SlideShare a Scribd company logo
1 of 72
Download to read offline
Maintenance Policies For A Deterioration System
Subject To Non-Self-Announcing Failures
Salih Tekin
N. Onur Bakır, Bü¸sra Kele¸s
TOBB University of Economics and Technology
Ankara, Turkey
June 6, 2014
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 1 / 35
System Maintenance
Maintenance is crucial to improve system availability performance with
a minimum cost, especially when
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 2 / 35
System Maintenance
Maintenance is crucial to improve system availability performance with
a minimum cost, especially when
I the system fails stochastically
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 2 / 35
System Maintenance
Maintenance is crucial to improve system availability performance with
a minimum cost, especially when
I the system fails stochastically
I the degree of deterioration, in addition to
failure, are known only through inspections
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 2 / 35
System Maintenance
Maintenance is crucial to improve system availability performance with
a minimum cost, especially when
I the system fails stochastically
I the degree of deterioration, in addition to
failure, are known only through inspections
I optional preventive maintenance is
suggested
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 2 / 35
System Maintenance
Maintenance is crucial to improve system availability performance with
a minimum cost, especially when
I the system fails stochastically
I the degree of deterioration, in addition to
failure, are known only through inspections
I optional preventive maintenance is
suggested
I maintenance is not perfect (except
replacement)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 2 / 35
Maintenance Categories in Literature
Table: Maintenance Categories in Literature
Maintenance policy System structure Maintenance degree
Age replacement
Block replacement
Periodic 3
Sequential
Control limit
Single-unit 3
Multi-unit
Perfect 3
Imperfect 3
Minimal 3
Worse
Maintenance cost Optimization criteria Modelling tools
Constant 3
Random
Complex
Minimize cost rate 3
Minimize availability
Minimize downtime
Renewal theory 3
Markov chain 3
Poisson process
Planning horizon Dependence System information
Infinite 3
Finite
Discrete
Continuous
Economic
Failure
Probability
State 3
Perfect 3
Imperfect
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 3 / 35
Outline
1 Introduction
2 Mathematical Formulations
3 Optimal Inspection Time and Optimal Policy
4 Numerical Example
5 References
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 4 / 35
Introduction
Problem Definition
System structure
For a stochastically failing system in which the degree of deterioration
are known only through inspections,
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
Introduction
Problem Definition
System structure
For a stochastically failing system in which the degree of deterioration
are known only through inspections,
I At each inspection epoch k⌧, k = {0, 1, 2, ...} the system occupies
one of three states: good (1), poor (2), failed (3) and the decision maker
chooses an available action: do nothing, repair and replace.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
Introduction
Problem Definition
System structure
For a stochastically failing system in which the degree of deterioration
are known only through inspections,
I At each inspection epoch k⌧, k = {0, 1, 2, ...} the system occupies
one of three states: good (1), poor (2), failed (3) and the decision maker
chooses an available action: do nothing, repair and replace.
S = {1, 2, 3}; State space
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
Introduction
Problem Definition
System structure
For a stochastically failing system in which the degree of deterioration
are known only through inspections,
I At each inspection epoch k⌧, k = {0, 1, 2, ...} the system occupies
one of three states: good (1), poor (2), failed (3) and the decision maker
chooses an available action: do nothing, repair and replace.
S = {1, 2, 3}; State space
T = {0, 1, 2, 3, ...}; Time horizon
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
Introduction
Problem Definition
System structure
For a stochastically failing system in which the degree of deterioration
are known only through inspections,
I At each inspection epoch k⌧, k = {0, 1, 2, ...} the system occupies
one of three states: good (1), poor (2), failed (3) and the decision maker
chooses an available action: do nothing, repair and replace.
S = {1, 2, 3}; State space
T = {0, 1, 2, 3, ...}; Time horizon
A = {do nothing, repair, replace}; Action space
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
Introduction
Problem Definition
System structure
For a stochastically failing system in which the degree of deterioration
are known only through inspections,
I At each inspection epoch k⌧, k = {0, 1, 2, ...} the system occupies
one of three states: good (1), poor (2), failed (3) and the decision maker
chooses an available action: do nothing, repair and replace.
S = {1, 2, 3}; State space
T = {0, 1, 2, 3, ...}; Time horizon
A = {do nothing, repair, replace}; Action space
It is called minor repair when the repair is performed at state 2 and major
repair when it is performed at state 3.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
Introduction
Problem Definition
Deterioration process
I The deterioration is modelled by three-state continuous time Markov
chain
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 6 / 35
Introduction
Problem Definition
Deterioration process
I The deterioration is modelled by three-state continuous time Markov
chain
If {Yt, t 0} represents the system state at time t, then we assume
that {Yt, t 0} is a continuous time Markov chain with transition
probabilities,
Pij(⌧) = P{Yt+⌧ = j | Yt = i}, 8 t, ⌧ 2 R (1)
when it is currently at state i = {1, 2, 3} at time t, will be at state j = {2, 3}
at time t + ⌧.
(Markovian assumption)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 6 / 35
Introduction
Problem Definition
Deterioration process
Good
(1)
Poor
(2)
Failed
(3)
𝐏𝟏𝟏(𝛕) 𝐏𝟐𝟐(𝛕) 𝐏𝟑𝟑(𝛕)
𝐏𝟏𝟑(𝛕)
𝐏𝟏𝟐(𝛕) 𝐏𝟐𝟑(𝛕)
Figure: Transition Probabilities
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 7 / 35
Introduction
Problem Definition
Deterioration process
Good
(1)
Poor
(2)
Failed
(3)
𝐏𝟏𝟏(𝛕) 𝐏𝟐𝟐(𝛕) 𝐏𝟑𝟑(𝛕)
𝐏𝟏𝟑(𝛕)
𝐏𝟏𝟐(𝛕) 𝐏𝟐𝟑(𝛕)
Figure: Transition Probabilities
I The event that causes a transition within an inspection period from state i
to j, where i 6= j, occur after an exponential amount of time with rate ij.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 7 / 35
Introduction
Problem Definition
Assumptions
I The time horizon considered is infinite and the system with good state
is put into service in time 0.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 8 / 35
Introduction
Problem Definition
Assumptions
I The time horizon considered is infinite and the system with good state
is put into service in time 0.
I Transition from state i = 1, 2 to state j = 2, 3 occurs according to a Poisson
Process with rate ij.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 8 / 35
Introduction
Problem Definition
Assumptions
I The time horizon considered is infinite and the system with good state
is put into service in time 0.
I Transition from state i = 1, 2 to state j = 2, 3 occurs according to a Poisson
Process with rate ij.
I The system state is monitored only through inspections and inspection time
is negligible.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 8 / 35
Introduction
Problem Definition
Assumptions
I The time horizon considered is infinite and the system with good state
is put into service in time 0.
I Transition from state i = 1, 2 to state j = 2, 3 occurs according to a Poisson
Process with rate ij.
I The system state is monitored only through inspections and inspection time
is negligible.
I Time to inspect is negligible; however, inspection cost is assumed to be a
monotonically non-increasing function of time interval between two
consecutive inspections [4] to prevent frequent inspections.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 8 / 35
Introduction
Problem Definition
Assumptions
I If the system is identified as good, "do nothing" action is required;
otherwise decision maker may choose other actions.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 9 / 35
Introduction
Problem Definition
Assumptions
I If the system is identified as good, "do nothing" action is required;
otherwise decision maker may choose other actions.
I Minor (Major) repair takes the system to state 1, with probability pmn (pmj),
and to state 2 with probability qmn = 1 pmn (qmj = 1 pmj) after a fixed time
tmn (tmj) with a fixed cost cmn (cmj).
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 9 / 35
Introduction
Problem Definition
Assumptions
I If the system is identified as good, "do nothing" action is required;
otherwise decision maker may choose other actions.
I Minor (Major) repair takes the system to state 1, with probability pmn (pmj),
and to state 2 with probability qmn = 1 pmn (qmj = 1 pmj) after a fixed time
tmn (tmj) with a fixed cost cmn (cmj).
I Replace always returns the system to state 1 with a fixed amount cost crpl
after a fixed maintenance time trpl.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 9 / 35
Introduction
Problem Definition
Assumptions
I If the system is identified as good, "do nothing" action is required;
