Electrical Power and Energy Systems 73 (2015) 350–360

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Rational Reliability Centered Maintenance Optimization for power

distribution systems
B. Yssaad a,⇑, A. Abene b
a
Research Laboratory: Industrial Engineering and Sustainable Development (IESD), University Center Ahmed Zabana of Relizane, Algeria
b
Euro-Mediterranean Institute of Environment and Renewable Energies (EIERE), University of Valenciennes and Hainaut Cambrai, ISTV Mount Huy, F-59304 Valenciennes, France

a r t i c l e i n f o a b s t r a c t

Article history: Reliability Centered Maintenance Method (RCM) currently used for maintenance optimization based on
Received 6 June 2014 the analysis of modes of dysfunctional system failures (FMEA), and after determining the criticality of the
Received in revised form 29 April 2015 system, an adequate maintenance plan is adopted. Following the succession of several applications of this
Accepted 5 May 2015
method, in several areas, along twenty years of experience, it was proved that the use of this approach
was incomplete for other analysis and evaluation criteria were completely ignored especially in electrical
systems. Our view is to apply this method rationally by introducing a whole dependability study (RAMS).
Keywords:
Ó 2015 Elsevier Ltd. All rights reserved.
Power distribution systems
Rational reliability centered maintenance
RRCM
Analysis of failure modes effects and
criticality
Optimization
Modeling
Costs

Introduction exiting the dreaded condition for each component of the electrical
network based on cost. And parallel with the modeling we will do a
In this work (extracted and summarized from my doctoral thesis) cost analysis of the life cycle to estimate the total cost of ownership
[1], we have applied a maintenance optimization approach based of the electricity networks of each analyzed element. Finally,
on techniques of modeling and simulation by applying the experi- everything is completed and simulated by the Monte Carlo
mental approach RCM that based on the functional and dysfunc- simulation in order to give important recommendations to keep
tional analysis of failure modes of the grid by using the FMEA the electrical system in the best circumstances [1–3].
model, and by using the SADT approach for decomposing hierar-
chically these different functional analysis tasks to classify the
material in order of criticality. Then we have modeled the system Presenting the new approach: Rational RCM
by using static methods (combinatorial, deductive, functional and
qualitative approaches) such as reliability diagram RBD and fault The implementation of a simplified maintenance plan
trees analysis FTA that identify the paths of weakness in the sys- comprises 4 steps. These steps call on many data whose supports
tem and determine these characteristics RAMS, by using data of are relating to production, quality, and maintenance. All
REX feedback that are collected from the old method (scheduled along these steps, involved groups must determine objectives
periodic preventive maintenance SPPM); then we determine the having priority and validate results of each phase in order to
characteristics of dependability and reliability indices (time, rate continue without over diversifying their work [1–4,10,12,13]
and cost) by dynamic methods (analytical, behavioral, quantitative (see Fig. 1).
and stochastic approaches) such as Markov graphs MG and Petri
network PSN, and therefore the determination of residence times,
the execution frequency of the program and the probability of Case studies: Application to the power distribution systems

To illustrate our view in this work, we will give a general

⇑ Corresponding author. Tel.: +213 550232405; fax: +213 46910098. description of the electrical distribution station located in the
E-mail address: [email protected] (B. Yssaad). region of RELIZANE Northwest of ALGERIA. The suggested

http://dx.doi.org/10.1016/j.ijepes.2015.05.015
0142-0615/Ó 2015 Elsevier Ltd. All rights reserved.
 B. Yssaad, A. Abene / Electrical Power and Energy Systems 73 (2015) 350–360 351

Table 1
Data of reliability.
Determination of site to study 1
Equipment Failure mode MTTF MTTR Estimated
(years) (h) average cost of
repair/year (x
103 AD)
2 Risk assessment 3 History of dependability
and feedback Electrical line EL Clippings 3.5 2.5 345
Analyses of functional Bus bars SB Short-circuit 8 4 210
failures Insulator IS Ageing 4 3 135
Determination of
characteristics and evaluation Priming
of RAMS reliability indicators Transformer Low insulation 5.5 3.5 240
Research of components and (λ, μ …) RBD, FTA, MG, PTR resistance
modes of significant failures PN Triggering power at the
(FMECA, HAZOP … etc.) time of operation
shedding
Noise: the necking
Event and Economic Analysis phenomena in
(maintenance costs) LCCA transformers
Research components and
Breaker CB Long and short 8 5 165
critical failure modes
interruptions
Arresters PL Overvoltage 8 2.5 75
Fuse F Overloads 2 1 75
Sectionalizer – Closing difficult or 7.5 5 30
SW – Opening difficult
Analysis and selection of maintenance tasks

Corrective
Materials and failures
maintenance
Not critical action Table 2
Critical Probability of occurrence (Safety).

