You are on page 1of 5

Reliability Engineering and System Safety 64 (1999) 127–131

Technical note

Maintenance free operating period – an alternative measure to MTBF and


failure rate for specifying reliability?
U. Dinesh Kumar a,*, J. Knezevic a, J. Crocker b,1
a
Centre for Management of Industrial Reliability, Cost and Effectiveness, Faculty of Engineering, University of Exeter, Exeter EX4 4QF, UK
b
Rolls Royce Military Aero Engines Limited, Customer Logistic Support, Bristol BS12 7QE, UK
Received 16 December 1997; accepted 2 June 1998

Abstract

The paper analyses the concept of maintenance free operating period (MFOP), the reliability requirement driven by the Ministry of
Defence (UK) for the next generation of future aircraft to be included in the fleet. Since the traditional reliability requirement MTBF (mean
operating time between failure) has several drawbacks, the immediate reaction would be to analyse the credibility of the new measure MFOP
against MTBF. The paper discusses various issues associated with MFOP. Two mathematical models are developed to predict the main-
tenance free operating period survivability (MFOPS), one using mission reliability approach and the other using alternating renewal theory.
The paper also analyses cost implications of MFOP to the customer and to the producer. 䉷 1999 Elsevier Science Ltd. All rights reserved.

Keywords: Failure rate; Maintenance free operating period (MFOP); Maintenance recovery period (MRP); Mean operating time between
failure (MTBF); Reliability; Renewal theory

1. Introduction is not beneficial either to customers or suppliers. Neither of


these measures adequately describe the reliability of the
Billions of dollars are spent by commercial and defence system. The main drawbacks of MTBF are:
industries every year as a direct consequence of the unrelia-
1. it is almost impossible to predict MTBF, if the time-to-
bility of their systems. The cost of unscheduled maintenance
failure distribution is not exponential;
by civil airline operators is in the order of one million
2. the methodology most widely used to predict MTBF and
pounds per aircraft per year [1]. Industries have started
failure rate is based on the exponential distribution. This
thinking whether they are specifying the reliability require-
distribution is used to model failure times (and supported
ment in the right way. For example, US Air Force [2] oper-
by Military Standards [4–6] like MIL-HDBK-217,
ating command prefer reliability requirements to be based
MIL-STD-1388 and British Defence Standard 00-41)
on mission and operational requirement rather than mean
primarily because of its mathematical friendliness and
time between maintenance, mean down time, availability,
the belief that Drenick’ limit theorem [7] is universally
etc. They argue that the old way relies on probabilistic,
applicable rather than any scientific reason.
design based measure that are more meaningful to statisti-
cians. The preferred way forces the technical community to Since MTBF has several drawbacks, The Royal Air Force
consider operational based requirements. In the past MTBF [8] are considering a new reliability metric, maintenance
(mean operating time between failure) or its reciprocal – the free operating period (MFOP), as the prime reliability and
‘failure rate’ – have been used by many customers as a maintainability requirement for their future generation
reliability specification without realising that in most aircraft. The SBAC’s (Society of British Aerospace
cases it is almost impossible to demonstrate. Knowles [3] Companies) ultra reliable aircraft project specifies MFOP
makes the point that specifying reliability in terms of MTBF as one of its important objectives. Now the ultimate goal for
many British Aero Space [8] industries is to reduce the
* Corresponding author. Tel.: +44-1392-263-612; Fax: +44-1392-263- dependence on MTBF and specify reliability as a probabil-
620; E-mail: d.k.unnikrishnan@exeter.ac.uk ity of time in service before failure. It is interesting to note
1
The views expressed in this paper are those of the author, and do not
necessarily reflect the opinion of the Rolls Royce Military Aero Engines that the Hockley and Appleton [8] have acknowledged the
Limited. concept of MFOP as a positive step for specifying reliability
0951-8320/99/$ - see front matter 䉷 1999 Elsevier Science Ltd. All rights reserved.
PII: S0 95 1 -8 3 20 ( 98 ) 00 0 48 - 9
128 U. Dinesh Kumar et al./Reliability Engineering and System Safety 64 (1999) 127–131