otherwise decision maker may choose other actions.
I Minor (Major) repair takes the system to state 1, with probability pmn (pmj),
and to state 2 with probability qmn = 1 pmn (qmj = 1 pmj) after a fixed time
tmn (tmj) with a fixed cost cmn (cmj).
I Replace always returns the system to state 1 with a fixed amount cost crpl
after a fixed maintenance time trpl.
I The system is not in working condition during repair or replace and hence,
the inspection cost does not occur.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 9 / 35
Introduction
Problem Definition
Assumptions
I If the system is identified as good, "do nothing" action is required;
otherwise decision maker may choose other actions.
I Minor (Major) repair takes the system to state 1, with probability pmn (pmj),
and to state 2 with probability qmn = 1 pmn (qmj = 1 pmj) after a fixed time
tmn (tmj) with a fixed cost cmn (cmj).
I Replace always returns the system to state 1 with a fixed amount cost crpl
after a fixed maintenance time trpl.
I The system is not in working condition during repair or replace and hence,
the inspection cost does not occur.
I The system state is known instantaneously after each repair.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 9 / 35
Introduction
Problem Definition
Maintenance Policies
Five available maintenance policies are proposed shown to emphasise how
the maintenance parameters affect the optimum inspection period and policy.
Table: Policy Description
Policy State 1 State 2 State 3
⇧1 Do nothing Do nothing Replacement
⇧2 Do nothing Do nothing Major repair
⇧3 Do nothing Minor repair Major repair
⇧4 Do nothing Minor repair Replacement
⇧5 Do nothing Replacement Replacement
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 10 / 35
Introduction
Problem Definition
Optimization criteria
Optimizing the maintenance policies is based on Renewal Theory.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 11 / 35
Introduction
Problem Definition
Optimization criteria
Optimizing the maintenance policies is based on Renewal Theory.
I A cycle ends (a renewal) after a maintenance of the failed system,
while the maintenance brings the system to "good" state.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 11 / 35
Introduction
Problem Definition
Optimization criteria
Optimizing the maintenance policies is based on Renewal Theory.
I A cycle ends (a renewal) after a maintenance of the failed system,
while the maintenance brings the system to "good" state.
I Then cost rate is the ratio of the expected total cost occurred during
the renewal cycle over the expected cycle length.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 11 / 35
Introduction
Problem Definition
Optimization criteria
Optimizing the maintenance policies is based on Renewal Theory.
I A cycle ends (a renewal) after a maintenance of the failed system,
while the maintenance brings the system to "good" state.
I Then cost rate is the ratio of the expected total cost occurred during
the renewal cycle over the expected cycle length.
I The objective is when to inspect the system periodically and what
action to be made by minimizing the expected cost rate in the long run.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 11 / 35
Mathematical Formulations
Mathematical Calculations
Transition probabilities
Let Xij be the holding time between state i = {1, 2, 3} and state j = {2, 3} .
By Poisson arrivals assumption leads to:
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 12 / 35
Mathematical Formulations
Mathematical Calculations
Transition probabilities
Let Xij be the holding time between state i = {1, 2, 3} and state j = {2, 3} .
By Poisson arrivals assumption leads to:
P11(⌧) = P{min{X12, X13} > ⌧} = e 1⌧
(2)
P12(⌧) = P{X12  ⌧, X12 < X13, X12 + X23 > ⌧}
=
8
>><
>>:
12 e 2⌧ ⌧, 1 2 = 0
12
1 2
e 2⌧ e 1⌧ , 1 2 6= 0
(3)
P13(⌧) = P{X12  ⌧, X12 < X13, X12 + X23  ⌧}
=
8
>><
>>:
1 e 1⌧
12 e 2⌧ ⌧, 1 = 2
1 e 1⌧ 12
1 2
e 2⌧ + e 1⌧ , 1 6= 2
(4)
where the total transition rate out of a state 1 is 1 = 12 + 13 and of state 2
is 2 = 23.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 12 / 35
Mathematical Formulations
Mathematical Calculations
Expected downtime
I How long ago could the failure have first been noticed
by an inspection?
1
Inspection time
kτ
At the time system fails
τ' Downtime
3
Inspection time
(k+1)τ
Figure: Downtime
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 13 / 35
Mathematical Formulations
Mathematical Calculations
Expected downtime
D1(⌧) : the expected downtime within an inspection period, ⌧ , of the system
initially found at state 1
D2(⌧) : the expected downtime within an inspection period, ⌧ , of the system
initially found at state 2
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 14 / 35
Mathematical Formulations
Mathematical Calculations
Expected downtime
D1(⌧) : the expected downtime within an inspection period, ⌧ , of the system
initially found at state 1
D2(⌧) : the expected downtime within an inspection period, ⌧ , of the system
initially found at state 2
D1(⌧) = DQ
(⌧) Q(⌧) + DH
(⌧) H(⌧) (5)
where
1
Inspection period, τ
32
Q(τ)
H(τ)
3
𝐏𝟑𝟑(τ)
Figure: Downtime
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 14 / 35
Mathematical Formulations
Mathematical Calculations
Expected downtime
DQ
(⌧) = ⌧ E[X13 | X13  ⌧, X13 < X12 ] =
⌧
1 e 1⌧
1
1
(6)
DH
(⌧) = ⌧ E[X12 | X12  ⌧, X12 < X13, X12 + X23  ⌧]
E[X23 | X12  ⌧, X12 < X13, X12 + X23  ⌧]
=
8
>>>><
>>>>:
⌧ 2
1
+e 1⌧
⇣
2
1
+⌧
⌘
1 e 1⌧
1e 1⌧ ⌧
, 1 2 = 0
( 1 2)
⇣
⌧ 1
1
1
2
⌘
1 2 1e 2⌧ + 2e 1⌧ +
e 2⌧
⇣
1
2
⌘
e 1⌧
⇣
2
1
⌘
1 2 1e 2⌧ + 2e 1⌧ , 1 2 6= 0
(7)
D2(⌧) = ⌧
1 e 2⌧
2
(8)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 15 / 35
Mathematical Formulations
Mathematical Calculations
Average cost rate
Let i(⌧) be the cost rate for ⇡i, i = 0, .., 5,
The first aim is to calculate each i(⌧) via optimizing the inspection
interval time ⌧.
The second is to determinate the optimal policy with the correspon-
ding minimum cost rate.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 16 / 35
Mathematical Formulations
Mathematical Calculations
Average cost rate
For 8 ⇡i, i = 1, .., 5, we define
Li
j(⌧) as the expected time length the system spends to state 1 from
state j = {1, 2, 3}, and
Ci
j(⌧) as the expected maintenance cost the system spends to state
1 from state j = {1, 2, 3}.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 17 / 35
Mathematical Formulations
Mathematical Calculations
Average cost rate
The cost rate of the system is then
i(⌧) =
Ci
1(⌧)
Li
1(⌧)
+ (⌧) (9)
where (⌧) is the inspection cost per time which is a non increasing
function of time interval between two consecutive inspections such that
lim
⌧!0
(⌧) = K < 1 ve lim
⌧!1
(⌧) = 0 (10)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 18 / 35
Mathematical Formulations
Mathematical Calculations
Average cost rate
Cycle Length
Li
1(⌧) =
1
1 P11(⌧)
⇥
⌧ + P12(⌧)Li
2(⌧) + P13(⌧)Li
3(⌧)
⇤
(11)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 19 / 35
Mathematical Formulations
Mathematical Calculations
Average cost rate
Cycle Length
Li
1(⌧) =
1
1 P11(⌧)
⇥
⌧ + P12(⌧)Li
2(⌧) + P13(⌧)Li
3(⌧)
⇤
(11)
where
Li
2(⌧) =
8
>>>><
>>>>:
1
1 P22(⌧)
⇥
⌧ + P23(⌧)Li
3(⌧)
⇤
i = 1, 2
tmn
pmn
i = 3, 4
trpl i = 5
(12)
Li
3(⌧) =
8
<
:
tmj + 1 pmj Li
2(⌧) i = 3, 4
trpl i = 5
(13)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 19 / 35
Mathematical Formulations
Mathematical Calculations
Average cost rate
Cycle Cost
Ci
1(⌧) =
1
1 P11(⌧)
⇥
P12(⌧)Ci
2(⌧) + P13(⌧) cdD1(⌧) + Ci
3(⌧)
⇤
(14)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 20 / 35
Mathematical Formulations
Mathematical Calculations
Average cost rate
Cycle Cost
Ci
1(⌧) =
1
1 P11(⌧)
⇥
P12(⌧)Ci
2(⌧) + P13(⌧) cdD1(⌧) + Ci
3(⌧)
⇤
(14)
where
Ci
2(⌧) =
8
>>><
>>>:
cdD2(⌧) + Ci
3(⌧), i = 1, 2
cmn+tmn[cd (⌧)]
pMn
, i = 3, 4
crpl + trpl [cd (⌧)], i = 5
(15)
Ci
3(⌧) =
8
>>><
>>>:
cmj + tmj [cd (⌧)]
+(1 pmj)Ci
2(⌧), i = 3, 4
crpl + trpl [cd (⌧)], i = 5
(16)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 20 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
For any ⌧ 2 (0, 1), the parameters in (11-16) are all positive and finite,
then the cost rate is both positive and finite.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 21 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
For any ⌧ 2 (0, 1), the parameters in (11-16) are all positive and finite,
then the cost rate is both positive and finite.
We need to examine the behaviour of the cost rate as
⌧ ! 0 (continuous inspection) and ⌧ ! 1 (never inspect)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 21 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Continuous inspection
I Continuously inspection ensures that the system state is always
known.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 22 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Continuous inspection
I Continuously inspection ensures that the system state is always
known.
I Failure is detected as soon as the system fails.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 22 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Continuous inspection
I Continuously inspection ensures that the system state is always
known.
I Failure is detected as soon as the system fails.
I The cost rate becomes
i(0) =
Ci
1(0)
Li
1(0)
+ K (17)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 22 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Continuous inspection
The cycle length leads to:
Li
1(0) =
12
1

