Beruflich Dokumente
Kultur Dokumente
Literaturhinweise
[1] Alsmeyer, G., Erneuerungstheorie. Analyse stochastischer Regenerationsschemata, Stuttgart, B.G.
Teubner, 1991
[2] Ascher, H.R., Feingold, H., Repairable System Reliability: Modeling, Inference, Misconceptions and
Their Causes, New York, Marcel Dekker, 1984
[3] Asmussen, S., Applied Probability and Queues, Second Edition, New York, Springer-Verlag, 2003.
[4] Aven, T., Reliability and Risk Analysis, London, Elsevier Applied Science, 1992
[5] Aven, T., Jensen, U., Stochastic models in reliability, New York, Springer-Verlag, 1999
[6] Barlow, R.E., Proschan, F., Mathematical Theory of Reliability, New York, Wiley, 1965. Reprinted:
Philadelphia, SIAM, 1996
[7] Barlow, R.E. Proschan, F., Statistical Theory of Reliability and Life Testing: Probabilistic Models, New
York, Holt, Rinehart and Winston, 1975. Reprinted with corrections: Silver Spring, MD, To Begin With,
1981
[8] Beichelt, F., Franken, P., Zuverlssigkeit und Instandhaltung: mathematische Methoden, Berlin, VEB
Verlag Technik, 1983
[9] Birolini, A., Reliability Engineering: Theory and Practice, Sixth Edition, Berlin, Springer-Verlag, 2010
[10] Cocozza-Thivent, C., Processus stochastiques et fiabilit des systmes, Berlin, Springer-Verlag, 1997
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
[11] Cox, D.R., Renewal Theory, London, Methuen & Co. Ltd, 1962
[12] Feller, W., An Introduction to Probability Theory and Its Applications, Volume II, Second Edition, New
York, Wiley, 1971
[13] Gnedenko, B.V., Belyayev, Y.K., Solovyew, A.D., Mathematical Methods of Reliability Theory, New
York, Academic Press, 1969
[14] Henley, E.J., Kumamoto, H., Reliability Engineering and Risk Assessment, Englewood Cliffs, Prentice
Hall, 1981
[15] Heyman, D.P., Sobel, M.J., Stochastic Models in Operations Research, Volume I, Stochastic Processes
and Operating Characteristics, New York, McGraw-Hill, 1982
[16] Resnick, S., Adventures in Stochastic Processes, Boston, Birkhuser, 1992 (4th printing, 2005).
[17] Ross, S.M., Stochastic Processes, Second Edition, New York, Wiley, 1996
[18] Villemeur, A., Reliability, Availability, Maintainability and Safety Assessment, Volume 1, Chichester,
Wiley, 1982
Fr eine vertiefende Behandlung der Zuverlssigkeit von Software wird auf folgende Literaturstellen
verwiesen:
BS 5760 Part 8:1998, Reliability of systems, equipment and components Part 8: Guide to assessment of
reliability of systems containing software.
IEC 61713:2000-06, Software dependability through the software life-cycle processes Application guide
56
1 CONTENTS
3 1 Scope ............................................................................................................................... 7
4 2 Normative references ....................................................................................................... 7
5 3 Definitions ........................................................................................................................ 8
6 4 Glossary of symbols and abbreviations ............................................................................. 9
7 4.1 Non-repairable items ............................................................................................... 9
8 4.2 Repairable items with zero time to restoration ....................................................... 10
9 4.3 Repairable items with non-zero time to restoration ................................................ 11
10 5 Assumptions ................................................................................................................... 14
11 5.1 General ................................................................................................................. 14
12 5.2 Assumptions for non-repairable items .................................................................... 14
13 5.3 Assumptions for repairable items ........................................................................... 15
14 6 Mathematical expressions .............................................................................................. 18
15 6.1 Non-repairable items ............................................................................................. 18
16 6.1.1 General ..................................................................................................... 18
17 6.1.2 Reliability [191-45-05] ................................................................................ 18
18 6.1.3 Instantaneous failure rate [191-45-06] ....................................................... 19
19 6.1.4 Mean failure rate [191-45-07] .................................................................... 20
20 6.1.5 Mean (operating) time to failure [191-45-11] .............................................. 21
21 6.2 Repairable items with zero time to restoration ....................................................... 22
22 6.2.1 General ..................................................................................................... 22
23 6.2.2 Reliability [191-45-05] ................................................................................ 22
24 6.2.3 Instantaneous failure intensity [191-45-08] ................................................ 23
25 6.2.4 Asymptotic failure intensity [191-45-10] ..................................................... 24
26 6.2.5 Mean failure intensity [191-45-09] .............................................................. 25
27 6.2.6 Mean (operating) time to failure [191-45-11] .............................................. 25
28 6.2.7 Mean operating time between failures [191-45-13] ..................................... 26
29 6.2.8 Mean up time [191-48-09] .......................................................................... 26
30 6.3 Repairable items with non-zero time to restoration ................................................ 26
31 6.3.1 General ..................................................................................................... 26
32 6.3.2 reliability [191-45-05] ................................................................................. 27
33 6.3.3 Instantaneous failure intensity [191-45-08] ................................................ 28
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
75
83 FOREWORD
84 1) The International Electrotechnical Commission (IEC) is a worldwide organization for standardization comprising
85 all national electrotechnical committees (IEC National Committees). The object of IEC is to promote
86 international co-operation on all questions concerning standardization in the electrical and electronic fields. To
87 this end and in addition to other activities, IEC publishes International Standards, Technical Specifications,
88 Technical Reports, Publicly Available Specifications (PAS) and Guides (hereafter referred to as IEC
89 Publication(s)). Their preparation is entrusted to technical committees; any IEC National Committee interested
90 in the subject dealt with may participate in this preparatory work. International, governmental and non-
91 governmental organizations liaising with the IEC also participate in this preparation. IEC collaborates closely
92 with the International Organization for Standardization (ISO) in accordance with conditions determined by
93 agreement between the two organizations.