Fig. 1. The operating profile of the system.

of future combat aircraft. The main objective of this paper is warranted maintenance-free operating period of so many
to develop mathematical models for predicting maintenance flying hours, miles or days resolves into deciding on an
free operating period for a system. The paper analyses the acceptable probability of survival offset against an
cost implications of MFOP and when one can use MFOP as increased cost. The manufacturer will need to determine
a reliability specification. To our best knowledge, this is the expected cost of providing the MFOP and hence
the first attempt to develop a mathematical model for increase the price to minimise the risk of losing money
maintenance free operating period. An example problem whilst keeping the price competitive. A MFOP (or cycles
is used to illustrate the mathematical model. of MFOP) is usually followed by a maintenance recovery
period (MRP). This is defined as the period during which the
appropriate scheduled maintenance is carried out.
2. Maintenance free operating period (MFOP) Consider a repairable system. The time to failure and the
time to repair follow some arbitrary distribution. The oper-
The concept of maintenance-free operating period ating profile of the system can be represented as shown in
(MFOP) is not new; it is essentially same as the warranty Fig. 1. In Fig. 1, it can be noted that the system fails at time
period. What is new is that the operators are considering points {T 1,T 3,T 5,…} and the system is repaired and put into
extending this concept throughout the life of the system. In operation at time points {T 2,T 4,T 6,…}. Assume that it is
practice, the contractor/manufacturer will be expected to required to have a MFOP of t mf life units and a MFOPS of
guarantee that no unscheduled maintenance activities will 95% (i.e. the probability of the system surviving t mf life
be required during each defined period of operation. To units without needing corrective maintenance). A necessary
achieve this with the required level of confidence will condition for the system to have a MFOP of t mf life units
almost certainly require full parts life tracking and an with probability of 0.95 is given by:
increased level of part and module exchange.  
• Maintenance-free operating period (MFOP), allows a MFOPS(tmf ¼ Pr d (T2i þ 1 ¹ T2i ) ⱖ tmf ⱖ 0:95: (1)
i¼0
period of operation (say t mf life units) during which an
item will be able to carry out all its assigned missions, If the random variables {T1 ¹ T0 ; T2 ¹ T1 ; T3 ¹ T2 ; …} are
without the operator being restricted in any way due to independent and identically distributed, then the operating
system faults or limitations, with the minimum of profile of the system can be modelled using alternating
maintenance. renewal process.
• Maintenance-free operating period survivability
(MFOPS) is then defined as the probability that the
item will survive for the duration of the MFOP. 3. MFOP prediction — mission reliability approach

In other words, it is the probability of not having any In this section, we develop a mathematical model based
unscheduled maintenance for a period of t mf life units on a mission reliability approach to predict MFOPS. Let us
given the current age of the item. It is also the probability consider a system with n components connected in series. If
that the item maintains its functionality at least for a period the reliability requirement is MFOP of t mf life units, then the
of t mf life units without the need for corrective maintenance corresponding probability given that all the components of
due to failure of a component of the system which results in the system are new is given by:
an overall critical failure of the system. However, it should
be noted that, during MFOP, the system is allowed to Yn
Rk (1 ⫻ tmf )
MFOPS(tmf , 1) ¼ , (2)
undergo any planned minimal maintenance. Also, the k¼1 Rk (0)
redundant components can fail during an MFOP without
forcing any corrective action. The idea here is to find the where R k(t mf) is the reliability of the kth component for (the
probability of any unplanned maintenance. Hockley and first) t mf life units. Eq. (2) gives the probability for the
Appleton [8] make a point that during MFOP the necessity system to have MFOP of t mf life units during the first
for maintenance should be kept to a minimum. They further cycle. For the second cycle, the expression for MFOPS is
went on to say that specifying reliability through MFOP will given by:
force the supplier to analyse various failure mechanism Y Rk (2 ⫻ tmf )
n
which will further help to improve the overall design of MFOPS(tmf , 2) ¼ : (3)
the system. However, Crocker [9] points out that, giving a k ¼ 1 k (1 ⫻ tmf )
R
U. Dinesh Kumar et al./Reliability Engineering and System Safety 64 (1999) 127–131 129