1
12
+ Li
2(0) +
13
1

1
13
+ Li
3(0) (18)
where
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 23 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Continuous inspection
The cycle length leads to:
Li
1(0) =
12
1

1
12
+ Li
2(0) +
13
1

1
13
+ Li
3(0) (18)
where
Li
2(0) =
8
>>><
>>>:
1
2
+ Li
3(0), i = 1, 2
tmn
pmn
, i = 3, 4
trpl, i = 5
(19)
Li
3(0) =
8
<
:
tmj + (1 pmj)Li
2(0), i = 3, 4
trpl, i = 5
(20)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 23 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Continuous inspection
The cycle maintenance cost leads to:
Ci
1(0) =
12
1
Ci
2(0) +
13
1
Ci
3(0) (21)
where
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 24 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Continuous inspection
The cycle maintenance cost leads to:
Ci
1(0) =
12
1
Ci
2(0) +
13
1
Ci
3(0) (21)
where
Ci
2(0) =
8
>>><
>>>:
Ci
3(0), i = 1, 2
cmn+tmn(cd K)
pmn
, i = 3, 4
crpl + trpl(cd K), i = 5
(22)
Ci
3(0) =
8
<
:
cmj + tmj(cd K) + (1 pmj) Ci
2(0), i = 3, 4
crpl + trpl(cd K), i = 5
(23)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 24 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Never inspect
When ⌧ ! 1, the failure of the system occurs with probability 1 which
leads that the cycle cost consists of only downtime cost and inspection
cost.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 25 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Never inspect
When ⌧ ! 1, the failure of the system occurs with probability 1 which
leads that the cycle cost consists of only downtime cost and inspection
cost.
The cost rate becomes
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 25 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Never inspect
When ⌧ ! 1, the failure of the system occurs with probability 1 which
leads that the cycle cost consists of only downtime cost and inspection
cost.
The cost rate becomes
lim
⌧!1
i(⌧) = lim
⌧!1
hcdD1(⌧)
⌧
+ (⌧)
i
= cd + lim
t!1
(⌧)
= cd (24)
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 25 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Never inspect
When ⌧ ! 1, the failure of the system occurs with probability 1 which
leads that the cycle cost consists of only downtime cost and inspection
cost.
The cost rate becomes
lim
⌧!1
i(⌧) = lim
⌧!1
hcdD1(⌧)
⌧
+ (⌧)
i
= cd + lim
t!1
(⌧)
= cd (24)
I i(⌧) is a continuous function and bounded on [0, min{ (0), i(1)}]
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 25 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Corollary 1
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 26 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Corollary 1
If there exists 9 ⌧ 2 [0, 1) such that i(⌧) < cd, then an optimal
inspection interval of each policy exists such that
⌧⇤
i 2 1
i ([0, min{ i(0), i(1)}])
otherwise , all policies converge each other (cd) in the long run.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 26 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Corollary 1
If there exists 9 ⌧ 2 [0, 1) such that i(⌧) < cd, then an optimal
inspection interval of each policy exists such that
⌧⇤
i 2 1
i ([0, min{ i(0), i(1)}])
otherwise , all policies converge each other (cd) in the long run.
Corollary 2
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 26 / 35
Optimal Inspection Time and Optimal Policy
Optimization Criteria
Corollary 1
If there exists 9 ⌧ 2 [0, 1) such that i(⌧) < cd, then an optimal
inspection interval of each policy exists such that
⌧⇤
i 2 1
i ([0, min{ i(0), i(1)}])
otherwise , all policies converge each other (cd) in the long run.
Corollary 2
If cd = 1 and other cost parameters set to zero, the average cost rate
gives the system unavailability.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 26 / 35
Numerical Example
Numerical Example
We present several numerical examples which illustrate how system
parameters (transition rates, maintenance cots and times) could affect the
optimum inspection period and the optimum maintenance policy as well.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 27 / 35
Numerical Example
Numerical Example
We present several numerical examples which illustrate how system
parameters (transition rates, maintenance cots and times) could affect the
optimum inspection period and the optimum maintenance policy as well.
Consider the deterioration model with inspection cost function as
(⌧) = Ke 0.1⌧
. Other input parameters are shown below of illustrated figures.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 27 / 35
Numerical Example
Numerical Example
Transition rates
Replacement
Minor
Repair
Major
Repair
Fixed cost 750 300 500
Time to
maintenance
1 0.5 2
Probability of repair-
success
1 0.9 0.6
Inspection cost
parameter
1000
Figure: Relationships between transition rates and inspection interval
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 28 / 35
Numerical Example
Numerical Example
Replacement cost
5100 2 2 2 2 2 3 3 3 3 3
4900 2 2 2 2 2 3 3 3 3 3
4700 3 3 2 2 2 3 3 3 3 3
4500 3 3 3 3 3 3 3 3 3 3
4300 3 3 3 3 3 3 3 3 4 4
4100 3 3 3 3 3 3 3 3 4 4
3900 3 3 3 3 3 4 4 4 4 4
3700 3 3 3 3 3 4 4 4 4 4
3500 4 4 3 3 3 4 4 4 4 4
3300 4 4 4 4 4 4 4 4 5 4
3100 5 4 4 4 4 4 4 5 5 5
2900 5 4 4 4 4 5 5 5 5 5
2700 5 5 4 4 4 5 6 5 5 5
2500 6 5 5 5 5 5 6 6 6 6
2300 6 6 6 6 6 6 6 6 6 6
2100 7 7 7 7 7 𝛑 𝟎 6 7 7 7 7
1900 7 7 7 7 7 𝛑 𝟏 7 7 7 7 7
1700 8 8 8 8 8 𝛑 𝟐 8 8 8 8 8
1500 9 9 9 9 9 𝛑 𝟑 9 9 9 9 9
1300 10 11 11 11 11 𝛑 𝟒 11 11 11 11 11
1100 12 12 13 13 13 𝛑 𝟓 12 13 13 13 13
900 15 15 15 15 15 15 16 16 16 16
700 19 20 20 20 20 20 20 21 21 21
500 26 27 27 27 27 27 28 28 28 28
300 38 39 39 39 39 38 39 40 40 40
100 60 65 66 66 66 60 66 66 66 66
75 66 75 75 75 75 66 75 76 76 76
50 76 103 108 108 108 76 102 108 108 108
250 500 1000 2500 5000 250 500 1000 2500 5000
Repla-
cement
Minor
Repair
Major
Repair
Fixed cost 𝐜 𝐫𝐩𝐥 300 500
Time to
maintenance
0.001 0.001 0.001
Probability of
repair-success
1 0.9 0.6
Inspection cost
parameter
1000
Repla-
cement
Minor
Repair
Major
Repair
Fixed cost 𝐜 𝐫𝐩𝐥 300 500
Time to
maintenance
1 0.5 2
Probability of
repair-success
1 0.9 0.6
Inspection cost
parameter
1000
Cost of replacement, 𝐜 𝐫𝐩𝐥
Cost of
downtime, 𝐜 𝐝
Figure: Relationships between downtime cost and replacement cost
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 29 / 35
Numerical Example
Numerical Example
Inspection cost
Cost of
downtime, 𝐜 𝐝
Cost of Inspection Parameter, K
Repla-
cement
Minor
Repair
Major
Repair
Fixed cost 1000 300 500
Time to
maintenance
0.001 0.001 0.001
Probability of
repair-success
1 0.9 0.6
Inspection cost
parameter
K
5100 0 1 2 13 25 0 0 3 13 26
4900 0 1 2 13 26 0 0 3 14 26
4700 0 1 2 14 27 0 0 3 15 27
4500 0 1 3 14 28 0 0 3 15 28
4300 0 1 3 15 29 0 0 3 16 29
4100 0 1 3 16 30 0 0 4 17 30
3900 0 2 3 17 31 0 0 4 17 31
3700 0 2 3 18 32 0 0 4 18 32
3500 0 2 3 19 33 0 0 4 19 34
3300 0 2 4 20 34 0 0 4 20 35
3100 0 2 4 21 36 0 0 5 22 36
2900 0 2 4 22 37 0 3 5 23 38
2700 0 2 4 24 39 0 3 5 24 39
2500 0 2 5 25 40 0 3 6 26 41
2300 0 3 6 27 42 0 3 6 28 43
2100 0 3 7 29 44 𝛑 𝟎 0 4 7 30 44
1900 0 3 7 32 46 𝛑 𝟏 0 4 7 32 46
1700 0 4 8 34 49 𝛑 𝟐 0 4 8 35 48
1500 0 4 9 37 50 𝛑 𝟑 0 5 10 37 51
1300 0 5 11 40 53 𝛑 𝟒 0 6 11 40 53
1100 0 6 13 43 56 𝛑 𝟓 0 7 13 44 56
900 2 7 15 47 59 0 8 16 48 59
700 2 10 20 52 64 0 10 21 53 64
500 3 14 27 58 69 4 15 28 60 69
300 5 24 39 67 77 6 25 40 67 77
100 20 55 66 89 98 21 55 66 89 98
75 31 65 75 97 106 32 65 76 97 106
50 72 98 108 128 135 73 99 108 126 135
100 500 1000 5000 10000 100 500 1000 5000 10000
Repla-
cement
Minor
Repair
Major
Repair
Fixed cost 1000 300 500
Time to
maintenance
1 0.5 2
Probability of
repair-success
1 0.9 0.6
Inspection cost
parameter
K
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 30 / 35
Numerical Example
Numerical Example
Inspection cost
Cost of major repair, 𝐜 𝐦𝐣 Maintenance time for major repair, 𝐭 𝐦𝐣
5100 3 3 3 3 3 2 2 3 3 3
4900 3 3 3 3 3 2 2 3 3 3
4700 3 3 3 3 3 2 3 3 3 3
4500 3 3 3 3 3 3 3 3 3 3
4300 4 4 4 4 4 3 3 4 4 4
4100 4 4 4 4 4 3 3 4 4 4
3900 4 4 4 4 4 3 3 4 4 4
3700 4 4 4 4 4 3 3 4 4 4
3500 4 4 4 4 4 3 3 4 4 4
3300 5 5 5 5 5 4 4 5 5 5
3100 5 5 5 5 5 4 4 5 5 5
2900 5 5 5 5 5 4 4 5 5 5
2700 5 5 5 5 5 4 5 5 5 5
2500 6 6 6 6 6 5 5 6 6 6
2300 6 6 6 6 6 5 5 6 6 6
2100 7 7 7 7 7 6 6 7 7 7
1900 7 7 8 8 8 𝛑 𝟏 7 7 7 8 8
1700 8 8 9 9 9 𝛑 𝟐 7 7 8 9 9
1500 9 9 10 10 10 𝛑 𝟑 9 9 9 10 10
1300 11 11 11 12 12 𝛑 𝟒 10 10 11 12 12
1100 13 13 13 14 14 𝛑 𝟓 12 12 13 14 13
900 16 16 17 18 18 15 15 16 17 18
700 20 21 22 23 23 20 20 21 22 24
500 27 28 30 32 33 27 28 28 29 31
300 38 40 43 48 50 39 40 40 41 42
100 62 66 99 >1000 >1000 66 66 66 67 68
75 68 76 >1000 >1000 >1000 75 75 76 76 77
50 79 108 >1000 >1000 >1000 108 108 108 109 109
300 500 1000 1500 1750 0.001 0.1 2 5 10
Repla-
cement
Minor
Repair
Major
Repair
Fixed cost 2000 300 𝐜 𝐦𝐫
Time to
maintenance
1 0.5 2
Probability of
repair-success
1 0.9 0.6
Inspection cost
parameter
1000
Repla-
cement
Minor
Repair
Major
Repair
Fixed cost 2000 300 𝟓𝟎𝟎
Time to
maintenance
1 0.5 𝐭 𝐦𝐫
Probability of
repair-success
1 0.9 0.6
Inspection cost
parameter
1000
Cost of
downtime, 𝐜 𝐝
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 31 / 35
Numerical Example
Numerical Example
Maintenance probabilities
Probability of Major
repair-success, 𝐩 𝐦𝐣
1
0,8
0,6
0,4
0,2
0,2 0,4 0,6 0,8 1 0,2 0,4 0,6 0,8 1
1
0,8
0,6
0,4
0,2
0,2 0,4 0,6 0,8 1 0,2 0,4 0,6 0,8 1
𝛑 𝟎
𝛑 𝟏
𝛑 𝟐
𝛑 𝟑
𝛑 𝟒
𝛑 𝟓
Probability of Minor
repair-success, 𝐩 𝐦𝐧
I. 𝐜 𝐫𝐩𝐥 = 𝟓𝟎𝟎 II. 𝐜 𝐫𝐩𝐥 = 𝟔𝟓𝟎
III. 𝐜 𝐫𝐩𝐥 = 𝟕𝟓𝟎 IV. 𝐜 𝐫𝐩𝐥 = 𝟏𝟎𝟎𝟎
Replace
ment
Minor
Repair
Major
Repair
Fixed
cost
𝐜 𝐫𝐩𝐥 150 500
Time to
maintenance
1 0.5 2
Probability
of repair-
success
1 𝐩 𝐦𝐧 𝐩 𝐦𝐣
Inspection cost
parameter
1000
Down Time Cost 400
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 32 / 35
Numerical Example
Numerical Example
Maintenance times
Maintenance time
for major repair, 𝐭 𝐦𝐣
Replace
ment
Minor
Repair
Major
Repair
Fixed
cost
750 150 500
Time to
maintenance
𝐭 𝐫𝐩𝐥 𝐭 𝐦𝐣 𝐭 𝐦𝐧
Probability
of repair-
success
1 0.9 0.6
Maintenance time
for minor repair,
𝐭 𝐦𝐧
Inspection cost
parameter
1000
Downtime Cost 400
𝛑 𝟎
𝛑 𝟏
𝛑 𝟐
𝛑 𝟑
𝛑 𝟒
𝛑 𝟓
I. 𝐭 𝐫𝐩𝐥 = 𝟎. 𝟎𝟏
II. 𝐭 𝐫𝐩𝐥 = 𝟎. 𝟏
III. 𝐭 𝐫𝐩𝐥 = 𝟎. 𝟓
IV. 𝐭 𝐫𝐩𝐥 = 𝟏
2,5
2
1,5
1
0,5
0,5 1 1,5 2 2,5 0,5 1 1,5 2 2,5
2,5
2
1,5
1
0,5
0,5 1 1,5 2 2,5 0,5 1 1,5 2 2,5
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 33 / 35
Numerical Example
Thank you!
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 34 / 35
References
References
Barlow, R., Hunter, L., Optimum preventive maintenance policies. Operations Research , 8 (1): 90-100, 1960
Barlow, R.E., Proschan, F., Mathematical theory of reliability. New York: Wiley 1965
Luss, H., Maintenance policies when deterioration can be observed by inspections. Operation Research, 24 (2): 359–366,
1976
Zuckerman, D., Inspection and replacement policies. Applied Probability, 17 (1): 168–177, 1980
Christer, A.H., Waller, W.M, Delay time models of industrial inspection maintenance models. Operational Research
Society, 35 (5): 401–406, 1984
Chelbi, A., Ait-Kadi, D., Replacement strategy for non self announcing failure equipment. Emerging Technologies and
Factory Automation 1, pp.423-430, 1994
Chelbi, A., Ait-Kadi, D., An optimal inspection strategy for randomly failing equipment. Reliability Engineering and System
Safety 63 (2), pp.127-131, 1999
Klutke, G.A., Yang, Y., The availability of inspected systems subject to shocks and graceful degradation. IEEE Transition
Reliability, 51 (3), pp.371-374, 2002
Wang, W., An inspection model for a process with two types of inspections and repairs. Reliability Engineering and
System Safety, 94 : 526–533, 2008.
Wang, W., An inspection model based on a three-stage failure process. Reliability Engineering and System Safety, 96
:838–848, 2011.
Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 35 / 35