Final selection and preparation of preventive maintenance Criteria Frequency (O) Value
Failure by 10 years Very rare failure 0.1
Maintenance optimization 4
Failure by 5 years Rare failure 0.2
Failure by 3 years Possible failure 0.3
Fig. 1. Steps of RCM rational approach. Failure by one year Frequent failure 0.4
Failure by 4–6 months Very common failure 0.5

L1: 60 KV L2 : 60 KV
PL Table 3
G
Severity (Availability).
Coupling
CB Criteria Severity (S) Value

30 KV Power interruption < 5 min Minor 0.1

SW Power interruption < 20 min Significant 0.2
SB1 25 min < Power interruption < 30 min Average 0.3
SB2 1 h < Power interruption < 2 h Major 0.4
5 h < Power interruption < 1 day Catastrophic 0.5
30 KV

Table 4
CT
Detectability (Maintainability).
VT
Criteria Detectability (D) Value
Time taken to detect the failure Obvious detection 1
PTR Failure early detectable Possible detection 2
Failure is difficult to detect Detection unlikely 3
Failure painfully detectable Detection improbable 4
Failure undetectable Cannot detect 5

Fig. 2. Distribution network station of RELIZANE.

Table 5
Partition equipment and action plan.
alternative for the Power Distribution Systems can be schematized
as follows (Fig. 2) [1–3]. Condition Consequence Maintenance action
imposed

Data reliability material according to the method SPPM C<R No problem, nothing to Correctives de maintenance
report
R<C<S Acceptable but! Special surveillance preventive/
These data are collected and calculated in collaboration with predictive maintenance
engineer’s service station maintenance of Subsidiary Power C>S Complete questioning – Systematic preventive maintenance
Distribution System PDS SONELGAZ of RELIZANE [1–3] of the study action
– New ameliorative study.
(see Tables 1–5).
352 B. Yssaad, A. Abene / Electrical Power and Energy Systems 73 (2015) 350–360

Risk assessment of power distribution systems by using the FMECA this step, we projected phases of RCM on the SADT model [1,2]
method (see Fig. 3).
The glossary in (Table 6) is essential in the SADT method
The FMECA (Failure Modes, Effects and Criticality) is a method because this approach is based on symbolization of all effects influ-
of predictive reliability analysis that identified the potential fail- encing its description.
ures whose consequences affect the proper functioning of the sys- The experimental results of the FMECA analysis is summarized
tem being studied, and then estimate the risks related to the in (Table 7) below, and therefore we moved to different simula-
occurrence of these failures to take action to provide maintenance tions and modeling to confirm this practical result and therefore
during the design, implementation or operation of the system [01, to come up with important recommendations to put at the disposal
32, 33]. of the maintenance department, whose purpose is to improve
management of the electrical systems [1–3].
Risk estimation
When FMEA is completed, an analysis of risk assessment is
The determination of optimal parameters failure rate k, repair rate l,
realized for all the previously determined failures. The criticality
RAMS, cost and reliability indicators
C or the risk priority number RPN is given by:
RPN ¼ S  O  D ð1Þ After realizing the FMECA analysis in the Relizane PDS post, we
could identify and rank the most critical components during the
where
trial period on electrical network (see Fig. 4).
O: Represents the probability or the frequency of occurrence
The FMEA analysis leads to two complementary descriptions of
that causes the failure to display and leads the considered potential
the network:
mode of failure. The rating scale ranges between 0.1 and 0.5.
S: Represents the severity. The rating scale ranges between 0.1
– A description formalized by functional block diagrams RBD and
and 0.5 based on the failure caused by the effects in terms of:
fault tree FTA whose purpose is to present the system architec-
ture and functional links between the different elements of the
– Response time
system and determine the characteristics RAMS.
– Quality power supply
– Safety of people and property

D: Represents the probability that the cause or failure mode

supposed to appear leads to the most serious effect, without
Table 6
detecting the failure previously. The rating scale ranges from 1 to 5. The glossary.
From these criticality indexes, it is possible to prioritize the
MPD: Maintenance Programming Documents; HD: Head of Division;
failures and identify those whose criticality level is greater than
RD: Reporting Document; HOM: Head of the Office of Maintenance;
a constant limit. It is contractually imposed as follows: [5–7,14]. WF: Working frequency; MOS: Method Office Service;
WP: Work Program; MT: Maintenance Team; Ob: Observation;
Modeling of approach RCM by using SADT method S GRTE: staff of GRTE; T and V:Tools and Vehicles;
EXP R: Experiences of Responsible; ISO 9000: Standard ISO 9000.

The functional analysis of the RCM approach describes, in visual NODE G1 TITLE The glossary NUMBER 1
and textual form, the role of network and/or the components. In

Optimzation of
the maintenance
by RCM
A0
WF ISO and
Periodicity
OHSAS
OB
Site is Failures are
MDP selected analyzed
Determination
of sites to study
Post
PDS A1
RAMS are
Analyses of determined
RD functional
failures Project is
valited
A2 Post PDS
Setting up is
TMP maintened
Study of the
A4 Optimization
safty
and experince
A3 feedback
A5

SG
Intervention
TV report
and update
MOS MT HD HOM

Node : 1 Title : Optimization of the maintenance by RCM N° : 2

Fig. 3. Modeling of approach RCM by using SADT method.

