8/8/2019 Probabilistic Inventory Models Raw Materials
1/37
Probabilistic inventory models of
raw materials
Topic objectives
Understand the nature of uncertainty in inventory
management Select and apply the appropriate probabilistic inventory
model
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
2/37
Uncertainties
In reality, demands many not be predictable with relative ease.
Average demands
can
be
adequate
in
some
situations,
but
considering the demands as a random variable can provide far more accura e an reasona e resu s.
Probabilistic inventory models assign a probability function to the demand variable which is characterized b a mean and variance.
When the demand is considered as a random variable, pro uc on can ace wo s ua ons, e er over s oc o ems or under stock.
Both of these situations are not desirable.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
3/37
Single period stochastic model A plant uses a certain raw material that is used to
manufacture its product.
The raw material is usable for a single period only, any leftovers are discarded.
The plant incurs a cost for any unit of demand that is not satisfied.
. This model is commonly known as the newsvendor model and
newsboy model.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
4/37
Define the following costs
cu : shortage cost (underage cost) The demand D is assumed to be a continuous nonnegative
random variable with a probability density function f(x) and
cumulative function
F(x).
be purchased at the beginning of the period.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
5/37
Define G(Q, D) to be the total underage and overage costs at the end of the period.
If it happens Q > D, positive inventory is realized; if Q < D, a shortage happens:
inventory = max(Q D, 0)
shortage = max(D
Q,
0) ,
G(Q, D) = c0 max(Q D, 0) + cu max(D Q, 0)
== )(),()),(()( dx x f xQG DQG E QG
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
0
8/8/2019 Probabilistic Inventory Models Raw Materials
6/37
Substitute in:
+=
0 0)()0),max(0),max(()( dx x f x QcQ xcQG
u
+=00
0 )(0),max()(0),max( dx x f x Qcdx x f Q xc u
Notice max(Q D, 0) > 0 only when Q > D and max(D Q, 0) > 0 only when D > Q:
Q
+=
Qu x x x c x x xc
00
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
+= Qu
Qu dx x f Qcdx x xf cdx x xf cdx xQc )()()()(
00
00
8/8/2019 Probabilistic Inventory Models Raw Materials
7/37
To prove G(Q) is convex:
)()()(0
0 =
dx x f cdx x f cdQ
QdG uQ
)(
))(1()(2
=QGd
Q F cQ F c uo
0)()(
02
+=
=
Q f cc
ccdQ
uo
u
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
8/37
The minimizer Q* of G(Q) is obtained by setting the first derivative of G(Q) to zero:
uo
c F ccQ F cQ F c=+
=0*
0*))(1(*)(
uo
u
ccc
Q F +=*)(
Since F(Q*) is the probability that the demand does not
* + u o u satisfying the demand during the period if Q* is ordered.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
9/37
The probability of shortage happening is
)(1)(1)( 0 Q F dx x f dx x f Q == The probability of overstock happening is
Q
0
x x =
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
10/37
Illustration
A laban producer buys a certain additive which has a shelf life of 40 days.
The demand for laban is uncertain, and hence manufacturing consumes this additive in uncertain amounts.
The total consumption during 40 days is normally distributed with mean = 11.73 tons and standard deviation = 4.74.
, amount of the additive can be sold for 1,000 riyals per ton as animal feed.
Each ton of the additive can generate a profit of 7,500 riyals. How many tons of the additive should be purchased at the
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
11/37
The cost of unused ton of the additive is 2,500 1,000 = 1,500 riyals, hence c = 1,500 riyals per ton.
The shortage cost is the opportunity cost; hence cu
= 7,500 2,500 = 5,000 riyals per ton.
The critical ratio is equal to 5,000/(1,500+5,000) = 0.77.
The laban producer
should
purchase
enough
tons
of
the
. .
From the normal distribution curve, the area under the curve and to the left of Q* is 0.77.
From the standard normal tables, z = 0.74. Using the standard normal distribution, z = (Q* )/ .
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
12/37
Therefore,
* =. . .