More Related Content

Viewers also liked (9)

mid
midmid
mid
 
Unidad educativa ‘’tomas oleas``
Unidad educativa ‘’tomas oleas``Unidad educativa ‘’tomas oleas``
Unidad educativa ‘’tomas oleas``
 
B enot1 kef1_psifiakoskosmos_tk
B enot1 kef1_psifiakoskosmos_tkB enot1 kef1_psifiakoskosmos_tk
B enot1 kef1_psifiakoskosmos_tk
 
KlientBoost and Kissmetrics Present: PPC Analytics [infographic]
KlientBoost and Kissmetrics Present: PPC Analytics [infographic]KlientBoost and Kissmetrics Present: PPC Analytics [infographic]
KlientBoost and Kissmetrics Present: PPC Analytics [infographic]
 
Taller3 de internet
Taller3 de internetTaller3 de internet
Taller3 de internet
 
Les TIC a la societat
Les TIC a la societatLes TIC a la societat
Les TIC a la societat
 
18.9. 15 ichthyodiversity of maimala river, kerala, india, jzbr 2015
18.9. 15 ichthyodiversity of maimala river, kerala, india, jzbr 201518.9. 15 ichthyodiversity of maimala river, kerala, india, jzbr 2015
18.9. 15 ichthyodiversity of maimala river, kerala, india, jzbr 2015
 
Evidencia 2 modulo 1
Evidencia 2 modulo 1Evidencia 2 modulo 1
Evidencia 2 modulo 1
 
