Reliability Engineering and System Safety 79 (2003) 301–308

www.elsevier.com/locate/ress

On preventive maintenance policy of a critical reliability level

for system subject to degradation
Y.X. Zhao*
Institute of Mechanical Engineering, Southwest Jiaotong University, Beiyuan 32-2-06#, Chengdu 610031, People’s Republic of China
Received 14 September 2000; revised 30 July 2002; accepted 8 October 2002

Abstract
Conventional preventive maintenance (PM) policies generally hold same time interval for PM actions and are often applied with known
failure modes. The same time interval will give unavoidably decreasing reliabilities at the PM actions for degradation system with imperfect
PM effect and the known failure modes may be inaccurate in practice. Therefore, field managers would prefer policy with an acceptable
reliability level to keep system often at a good state.
A PM policy with the critical reliability level is presented to address the preference offield managers. Through assuming that system after a PM
action starts a new failure process, a parameter so-called degradation ratio is introduced to represent the imperfect effect. The policy holds a law
that there is same number of failures in the time intervals of various PM cycles, and same degradation ratio for the system reliability or benefit
parameters such as the optimal time intervals and the hazard rates between the neighboring PM cycles. This law is valid to any of the failure
modes that could be appropriately referred as a ‘general isodegrading model’, and the degradation ratio as a ‘general isodegrading ratio’. In
addition, life cycle availability and cost functions are derived for system with the policy. An analysis of the field data of a loading and unloading
machine indicates that the reliability, availability and cost in life cycle might be well modeled by the present theory and approach.
q 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Preventive maintenance; Degradation; Critical reliability; Life cycle; Availability; Cost

