RELI ABI LI TY ENGI NEERI NG UNI T

ASST4403
LECTURE 32 SPARE PARTS CALCULATI ONS
1
Learning outcomes Learning outcomes
Some basic models about spare parts calculations Some basic models about spare parts calculations
Determine optimal stock holdings for preventive
replacement spares replacement spares
Determine optimal stock holdings for insurance
(emergency) spares
2
Main considerations of spares provisioning
1. Failure rate determines quantity and perhaps location
of spares. p
2. Acceptable probability of stockout fixes spares level.
3. Turnaround of second-line repair affects lead time.
4 Cost of each spare affects spares level and hence 4. Cost of each spare affects spares level and hence
item 2.
5 Standardization and least replaceable assembly (LRA) 5. Standardization and least replaceable assembly (LRA)
affects number of different spares to be held.
6. Lead time on ordering effectively part of second-line
repair time.
3
Spare parts provisioning models Spare parts provisioning models
Given a preventive replacement policy (either interval or p p p y (
age) applied on a regular basis, what is the expected
number of spare parts required over a certain time period?
How many critical items (with/without PR) should be
stocked in order to assure the following? g
Availability of stock
Interval availability
Minimised cost
Equipment availability
4
Preventive replacement spares
A model to predict the expected number of spare parts
i d ti i d required over a time period
T the planning horizon e.g., 1 year
E[N(T, t
p
)] expected number of spare parts required
over T, when preventive replacement occurs at time t
p
over T, when preventive replacement occurs at time t
p
5
The interval/age replacement situation
For interval replacement:
E[N(T, t
p
)] =number of preventive replacements in
p
(0,T) + expected number of failures in (0,T)
E(t
p
) = |1 +E(t
p
- i - 1)] _ (t)Jt, foi t
p
1
i+1
i
t
p
-1
i=u
anu E(u) = u
For age replacement, the expected cycle length is
6
Numeric example p
Failures follow N(5,1) (weeks) and T=12 months=52 weeks
The optimal interval is 4 weeks
Solution: Solution:
We need to find H(4). Note H(0)=0
_
i+1
1-1
_
1
Note for normal distribution N(5 1) to be integrated in (0 1)
E(1) = |1 + E(1 -i - 1)] _ (t)Jt
i
i=u
= |1 + E(u)] _ (t)Jt
u