Electricos
ElectricosElectricos
Electricos
 

Busra_Keles

  • 1. Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing Failures Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s TOBB University of Economics and Technology Ankara, Turkey June 6, 2014 Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 1 / 35
  • 2. System Maintenance Maintenance is crucial to improve system availability performance with a minimum cost, especially when Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 2 / 35
  • 3. System Maintenance Maintenance is crucial to improve system availability performance with a minimum cost, especially when I the system fails stochastically Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 2 / 35
  • 4. System Maintenance Maintenance is crucial to improve system availability performance with a minimum cost, especially when I the system fails stochastically I the degree of deterioration, in addition to failure, are known only through inspections Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 2 / 35
  • 5. System Maintenance Maintenance is crucial to improve system availability performance with a minimum cost, especially when I the system fails stochastically I the degree of deterioration, in addition to failure, are known only through inspections I optional preventive maintenance is suggested Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 2 / 35
  • 6. System Maintenance Maintenance is crucial to improve system availability performance with a minimum cost, especially when I the system fails stochastically I the degree of deterioration, in addition to failure, are known only through inspections I optional preventive maintenance is suggested I maintenance is not perfect (except replacement) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 2 / 35
  • 7. Maintenance Categories in Literature Table: Maintenance Categories in Literature Maintenance policy System structure Maintenance degree Age replacement Block replacement Periodic 3 Sequential Control limit Single-unit 3 Multi-unit Perfect 3 Imperfect 3 Minimal 3 Worse Maintenance cost Optimization criteria Modelling tools Constant 3 Random Complex Minimize cost rate 3 Minimize availability Minimize downtime Renewal theory 3 Markov chain 3 Poisson process Planning horizon Dependence System information Infinite 3 Finite Discrete Continuous Economic Failure Probability State 3 Perfect 3 Imperfect Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 3 / 35
  • 8. Outline 1 Introduction 2 Mathematical Formulations 3 Optimal Inspection Time and Optimal Policy 4 Numerical Example 5 References Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 4 / 35
  • 9. Introduction Problem Definition System structure For a stochastically failing system in which the degree of deterioration are known only through inspections, Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
  • 10. Introduction Problem Definition System structure For a stochastically failing system in which the degree of deterioration are known only through inspections, I At each inspection epoch k⌧, k = {0, 1, 2, ...} the system occupies one of three states: good (1), poor (2), failed (3) and the decision maker chooses an available action: do nothing, repair and replace. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
  • 11. Introduction Problem Definition System structure For a stochastically failing system in which the degree of deterioration are known only through inspections, I At each inspection epoch k⌧, k = {0, 1, 2, ...} the system occupies one of three states: good (1), poor (2), failed (3) and the decision maker chooses an available action: do nothing, repair and replace. S = {1, 2, 3}; State space Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
  • 12. Introduction Problem Definition System structure For a stochastically failing system in which the degree of deterioration are known only through inspections, I At each inspection epoch k⌧, k = {0, 1, 2, ...} the system occupies one of three states: good (1), poor (2), failed (3) and the decision maker chooses an available action: do nothing, repair and replace. S = {1, 2, 3}; State space T = {0, 1, 2, 3, ...}; Time horizon Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
  • 13. Introduction Problem Definition System structure For a stochastically failing system in which the degree of deterioration are known only through inspections, I At each inspection epoch k⌧, k = {0, 1, 2, ...} the system occupies one of three states: good (1), poor (2), failed (3) and the decision maker chooses an available action: do nothing, repair and replace. S = {1, 2, 3}; State space T = {0, 1, 2, 3, ...}; Time horizon A = {do nothing, repair, replace}; Action space Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
  • 14. Introduction Problem Definition System structure For a stochastically failing system in which the degree of deterioration are known only through inspections, I At each inspection epoch k⌧, k = {0, 1, 2, ...} the system occupies one of three states: good (1), poor (2), failed (3) and the decision maker chooses an available action: do nothing, repair and replace. S = {1, 2, 3}; State space T = {0, 1, 2, 3, ...}; Time horizon A = {do nothing, repair, replace}; Action space It is called minor repair when the repair is performed at state 2 and major repair when it is performed at state 3. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 5 / 35
  • 15. Introduction Problem Definition Deterioration process I The deterioration is modelled by three-state continuous time Markov chain Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 6 / 35
  • 16. Introduction Problem Definition Deterioration process I The deterioration is modelled by three-state continuous time Markov chain If {Yt, t 0} represents the system state at time t, then we assume that {Yt, t 0} is a continuous time Markov chain with transition probabilities, Pij(⌧) = P{Yt+⌧ = j | Yt = i}, 8 t, ⌧ 2 R (1) when it is currently at state i = {1, 2, 3} at time t, will be at state j = {2, 3} at time t + ⌧. (Markovian assumption) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 6 / 35
  • 17. Introduction Problem Definition Deterioration process Good (1) Poor (2) Failed (3) 𝐏𝟏𝟏(𝛕) 𝐏𝟐𝟐(𝛕) 𝐏𝟑𝟑(𝛕) 𝐏𝟏𝟑(𝛕) 𝐏𝟏𝟐(𝛕) 𝐏𝟐𝟑(𝛕) Figure: Transition Probabilities Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 7 / 35
  • 18. Introduction Problem Definition Deterioration process Good (1) Poor (2) Failed (3) 𝐏𝟏𝟏(𝛕) 𝐏𝟐𝟐(𝛕) 𝐏𝟑𝟑(𝛕) 𝐏𝟏𝟑(𝛕) 𝐏𝟏𝟐(𝛕) 𝐏𝟐𝟑(𝛕) Figure: Transition Probabilities I The event that causes a transition within an inspection period from state i to j, where i 6= j, occur after an exponential amount of time with rate ij. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 7 / 35
  • 19. Introduction Problem Definition Assumptions I The time horizon considered is infinite and the system with good state is put into service in time 0. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 8 / 35
  • 20. Introduction Problem Definition Assumptions I The time horizon considered is infinite and the system with good state is put into service in time 0. I Transition from state i = 1, 2 to state j = 2, 3 occurs according to a Poisson Process with rate ij. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 8 / 35
  • 21. Introduction Problem Definition Assumptions I The time horizon considered is infinite and the system with good state is put into service in time 0. I Transition from state i = 1, 2 to state j = 2, 3 occurs according to a Poisson Process with rate ij. I The system state is monitored only through inspections and inspection time is negligible. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 8 / 35