1. Introduction [4] or the consideration with minimal repair by Barlow and

Preventive maintenance (PM) is a necessary activity to
restore or keep the function of a repairable system in good
state. How to assess the effect of this activity and how to
arrange it properly for addressing one’s satisfactions have
been long focused in practice since 1960s [1].
It has been revealed that maintenance effects can be
subdivided into a perfect, a non-effect, and an imperfect
[2,3]. A perfect effect restores the system to good-as-new, a
non-effect to bad-as-old, and an imperfect effect to partly
good. For system subjected to age/degradation, the
imperfect should be generally for PM, and the perfect and
non-effects are extreme. Present paper will discuss the
imperfect.
Conventional replacement/PM policies [5 – 10] are
derived using non-decreasing hazard rate functions on a
basis of the imperfect effect model by Brown and Proschan
* Tel.: þ86-28-87602465; fax: þ86-28-87609007.
E-mail address: yxzhao@home.swjtu.edu.cn (Y.X. Zhao).
0951-8320/03/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 5 1 - 8 3 2 0 ( 0 2 ) 0 0 2 0 1 - 6
Hunter [1]. They hold generally a same time interval T for
replacement/PM actions and are often applied with known
failure modes. The age-T policy will give unavoidably
decreasing reliabilities at the PM actions for degradation
system with imperfect PM effect and the known failure
modes may be inaccurate in practice for complicated
system.
An interesting phenomenon worthy to pay attention is
that the field managers preferred the PM opportunities being
derived from an acceptable specified reliability level
[11 –13]. This phenomenon is known in our field investi-
gation into the loading and unloading machinery used at
Chinese railway terminals. Major causes may be included
as:
1. Proper PM actions should be at best derived from the
inspection in service. Under this case, unless there is a
fine support system for decisions making, the known
failure models can be seldom quantified accurately in
practice for the complicated repairable system.
302 Y.X. Zhao / Reliability Engineering and System Safety 79 (2003) 301–308
2. Following the management rules with T-age PM policy
[14], the machinery exhibit increasing hazard rates at the
PM actions due to the imperfect PM effect. Correspond-
ingly, the reliabilities at the PM actions show a gradual
decreasing.
3. To keep the machinery operated at a good technical state
with reliability above an acceptable reliability level, the
field managers took often unconsciously the measure-
ment of shortening the T intervals as the PM action
increasing.
However, there is no effort addressing the preference of
field managers. Key issues are how to quantify the imperfect
effect and how to arrange PM actions properly to address
one’s satisfactions. These should be very difficult tasks.
Some good efforts have made in the proportional hazard
(PH) model [15 – 17] and the work by Soares and Garbatov
[18]. The PH model uses the proportional age reduction
factor to the baseline of hazard rate or to the operation time.
Considering the reliability of the ship hull girder with age
failure mechanism after repair smaller than the initial value
for new, Soares and Garbatov introduced a reduction factor
for piece of time to represent the ‘recovered’ time of a repair
action.
Through assuming that system after PM action starts a
new failure process, a parameter so-called degradation ratio
should be introduced to represent the imperfect effect of PM
on the system reliability, availability, or benefit between
neighboring PM cycles. This paper tries to develop a PM
policy with a critical reliability level to meet the preference
of field managers. Aim is addressed on that the policy is
valid to arbitrary failure process/mode. Relative approaches
are also explored for predicting the life cycle availability
and cost under this policy.
2. Preventive maintenance policy
2.1. Degradation viewpoints
PM cost and effect in general have a non-linear relation.
The cost per unit time may increase rapidly when the effect
Fig. 1. Hazard rates of PM cycles for the system subject to degradation (hi ðtÞ is hazard rate function in the PM cycle i; oi is start point of the PM cycle i;
i ¼ 1; 2; …; np ; and np is number of PM cycles for the system in life cycle).
is over a critical level. Therefore, there are few of
preferences to cost much more to restore the system good-
as-new or well-than-new, i.e. PM effect is generally
imperfect.
From viewpoint of the thermodynamics, the components
in operation are subject to irreversible energy dissipation
due to fatigue, wear, corrosion, etc. up to be replaced. The
system in a PM cycle is physically subject to an irreversible
degradation due to causes:
1. The components not repaired in PM actions undergo an
irreversible, continuous and cumulative damage.
2. Some damaged components are repaired but not replaced
in PM actions; the existing cumulative damage cannot be
removed.
3. Although some damaged components are replaced in
PM actions, the cumulative wear of adjacent and non-
replaced components may still exist to worsen the
stress conditions of the relative parts.
As shown in Fig. 1, the degradation results in a gradual
increase of the hazard rates in the time intervals with the PM
cycling.
2.2. Policy
It is here assumed that:
1. The system undergoes relatively constant conditions of
stress, environment and maintenance in life cycle.
2. A corrective repair (CR) restores locally the system
good-as-before the failure. This implies that the failure
process during a PM cycle can be reasonably described
as a continuous failure mode.
3. A PM action has an imperfect effect on the system in
general. This implies that there is difference between
the failure processes before and after the PM action. It
is reasonably to consider that after PM action the
system begins a new failure process. As shown in
Fig. 1, the failure rates at the PM actions exhibit an
increase as PM cycling.
 Y.X. Zhao / Reliability Engineering and System Safety 79 (2003) 301–308 303

Consistency with the mentioned preference of field From Eq. (2) it can be obtained
managers, and combining the considerations in the existent
nf1 ¼ nf2 ¼ · · · ¼ nfi ¼ nf ; i ¼ 1; 2; …; np ð3Þ
policies [1 – 10], a PM policy based on a systematic
engineering should be presented with following consider- where nf is number of failures in a PM cycle. This equation
ations: indicates that there is same number of failures in the time
intervals of various PM cycles for a system with the present
1. PM actions should be taken at an acceptable critical PM policy of critical reliability level. It is valid to arbitrary
reliability level Rc as shown in Fig. 2. This level can keep failure modes.
the system operate constantly at a good technical state Taking into account the repair time of both CR and PM
with an acceptable operational cost per unit time, or an actions, an availability function for the system in PM cycle i
acceptable availability. should be defined as [19]
2. The time interval in a PM cycle should be determined by
t
a maximum availability, or a minimum cost per unit time Ai ðtÞ ¼ ðt ; i ¼ 1; 2; …; np ð4Þ
in the interval. t þ tc hi ðtÞdt þ tp
3. Number of PM cycles for the system in life cycle 0