Table 7
Analysis of failure modes effects and criticality FMECA.

[PDS of R in A]a Analysis of failure modes effects and criticality FMECA

Electrical distribution station located in the region of RELIZANE Northwest of ALGERIA Date: 20/05/2013

Element Function Failure mode Cause Effect Detection Criticality Maintenance plan adopted: Action to be taken

R = 0.540 and S = 0.999

O S D C The test

Electrical Line EL Electrical – Contact of two Electrodynamics force of the wind – Line break Visual 0.5 0.5 4 1 C>S – Corrective action: If the fault is minor
transport lines – Preventive systematic maintenance action
– A new study, using cables of the same material
(replacement)
Insulator out of glass, or Insulation – Aging – Partial or complete discharge of the arc in – Perforation Visual 0.5 0.5 4 1 C>S – Corrective action: it can be covered with grease or washed
ceramics IS its volume regularly
– Priming – Over voltage (lightning) streamer surface – Preventive systematic maintenance action
by an electric arc
Bus Bars SB Selection of lines Short-circuit – Air pollution (depositing a layer of pollution – A new study, using composite insulator (replacement)
on the surface of the insulator, progressive

B. Yssaad, A. Abene / Electrical Power and Energy Systems 73 (2015) 350–360

wetting layer of pollution, development of
dry lands and appearance of partial arcs
– Extension of partial arcs if conditions are – Visual 0.5 0.4 3 0.6 R<C<S – Preventive systematic maintenance action
favurable to the insulator boot) (dusty Deformation
environments near-desert or salted ones)
Power Transformer PTR Used to raise or 0.5 0.5 4 1 C>S Special surveillance, conditional/preventive maintenance
lower the – Review the fasteners and clamping panels and supports
voltage level of – Check the mechanical condition of the transformer
an alternative – Noise – The vibration of the metal envelope, the – Excessive – Noise – Measurements of noise
power source pollutions: sets of bars and the maker badge heating
phenomena of
constriction in
the transformers
– Low isolation – An abnormal dust contamination. – An – – Visual – Check the timing and balance coils’ hypertension
resistance incrassate by fatty vapor’s and dust – The Deformation
electronic converters are badly connected or
the supply voltage does not conform. The
fuses are badly gauged. The transformers of
current (TC) are saturated by the currents
with interlocking. Starting with the mass or
between whorls of rolling up of the
transformer
– Triggering of – No voltage – Heat – Review the determination of the plate
the transformer
to the powering
at the time of the
Unballasting
operation
– Inhale the dust and clean all accessible parts then blow the
transformer to the ‘nitrogen or dry air
– Use a cold degreaser to clean the resin and blocking
– Check connection and power with specified voltages on the
case of electronic converters
– Change the fuse rating
– Revise the definition of current transformers
– Perform a visual inspection and diagnosis with measurement
of insulation resistance of windings
– In the critical case, if the Buchholz relay is open? Then the
replacement of the transformer will be required
Arrester PL Protection Long and short – The lightning 0.2 0.1 2 0.04 C<R Corrective maintenance
interruptions – Maneuver of equipment – Breakdown – Visual – To maintain the lightening protectors
of materials
– Failure isolation – Control of the default isolation of the earth electrodes
– choice of the level of insulation
Circuit Breaker CB Protection and Over voltage – Food network – Stop of – Noise 0.5 0.5 4 0.5 C>S – Preventive systematic maintenance action
control equipment
– Reset – Stop of
installation
– Internal Failures – Production – Visual – New study (Improvement)

353
loss

(continued on next page)

354 B. Yssaad, A. Abene / Electrical Power and Energy Systems 73 (2015) 350–360

Reliability block diagrams

– To inhale the dust and clean all accessible parts then blow
This is an almost direct representation of the functional vision
system, through a series parallel juxtaposition, or mixed blocks
associated with basic entities of the system. In this representa-
tion, the blocks represent elements (hardware or events) whose
Maintenance plan adopted: Action to be taken

failures causes a breakdown of the system. The reliability dia-

the transformer to the ‘nitrogen or air dry

gram is a graph without circuit. It admits an input and output.
The vertices, called blocks represent the system elements. Its arcs
are the relationships between the various elements [1,2,9] (see
– Editing architecture system

Fig. 5).
This model is regulated by the resulting equation simulation of
the reliability from following Monte Carlo simulation.
– Replacement
Replacement

Y
n Y
p
RðtÞ ¼ ½1  ½1  Rij ðtÞ ð2Þ
1¼1 j¼1
S = 0.999

The test

R<C<S
R<C<S

The dynamic fault tree DFT

The fault tree identifies the paths of the system weakness, its
purpose is to determine the initial reason of the occurrence
of an unwanted event and the probability that it occurs. It
0.6
0.6
C

proceeds by a combination of faults leading to unwanted or feared

and
Date: 20/05/2013

4
3

event.
Fig. 6, above, shows the shaft events example of the studied sys-
0.3
0.4
Criticality

R = 0.540

tem which leads to the of power supply loss.