  • 22. Introduction Problem Definition Assumptions I The time horizon considered is infinite and the system with good state is put into service in time 0. I Transition from state i = 1, 2 to state j = 2, 3 occurs according to a Poisson Process with rate ij. I The system state is monitored only through inspections and inspection time is negligible. I Time to inspect is negligible; however, inspection cost is assumed to be a monotonically non-increasing function of time interval between two consecutive inspections [4] to prevent frequent inspections. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 8 / 35
  • 23. Introduction Problem Definition Assumptions I If the system is identified as good, "do nothing" action is required; otherwise decision maker may choose other actions. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 9 / 35
  • 24. Introduction Problem Definition Assumptions I If the system is identified as good, "do nothing" action is required; otherwise decision maker may choose other actions. I Minor (Major) repair takes the system to state 1, with probability pmn (pmj), and to state 2 with probability qmn = 1 pmn (qmj = 1 pmj) after a fixed time tmn (tmj) with a fixed cost cmn (cmj). Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 9 / 35
  • 25. Introduction Problem Definition Assumptions I If the system is identified as good, "do nothing" action is required; otherwise decision maker may choose other actions. I Minor (Major) repair takes the system to state 1, with probability pmn (pmj), and to state 2 with probability qmn = 1 pmn (qmj = 1 pmj) after a fixed time tmn (tmj) with a fixed cost cmn (cmj). I Replace always returns the system to state 1 with a fixed amount cost crpl after a fixed maintenance time trpl. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 9 / 35
  • 26. Introduction Problem Definition Assumptions I If the system is identified as good, "do nothing" action is required; otherwise decision maker may choose other actions. I Minor (Major) repair takes the system to state 1, with probability pmn (pmj), and to state 2 with probability qmn = 1 pmn (qmj = 1 pmj) after a fixed time tmn (tmj) with a fixed cost cmn (cmj). I Replace always returns the system to state 1 with a fixed amount cost crpl after a fixed maintenance time trpl. I The system is not in working condition during repair or replace and hence, the inspection cost does not occur. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 9 / 35
  • 27. Introduction Problem Definition Assumptions I If the system is identified as good, "do nothing" action is required; otherwise decision maker may choose other actions. I Minor (Major) repair takes the system to state 1, with probability pmn (pmj), and to state 2 with probability qmn = 1 pmn (qmj = 1 pmj) after a fixed time tmn (tmj) with a fixed cost cmn (cmj). I Replace always returns the system to state 1 with a fixed amount cost crpl after a fixed maintenance time trpl. I The system is not in working condition during repair or replace and hence, the inspection cost does not occur. I The system state is known instantaneously after each repair. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 9 / 35
  • 28. Introduction Problem Definition Maintenance Policies Five available maintenance policies are proposed shown to emphasise how the maintenance parameters affect the optimum inspection period and policy. Table: Policy Description Policy State 1 State 2 State 3 ⇧1 Do nothing Do nothing Replacement ⇧2 Do nothing Do nothing Major repair ⇧3 Do nothing Minor repair Major repair ⇧4 Do nothing Minor repair Replacement ⇧5 Do nothing Replacement Replacement Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 10 / 35
  • 29. Introduction Problem Definition Optimization criteria Optimizing the maintenance policies is based on Renewal Theory. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 11 / 35
  • 30. Introduction Problem Definition Optimization criteria Optimizing the maintenance policies is based on Renewal Theory. I A cycle ends (a renewal) after a maintenance of the failed system, while the maintenance brings the system to "good" state. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 11 / 35
  • 31. Introduction Problem Definition Optimization criteria Optimizing the maintenance policies is based on Renewal Theory. I A cycle ends (a renewal) after a maintenance of the failed system, while the maintenance brings the system to "good" state. I Then cost rate is the ratio of the expected total cost occurred during the renewal cycle over the expected cycle length. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 11 / 35
  • 32. Introduction Problem Definition Optimization criteria Optimizing the maintenance policies is based on Renewal Theory. I A cycle ends (a renewal) after a maintenance of the failed system, while the maintenance brings the system to "good" state. I Then cost rate is the ratio of the expected total cost occurred during the renewal cycle over the expected cycle length. I The objective is when to inspect the system periodically and what action to be made by minimizing the expected cost rate in the long run. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 11 / 35
  • 33. Mathematical Formulations Mathematical Calculations Transition probabilities Let Xij be the holding time between state i = {1, 2, 3} and state j = {2, 3} . By Poisson arrivals assumption leads to: Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 12 / 35
  • 34. Mathematical Formulations Mathematical Calculations Transition probabilities Let Xij be the holding time between state i = {1, 2, 3} and state j = {2, 3} . By Poisson arrivals assumption leads to: P11(⌧) = P{min{X12, X13} > ⌧} = e 1⌧ (2) P12(⌧) = P{X12  ⌧, X12 < X13, X12 + X23 > ⌧} = 8 >>< >>: 12 e 2⌧ ⌧, 1 2 = 0 12 1 2 e 2⌧ e 1⌧ , 1 2 6= 0 (3) P13(⌧) = P{X12  ⌧, X12 < X13, X12 + X23  ⌧} = 8 >>< >>: 1 e 1⌧ 12 e 2⌧ ⌧, 1 = 2 1 e 1⌧ 12 1 2 e 2⌧ + e 1⌧ , 1 6= 2 (4) where the total transition rate out of a state 1 is 1 = 12 + 13 and of state 2 is 2 = 23. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 12 / 35
  • 35. Mathematical Formulations Mathematical Calculations Expected downtime I How long ago could the failure have first been noticed by an inspection? 1 Inspection time kτ At the time system fails τ' Downtime 3 Inspection time (k+1)τ Figure: Downtime Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 13 / 35
  • 36. Mathematical Formulations Mathematical Calculations Expected downtime D1(⌧) : the expected downtime within an inspection period, ⌧ , of the system initially found at state 1 D2(⌧) : the expected downtime within an inspection period, ⌧ , of the system initially found at state 2 Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 14 / 35
  • 37. Mathematical Formulations Mathematical Calculations Expected downtime D1(⌧) : the expected downtime within an inspection period, ⌧ , of the system initially found at state 1 D2(⌧) : the expected downtime within an inspection period, ⌧ , of the system initially found at state 2 D1(⌧) = DQ (⌧) Q(⌧) + DH (⌧) H(⌧) (5) where 1 Inspection period, τ 32 Q(τ) H(τ) 3 𝐏𝟑𝟑(τ) Figure: Downtime Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 14 / 35
  • 38. Mathematical Formulations Mathematical Calculations Expected downtime DQ (⌧) = ⌧ E[X13 | X13  ⌧, X13 < X12 ] = ⌧ 1 e 1⌧ 1 1 (6) DH (⌧) = ⌧ E[X12 | X12  ⌧, X12 < X13, X12 + X23  ⌧] E[X23 | X12  ⌧, X12 < X13, X12 + X23  ⌧] = 8 >>>>< >>>>: ⌧ 2 1 +e 1⌧ ⇣ 2 1 +⌧ ⌘ 1 e 1⌧ 1e 1⌧ ⌧ , 1 2 = 0 ( 1 2) ⇣ ⌧ 1 1 1 2 ⌘ 1 2 1e 2⌧ + 2e 1⌧ + e 2⌧ ⇣ 1 2 ⌘ e 1⌧ ⇣ 2 1 ⌘ 1 2 1e 2⌧ + 2e 1⌧ , 1 2 6= 0 (7) D2(⌧) = ⌧ 1 e 2⌧ 2 (8) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 15 / 35
  • 39. Mathematical Formulations Mathematical Calculations Average cost rate Let i(⌧) be the cost rate for ⇡i, i = 0, .., 5, The first aim is to calculate each i(⌧) via optimizing the inspection interval time ⌧. The second is to determinate the optimal policy with the correspon- ding minimum cost rate. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 16 / 35