should be derived from an acceptable level of the where tc is expected time for CR actions, and tp is expected
availability, or the cost per unit time in life cycle. time for PM actions. From the consideration of PM policy to
maximize the system availability, the optimal time interval
Quantifying of this policy will be addressed in the
Ti should meet a condition of dAi ðtÞ=dtlt¼Ti ¼ 0: The
following effort.
differential result of Eq. (4) gives:
2.3. Quantifying tp ðT1
Ti hi ðTi Þ ¼ þ hi ðtÞdt; i ¼ 1; 2; …; np ð5Þ
tc 0
For a degradation system with the PM policy of critical
reliability level Rc, a reliability equation at the critical According to Eqs. (2) and (3), Eq. (5) can be rewritten as:
reliability level can be constructed tp
Ti hi ðTi Þ ¼ þ nf ; i ¼ 1; 2; …; np ð6Þ
tc
 ð T1   ð T2 
exp 2 h1 ðtÞdt ¼ exp 2 h2 ðtÞdt ¼ · · · Therefore, the optimal time interval and the hazard rate
0 0
 ð Ti  between neighboring PM cycles have a relation:
ð1Þ
¼ exp 2 hi ðtÞdt ; i ¼ 1; 2; …; np Ti hi ðTi Þ ¼ Tiþ1 hiþ1 ðTiþ1 Þ; i ¼ 1; 2; …; np 2 1 ð7Þ
0
A degradation ratio relation can be then obtained by
where i is ordinal of the PM cycles, i ¼ 1; 2; …; np ; np is converting Eq. (7):
number of PM cycles for the system in life cycle; hi ðtÞ;
Ti 2 Tiþ1 h ðT Þ 2 hi ðTi Þ
hazard rate function in PM cycle i; Ti is the time interval in ¼ iþ1 iþ1 ;
PM cycle i. Eq. (1) can be rewritten as Ti hiþ1 ðTiþ1 Þ
ð8Þ
ð T1 ð T2 ð Ti i ¼ 1; 2; …; np 2 1
h1 ðtÞdt ¼ h2 ðtÞdt ¼ · · · ¼ hi ðtÞdt;
0 0 0 ð2Þ Extrapolating the consideration of PH models [15 – 17] and
the work by Soares and Garbatov [18], introducing ratio
i ¼ 1; 2; …; np parameters rT ¼ ðTi 2 Tiþ1 Þ=Ti and rh ¼ ½hiþ1 ðTiþ1 Þ 2

Fig. 2. Reliability functions for a system with a PM policy of critical reliability level Rc.
304 Y.X. Zhao / Reliability Engineering and System Safety 79 (2003) 301–308

hi ðTi Þ=hiþ1 ðTiþ1 Þ; it yields: The cumulative availability function for the system should
be expressed as
rT ¼ rh ¼ r ð9Þ
TP0 ðtÞ
P
A ðtÞ ¼ P ð15Þ
This equation represents that there is a same degradation T w ðtÞ
ratio of the optimal time intervals and the hazard rates
between neighboring PM cycles for the system with the where TP0 ðtÞ is cumulative function of operational time,
present PM policy of critical reliability level. and TPw ðtÞ is cumulative function of the entire time since
Summing up the above, a reliability degradation law can the system is going to operation, in which it consists of
be deduced that there is same number of failures in the time
operational, CR and PM time. TP0 ðtÞ and TPw ðtÞ can be
intervals of various PM cycles, and a same degradation ratio
of the optimal time intervals and the hazard rates between calculated by equations, respectively
neighboring PM cycles for the system with present PM TP0 ðtÞ ¼ T1 f ðr; i 2 1Þ þ t ð16Þ
policy of critical reliability level. This law is valid to any of
the failure modes in the time intervals of various PM cycles.
TPw ðtÞ ¼ T1 f ðr; i 2 1Þ þ t þ tc ðnf 2 1Þði 2 1Þ þ tp i ð17Þ
Therefore, it can be appropriately referred as a ‘general
isodegrading model’ and the ratio as a ‘general isodegrading
ratio’ [11 –13]. where function f ðr; i 2 1Þ is defined as:
What worthy noted issue is that the ratio should be an
integrated parameter. In practice, this parameter must be a 1 2 ð1 2 rÞi21
f ðr; i 2 1Þ ¼ ð18Þ
reasonable reflection of the imperfect effect of PM actions, r
which is relative to the operational conditions of stress, Up to ith PM action, the cumulative availability should be
environment and maintenance. When the ratio is equal to evaluated by equation:
unity and zero, they represent, respectively, the repair
effects of no effect and restoring the system good-as-new. T1 f ðr; iÞ
APi ¼ ð19Þ
T1 f ðr; iÞ þ tc iðnf 2 1Þ þ tp i