In general, for ith cycle, the probability the system will have appropriate maintenance actions are carried out. In this
MFOP of t mf life units is given by: section, we derive MFOPS for a repairable item using
Y renewal theory allowing a maintenance recovery period
Rk (i ⫻ tmf )
n
MFOPS(tmf , i) ¼ : (4) after an MFOP. Assume that it is required to have a
k¼1 Rk ([i ¹ 1] ⫻ tmf ) MFOPS of t mf life units. Then it will be interesting to find
In Eq. (4), MFOPS(t mf,i) represent the probability that the MFOPS during a stated period of time T along with the
system will survive ith cycle of MFOP given that it survives maintenance recovery period. Consider a repairable item,
(i ¹ 1) cycles. Let us assume that it is required to achieve a assume that:
MFOP of t mf life units with MFOPS ¼ a. The following 1. The time to failure distribution of the item follows
procedure can be used to find the number of cycles the arbitrary distribution with density function represented
system satisfies MFOP with probability ⱖ a. Note that in by f(t).
deriving the above mathematical expressions, we assume 2. Maintenance recovery time of the item follows some
independence between time-to-failure distribution of arbitrary distribution with density function represented
various items of the system. This assumption may not be by g(t).
valid in some cases. In that case, one has to derive
alternative mathematical model. The item can be in two states {1,0}. Where 1 is up state
and 0 is the down state. Let P 1(T) be the probability that the
3.1. Procedure to calculate the number of cycles the system item will have t mf hours of maintenance free operating
satisfies the required MFOPS period through out the mission T. Maintenance is carried
out as soon as the item fails. The expression for P 1(T) can
Step 1. Set i ¼ 1.
be written as:
Step 2. Calculate Y
n
Rk (i ⫻ tmf )
MFOPS(tmf , i) ¼ : T
Z
Rk ([i ¹ 1] ⫻ tmf )
k¼1
P1 (T) ¼ R(tmf þ f (u l tmf )P0 (T ¹ u) du, (7)
Step 3. If 0
Yi
k ¼ 1 MFOPS(tmf , i)
T
ⱕ a; Z
P0 ¼ g(v)P1 (T ¹ v) dv, (8)
then Go To Step 5. 0
Step 4. i ¼ i þ 1, Go To Step 2.
Step 5. Number of cycles is i ¹ 1. where f(u l t mf) is the probability that the system fails at
Step 6. Stop. time u, given that it has survived up to time t mf. The
above system of integral equations can be solved by
numerical approximation [10] for any time to failure
3.2. MFOP of items with Weibull distributed failure times distribution like exponential, normal, Weibull, log-
normal, etc. to evaluate the value of P 1(T). Here, P 1(T)
For a component with a failure mode which can be gives the probability that the system will operate for
modelled by the Weibull distribution the probability of at least t mf hours before it fails during T hours of
surviving t mf units of time given that the item has survived operation.
t units of the time is given by:
!
tb ¹ (t þ tmf )b
MFOPS(tmf ) ¼ exp ¹ , (5) 5. Relationship of MFOP duration to degree of ageing
hb
Fig. 2 shows how the 95% confidence level of MFOP
where h is the scale parameter and b is the shape parameter
varies with age of the item for Weibull distributions with
of the Weibull distribution.
different shapes (b) but same MTTF of 1000 h. The
The MFOP period for a given level of confidence can be
important point is to note with these lines are that for b ⬍
calculated by rearranging the above equation as follows:
1 they are convex starting from zero, for b ¼ 1, it is
n  o1=b
tmf ¼ tb ¹ hb ln MFOPS(tmf ) ¹ t: (6) horizontal and for b ⬎ 1 they are concave tending towards
zero (as the age of the item tends to infinity). This means
for items with constant failure rate the MFOP probability
will remain same throughout the life of the component.
4. MFOP prediction – renewal theory approach For components which wear out, the probability of
achieving the desired MFOP will decrease as they age
After a certain number of cycles of MFOP, there will be a almost certainly to a point where these components will
maintenance recovery period (MRP) [3]. MRP is related to have to be replaced prematurely to maintain the desired
MFOP’s and is defined as the down time during which MFOP.
130 U. Dinesh Kumar et al./Reliability Engineering and System Safety 64 (1999) 127–131