  • 40. Mathematical Formulations Mathematical Calculations Average cost rate For 8 ⇡i, i = 1, .., 5, we define Li j(⌧) as the expected time length the system spends to state 1 from state j = {1, 2, 3}, and Ci j(⌧) as the expected maintenance cost the system spends to state 1 from state j = {1, 2, 3}. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 17 / 35
  • 41. Mathematical Formulations Mathematical Calculations Average cost rate The cost rate of the system is then i(⌧) = Ci 1(⌧) Li 1(⌧) + (⌧) (9) where (⌧) is the inspection cost per time which is a non increasing function of time interval between two consecutive inspections such that lim ⌧!0 (⌧) = K < 1 ve lim ⌧!1 (⌧) = 0 (10) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 18 / 35
  • 42. Mathematical Formulations Mathematical Calculations Average cost rate Cycle Length Li 1(⌧) = 1 1 P11(⌧) ⇥ ⌧ + P12(⌧)Li 2(⌧) + P13(⌧)Li 3(⌧) ⇤ (11) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 19 / 35
  • 43. Mathematical Formulations Mathematical Calculations Average cost rate Cycle Length Li 1(⌧) = 1 1 P11(⌧) ⇥ ⌧ + P12(⌧)Li 2(⌧) + P13(⌧)Li 3(⌧) ⇤ (11) where Li 2(⌧) = 8 >>>>< >>>>: 1 1 P22(⌧) ⇥ ⌧ + P23(⌧)Li 3(⌧) ⇤ i = 1, 2 tmn pmn i = 3, 4 trpl i = 5 (12) Li 3(⌧) = 8 < : tmj + 1 pmj Li 2(⌧) i = 3, 4 trpl i = 5 (13) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 19 / 35
  • 44. Mathematical Formulations Mathematical Calculations Average cost rate Cycle Cost Ci 1(⌧) = 1 1 P11(⌧) ⇥ P12(⌧)Ci 2(⌧) + P13(⌧) cdD1(⌧) + Ci 3(⌧) ⇤ (14) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 20 / 35
  • 45. Mathematical Formulations Mathematical Calculations Average cost rate Cycle Cost Ci 1(⌧) = 1 1 P11(⌧) ⇥ P12(⌧)Ci 2(⌧) + P13(⌧) cdD1(⌧) + Ci 3(⌧) ⇤ (14) where Ci 2(⌧) = 8 >>>< >>>: cdD2(⌧) + Ci 3(⌧), i = 1, 2 cmn+tmn[cd (⌧)] pMn , i = 3, 4 crpl + trpl [cd (⌧)], i = 5 (15) Ci 3(⌧) = 8 >>>< >>>: cmj + tmj [cd (⌧)] +(1 pmj)Ci 2(⌧), i = 3, 4 crpl + trpl [cd (⌧)], i = 5 (16) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 20 / 35
  • 46. Optimal Inspection Time and Optimal Policy Optimization Criteria For any ⌧ 2 (0, 1), the parameters in (11-16) are all positive and finite, then the cost rate is both positive and finite. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 21 / 35
  • 47. Optimal Inspection Time and Optimal Policy Optimization Criteria For any ⌧ 2 (0, 1), the parameters in (11-16) are all positive and finite, then the cost rate is both positive and finite. We need to examine the behaviour of the cost rate as ⌧ ! 0 (continuous inspection) and ⌧ ! 1 (never inspect) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 21 / 35
  • 48. Optimal Inspection Time and Optimal Policy Optimization Criteria Continuous inspection I Continuously inspection ensures that the system state is always known. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 22 / 35
  • 49. Optimal Inspection Time and Optimal Policy Optimization Criteria Continuous inspection I Continuously inspection ensures that the system state is always known. I Failure is detected as soon as the system fails. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 22 / 35
  • 50. Optimal Inspection Time and Optimal Policy Optimization Criteria Continuous inspection I Continuously inspection ensures that the system state is always known. I Failure is detected as soon as the system fails. I The cost rate becomes i(0) = Ci 1(0) Li 1(0) + K (17) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 22 / 35
  • 51. Optimal Inspection Time and Optimal Policy Optimization Criteria Continuous inspection The cycle length leads to: Li 1(0) = 12 1  1 12 + Li 2(0) + 13 1  1 13 + Li 3(0) (18) where Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 23 / 35
  • 52. Optimal Inspection Time and Optimal Policy Optimization Criteria Continuous inspection The cycle length leads to: Li 1(0) = 12 1  1 12 + Li 2(0) + 13 1  1 13 + Li 3(0) (18) where Li 2(0) = 8 >>>< >>>: 1 2 + Li 3(0), i = 1, 2 tmn pmn , i = 3, 4 trpl, i = 5 (19) Li 3(0) = 8 < : tmj + (1 pmj)Li 2(0), i = 3, 4 trpl, i = 5 (20) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 23 / 35
  • 53. Optimal Inspection Time and Optimal Policy Optimization Criteria Continuous inspection The cycle maintenance cost leads to: Ci 1(0) = 12 1 Ci 2(0) + 13 1 Ci 3(0) (21) where Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 24 / 35
  • 54. Optimal Inspection Time and Optimal Policy Optimization Criteria Continuous inspection The cycle maintenance cost leads to: Ci 1(0) = 12 1 Ci 2(0) + 13 1 Ci 3(0) (21) where Ci 2(0) = 8 >>>< >>>: Ci 3(0), i = 1, 2 cmn+tmn(cd K) pmn , i = 3, 4 crpl + trpl(cd K), i = 5 (22) Ci 3(0) = 8 < : cmj + tmj(cd K) + (1 pmj) Ci 2(0), i = 3, 4 crpl + trpl(cd K), i = 5 (23) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 24 / 35
  • 55. Optimal Inspection Time and Optimal Policy Optimization Criteria Never inspect When ⌧ ! 1, the failure of the system occurs with probability 1 which leads that the cycle cost consists of only downtime cost and inspection cost. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 25 / 35
  • 56. Optimal Inspection Time and Optimal Policy Optimization Criteria Never inspect When ⌧ ! 1, the failure of the system occurs with probability 1 which leads that the cycle cost consists of only downtime cost and inspection cost. The cost rate becomes Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 25 / 35
  • 57. Optimal Inspection Time and Optimal Policy Optimization Criteria Never inspect When ⌧ ! 1, the failure of the system occurs with probability 1 which leads that the cycle cost consists of only downtime cost and inspection cost. The cost rate becomes lim ⌧!1 i(⌧) = lim ⌧!1 hcdD1(⌧) ⌧ + (⌧) i = cd + lim t!1 (⌧) = cd (24) Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 25 / 35
  • 58. Optimal Inspection Time and Optimal Policy Optimization Criteria Never inspect When ⌧ ! 1, the failure of the system occurs with probability 1 which leads that the cycle cost consists of only downtime cost and inspection cost. The cost rate becomes lim ⌧!1 i(⌧) = lim ⌧!1 hcdD1(⌧) ⌧ + (⌧) i = cd + lim t!1 (⌧) = cd (24) I i(⌧) is a continuous function and bounded on [0, min{ (0), i(1)}] Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 25 / 35
  • 59. Optimal Inspection Time and Optimal Policy Optimization Criteria Corollary 1 Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 26 / 35
  • 60. Optimal Inspection Time and Optimal Policy Optimization Criteria Corollary 1 If there exists 9 ⌧ 2 [0, 1) such that i(⌧) < cd, then an optimal inspection interval of each policy exists such that ⌧⇤ i 2 1 i ([0, min{ i(0), i(1)}]) otherwise , all policies converge each other (cd) in the long run. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 26 / 35
  • 61. Optimal Inspection Time and Optimal Policy Optimization Criteria Corollary 1 If there exists 9 ⌧ 2 [0, 1) such that i(⌧) < cd, then an optimal inspection interval of each policy exists such that ⌧⇤ i 2 1 i ([0, min{ i(0), i(1)}]) otherwise , all policies converge each other (cd) in the long run. Corollary 2 Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 26 / 35
  • 62. Optimal Inspection Time and Optimal Policy Optimization Criteria Corollary 1 If there exists 9 ⌧ 2 [0, 1) such that i(⌧) < cd, then an optimal inspection interval of each policy exists such that ⌧⇤ i 2 1 i ([0, min{ i(0), i(1)}]) otherwise , all policies converge each other (cd) in the long run. Corollary 2 If cd = 1 and other cost parameters set to zero, the average cost rate gives the system unavailability. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 26 / 35
  • 63. Numerical Example Numerical Example We present several numerical examples which illustrate how system parameters (transition rates, maintenance cots and times) could affect the optimum inspection period and the optimum maintenance policy as well. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 27 / 35