3. Availability function in life cycle
Combining Eqs. (8) and (9), the optimal time interval Ti
in PM cycle i can be predicted by T1 and r parameters:
Ti ¼ T1 ð1 2 rÞi21
At the same time, combining Eqs. (3) and (10), the mean
time between failures (MTBF) MTBFi in PM cycle i has a
relation with the parameters MTBF1 and r:
MTBFi ¼ MTBF1 ð1 2 rÞi21
Assuming that the system is replaced just before a PM
action, the cumulative availability of system in life cycle
can be obtained by equation:
T1 f ðr; np Þ
ð10Þ AP ¼ ð20Þ
T1 f ðr; np Þ þ tc np ðnf 2 1Þ þ tp ðnp 2 1Þ
Obviously, the cumulative availability is subjected to an
irreversible decreasing process. The number of PM cycles
np for the system in life cycle may be derived from
ð11Þ consideration of an acceptable availability.

In addition, from Eqs. (3), (8) and (9) the ratio r can be
evaluated by equation: 4. Life cycle cost function

MTBF1 2 MTBF2 Assuming that PM action is taken just at the time a failure
r¼ ð12Þ
MTBF1 occurs, cost function in PM cycle i can be expressed as
ðt 
Assuming that PM action is taken just at the time a failure c0 t þ cc tc hi ðtÞdt 2 1 þ cp tp
occurs, from Eq. (3) the system has a cumulative CR time ðt 0 
ci ðtÞ ¼ ð21Þ
tPc in each PM cycle as:
t þ tc hi ðtÞdt 2 1 þ tp
0
tPc ¼ tc ðnf 2 1Þ ð13Þ
where c0, cc and cp are expected costs per unit operational
Then, the expected availability in PM cycle i can be time, CR time, and PM time, respectively. Expected cost in
predicted by formula: PM cycle i can be then evaluated by:

Ti c0 Ti þ cc tc ðnf 2 1Þ þ cp tp
AEi ¼ P ð14Þ cEi ¼ ð22Þ
Ti þ t c þ tp Ti þ tc ðnf 2 1Þ þ tp
 Y.X. Zhao / Reliability Engineering and System Safety 79 (2003) 301–308 305

Cumulative cost function of the system should be given as 3-PWD, the optimal time interval T1 can be evaluated by
equation:
CPw ðtÞ
P
c ðtÞ ¼ P ð23Þ  1=h1
T w ðtÞ tp
T1 ¼ u1 þm1 ð28Þ
ðh1 2 1Þtc
where CPw ðtÞ is the cumulative cost in the entire time TPw ðtÞ The evaluated result of T1 is 2670.3 h. In addition, the
since the system is going to operation, which can be parameter MTBF in this time interval can be calculated by
evaluated by equation: equation
CPw ðtÞ ¼ cc tc ðnf 2 1Þði 2 1Þ þ cp tp i þ c0 T1 f ðr; i 2 1Þ þ c0 t  
1
MTBF1 ¼ u1 G 1 þ þ m1 ð29Þ
ð24Þ h1

Up to ith PM action, the cumulative cost should be evaluated
by equation:
c0 T1 f ðr; iÞ þ cc tc iðnf 2 1Þ þ cp tp i
cPi ¼
T1 f ðr;iÞ þ tc iðnf 2 1Þ þ tp i
Assuming that the system is replaced just before a PM action,
cumulative cost for the system in life cycle can be calculated
by:
c0 T1 f ðr; np Þ þ cc tc np ðnf 2 1Þ þ cp tp ðnp 2 1Þ
cP ¼
T1 f ðr; np Þ þ tc np ðnf 2 1Þ þ tp ðnp 2 1Þ
Apparently, the cumulative cost per unit time for a
degradation system shows an irreversible increase process.
The number of PM cycle np for the system in life cycle may be
also determined from a consideration of an acceptable cost
level.
where GðxÞ is Gamma function. The calculated result is
87.64 h. The number of failures in the time interval can be
then obtained by the formula:
ð25Þ T1
nf ¼ ð30Þ
MTBF1
The result is 30.47, and 30 is applied for the subsequent
predictions.
When nf parameter is available, the successive PM
actions are carried out by a rule of same number of failures
ð26Þ nf in each PM cycle. Now, six PM cycles have passed. The
test data of time between failures (TBF), ti;j ; are given in
Table 1. The statistical parameters of 3-PWD fitting the test
data are shown in Table 2. It is known that:
Table 1
Test data of the TBF ti;j of the machine (dimension in hours)

