Mathematical Modelling of The Effect of Human Learning in The Finite Production Inventory Model

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

Mathematical modelling of the effect of human

learning in the finite production inventory model


Moueen K. Salameh, Mohamed-Asem U. Abdul-Malak, and Mohamad Y. Jaber

American University of Beirut, Beirut, Lebanon

It is a natural phenomenon that the performance of a person engaged in a repetitive task improves with time.
In the manufacturing engineering literature, lx2 this phenomenon is referred to as the learning curve, which is
a decrease in the efSort required for producing each unit in a repetitive manufacturing operation. Factors
contributing to this improved performance include more effective use of tools and machines, increased
familiarity with operation tasks and work environment, and enhanced management eficiency.
Deterministic production inventory models found in the literature 3*4 have been formulated to show the
effects of numerous variables or a combination of variables on model outputs. Variables considered include
stock shortages, safety stock, and level of customer patience. More recently, a finite production inventory
model was formulated taking into consideration the effect of machine unavailability on optimum production
quantity. 5
Because the learning phenomenon is believed to have economic and decision-making implications in an
inventory management system,6 it is the objective of the study presented herein to investigate the effect of
learning on both optimum production quantity and minimum total inventory system cost.

Keywords: learning, production, inventory, order quantity

Methodology man-hour input and cumulative production of the form:2


Consider a case where the effect of learning is evident. tj = ay (3)
The assumption will be that a single item manufactured
where
in batches will be produced at an increasing rate, but it
will be consumed at a constant rate, /I units per unit time. tj = the time to produce thejth unit,
Define TI as the time to build a maximum inventory level a = constant, numerically theoretical time required
of z units and T, as the time required to consume the to produce the first unit, and
inventory achieved by the end of time Tl. The cumulative S = a negative slope average, reflecting the decrease
amount produced during period Tl is defined as Y units. in the production time required per unit.
The inventory level can be expressed as a function of
time, 4(t), as follows: The methodology used in this study is to integrate the
above learning relationship into the finite production
4(t) = Y - Pt for O<t<T, (1) inventory model, and, by so doing, to determine the
optimal order quantity y,, for period c = 1, 2, 3,. . ., etc.,
and that minimizes the total inventory system cost per unit
for T,<tlT, time for a production cycle c. This can be achieved by
4(t) = -Bt + Bto (2)

To, the cycle time, is the sum of T, and T2. Figure I

z>
illustrates the hypothesized variation in the inventory G(t)
level given by expressions (1) and (2).
A typical learning curve would show a progressive z
improvement in productivity, but at a diminishing rate. units
This suggests an exponential relationship between direct

<--T1- > 1<--Tz-> 1 time


Address reprint requests to Dr. Salameh at the American University of
Beirut, Beirut, Lebanon. < TO ‘I
Received 27 July 1992; revised 19 April 1993; accepted 27 April 1993 Figure 1. Variation in inventory level.

0 1993 Butterworth-Heinemann Appl. Math. Modelling, 1993, Vol. 17, November 613
Effect of learning on the finite production inventory model: M. K. Salameh

attempting to trade off labor and material cost, fixed Substituting ayS+ l/s+ 1 for Tl gives

s
setup cost, and inventory holding cost per unit time so
that their sum will be minimum. oTo
4(t) dt = a
s+2y
cs+2) BT:
2
) (Y - PA2
2P -
and by further substitution of T,, the equation is reduced
to
Learning-modified inventory model
Define t(y) as the expected cumulative time to produce
Y units. Using expression (3), t(y) can be expressed as
t(y) = t, + t, + t, + .‘. + ty
Therefore, the holding cost is given by

ay
1
s+2
=a+a2”+a3”+~..+aY”
(S + l)(S + 2)
= a[ 1” + 2” + 3” + . . . + y”]
The expected total system cost equation can then be
= a i X” (4) expressed as
.X=1
lay” + 1
An approximation can be obtained by treating expres- TC(Y) = s+l + d, y + k
sion (3) as a continuous function rather than a discrete

1
one. With suitable limits, t(y) is then given by
rv ,,,s+ 1 (11)
L
t(Y)=
0 J
a_?(” dX =
s+l
(5)
and the expected total cost per unit time (TCU(y))
can be obtained by dividing expression (10) by the cycle
Solving for y, the following is obtained:
time To as shown:

Y=
[
s+l t (lP+l)
a 1
For 0 < t 5 Tr and by substitution in expression (l), the
(6) TCU(y) = TC(y)/T, = TC(y) f