0.5
0.5
O

– A behavioral description, Markov graph and Petri nets whose

purpose is to describe the sequence of different states and
determine the characteristics of the dependability and reliabil-
Detection

– Visual

ity indicators.
– Noise
– Heat

Visual

Application of Markov graphs

The Markov chain method or Space States were developed in
Dysfunction
Electronic

Blocking

the 50’s for reliability analysis of reparable systems [1,9,11]. This

– Power

Fusion
Effect
Analysis of failure modes effects and criticality FMECA

method is to represent the dynamic behavior of a system by a

set of components that may be present in finite, operating states
and failures so as to calculate the probabilities of various system
states occurrence and the characteristics of dependability (MTTF,
MTBF, MTTR, etc.) and asymptotic availability [2,3].
To avoid such large number of differential equations with the
complexity of the resolution out the fact that these equations are
highly interdependent with each other on one hand, and having a
– Permutations sources

– Permutations sources

single equation with a coherent result, on the other hand, we

– Internal Failures

minimized the Markov model corresponding to our study as

– Food network
Electrical distribution station located in the region of RELIZANE Northwest of ALGERIA

follows:
Hence, the system of equations state space of the system ele-
PDS of R in A: Power Distribution System of RELIZANE in ALGERIA.
– Reset
Cause

ments or the linear differential equations Chapman–Kolmogorov

system is as follows:
8
dP½P 0 ðtÞ
Failure mode

>
Over current

> dt ¼ ðk1 þ k2 þ k3 þ k4 þ k5 þ k6 þ k7 ÞP0 ðtÞ

>
– Opening

>
– Closing

>
> !
difficult

difficult

< l1 P1 ðtÞ þ l2 P2 ðtÞ þ l3 P3 ðtÞ þ l4 P4 ðtÞ þ l5 P5 ðtÞ

þ ð3Þ
or

>
> þl6 P6 ðtÞ þ l7 P7 ðtÞ
>
>
>
>
:
P1 þ P2 þ P3 þ P4 þ P5 þ P6 þ P7 ¼ 1  P0
– Electrical
– Galvanic

separation
Protection
Function

isolator

Modeling a system with independent components. As each state is

subject to a similar equation to the other, all of these equations
can be collected in a matrix form, which is the basic formula of
homogeneous Markov processes (system of differential equations
Table 7 (continued)

of Chapman–Kolmogorov) [1–3]:
Sectionalizers SW
[PDS of R in A]a

(
dPðtÞ
dt
¼ M  Pi ðtÞ
P ð4Þ
Element

Fuses F

n
i¼0 P i ¼ 0
a

where
 B. Yssaad, A. Abene / Electrical Power and Energy Systems 73 (2015) 350–360 355

ReliaSoftBlockSim7 -
FAILURE CRITICALITY INDEX (FCI) The resolution and the result of this equation are given in the
26,93 AVAILABILI
100
Appendix A.
2154 91
Stochastic Petri nets
83
16,16
5 To represent the operation of the maintenance optimization of
our studied processes system, we used the graphical modeling by
10,77 using SPN and it all depends on the Markov graph Fig. 7.

5,38 Life cycle cost analysis LCCA of the electrical network

In general, the LCCA model uses the following typical inputs
0,00
EL IS PTR CB SB F SW PL of those generated by functional services such as (research and
development, production and construction, operation and main-
Fig. 4. The critical equipment. tenance, retirement and disposal). Pressing the discount rate,
the assumption can be made on the basis of relevant cost esti-
mates for the problem in question. The discount rate is the rate
M = [Pi,j]i,j2 E: transition rate matrix (or generator matrix) of
of change in the fair value of money over time, taking into
process Markov
account the fluctuations in investment interest rates and infla-
P(t) = (P1(t), P2(t), . . .): Probability vector occupancy states.
tion. Projections, of individual costs for each alternative, should
be discounted to present value. [1] The LCCA is determined by
Automatic construction of the generator matrix. For a large system, it
using the discounted multi-year basis following formula
is difficult to manually construct the generator matrix, and an
[1,2,8,9,11]:
automatic construction becomes necessary.
 n
This is done from the relationship 5 by using the Kronecker 1
CV ¼ ð9Þ
sum. 1 þ Ty
M ¼ M1  M2 . . .  Mn ð5Þ where
where CV: present value, Ty: time.And therefore, the maintenance life
Mi  Mj ¼ Mi  I þ I  Mj, for i, j = 1, . . . , n. I is the identity cycle cost analysis LCCA is the sum of each year costs for LCA. It
matrix. is calculated and multiplied by the factor of the current value CV.
This gives the current value of the sum of the costs LCA, CV by
Determination of the transition matrix M. If we correspond respec- LCA for each year that is, then, summarized over the years in the
tively the system components (SB, L, PL, CB, DS, IS, PT and F) by total amount of the present value of the cost CV, CV by LCA.
(P0, P1, P2, P3, P4, P5, P6 and P7), the transition matrix of the X
n