Ordinal j t1;j t2;j t3;j t4;j t5;j T6;j

5. Example 1 66.0 94.0 106.0 86.0 83.0 72.5

2 209.0 77.5 58.0 33.5 128.0 30.5
A field test of a loading and unloading machine has been 3 69.5 192.5 171.0 136.5 90.0 113.0
made to verify the above theory and approach. The expected 4 111.0 61.5 93.0 89.0 89.0 38.5
CR time tc, PM time tp, cost per unit operational time c0, 5 80.0 103.0 154.0 67.0 142.5 129.0
6 187.5 37.0 94.0 156.0 78.5 71.5
cost per unit corrective time cc, and cost per unit PM time cp 7 112.5 168.0 35.0 85.0 79.0 116.0
are obtained beforehand from similar machines. They are 8 39.0 50.0 150.0 140.0 32.0 76.0
7 h, 700 h, 34 ¥/h, 82 ¥/h, and 68 ¥/h, respectively. 9 182.5 71.5 46.5 77.5 125.0 82.0
In the test, random degradation process of the machine in 10 53.0 119.5 99.0 65.0 82.0 81.0
the first PM cycle is statistically tested in service. Using the 11 77.0 118.0 107.0 37.0 50.0 46.5
12 117.0 51.0 65.5 90.5 62.5 45.5
similar approach in Ref. [20], the total effects of seven 13 87.0 108.0 51.0 98.0 71.5 52.0
commonly used distributions fitting the data are compared. 14 101.0 80.5 56.5 96.5 60.0 45.5
The seven distributions are Weibull three- and two- 15 127.5 173.0 70.5 53.5 35.0 55.5
parameter, normal, lognormal, extreme minimum value, 16 83.5 104.0 56.0 52.5 49.0 45.0
extreme maximum value, and exponential. The results 17 44.0 42.0 42.5 28.0 60.0 35.5
18 119.0 110.0 59.5 43.5 48.0 47.5
indicate that the process can be well modeled by the three- 19 67.5 63.0 68.0 60.0 37.5 54.0
parameter Weibull distribution (3-PWD). The probability 20 129.0 61.5 97.0 52.5 51.0 26.0
distribution function can be modeled 21 59.0 77.0 73.0 64.5 58.0 41.5
    22 52.0 60.5 85.0 51.5 27.0 37.5
t 2 m1 h1 23 31.0 45.5 70.5 40.0 49.0 39.5
P1 ðtÞ ¼ 1 2 exp 2 ð27Þ
u1 24 55.0 65.0 39.5 55.0 40.5 35.0
25 47.0 74.5 56.5 62.5 56.0 39.0
where m1 ; u1 and h1 represent the location, scale and shape 26 54.0 30.0 29.0 47.5 44.5 75.0
parameters of 3-PWD in the first PM cycle. They are 26.44, 27 66.0 55.0 45.5 42.5 40.0 57.5
67.59, and 1.4649, respectively. 28 83.0 49.0 48.0 44.5 42.0 65.5
29 65.0 51.5 41.5 39.0 37.0 55.5
The test data are checked constantly to maximize the
30 48.0 44.0 47.0 44.0 41.5 33.0
availability by Eq. (4). When the data are modeled using
306 Y.X. Zhao / Reliability Engineering and System Safety 79 (2003) 301–308

Table 2
Parameters of the 3-PWD fitting the test data in Table 1

Ordinal of Mean value (h) Standard Coefficient of 3-PWD

PM cycles i deviation (h) variation
m u h rxy a

1 87.42 44.26 0.5063 26.44 67.59 1.4649 0.987904

2 81.25 40.46 0.4980 25.96 61.21 1.4520 0.987919
3 73.85 35.38 0.4791 25.46 53.58 1.4516 0.987958
4 67.95 31.84 0.4685 24.96 47.51 1.4361 0.987727
5 62.97 28.75 0.4566 24.29 42.71 1.4308 0.987747
6 58.07 25.56 0.4402 23.55 38.14 1.4377 0.987878
a
Linear relationship coefficient of the 3-PWD fitting the data in Table 1.

