Corrective

Corrective
Maintenance
Maintenance
• No preventive maintenance or
preventive replacement will be
Corrective Maintenance and conducted.
Spare Parts Decisions • Repair and maintenance will be
conducted whenever a failure occurs.
• This strategy may be the cheapest one
for non-critical components, devices,
and systems.

Relevant
Relevant Decisions
Decisions General
General Assumptions
Assumptions

• What is the expected number of • There are several units of such a

failures in a certain time interval? device that are used in the
• What is the optimal level of spare organization.
parts inventory? • Their failures are independent.
• How often and how much should one • Whenever there is a failure, a spare
order such spare parts? part is used to replace the device.
• How to make these decisions for • We don’t like to see a stockout.
different lifetime distributions?

Data
Data Needed
Needed
• How expensive is each item?
• What is the average lifetime?
• What is the lifetime distribution? The
The Exponential
Exponential Distribution
Distribution
• How long does it take to order and get
one?
• How high is the ordering cost?
• How high is the storage cost for spares?

1
 Exponential
Exponential Distribution
Distribution Poisson
Poisson Distribution
Distribution
x −ρ
• f(t) • f ( x)= ρ e x = 0, 1, 2, …
x!
• E(X) = ρ, Var(X) = ρ
• h(t) • If a device has exponential lifetime
distribution, and each failed unit is
replaced by a new one right away with
• The memoryless properly negligible time, then the number of
replacements in any interval of length t
follows the Poisson distribution with ρ =
λt.

When
When N
N Units
Units Are
Are Used
Used Example
Example 3.1
3.1

• The number of failure replacements • You are looking after 20 units of a

needed in interval t follows the certain device. It is used 24 hours a
Poisson distribution with ρ = Nλt. day and has a constant failure rate of
• When N is large (greater than 30), the 0.004/hr. You are considering
Poisson distribution here can be ordering spare parts once a month (30
approximated by the normal days). What is the quantity to order to
distribution. ensure the probability of running out
of spare parts in a month to be less
than 1%? (see Excel file too).

Use
Use the
the Normal
Normal Approximation
Approximation

2
 Practical
Practical Issues
Issues

• Assume the starting inventory is zero.

Order 82 units for the first month.
What could happen during this
month? What do we do
correspondingly?
• What should we do in the following
months?
• Lead time considerations?

Annual
Annual Cost
Cost of
of this
this Practice?
Practice? Optimal
Optimal Ordering
Ordering Policy
Policy

• Ordering cost: $1000 • How many times to order? Or what is

• Storage cost: $0.05/unit/hour the ordering interval?

• Shortage cost: $0 • What are the issues? How to find such

a policy?
• What are the assumptions? Are they
valid?
• Remember: Exponential Distribution

Weilbull
Weilbull Distribution
Distribution

• PDF f(x):
⎧β
β
⎛x⎞
−⎜ ⎟
⎪ β x β-1e ⎝ α ⎠ , x>0
f (x) = ⎨ α
The
The Weibull
Weibull Distribution
Distribution ⎪0 ,
⎩ Otherwise
• Mean value: E(X) = αΓ(1+1/ β)
• Variance:
Var(X) = α2{Γ(1+2/ β)–[Γ(1+1/ β)]2}

3
 Weilbull
Weilbull Distribution
Distribution If
If Only
Only One
One Device
Device Is
Is Used
Used

• Failure rate function h(x): • Suppose β = 2, α=1000, 24 hr/day

• h(x) = β xβ-1/αβ • Run to failure Weibull Probability Density Function

0.001000

• β< 1: • How many

0.000900
0.000800

pdf function f(x)

0.000700

failures a year?
0.000600

• β = 1:
0.000500
0.000400
0.000300
0.000200

• β > 1: • How many to 0.000100

0.000000
0 500 1000 1500 2000 2500 3000 3500

order each year? Time (x)

• What is the probability of shortage?

One
One Device
Device Answers
Answers
• Analytical solutions would be very complicated.
• Use computer simulation
• Generate a random variable value x1 following the pdf
function, x2, etc, until the sum is longer than a year’s
operation time, now you have N1. Repeat until you reach
Nn. From these Ni values, you find the probability
distribution function of N, the # of failures in a year.
• From this distribution function, you can find the
probability that the number of failures is greater than a
specified value. You can then find the annual order
quantity to ensure a certain shortage probability.
• See next page for illustration

One
One Device
Device Situation
Situation Consider
Consider nn Devices
Devices

• Relevant decision questions: • Suppose n is relatively large: > 30

• What is the cost of the current • Let Yi denote the number of failures
practice? that may be experienced by device i in
• What is the optimal order quantity? a certain time interval

• What is the optimal order interval? • Y = Y1 + Y2 + … + Yn, may be

approximated by the normal
• What if more than one device is used? distribution
We can use the same approach.

4
 Example
Example

• Consider 30 devices. The lifetime of

each one follows the Weibull
distribution with α= 3000, β=3. We
use the run to failure strategy. Consider
ordering once a year. Ordering cost =
$500. Inventory cost = $0.012/unit/hr.
Shortage probability < 2%. How many
to order? What is the total cost?

Discussions
Discussions

• Assumptions: Weibull distribution?

• Other issues to be considered?