Fig. 2. MFOPs for various shape parameter values of Weibull distribution.

6. Cost factors associated with MFOP parameter values of various items of the system are given
in Table 1.
To maintain a specified MFOP probability it may be Now using Eq. (4), the MFOPS(t mf,i), that is the prob-
required to replace the components prematurely (or insert ability that the system will survive ith cycle of MFOP, given
inspection). Discarding components before they fail will that it survives (i ¹ 1) cycles, is given by:
inevitably cost money (however, it reduces the cost due to Y Rk (i ⫻ tmf )
4
unscheduled maintenance). Besides the increase in the num- MFOPS(tmf , i) ¼ , (9)
ber of spares of these particular components, there is likely k¼1 Rk ([i ¹ 1] ⫻ tmf )
to be an increase in the number of LRI (line replaceable where
item) removals. In general, each component will have dif-
ferent failure distribution parameters and may have different R1 (i ⫻ tmf ) ¼ exp( ¹ 0:001 ⫻ i ⫻ tmf ),
ages. Thus, the times when they will become due for repla-    
cement (to maintain the required confidence level) will be i ⫻ tmf 3
R2 (i ⫻ tmf ) ¼ exp ¹ ,
different. This will effectively generate an LRI removal 1200
every time an item becomes critical (i.e. required to be  
1500 ¹ i ⫻ tmf
replaced). In practice, this number could be reduced if all R3 (i ⫻ tmf ) ¼ F ,
the components which will become critical in a given period 200
are replaced at the same time but, this will inevitably reduce   2:1 
their ‘useful’ life. The optimum length of this period will be i ⫻ tmf
R4 (i ⫻ tmf ) ¼ exp ¹ ,
a compromise between the cost of an LRI removal and the 1400
cost of the ‘lost’ life. Each such removal will require the where F(·) is the standard normal variate. Substituting the
LRI to be stripped, rebuilt and probably tested. Depending
above expressions in Eq. (9), one can get the value of
on the time to recover an LRI, the fact that there is an
MFOPS(t mf,i) for different values of i. Fig. 4 shows
increase in the number may lead to a need for more spare
MFOPS(t mf,i) for different cycles starting from the first
LRIs to maintain the system availability. Thus, use of cycle where t mf ¼ 50 h. It is easy to see from Fig. 4 that
MFOP will require careful analysis of all these factors. one has to carry out maintenance recovery after 4 cycles
(after 200 h) if the required confidence level, a, is greater
than 0.94 and one has to carry our maintenance recovery
7. Example after 11 cycles if a ⬎ 0.90 (after 550 h).
In this section, we consider an example to illustrate the
mathematical model based on mission reliability approach. Table 1
Consider a system with four items connected in series as Time-to-failure distribution of items of the system described in Fig. 3
shown in Fig. 3. The time-to-failure distribution and the Item Distribution Parameter values

Item 1 Exponential l ¼ 0.001 h ¹1


Item 2 Weibull h ¼ 1200 h, b ¼ 3
Item 3 Normal m ¼ 1500 h, d ¼ 200
Item 4 Weibull h ¼ 1400 h, b ¼ 2.1
Fig. 3. A series system with four items.
U. Dinesh Kumar et al./Reliability Engineering and System Safety 64 (1999) 127–131 131

Fig. 4. MFOPS(50,i) for different cycles.