Fig. 3. Curves of the hazard rate, reliability, availability and cost of the machine.
 Y.X. Zhao / Reliability Engineering and System Safety 79 (2003) 301–308 307

Fig. 4. Cumulative availability and cost curves of the machine.

Table 3
Predicted values of the parameters T, MTBF, AE, cE, AP, cP, and r

Ordinal of PM cycle i T (h) MTBF (h) nf r AE cE (¥/h) AP cP (¥/h)

1 2670.3 87.64 30.47 – 0.7466 53.88 0.7466 53.88

2 2548.3 81.45 31.29 0.0706 0.7364 54.13 0.7416 54.00
3 2236.9 74.03 30.21 0.0910 0.7121 54.67 0.7325 54.21
4 2115.9 68.10 31.07 0.0802 0.6991 54.99 0.7248 54.39
5 1947.0 63.09 30.86 0.0735 0.6817 55.39 0.7172 54.57
6 1690.9 58.18 29.06 0.0779 0.6535 55.99 0.7083 54.76

1. The parameter MTBF shows a decrease with PM cycling.
This indicates that the machine is subject to a
degradation process.
2. The linear relationship coefficients are all greater than
0.987. This indicates that 3-PWD is an appropriated
model of the data.
Curves of hazard rate, reliability, availability and cost in
various PM cycles are shown in Fig. 3. The cumulative
availability and cost curves are exhibited in Fig. 4. It can be
seen that:
2. The critical reliability level makes a decrease of the
optimal time intervals as PM cycling.
3. The availabilities at the PM actions show an irreversible
decrease as PM cycling, and the costs per unit time at the
PM actions give an increase as PM cycling due to the
degradation.
4. The cumulative availability and cost functions wave
relatively small after first PM cycle. The values at PM
actions still appear, respectively, a decrease for the
availabilities and an increase for the costs as the PM cycling.

1. The hazard rates at the PM actions exhibit a gradual Predicted and test values of the parameters T, MTBF, AE,
increase as the PM cycling. This indicates that the cE, AP, cP, and r are given, respectively, in Tables 3 and 4.
machine is subject to a degradation process. It is known that:

Table 4
Test values of the parameters T, MTBF, AE cE, AP, cP, and r

Ordinal of PM cycle i T (h) MTBF (h) R AE cE (¥/h) AP cP (¥/h)