Q* = 15 tons The probability of shortage is
0.24510.75491)6899.0(174.4
73.11151 ===
F F
The probability of overstock isF(0.6899) = 0.7549
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
13/37
Multiple period stochastic model When multiple periods are considered, excess items incur
inventory holding cost. The multiple period model is used when items have extended
shelf lives (non perishable). In this case, uo is set to the inventory holding cost per unit per
period. * , ,
ordered at the beginning of the following period.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
14/37
Stochastic model with lead time If the lead time is > 0, then the demand during + 1 periods
should be considered. The mean and the standard deviation of the demand should
be replaced by ( + 1) and ( + 1)1/2 .
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
15/37
Continuous review stochastic demand model with lead
This model is formally referred to as lot size reorder point model.
It is a continuous review; when the level of stock reaches R, an order is places, so it is called (Q, R) model.
Manufacturing is vulnerable to shortage only during the lead time (waiting for the order to arrive), hence the total demand durin the lead time is considered to be a random variable.
The lot size Q is selected so that the setup cost and holding cost are minimum and R is selected that will reduce the expec e s or age cos .
Assume the total demand during the lead time has mean and variance 2.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
When the demand rate is , then = .
8/8/2019 Probabilistic Inventory Models Raw Materials
16/37
The expression of cost that is derived for this case is only an approximation; hence the application of the model in real production planning will require making reasonable assumptions.
n e erm n ng , s mo e s approx ma e y e model.
Hence the demand er eriod demand rate is iven b . The level of inventory at the end of the cycle is
s = R The quantity s is called the safety stock.
The total inventory in a cycle is
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
s 2+
8/8/2019 Probabilistic Inventory Models Raw Materials
17/37
The setup cost is K. ,
short during the lead time has to be estimated. Shortage happens when the total demand during the lead
time exceeds R; hence the expected shortage units is
= dx x f R xn )()( The total cost in a cycle is
p : shortage cost per unit
2
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
18/37
The average total cost is
The quantities Q and R are selected in order to minimize G:
n p, 2=
0/ /2/22
Qn pQ K hG
==
2//) (
2
2 hQn p K =+
( ))1(
2 pn K Q
+=
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
19/37
0/))(( =+=
Qdx x f p hG
))(1( = hQ R F p
)(1 = p
F
)2(11 =
p F R
e so u on w requ re era ve y so v ng an un convergence is reached, with any reasonable starting value for Q.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
20/37
The standardized loss function is given by
(t) : standard normal density function
)(1)()()()( z z z dt t z t z L z
==
If D is a normal random variable with and 2, then the
expected number of shortage units is
)()( dx x f R xn R
=
)(
z L
dt t t R z
=
= =
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
21/37
Illustration
A manufacturing plant requires a specific raw material, with a unit cost of 100 riyals.
The vendor delivers the requested raw material quantity in 6 months.
The fixed ordering cost is 500 riyals, and inventory interest rate is 20% annual.
250 riyals per output unit when the raw material is in short.
The demand during the lead time is normally distributed with mean of 100 units and standard deviation of 25 units.
How many units of this raw material should be ordered every
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
22/37
The inventory holding cost h = 0.2 100 = 20 riyals. .
The mean
demand
in
6
months
(lead
time)
is
100
units;
then
the demand rate (the demand per year) is = 2100 = 200 units.
Initially, Q
is
set
to
any
reasonable
quantity,
such
as
200
units
.
The value of z is computed:20020
11
From the standard normal table, z = 1.41.
.200250
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
Hence, R = 251.41+100 = 135.
8/8/2019 Probabilistic Inventory Models Raw Materials
23/37
From the L tables, L(1.41) = 0.0359. = = . . .
The modified
value
of
Q
is
computed:
0.89752505002002 +
The rest of the calculations are
20==
Q 1Qh/p z R L n200 0.9200 1.41 135 0.0359 0.8975121 0.9516 1.66 142 0.0201 0.5025
112 0.9552 1.70 142 0.0183 0.4575
111 0.9556 1.70 143 0.0183 0.4575
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
.
8/8/2019 Probabilistic Inventory Models Raw Materials
24/37
The safety stock is= = = .