Note for normal distribution N(5,1) to be integrated in (0,1)
it is equivalent to finding the value of standard normal
distribution in (-5,-4), that is
_ (t)Jt
1
u
= u(-4) - u(-S) = u - u = u
( ) | ( )] _ ( )
1
E(1) = |1 + E(u)] _ (t)Jt
u
= u
Numeric example (contd)
.
Similarly
So
i+1
2-1
_ (t)Jt
2
1
= u(-S) - u(-4) = u.uu1SS - u = u.uu1SS
So
E(2) = |1 +E(2 - i - 1)] _ (t)Jt
i+1
i
i=u
|1 + E(2 u 1)] _ (t)Jt
1
+ |1 + E(2 1 1)] _ (t)Jt
2
= |1 + E(2 - u - 1)] _ (t)Jt
u
+ |1 + E(2 - 1 - 1)] _ (t)Jt
1
= |1 + E(1)]|u(-4) - u(-S)] + |1 + E(u)]|u(-S) - u(-4)]
= (1 + u) u + (1 + u) u.uu1SS = u.uu1SS
H(4) =0.16
Now we need to find, where | is the ordinate, u=CDF
Numeric example
Failure = N(5,1), i.e., =5 weeks and o=1 week. For t
p
=1
Calculating directly or from the normal ordinate table
_
t
] _
1 S
] | _
t
p
-
o
] = | _
1 - S
1
] = |(-4) = u.uuu1
_ t(t)Jt
t
p
= -1 u.uuu1 + S u.uuuuS = u.uuuuS
Similarly for t
p
=4, we have =0.551
-
Numeric example p
Failures follow N(5,1) (weeks) and T=12 months=52 weeks
The optimal interval is 4 weeks
Solution: Solution:
With interval replacement, we have earlier
For age replacement policy we obtain as earlier For age replacement policy, we obtain as earlier
=0.84
Note age replacement requires less spares Note age replacement requires less spares
I ( ) Insurance (emergency) spares
Concerned with the factual demand for spares by Concerned with the factual demand for spares by
the failure replacement, not preventive replacement
The problemis in general howto balance the risk of The problem is in general how to balance the risk of
running out stock of a spare part against the cost of
overstocking and over purchasing the spares overstocking and over purchasing the spares
The stocked spare parts will be optimal with regards
to one of the objectives to one of the objectives
Availability of stock
Interval availability Interval availability
Minimised cost
Equipment availability
11
Series of events (point processes)
Discrete events that occur randomly in a continuum such
as time are called stochastic point processes as time are called stochastic point processes
The arrival values x
1
, x
2
, , x
n
are the values of the
independent variables e.g. time from x=0 at which each
event occurs
The inter-arrival values X
1
, X
2
, , X
n
are the intervals
between successive events 1, 2, ,n from x=0.
Considering time as the independent variable, we use
the concept of counting process.
We denote by N(t) the number of events occurring in
(0,t). Its expected value E[N(t)] is essential. Another (0,t). Its expected value E[N(t)] is essential. Another
important concept is the time until the k
th
event S(k).
12
13
Model inputs/assumptions p / p
m independent components in a group or fleet m independent components in a group or fleet
T length of a time period under consideration
mean time to failure of component
o standard deviation of the time to failure o standard deviation of the time to failure
N(T,m) total number of failures in interval (0,T)
S(k,m) time until the r
th
failure
Consider only the nonrepairable components Consider only the nonrepairable components
Disregard the time it takes for the replacement
14
Model development
By relating N(T,m) to S(k,m), the probability of not running out y g ( , ) ( , ), p y g
of stock of size k, for a fleet of m components in a time period
T, is
For a probability target and failure distribution, we can
determine a stock size k or how long a stock of size k will last, g
i.e., maximum T for a given stock size k.
Solutions to the above problems involve the distribution of Solutions to the above problems involve the distribution of
N(T,m) or S(k,m),which is complicated. Approximation is often
suggested
15
Special case when T is large p g
When the time period T is large compared to /m, we
l di t ib ti t i t can use normal distribution to approximate
S(k,m) has an approximate normal distribution with ( )
mean k/m and variance o
2
k/m for large k
If we demand an availability of stock p, then we can find
k by k by
where Z
p
is the value corresponding to cumulative
standard normal distribution value being 1p. g p
16
Numeric example Numeric example
Failures follow N(5,1) (weeks) and T=12 months=52
k W th fl t i 40 d i weeks. We assume the fleet size m=40 and require
the probability of having spares in stock in T=52
weeks p=95% weeks, p=95%
For p=0.95, 1p=0.05, which gives us Zp=1.65
17
Special case by a Poisson
di t ib ti i ti distribution approximation
If th f il di t ib ti i ti l ith b i th If the failure distribution is exponential with a being the
expected number of failures in (0,T) and assume a not a
large number then large number, then
from which we can solve the k for a specified probability
18
Applying Poisson distribution
to spare provisioning to spare provisioning
If the underlying failure distributions are exponential , the
b f f il N(T ) f ll tl P i number of failures N(T,m) follows exactly Poisson process
for any m, i.e.
a
i
e
i
a
i m T N P

= =
!
) ) , ( (
Where a= expected number of failures in [0,T]. For m=1,
a=T/ and for m components, a=mT/ (a is assumed not to
b l )
i!
be very large)
We calculate k such that
Th b i d l f k ill b h i i k l l

> = > + = s
k
i
a
i
p e
i
a
T m k S P k m T N P
0
!
) ) , 1 ( ( ) ) , ( (
The obtained value of k will be the minimum stock level
that ensures a reliability of p (probability of not having a
stock-out i e no demand when there is no spare in stock) stock-out, i.e. no demand when there is no spare in stock).
19
Spare part requirement
monograph monograph
20
Example p
Equipment has K=20 parts, operating 24 h per day
=0.1 /1000h
Spares to be calculated for T=3 month interval Spares to be calculated for T 3 month interval
P=95% availability of having spare when required
Answer:
Either calculate KT=4 32 to locate on scale (5) then go to Either calculate KT 4.32 to locate on scale (5), then go to
P=95% on the far-right vertical, thus finding S = wanted
number of spares on scale (7), i.e., 8
Or 1) locate K on the far-left vertical scale, go to scale (2)
to find (cross index (3)), then locate T on scale (4) and to find (cross index (3)), then locate T on scale (4) and
project to cross point on (3) to extend to (5) to find KT,
then follow the steps in above bullet
21