1 2622.5 87.42 – 0.7439 53.93 0.7439 53.93

2 2437.5 81.25 0.0705 0.7297 54.26 0.7370 54.09
3 2215.5 73.85 0.0911 0.7104 54.70 0.7287 54.28
4 2038.5 67.95 0.0799 0.6930 55.11 0.7206 54.47
5 1889.0 62.97 0.0733 0.6766 55.49 0.7127 54.65
6 1742.0 58.07 0.0778 0.6586 55.90 0.7050 54.83
308 Y.X. Zhao / Reliability Engineering and System Safety 79 (2003) 301–308
1. Predicted values of the number of failures nf, with respect
to the optimal time intervals in various PM cycles to
maximize the availabilities, have an average value
around 30 with an error of 1.29. This indicates that the
degradation law of same number of failures in each PM
cycle is valid to meet the PM policy.
2. Degradation ratio values are round 0.08 having an error
of 0.0111. This indicates that the degradation law of
isodegradation ratio between neighboring PM cycles is
valid to meet the PM policy.
3. A few of differences between the predicted and test
values of the parameters indicate that present degra-
dation law and approach are available.
6. Conclusions
For a repairable system subject to an intrinsic degradation
with T-age PM policy, it is unavoidable to exhibit a decreasing
of reliabilities as the PM action increasing. Therefore, it is
preferred by field managers to hold an acceptable reliability
level to keep the system often at good technical state.
A PM policy has been suggested for degradation system
with an acceptable reliability level. It consists of the
following considerations:
1. PM actions should be taken at an acceptable critical
reliability level Rc. This level can keep the system operates
at a good technical state with an acceptable operational cost
per unit time, or an acceptable availability.
2. Optimal time interval in a PM cycle should be determined
by a maximum availability, or a minimum cost per unit
time.
3. Number of PM cycles for the system in life cycle should be
derived from an acceptable level of the availability or the
cost per unit time in life cycle.
Through assuming that system after a PM action starts a
new failure process, a parameter so-called degradation ratio
is introduced to represent the imperfect effect. The policy
holds a law of same number of failures for the time intervals
of various PM cycle and a same degradation ratio for the
system reliability or benefit parameters such as optimal time
intervals and hazard rates between neighboring PM cycles. In
addition, this law is valid to any of the failure modes.
Therefore, the law should be appropriately referred as a
general isodegrading model and the degradation ratio a
general isodegrading ratio.
Life cycle availability and cost functions have been
deduced for the system with the PM policy. The number of
PM cycles for the system in life cycle may be derived from
an acceptable cumulative availability or an acceptable
cumulative cost in life cycle.
The analysis results of the field-test data of a loading and
unloading machine indicates that the availability and cost of
the machine with the suggested PM policy have been well
predicted and therefore, the present theory and approach are
available.
Acknowledgements
Present research is supported by the Postdoctoral
Foundation of China and the Provincial Special Foundation
for Applied Basic Research of Sichuan.
References
[1] Barlow RE, Hunter LC. Optimal preventive maintenance policies.
Oper Res 1960;8:90–100.
[2] Ascher A, Feinggold H. Repairable system reliability. New York:
Marcel Dekker; 1984.
[3] Martorell S, Sanchez A, Serradell V. Age-dependent reliability model
considering effects of maintenance and working conditions. Reliab
Engng Syst Safety 1999;64:19–31.
[4] Brown M, Proschan F. Imperfect repair. J Appl Probab 1983;20:
851 –9.
[5] Murthy DNP, Nguyen DG. Optimal age-policy with imperfect
maintenance. IEEE Trans Reliab 1981;30:80 –1.
[6] Canfield RV. Cost optimization of periodic preventive maintenance.
IEEE Trans Reliab 1986;35:78–81.
[7] Abdel-Hameed M. An imperfect maintenance model with block
replacements. Appl Stochast Models Data Anal 1987;3:63–72.
[8] Nakagawa T. Optimum policies for a system with imperfect
maintenance. IEEE Trans Reliab 1987;36:631 –3.
[9] Nakagawa T. Sequential imperfect preventive maintenance policies.
IEEE Trans Reliab 1988;37:295– 8.
[10] Chun YH. Optimal number of periodic preventive maintenance
operations under warranty. Reliab Engng Syst Safety 1992;37:
223 –5.
[11] Zhao YX. The theory of loading and unloading machinery
isodegradation and its application. MS Thesis. Southwest Jiaotong
University, Chengdu; 1993. In Chinese.
[12] Zhao YX, Yang B, Wang JN, Ding GF. Concurrent prediction of
crane operational reliability. Research report. Project of the
Postdoctoral Foundation of China. Chengdu: Southwest Jiaotong
University; 2000. In Chinese.
[13] Zhao YX, Yang B, Ding GF, Wang JN, Gao Q, Zhang ZP.
Degradation theory-based machine reliability design and software.
Research report. Provincial Special Foundation for Applied Basic
Research of Sichuan. Chengdu: Southwest Jiaotong University; 2000.
[14] Minister of Railway of China, Management rules of loading and
unloading machinery. Beijing: China Railway Press; 1982. in
Chinese.
[15] Kumer D, Klefsjö B. Proportional hazards model: a review. Reliab
Engng Syst Safety 1994;44:177–88.
[16] Shin I, Lim TJ, Lie CH. Estimating parameters of intensity function
and maintenance effect for repairable unit. Reliab Engng Syst Safety
1996;54:1–10.
[17] Ansell JI, Phillips MJ. Practical aspects of modeling of repairable
systems data using proportional hazard models. Reliab Engng Syst
Safety 1997;58:165–71.
[18] Soares CG, Garbatov Y. Fatigue reliability of the ship hull girder
accounting for inspection and repair. Reliab Engng Syst Safety 1996;
51:341–51.
[19] He GV. Reliability design. Beijing: Machine Press; 1993. in Chinese.
[20] Zhao YX, Gao Q, Wang JN. An approach for determining an
appropriate assumed distribution of fatigue life under limited data.
Reliab Engng Syst Safety 2000;67:1–7.