94 2) The formal decisions or agreements of IEC on technical matters express, as nearly as possible, an international
95 consensus of opinion on the relevant subjects since each technical committee has representation from all
96 interested IEC National Committees.
97 3) IEC Publications have the form of recommendations for international use and are accepted by IEC National
98 Committees in that sense. While all reasonable efforts are made to ensure that the technical content of IEC
99 Publications is accurate, IEC cannot be held responsible for the way in which they are used or for any
100 misinterpretation by any end user.
101 4) In order to promote international uniformity, IEC National Committees undertake to apply IEC Publications
102 transparently to the maximum extent possible in their national and regional publications. Any divergence
103 between any IEC Publication and the corresponding national or regional publication shall be clearly indicated in
104 the latter.
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
105 5) IEC itself does not provide any attestation of conformity. Independent certification bodies provide conformity
106 assessment services and, in some areas, access to IEC marks of conformity. IEC is not responsible for any
107 services carried out by independent certification bodies.
108 6) All users should ensure that they have the latest edition of this publication.
109 7) No liability shall attach to IEC or its directors, employees, servants or agents including individual experts and
110 members of its technical committees and IEC National Committees for any personal injury, property damage or
111 other damage of any nature whatsoever, whether direct or indirect, or for costs (including legal fees) and
112 expenses arising out of the publication, use of, or reliance upon, this IEC Publication or any other IEC
113 Publications.
114 8) Attention is drawn to the Normative references cited in this publication. Use of the referenced publications is
115 indispensable for the correct application of this publication.
116 9) Attention is drawn to the possibility that some of the elements of this IEC Publication may be the subject of
117 patent rights. IEC shall not be held responsible for identifying any or all such patent rights.
118 International Standard IEC 61703 has been prepared by IEC technical committee 56:
119 Dependability.
121
122 Full information on the voting for the approval of this standard can be found in the report on
123 voting indicated in the above table.
124 This publication has been drafted in accordance with the ISO/IEC Directives, Part 2.
126 The committee has decided that the contents of this publication will remain unchanged until
127 the stability date indicated on the IEC web site under "http://webstore.iec.ch" in the data
128 related to the specific publication. At this date, the publication will be
129 reconfirmed,
130 withdrawn,
131 replaced by a revised edition, or
132 amended.
133
134 The National Committees are requested to note that for this publication the stability date
135 is ....
136 THIS TEXT IS INCLUDED FOR THE INFORMATION OF THE NATIONAL COMMITTEES AND WILL BE
137 DELETED AT THE PUBLICATION STAGE .
138
139
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`
140 INTRODUCTION
141 IEC 60050-191, Dependability terminology, provides definitions for dependability and its
142 influencing factors, reliability, availability, maintainability and maintenance support, together with
143 definitions of other related terms commonly used in this field. Some of these terms are
144 measures of specific dependability characteristics, which can be expressed mathematically.
145 This standard, used in conjunction with IEC 60050-191, provides practical guidance essential
146 for the quantification of those performance measures. For those requiring further information,
147 for example on detailed statistical methods, reference should be made to the IEC 60605
148 series of standards.
149 Annex A provides a diagrammatic explanation of the relationships between some basic
150 dependability terms, related random variables, probabilistic descriptors and modifiers.