So, when
the
lot
is
received,
the
level
of
inventory
is
Q + s = 111 + 43 = 154 units
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
25/37
Determining Q and R from a service level
If the shortage cost p is difficult to come by, other indicators of acceptable service are used.
There are two particular indicators in use; shortage probability (type 1) and proportion of demand that is met
ype . In type 1, the probability of shortage not happening during
the lead time is s ecified. The specification of determines the quantity R; hence Q and
R are separately calculated. To determine R, set F(R) = . The lot size is specified by Q = EOQ.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
26/37
Illustration
Manufacturing has determined that the best service level is when the probability of no shortage is 0.98.
The total demand during the lead time is normally distributed with mean 100 units and standard deviation of 25 units.
The inventory holding cost is 20 riyals per unit per year, and order setup cost is 500 riyals.
. Then, R is selected so that
100 R From the z tables,
.25 =
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
25.15105.225 ==
R
8/8/2019 Probabilistic Inventory Models Raw Materials
27/37
The demand rate is = 100 2 = 200 units per year.
1005002002 ==Q
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
28/37
In type 2, the proportion of the demand that is met is specified.
The average fraction of the demand that is in short is n/Q. Hence, R is selected so that n/Q = 1. Similarly, Q is set to EOQ.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
29/37
Illustration
Manufacturing has determined that the best service level is when the proportion of the demand that is met is 0.98.
The total demand during the lead time is normally distributed with mean 100 units and standard deviation of 25 units.
The inventory holding cost is 20 riyals per unit per year, and order setup cost is 500 riyals.
. The demand rate is = 100 2 = 200 units per year.
The lot
size
is
10020
5002002 ==Q
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
30/37
Then, R is selected so that
2
.
=
=
n
n
From the L tables for L(z) = 2/25 = 0.08, z = 1.02. Hence, R is
5.12510002.125 =+= R
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
31/37
Exact Q under type 2 policy A more accurate value of Q can be found if desired. ,
( ))(1 R F Qh
p = Substituting p in the Q equation leads to
22 n K n)(1)(1
++
= R F h R F
s equat on an t e equat on n = are so ve iteratively until convergence is reached.
The uantit can be initiall set to EO .
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
32/37
Illustration
Manufacturing has determined that the best service level is when the proportion of the demand that is met is 0.98.
The total demand during the lead time is normally distributed with mean 100 units and standard deviation of 25 units.
The inventory holding cost is 20 riyals per unit per year, and order setup cost is 500 riyals.
. The demand rate is = 100 2 = 200 units per year.
The initial
value
of
Q
is
10020
5002002 ==Q
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
33/37
The expected shortage units is
The value
of
R
is
.
12610025/225 1 =+= L The next estimate of Q is
2
114
25
1001261
)220
2005002
25
1001261
2 =
++
=Q
The next value of R is 124; and Q = 114.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
and R = 124.
8/8/2019 Probabilistic Inventory Models Raw Materials
34/37
Imputed shortage cost The use of a service level criterion does not necessitate the
specification of a shortage cost. For a given service level or , there is a corresponding value
for p in the continuous review stochastic demand model, w c w g ve e same an .
Since there is no shortage cost specified, the value of p is called the im uted shorta e cost.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
35/37
Illustration
For a certain manufacturer, Q = 100, R = 151, = 0.98, = 200, = 25, and h = 2.
The imputed shortage cost is2100 ==
( )
)151(1002 F
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
36/37
Scaling of lead time demand It may be possible that the demand is expressed per period
such as a week or month. It is necessary to scale the lead time demand if the lead time
is different than the period of the demand. If the demand follows a normal distribution, then the demand
during the lead time is the sum of the demand normal random variables.
Assume the per period demand has a mean of and variance 2.
If the lead time is , then the demand during the lead time has a normal distribution function with mean = and variance 2 = 2.
Dr Muhammad AlSalamah, Industrial Engineering, KFUPM
8/8/2019 Probabilistic Inventory Models Raw Materials
37/37
Illustration
The demand for a certain raw material in a week is normally distributed with mean 34 and variance 8.
The lead time is 5 weeks. The total demand during the lead time is normally distributed
with
mean = 34
5 = 170 units = =