Therefore,
expected inventory level while in production can be
written as
TCU(y) = z +d,P+kg+F
Y
O(t) = s:-’ t (l’s+l)
- pt (7)
[ 1 ha$?y” + ’
(12)
As is customary, K is the setup cost ($/production - (S + l)(S + 2)
cycle), h is the inventory carrying cost ($/unit/unit time), To show the convexity of the total cost function, the
d, is the material cost ($/unit), and 1 is the labor cost first and second derivatives are obtained as shown below:
($/unit time). The expected total cost to produce y units
in the first cycle (TC(y)) is the sum of the expected dTCU(y) Slafiy”-l _ kp + h
production cost (PC(y)) and the expected inventory cost T’(Y) =
dY = S+l y2 2
(HC(y)). PC(Y) is expressed as
PC(y) = It +d,y+K -~ WY” (13)
(S+ 2)
and by substituting t from expression (5), PC(y) becomes
d2TCU(y) S(S - 1)1~/3y”-~
sly”+ 1 T”(y) = +2kp
PC(Y) = s+l + drny+ K dy2 = s+1 Y3
_ Shajy”- ’
HC(y) is expressed as (14)
s+2
HC(y) = h x (average inventory level per cycle)
Because it appears from expression (14) that T”(y) > 0
The average inventory level per cycle is the area under for all values of y, TCU(y) is therefore a unimodal
the inventory function shown in Figure 1. It can be function with its minimum at y = y’, where T’(y*) = 0.
expressed as follows:

s+1(Ii.7
+1)
Qs+w+u
~~n~(~)dt~~o=‘[[S:_!t~‘s+il-~t]dt+~

p-f
p
Model verification

=L-1 a s+2 -7+2p


In this section, the Newton-Raphson
solve the nonlinear,
method is used to
algebraic total cost model expressed
in (12). The usefulness of the model is thus demonstrated
S+l in a numerical example.

614 Appl. Math. Modelling, 1993, Vol. 17, November


Effect of learning on the finite production inventory model: M. K. Salameh

Numerical example through 9. The tabulated results indicate that the opti-
mum production quantity levels off after the seventh
To illustrate the solution procedure, an inventory situa-
cycle, which is a stage that corresponds to the maximum
tion with the following parameters is considered:
learning on this production operation. Also included in
Time to produce the first unit a= 0.0625 days Table 2 are the results computed for T,, the time required
Consumption rate p = 12 units/day to produce y units, and z, the cycle maximum inventory
Direct material cost d, = $lOO/unit level. It is to be noted that the TI values decrease with
Labor cost I= $lO/hr each additional cycle the same way the optimum quanti-
Slope constant s = -0.1 ties decrease, whereas the z values increase with each
Holding cost h= 0.2 $/unit/day additional cycle until they level off. The decrease in the
Fixed setup cost k= $200 optimum quantity and required production time and the
increase in the maximum inventory level are more drastic
It is required to determine the optimum production between the first and second cycles than in the later ones.
quantities for cycles 1,2, 3, . . . . Finally, the effect of learning is most evident when
Using expressions (12) and (13), the optimum produc- comparing the optimum production quantity for cycle 1
tion quantity for cycle 1, obtained by applying the computed without considering the learning effect with
Newton-Raphson method, is yT = 216 units as shown in that computed using the learning-modified model. In this
Table 1. The time required to produce the first unit in example, the comparison reveals a decrease in the opti-
the second cycle, unit number 217, can then be calculated mum production quantity of about 30%, calculated as
as follows:
t 217 = 0.0625 (217)-O.’ = 0.0365 days
y* (no learning) = [$+J”
By reapplying the Newton-Raphson method using an
a value of 0.0365 days, the optimum production quantity
for the second cycle is found to be 184 units. Table 2 (15)
shows the optimum quantities computed for cycles 1
where r (the replenishment rate) = I
a

Table 1. Computation of optimum = 110.0625 = 16


yfor cycle 1

Y T’(Y) T”(Y)
Substituting

y*
in expression (15) gives
2(200)12 112
20
30
-5.99
-2.65
-1.17
0.60
0.18
0.05
(no learning) =

The ratio of 2161310 represents


0.2(1 - 12/16) 1 = 310 units

a decrease of about 30%.


45
66 -0.51 0.02
97 -0.21 0.01
136 -0.08 0.00 Summary
177 -0.03 0.00
206 -0.01 0.00 This paper presents a modified production inventory
215 -0.00 0.00 model that incorporates the effect of learning, which may
216 -0.00 0.00 be achieved during production, on the inventory system
216 -0.00 0.00 total cost per unit time and optimum production quant-
216 -0.00 0.00
ity. The modified model is described, and its usefulness
is demonstrated through a numerical example.

Table 2. Optimum quantities


References
Time to Fabrycky, W. J. and Torgersen, P. E. Operations Economy.
produce Maximum Prentice-Hall, Englewood Cliffs, NJ, 1966
Cycle c Alpha y* Y units inventory 2 Jelen, F. C. Cost and Optimization Engineering. McGraw-Hill,
New York, 1970
1 0.0625 216 8.750 111 Aggarwal, S. C. A review of current inventory theory and its
2 0.0365 184 4.425 131 applications. ht. J. Production Rex No. 12, 1974
3 0.0343 182 4.118 132 Hax, A. C. and Cardea, D. Production and Inventory Management.
4 0.0331 180 3.943 133 Prentice-Hall, Englewood Cliffs, NJ, 1984
5 0.0322 180 3.822 134 Abboud, N. E. and Salameh, M. K. How machine unavailability
6 0.0315 179 3.731 134 affects optimal production quantity for the finite production in-
7 0.0310 178 3.657 135 ventory model. Production and Inventory Management. 1987, 28(l)
8 0.0305 178 3.596 135 Peterson, R. and Silver, E. A. Decision Systems for Inventory
9 0.0301 178 3.544 135 Management and Production Planning. John Wiley & Sons, New
York, 1979

Appt. Math. Modelling, 1993, Vol. 17, November 615

You might also like