8. Conclusion policy can only work if the producer, operator and


suppliers recognise that some components wear out with
The concept of MFOP is acknowledged by many aero- use, that not all failures are independent of the age of the
space industries in UK as a large step for future reliability component as MIL-HDBK-217 and MIL-STD-1388 imply.
specifications. Major projects like the ultra reliable aircraft They will also need to recognise that nothing in this world is
(to be implemented by the Society of British Aerospace free such that to achieve the required MFOP the suppliers
Companies) and future offensive aircraft (FOA) have recog- will inevitably have to increase their prices or the cost of
nised MFOP as a suitable reliability measure for replacing support.
MTBF as the latter has many drawbacks. The main advan-
tage of MFOP is that it tracks behaviour of the system
throughout the life of the system. Also MFOP will force Acknowledgements
the suppliers to analyse various failure mechanism which
will further help to improve the reliability and overall design We thank Mr. Ian Knowles, Ministry of Defence (UK) for
of the item. Though, a MFOP requirement is a good way of his useful suggestions. We also thank the referee for his
specifying ‘reliability’ but may prove expensive in certain constructive comments.
types of operation (e.g. training). In this paper, two
mathematical models are developed one based on mission
reliability approach and the other based on alternating References
renewal theory. To our best knowledge this is the first
time an analytical expression for MFOP prediction is [1] Ultra Reliable Aircraft Consortium: Ultra reliable aircraft — Pilot
phase report, 1997.
reported. The models will help reliability engineers and
[2] USAF R&M 2000 Process. United States Air Force report, ed. 1.
practitioners to predict and specify MFOP probability. In Washington, DC: USAF, 1987.
our models we assume that the time-to-failure of different [3] Knowles DI. Should we move away from acceptable failure rate.
items to be independent of each other. Future models should Communications in Reliability, Maintainability and Supportability,
consider items with dependent time-to-failure distribution 1995;2(1):23–28.
[4] MIL-HDBK-217. Reliability prediction of electronic equipment, ed.
and also the impact of maintenance on reliability function.
F. Department of Defence, 1991.
However, Eq. (4) is still valid if the time-to-failure is depen- [5] MIL-STD-1388-B. Department of Defence requirement for a logistic
dent except that one has to derive R k(t mf,i) considering support analysis record, ed. B, 1991.
dependence between time-to-failure distribution. [6] British Defence Standard 00-41 Part 2: Reliability apportionment,
If the majority of the failures are non-age related (or so modelling and calculation, 1989.
[7] Drenick RF. The failure law of complex equipment. Journal of Society
assumed) then there will be very little chance of improving
Of Industrial Applied Mathematics, 1960;8:680.
the MFOP probability or the duration of MFOP. In this case, [8] Hockley CJ, Appleton DP. Setting the requirements for the Royal Air
MFOP will not have any advantage compared to that of Force’s next generation aircraft. Annual Reliability and Maintainabil-
MTBF. To determine if any of the major causes of failure ity Symposium, 1997.
are age-related, it will be necessary to keep a full record of [9] Crocker J. Maintenance free operating period — Is this the way for-
ward? In: Proceedings of the 7th International M.I.R.C.E Symposium.
the ages of each occurrence of each of these components
Exeter, 1997.
and carry out an analysis of the times to failure. The actual [10] Gopalan MN, Dinesh Kumar U. Approximate analysis of n-unit cold
economics of MFOP’s would need to consider the above as standby systems. Microelectronics and Reliability, 1996;36(4):505–
well as the parts costs and maintenance costs. An MFOP 509.

You might also like