Markov model characterizing Fig. 7 is given by the following LCCA ¼ LCA þ C V i  LCAi ð10Þ
i¼1
equations:

2 dP 0
3
2
6
dt
dP1
7 ðk0 þ k1 þ k2 þ k3 þ k4 þ k5 þ k6 þ k7 Þ l1 l2 l3 l4 l5 l6 l7 32 P0 3
6 7 6 76 P1 7
6 dt 7 6 k1 l1 0 0 0 0 0 0 76 7
6 dP2 7 6 76 7
6 7 6 k 0 l2 0 0 0 0 0 76 P2 7
6 dt 7 6 2
76 7
6 dP3 7 6 76 P 7
6 7 6 k3 0 0 l3 0 0 0 0 76 3 7
6 dt 7¼6 76 7 ð6Þ
6 dP4 7 6 k 0 0 0 l4 0 0 0 76 P4 7
6 dt 7 6 4
76 7
6 7 6 76 P 7
6
6
dP5 7 6
7 6 k5 0 0 0 0 l5 0 0 76 5 7
6 dt
7 4 76 7
6 dP6 7 k6 0 0 0 0 0 l6 0 54 P6 5
4 dt 5
dP7
k7 0 0 0 0 0 0 l7 P7
dt

P1 þ P2 þ P3 þ P4 þ P5 þ P6 þ P7 ¼ 1  P0 ð7Þ The following figure shows us the evolution and distribution

expenses for equipment distribution system of electricity and the
To solve this equation, we must replace Eq. (7) in the equation
percentage of maintenance costs made by each member prior to
of state space 6. We have:

2 3 2
0 ðk0 þ k1 þ k2 þ k3 þ k4 þ k5 þ k6 þ k7 Þ l1 l2 l3 l4 l5 l6 l7 3 2 P 0 3
607 6 l1 0 0 0 0 0 0 7 6 7
6 7 6 k1 76 P 1 7
6 7 6 76 7
607 6 k2 0 l2 0 0 0 0 0 76 P 2 7
6 7 6 76 7
607 6 0 0 l3 0 0 0 0 7 6 7
6 7 6 k 3 76 P 3 7
6 7¼6 76 7 ð8Þ
607 6 k4 0 0 0 l4 0 0 0 76 P 4 7
6 7 6 76 7
607 6 0 0 0 0 l5 0 0 7 6 7
6 7 6 k5 76 P 5 7
6 7 6 76 7
405 4 k6 0 0 0 0 0 l6 0 54 P 6 5
1 k7 0 0 0 0 0 0 l7 P7
356 B. Yssaad, A. Abene / Electrical Power and Energy Systems 73 (2015) 350–360

PL3
PT3
DS23
CB3 CT3 DS35
1/2.3 2/2.3
CBB DSB VT3
DS13 Node
DS34
S3
Line B DSB VTB
PL2
PT2

PLB SB/ISB DS22 CB2 CT2

DS25 END
Starting 2/3 1/2.2 2/2.2 VT2 Block
Block
PLA SB/ISA DS12 Node
DS24
S2

Line A DSA VTA Node

DS21 DS14
S1
1/2.1 2/2.1
VT1
CBA DSA DS11
CB1 CT1 DS15

PT1
PL1

Fig. 5. Reliability Block diagrams.

Failing system
Post VHV

OR

Choice of
power line AND

VT
Line in or / out Line in or / out Line in or / out
servce servce servce

OR OR OR
OR OR

OR OR OR OR OR OR
OR SB/ISA OR SB/ISB

Choice of Node Choice of Node

1/2.1 AND Node Choice of AND
Line Line power line S1 1/2.2 power line AND S3
AND AND S2 1/2.3 power line
A B
2/2.1 2/2.2
2/2.3

OR DS14 OR DS24 OR DS34

OR DSA PLA OR DSB PLB
DS11 DS21 DS12 DS22 DS13 DS23

CB1 OR CB2 OR CB3 OR

CBA AND CBB AND

CT1 AND CT3 AND

CT2 AND

VTA DSA VTB DSB

OR VT1 OR VT3
OR VT2

PT1 AND PT2 AND PT3 AND

PL1 DS15 PL2 DS25 PL3 DS35

Fig. 6. Ddynamic fault tree DFT.

