] Veinott, Jr, A. F., "The Status of Mathematical Inventory Theory,"
Management Science, 12, 745-777 (1966).
(22) Wagner, H. M., "Research Portfolio for Inventory Management and
Production Planning Systems," Operations Research, 28. 225-475
(1980).
[23) Wagner, H. M. and Whitin, T. M., Analysis of Inventory Systems,
Prentice-Hall, New Jersey (1963).
[24] Wall, T. W., Greening, D. and Woolsey, R. E. D.• "Solving Com-
plex Chemical Equilibrium using a Geometric-Programming based
Technique," Opera/ions Research, 34, 345-355 (1986).
125) Whitin, T. M., "Inventory Control Research: a Survey," Manage-
ment Science, 1,32-40 (1954).
(26] Worrall, B. M. and Hall, M. A., "The Analysis of an Inventory Con-
trol Model using PosynornialGeometric Programming," International
Journal of Production Research, 20, 657-667 (1982).
r-I::::;O (A8)
)..(r-I) = 0 (A9)
can be solved to find the optimal values of the decision var-
iables.
However, to obtain a closed-form optimal solution to this
relatively simple problem by calculus is not easy because
it requires solving a system of non-linear equations and in-
equalities (A2)-(A9). For a more involved problem, for ex-
ample, one in which all the parameters a, b and c are
fractional values, we not only have to rely on GP to sim-
plify the computation but also to obtain a closed-form op-
timal solution.
T. C. Edwin Cheng is Professor of Operations Management in the Faculty
of Management, University of Manitoba. He obtained a B.Sc. (Hons.) in
Industrial Engineering from the University of Hong Kong, an M.Sc. in
Industrial Management from the University of Birmingham and a Ph.D.
in Operations Research from the University of Cambridge. His research
interests are in scheduling, simulation and inventory management. He has
published over fifty papers in various international journals. He is a reg-
istered professional engineer in Manitoba and is a member of CAPIC. UE
and TlMS.

Appendix: Calculus Solution

In order to find the optimal solution to the example prob-

lem using the calculus-based optimization technique, we first
construct the Lagrangian function associated with the prob-
lem as follows:

F(D,q,r) = C(D,q,r) + )..(r-l)

= SDq-lr-t + ao- tr- 1 / 2 + Hqrll + )..(r-I) . (AI)

Differentiating (A I) with respect to D. q and r respectively

yields the following set of Kuhn-Tucker optimality condi-
tions (BeightJer et al. [I]).

28 lIE Transactions, March 1991