  • 64. Numerical Example Numerical Example We present several numerical examples which illustrate how system parameters (transition rates, maintenance cots and times) could affect the optimum inspection period and the optimum maintenance policy as well. Consider the deterioration model with inspection cost function as (⌧) = Ke 0.1⌧ . Other input parameters are shown below of illustrated figures. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 27 / 35
  • 65. Numerical Example Numerical Example Transition rates Replacement Minor Repair Major Repair Fixed cost 750 300 500 Time to maintenance 1 0.5 2 Probability of repair- success 1 0.9 0.6 Inspection cost parameter 1000 Figure: Relationships between transition rates and inspection interval Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 28 / 35
  • 66. Numerical Example Numerical Example Replacement cost 5100 2 2 2 2 2 3 3 3 3 3 4900 2 2 2 2 2 3 3 3 3 3 4700 3 3 2 2 2 3 3 3 3 3 4500 3 3 3 3 3 3 3 3 3 3 4300 3 3 3 3 3 3 3 3 4 4 4100 3 3 3 3 3 3 3 3 4 4 3900 3 3 3 3 3 4 4 4 4 4 3700 3 3 3 3 3 4 4 4 4 4 3500 4 4 3 3 3 4 4 4 4 4 3300 4 4 4 4 4 4 4 4 5 4 3100 5 4 4 4 4 4 4 5 5 5 2900 5 4 4 4 4 5 5 5 5 5 2700 5 5 4 4 4 5 6 5 5 5 2500 6 5 5 5 5 5 6 6 6 6 2300 6 6 6 6 6 6 6 6 6 6 2100 7 7 7 7 7 𝛑 𝟎 6 7 7 7 7 1900 7 7 7 7 7 𝛑 𝟏 7 7 7 7 7 1700 8 8 8 8 8 𝛑 𝟐 8 8 8 8 8 1500 9 9 9 9 9 𝛑 𝟑 9 9 9 9 9 1300 10 11 11 11 11 𝛑 𝟒 11 11 11 11 11 1100 12 12 13 13 13 𝛑 𝟓 12 13 13 13 13 900 15 15 15 15 15 15 16 16 16 16 700 19 20 20 20 20 20 20 21 21 21 500 26 27 27 27 27 27 28 28 28 28 300 38 39 39 39 39 38 39 40 40 40 100 60 65 66 66 66 60 66 66 66 66 75 66 75 75 75 75 66 75 76 76 76 50 76 103 108 108 108 76 102 108 108 108 250 500 1000 2500 5000 250 500 1000 2500 5000 Repla- cement Minor Repair Major Repair Fixed cost 𝐜 𝐫𝐩𝐥 300 500 Time to maintenance 0.001 0.001 0.001 Probability of repair-success 1 0.9 0.6 Inspection cost parameter 1000 Repla- cement Minor Repair Major Repair Fixed cost 𝐜 𝐫𝐩𝐥 300 500 Time to maintenance 1 0.5 2 Probability of repair-success 1 0.9 0.6 Inspection cost parameter 1000 Cost of replacement, 𝐜 𝐫𝐩𝐥 Cost of downtime, 𝐜 𝐝 Figure: Relationships between downtime cost and replacement cost Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 29 / 35
  • 67. Numerical Example Numerical Example Inspection cost Cost of downtime, 𝐜 𝐝 Cost of Inspection Parameter, K Repla- cement Minor Repair Major Repair Fixed cost 1000 300 500 Time to maintenance 0.001 0.001 0.001 Probability of repair-success 1 0.9 0.6 Inspection cost parameter K 5100 0 1 2 13 25 0 0 3 13 26 4900 0 1 2 13 26 0 0 3 14 26 4700 0 1 2 14 27 0 0 3 15 27 4500 0 1 3 14 28 0 0 3 15 28 4300 0 1 3 15 29 0 0 3 16 29 4100 0 1 3 16 30 0 0 4 17 30 3900 0 2 3 17 31 0 0 4 17 31 3700 0 2 3 18 32 0 0 4 18 32 3500 0 2 3 19 33 0 0 4 19 34 3300 0 2 4 20 34 0 0 4 20 35 3100 0 2 4 21 36 0 0 5 22 36 2900 0 2 4 22 37 0 3 5 23 38 2700 0 2 4 24 39 0 3 5 24 39 2500 0 2 5 25 40 0 3 6 26 41 2300 0 3 6 27 42 0 3 6 28 43 2100 0 3 7 29 44 𝛑 𝟎 0 4 7 30 44 1900 0 3 7 32 46 𝛑 𝟏 0 4 7 32 46 1700 0 4 8 34 49 𝛑 𝟐 0 4 8 35 48 1500 0 4 9 37 50 𝛑 𝟑 0 5 10 37 51 1300 0 5 11 40 53 𝛑 𝟒 0 6 11 40 53 1100 0 6 13 43 56 𝛑 𝟓 0 7 13 44 56 900 2 7 15 47 59 0 8 16 48 59 700 2 10 20 52 64 0 10 21 53 64 500 3 14 27 58 69 4 15 28 60 69 300 5 24 39 67 77 6 25 40 67 77 100 20 55 66 89 98 21 55 66 89 98 75 31 65 75 97 106 32 65 76 97 106 50 72 98 108 128 135 73 99 108 126 135 100 500 1000 5000 10000 100 500 1000 5000 10000 Repla- cement Minor Repair Major Repair Fixed cost 1000 300 500 Time to maintenance 1 0.5 2 Probability of repair-success 1 0.9 0.6 Inspection cost parameter K Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 30 / 35
  • 68. Numerical Example Numerical Example Inspection cost Cost of major repair, 𝐜 𝐦𝐣 Maintenance time for major repair, 𝐭 𝐦𝐣 5100 3 3 3 3 3 2 2 3 3 3 4900 3 3 3 3 3 2 2 3 3 3 4700 3 3 3 3 3 2 3 3 3 3 4500 3 3 3 3 3 3 3 3 3 3 4300 4 4 4 4 4 3 3 4 4 4 4100 4 4 4 4 4 3 3 4 4 4 3900 4 4 4 4 4 3 3 4 4 4 3700 4 4 4 4 4 3 3 4 4 4 3500 4 4 4 4 4 3 3 4 4 4 3300 5 5 5 5 5 4 4 5 5 5 3100 5 5 5 5 5 4 4 5 5 5 2900 5 5 5 5 5 4 4 5 5 5 2700 5 5 5 5 5 4 5 5 5 5 2500 6 6 6 6 6 5 5 6 6 6 2300 6 6 6 6 6 5 5 6 6 6 2100 7 7 7 7 7 6 6 7 7 7 1900 7 7 8 8 8 𝛑 𝟏 7 7 7 8 8 1700 8 8 9 9 9 𝛑 𝟐 7 7 8 9 9 1500 9 9 10 10 10 𝛑 𝟑 9 9 9 10 10 1300 11 11 11 12 12 𝛑 𝟒 10 10 11 12 12 1100 13 13 13 14 14 𝛑 𝟓 12 12 13 14 13 900 16 16 17 18 18 15 15 16 17 18 700 20 21 22 23 23 20 20 21 22 24 500 27 28 30 32 33 27 28 28 29 31 300 38 40 43 48 50 39 40 40 41 42 100 62 66 99 >1000 >1000 66 66 66 67 68 75 68 76 >1000 >1000 >1000 75 75 76 76 77 50 79 108 >1000 >1000 >1000 108 108 108 109 109 300 500 1000 1500 1750 0.001 0.1 2 5 10 Repla- cement Minor Repair Major Repair Fixed cost 2000 300 𝐜 𝐦𝐫 Time to maintenance 1 0.5 2 Probability of repair-success 1 0.9 0.6 Inspection cost parameter 1000 Repla- cement Minor Repair Major Repair Fixed cost 2000 300 𝟓𝟎𝟎 Time to maintenance 1 0.5 𝐭 𝐦𝐫 Probability of repair-success 1 0.9 0.6 Inspection cost parameter 1000 Cost of downtime, 𝐜 𝐝 Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 31 / 35
  • 69. Numerical Example Numerical Example Maintenance probabilities Probability of Major repair-success, 𝐩 𝐦𝐣 1 0,8 0,6 0,4 0,2 0,2 0,4 0,6 0,8 1 0,2 0,4 0,6 0,8 1 1 0,8 0,6 0,4 0,2 0,2 0,4 0,6 0,8 1 0,2 0,4 0,6 0,8 1 𝛑 𝟎 𝛑 𝟏 𝛑 𝟐 𝛑 𝟑 𝛑 𝟒 𝛑 𝟓 Probability of Minor repair-success, 𝐩 𝐦𝐧 I. 𝐜 𝐫𝐩𝐥 = 𝟓𝟎𝟎 II. 𝐜 𝐫𝐩𝐥 = 𝟔𝟓𝟎 III. 𝐜 𝐫𝐩𝐥 = 𝟕𝟓𝟎 IV. 𝐜 𝐫𝐩𝐥 = 𝟏𝟎𝟎𝟎 Replace ment Minor Repair Major Repair Fixed cost 𝐜 𝐫𝐩𝐥 150 500 Time to maintenance 1 0.5 2 Probability of repair- success 1 𝐩 𝐦𝐧 𝐩 𝐦𝐣 Inspection cost parameter 1000 Down Time Cost 400 Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 32 / 35
  • 70. Numerical Example Numerical Example Maintenance times Maintenance time for major repair, 𝐭 𝐦𝐣 Replace ment Minor Repair Major Repair Fixed cost 750 150 500 Time to maintenance 𝐭 𝐫𝐩𝐥 𝐭 𝐦𝐣 𝐭 𝐦𝐧 Probability of repair- success 1 0.9 0.6 Maintenance time for minor repair, 𝐭 𝐦𝐧 Inspection cost parameter 1000 Downtime Cost 400 𝛑 𝟎 𝛑 𝟏 𝛑 𝟐 𝛑 𝟑 𝛑 𝟒 𝛑 𝟓 I. 𝐭 𝐫𝐩𝐥 = 𝟎. 𝟎𝟏 II. 𝐭 𝐫𝐩𝐥 = 𝟎. 𝟏 III. 𝐭 𝐫𝐩𝐥 = 𝟎. 𝟓 IV. 𝐭 𝐫𝐩𝐥 = 𝟏 2,5 2 1,5 1 0,5 0,5 1 1,5 2 2,5 0,5 1 1,5 2 2,5 2,5 2 1,5 1 0,5 0,5 1 1,5 2 2,5 0,5 1 1,5 2 2,5 Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 33 / 35
  • 71. Numerical Example Thank you! Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 34 / 35
  • 72. References References Barlow, R., Hunter, L., Optimum preventive maintenance policies. Operations Research , 8 (1): 90-100, 1960 Barlow, R.E., Proschan, F., Mathematical theory of reliability. New York: Wiley 1965 Luss, H., Maintenance policies when deterioration can be observed by inspections. Operation Research, 24 (2): 359–366, 1976 Zuckerman, D., Inspection and replacement policies. Applied Probability, 17 (1): 168–177, 1980 Christer, A.H., Waller, W.M, Delay time models of industrial inspection maintenance models. Operational Research Society, 35 (5): 401–406, 1984 Chelbi, A., Ait-Kadi, D., Replacement strategy for non self announcing failure equipment. Emerging Technologies and Factory Automation 1, pp.423-430, 1994 Chelbi, A., Ait-Kadi, D., An optimal inspection strategy for randomly failing equipment. Reliability Engineering and System Safety 63 (2), pp.127-131, 1999 Klutke, G.A., Yang, Y., The availability of inspected systems subject to shocks and graceful degradation. IEEE Transition Reliability, 51 (3), pp.371-374, 2002 Wang, W., An inspection model for a process with two types of inspections and repairs. Reliability Engineering and System Safety, 94 : 526–533, 2008. Wang, W., An inspection model based on a three-stage failure process. Reliability Engineering and System Safety, 96 :838–848, 2011. Salih Tekin N. Onur Bakır, Bü¸sra Kele¸s (TOBB University of Economics and Technology Ankara, Turkey)Maintenance Policies For A Deterioration System Subject To Non-Self-Announcing FailuresJune 6, 2014 35 / 35