 B. Yssaad, A. Abene / Electrical Power and Energy Systems 73 (2015) 350–360 357

The result of modeling and simulation of the three approaches:

CB(P4) SPPM, RCM and RRCM
PL(P 2 )
λ2 μ3 λ3 DS(P5) Through this study, carried out on the optimization of
μ4
maintenance of PDS station of Relizane by the rationalized RCM
μ2
λ1 λ4 model, based on the twinning of the experimental, modeling and
SB(P0)
μ1 μ5
simulation, it appears quite, clearly, that the obtained result by
L(P1)
IS(P6) the application of this model is better while compared with the
λ7 μ7 μ6 λ5 use of the RCM model and much better compared to the old model
λ6 SPPM. Then,
F(P3) PT(P7) The results for the number of preventive maintenance (MP) are
presented in Fig. 11. The situation of steady state is reached after
about three replacements. In the analyzed case, the stability is
Fig. 7. Markov graph minimized MG. reached after 27 years of service and that after the third MTBF.
However, as shown in Fig. 12, the expenses incurred in the stable
the adoption and implementation of a maintenance plan opti- situation are not considered as an interest from today’s point of
mized. Thus we see clearly that the majority of maintenance costs view, because the present value (PV) of these costs converges to
are absorbed by the power transformer, power lines and bus bars. zero after a few decades. Thus, the (VA) costs for (MC, MP) and
In addition, a cost of materials improved life cycle of the electrical inspections are not summarized in the total cost (TC) in the first
system is an essential task for the reliability, availability and safety decades.
of the power system (see Figs. 8–10).

Fig. 8. Stochastic Petri nets.

Fig. 9. The correspondence MG and SPN.

358 B. Yssaad, A. Abene / Electrical Power and Energy Systems 73 (2015) 350–360

Cost for distribution equipment maintenance before Thus, the expected number of preventive maintenance per year
optimization (following SPPM) in Markov model and simulated by the Monte Carlo simulation is
L
2% shown in the following figure:
11% SB
5% 24% Finally, the graph shows clearly that the optimal intervention
5% IS
strategy is given by the point (CVCT  300,103 AD and
Tr
12% t2  10 years, 3 months). Inducers of different maintenance costs
CB
15% and total costs are shown relative to the optimum value for t2 in
PL
17% 9% F the following Fig. 13.
DS
charges

Fig. 10. Distribution equipment maintenance costs before optimization for a year.

Reliability
The expected annual costs for PM
1.0
SPPM
500 0.8
RCM
0.8
RRCM
400 0.7
Costs of PM

0.6
300
R (t) 0.5
0.4
200
0.3
Not remitted
100 Remitted (PA) 0.2
0.1
0 0.0
0 13 26 39 52 65 78 91 104 117 130
10 15 20 25 30
TIME (Months)
Years (n)
Fig. 14. Reliability curve before and after optimization.
Fig. 11. The expected annual costs for PM.

Unreliability
1.0
The expected number of PM per year
SPPM
0.8
RCM
0,05 0.8
RRCM
0.7
0,04
0.6
F (t) 0.5
0,03
NPM

0.4

0,02 0.3
0.2
0,01 0.1
Markov model
Monte Carlo Simulation 0.0
0,00 0 13 26 39 52 65 78 91 104 117 130
10 15 20 25 30 35 TIME (Months)
Year (n)
Fig. 15. Unreliability curve before and after optimization.
Fig. 12. The expected number of PM.

Failure rate λ
The current value of the total cost 4.0E-3
SPPM
3.6E-3
5000 RCM
3.2E-3
RRCM
CV of Cost * 100AD

4000 2.8E-3
Optimal 2.4E-3
3000
r (t) 2.0E-3
Total costs
2000 Inspection costs 1.6E-3
Costs CM
Costs PM 1.2E-3
1000 8,0E-3
4,0E-3
0
0.0
0 1 2 3 4 5 0 13 26 39 52 65 78 91 104 117 130

Year (n)).10 TIME (Months)

Fig. 13. The current value of the total cost. Fig. 16. Failure rate curve before and after optimization.
 B. Yssaad, A. Abene / Electrical Power and Energy Systems 73 (2015) 350–360 359

The graphs of reliability The next figure shows the failure rate k of the three approaches.
It confirms that the failure rate starts at an early date from the 25th
Figs. 14 and 15, below, show the curves of the reliability and month and it increases greatly by applying the method SPPM. But
unreliability of the equipment of the three approaches for a trial it is within the standards by the application of RCM model, and it
period of 10 years. So we see clearly that the power system reliabil- remains still be better by adopting RRCM approach (see Fig. 16)
ity decreases below 50% from the 27th month to the SPPM method, From the average availability viewpoint and according to the
but for the RCM model, it decreases from the 49th month and 60th previous figure (Fig. 17), it is quite evident that the electrical sys-
month for the model RRCM. This, in, turn gives a ratio improve- tem will remain available for a very good period by applying our
ment of 1.81 and of 2.22 of the RCM and RRCM compared to the approach.
SPPM. Fig. 18, below, shows the maintenance cost vs time. So we,
clearly, see the big difference in spending between the three
approaches.
Mean Availability
1.0 Through these reliability curves, we can clearly feel the differ-
0.8
SPPM ence between these three approaches, hence, improving reliability
0.8
RCM and availability of an overall report (practice and simulation) of
0.7
RRCM 1.81 of the RCM model and of 2.22 of the RRCM approach com-
pared to the old SPPM method and reduced maintenance cost of
0.6
34.58% for a mission time of four years, see Table 8.
A (t) 0.5
0.4
0.3
Conclusion
0.2
This paper presents the application of the new RRCM model to
0.1
0.0
optimize the maintenance management of electrical power equip-
0 13 26 39 52 65 78 91 104 117 130 ment. From the various results obtained by experimental applica-
TIME (Months) tion and the different simulations of the model, we can conclude
that the effective implementation of the maintenance optimization
Fig. 17. Failure rate curve before and after optimization. by such an approach will therefore improve the reliability and
availability of the electrical system, increase the persons’ safety
and property and the environment and reduce overall operating
and maintenance costs. The main recommendations that we used
Cost over time
35000.00
is the re-integration of this maintenance strategy, including an
31500.00
SPPM optimal and rational implementation of the RCM approach. This
28000.00
RCM work showed the feasibility of conducting an optimization of the
24500.00
RRCM maintenance by using this method in power electrical system.
21000.00 And therefore, it will be necessary to take into account the follow-
Cost (t) 17500.00 ing recommendations: A complete renovation of the system, since
14000.00 most of the equipment is very old and comes from the colonial per-
10500.00 iod, especially the electric lines and transformers, energize the
7000.00 maintenance department, completely change the power lines
3500.00 using the same material, rebuild or replace power transformers
0.0 by other more powerful and robust and finally, change the old
0 13 26 39 52 65 78 91 104 117 130
insulators completely by other composite material to resist mois-
TIME (Months)
ture and salinity of this area. [1–3].
Fig. 18. Cost curve before and after optimization.
Acknowledgments

Table 8 We thank the entire team of the station distribution electrical of

Summary of results obtained. Relizane Algeria, for the great support and joint work team. Special
thanks to the reviewers and the entire IJEPES journal staff.
Procedure adopted E3P RCM RRCM
Mission time = 48 months (4 years)
Reliability (R) 0.054 0.709 0.872 Appendix A. Program of graph Markov
Maintainability (M) 0.945 0.290 0.125
Mean availability (A) 0.726 0.9536 0.985
Safety (S) 0.999
% Solving the differential equation
Unreliability (F) 0.945 0.291 0.128
Duration at 85% 303.054 621.286 668.245 dP/dt=M⁄P(t)(Markov Graph) by the method of: Rung
Time reliability and availability 85% 361 462 491 Kutta
Cost (Million AD) 280 112 83 clear all
Mean time to first failure (MTTF) 71.42 200 263.157
close all
Failure rate (k) 0.014 0.005 0.0038
Mean downtime MDT 6.075 0.335 0.225
% construction of the matrix M
Mean uptime MUT 10.625 51.965 58.348 lamda1=1/2.53;lamda2=1/2.49;lamda3=1/5.14;lamda4=
Mean time between two consecutive failures 16.7 52.3 58.1 1/5.27;lamda5=1/2.82;lamda6=1/3.38;
(MTBF) beyond MTTF lamda7=1/1.2;mu1=1/3.42;mu2=1/7.93;mu3=1/7.79;m
Repair rate (l) 0.2454 2.985 3.564
u4=1/7.65;mu5=1/3.98;mu6=1/5.31;mu7=1/2.21;
Mean time to repair (MTTR) 4.075 0.335 0.280

(continued on next page)

360 B. Yssaad, A. Abene / Electrical Power and Energy Systems 73 (2015) 350–360

0:0388 0:0008 0:0008 0:0004 0:0004 0:0007 0:0006 0:9574

M=[-(lamda1+lamda2+lamda3+lamda4+lamda5+lamda6+
lamda7) mu1 mu2 mu3 mu4 mu5 mu6 mu7; 0:0671 0:0029 0:0030 0:0014 0:0014 0:0026 0:0022 0:9195
lamda1 -mu1 0 0 0 0 0 0;lamda2 0 -mu2 0 0 0 0 0;lamda3 0 0:0876 0:0058 0:0060 0:0029 0:0028 0:0053 0:0044 0:8851
0 -mu3 0 0 0 0;lamda4 0 0 0 -mu4 0 0 0; 0:1022 0:0094 0:0098 0:0047 0:0046 0:0085 0:0071 0:8538
lamda5 0 0 0 0 -mu5 0 0;lamda6 0 0 0 0 0 -mu6 0;lamda7 0 0:1125 0:0133 0:0139 0:0067 0:0066 0:0120 0:0101 0:8248
0 0 0 0 0 -mu7] 0:1196 0:0174 0:0184 0:0089 0:0087 0:0158 0:0134 0:7978
variabl=[0 0 0 0 0 0 0 1]’; 0:1244 0:0217 0:0230 0:0112 0:0109 0:0197 0:0167 0:7724
tempfin=1.5; P ¼ 0:1275 0:0260 0:0278 0:0135 0:0131 0:0236 0:0201 0:7485
h=0.1; 0:1292 0:0302 0:0326 0:0158 0:0154 0:0275 0:0235 0:7259
k=1; 0:1300 0:0344 0:0373 0:0181 0:0176 0:0314 0:0268 0:7043
temps=0; 0:1302 0:0385 0:0421 0:0204 0:0198 0:0352 0:0301 0:6838
while temps<= tempfin 0:1298 0:0425 0:0467 0:0226 0:0220 0:0388 0:0334 0:6641
% 0:1291 0:0463 0:0513 0:0248 0:0242 0:0424 0:0366 0:6453
s1 = feval(’fonc1’, temps,variable,A); 0:1282 0:0500 0:0558 0:0270 0:0263 0:0459 0:0397 0:6272
s2 = feval(’fonc1’, temps+h/2,variable+h/2⁄s1,A);
0:1271 0:0535 0:0602 0:0291 0:0284 0:0492 0:0427 0:6099
s3 = feval(’fonc1’, temps+h/2,variable+h/2⁄s2, A);
s4 = feval(’fonc1’, temps+h,variable+h⁄s3, A);
variable1=variable+h⁄(s1+s2⁄2+2⁄s3+s4)/6; References
P(:,k)=variable1;
temp(k)=temps; [1] Yssaad B, Khiat M, Chaker A. Reliability centered maintenance optimization for
k=k+1; power distribution systems. Int J Electr Power Energy Syst IJEPES
2014;55:108–15.
temps=temps+h; [2] Yssaad B. Optimization of the maintenance based on the approach of the
variable=variable1; techniques of modeling and simulation: application to the networks of
end transport and interconnection. Doctoral thesis; January 2014.
[3] Yssaad B, Khiat M, Chaker A. Reliability centered asset maintenance
P=P’; optimization for power distribution systems. Int Rev Model Simul IREMOS
P0=P(:,1); 2013;6(1) (Part A).
P1=P(:,2); [4] da Rosa Mauro A, Leite da Silva Armando M, Miranda Vladimiro. Multi-agent
systems applied to reliability assessment of power systems. IJEPES 2012;42(1).
P2=P(:,3);
[5] Nordgård Dag Eirik. A framework for risk-informed decision support in
P3=P(:,4); electricity distribution companies utilizing input from quantitative risk
P4=P(:,5); assessment. IJEPES 2012;43(1).
[6] Chang Liang, Wu Zhigang. Performance and reliability of electrical power grids
P5=P(:,6);
under cascading failures. IJEPES 2011;33(8).
P6=P(:,7); [7] Arabian Hoseynabadi Hooman. Failure modes and effects analysis (FMEA) for
P7=P(:,8); wind turbines. IJEPES-D-09-00041; September 2010.
The fon1 function is given by the following program: [8] Life-Cycle Cost Analysis (LCCA). Sieglinde Fuller NationalInstitute of Standards
and Technology (NIST) [last updated 28.06.10].
function yc=fonc1(temps,variable,M); [9] Dinesh Kumar U. Tutorials on life cycle costing and reliability engineering
% course material. Indian Institute of Management Bangalore; 2005.
yc=M⁄variable; [10] Despujols A. Reliability Centred Maintenance optimization. Technical
engineer. file MT9310; 2004.
[11] Bertling L, Eriksson R, Allan RN. Relation between preventive maintenance and
reliability for a cost-effective distribution system. In: IEEE Porto power tech
Result of simulation: conference; 2001.
[12] Moubray J. Reliability-centered maintenance. Butterworth-Heinemann; 1991.
2:6650 0:2924 0:1261 0:1284 0:1307 0:2513 0:1883 0:4525 [13] Bertling L. Reliability centred maintenance for electric power distribution.
0:3953 0:2924 0 0 0 0 0 0 [14] Reliasoft Cooperation. FMEA Software, Reliasoft Simbloc, RCM5, Xfmea.
0:4016 0 0:1261 0 0 0 0 0 <http://www.reliasoft.com/xfmea/index.htm/>.
0:1946 0 0 0:1284 0 0 0 0
M¼ >> P
0:1898 0 0 0 0:1307 0 0 0
0:3546 0 0 0 0 0:2513 0 0
0:2959 0 0 0 0 0 0:1883 0
0:8333 0 0 0 0 0 0:4525