152 Annex C compares some dependability measures for continuously operating items.
154 The bibliography gives references for the mathematical basis of this standard, in particular,
155 the mathematical material is based on references [2], [6], [8], [9], [13], [14] and [18]; the
156 renewal theory (renewal and alternating renewal processes) may be found in [6], [8], [9], [10],
157 [11], [13], [15] and [17]; and more advanced treatment of renewal theory may be found in
158 references [1], [3], [12] and [16].
159
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
160
167 1 Scope
168 This International Standard provides mathematical expressions for reliability, availability,
169 maintainability and maintenance support measures defined in IEC 60050-191. The following
170 classes of items are considered separately in this standard:
174 In order to keep the mathematical formulae as simple as possible, the following basic
175 mathematical models are used to quantify dependability measures:
180 To facilitate location of the full definition, the IEC 60050-191 reference for each term is shown
181 [in brackets] immediately following each term, for example:
184 This standard is mainly applicable to hardware dependability, but many terms and their
185 definitions may be applied to items containing software. Some of the software dependability
186 aspects are explained in Annex D.
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
187 2 Normative references
188 The following referenced documents are applicable to this standard. For dated references,
189 only the edition cited applies. For undated references, the latest edition of the referenced
190 documents (including any amendments) applies.
193 ISO 3534-1:2006, Statistics Vocabulary and symbols Part 1: General statistical terms and
194 terms used in probability
195
196 3 Definitions
197 For the purpose of this International Standard, the terms and definitions given in
198 IEC 60050-191:XXXX and ISO 3534-1 apply.
199 In addition, the following terms and definitions, which do not appear in IEV-191, are used in
200 order to facilitate the presentation of mathematical expressions for other IEV-191 terms.
201 3.1
202 instantaneous restoration intensity
203 restoration intensity
204 limit, if it exists, of the quotient of the mean number of restorations (191-46-23) of a repairable
205 item (191-41-11) within time interval (t, t + t), and t, when t tends to zero
E [ N R (t + t ) N R (t )]
206 v(t ) = lim
t 0 + t
207 where
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
208 N R (t) is the number of restorations in the time interval (0, t);
209 E denotes the expectation.
210 NOTE 1 to entry: The unit of measurement of instantaneous failure rate is the unit of time to the power 1.
211 NOTE 2 to entry: Other term used for restoration intensity is restoration frequency.
212 [[Document to be checked for usage of this term, and harmonized with IEV-191 Ed 2]]
213 3.2
214 up-time distribution function
215 function giving, for every value of t, the probability that an up time will be less than, or equal
216 to, t
217 NOTE 1 to entry: If the up time is (strictly) positive and continuous random variable, then F U (0) = 0 and
t
218 FU (t ) = 1 exp
0 U ( x) dx
219 where U (t) is the instantaneous up-time hazard rate function.
220 NOTE 2 to entry: If the up time is exponentially distributed, then
221 F U (t) = 1 exp(t/MUT)
222 where MUT is the mean up time.
223 In this case, the reciprocal of MUT is denoted by U :
224 U = 1/MUT
225 3.3
226 instantaneous up-time hazard rate function
227 up-time hazard rate function
228 U (t)
229 limit, if it exists, of the quotient of the conditional probability that the up-time will end within
230 time interval (t, t + t) and t, when t tends to zero, given that the up-time started at t = 0
231 and had not been finished before time t
232 NOTE 1 to entry: The instantaneous up-time hazard rate function is expressed by the formula:
1 FU (t ) FU (t + t ) f U (t )
233 U (t ) = lim =
t 0 t 1 FU (t ) 1 FU (t )
234 where F U (t) is up-time distribution function and f U (t) is the probability density at the up-time.
235 NOTE 2 to entry: If the up time is exponentially distributed, then the instantaneous up-time hazard rate function is
236 constant in time and is denoted by U .
237 NOTE 3 to entry: The unit of measurement of instantaneous up-time hazard rate function is the unit of time to the
238 power 1.
239 3.4
240 continuously operating item
241 COI
242 item for which operating time (191-42-05) is equal to its enabled time (191-42-17)
243 3.5
244 intermittently operating item
245 IOI
246 item for which operating time (191-42-05) is less than its enabled time (191-42-17)
248 The symbols given below are widely used and recommended but are not mandatory. For
249 consistency of presentation, the notation in this document may differ from that used in a
250 referenced document.
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
R(t) Reliability function, i.e. the probability of survival until time t:
R(t) = R(t1, t2) for t 1 = 0 and t 2 = t
R (t ) Point estimate of the reliability function at time t
R(t 1 , t 2 ) Reliability for the time interval (t 1 , t 2 )
R ( t, t + x | t ) Conditional reliability for the time interval (t, t + x), assuming that the
item survived to time t
TTF i Observed time to failure of item i
f(t) Probability density function of the (operating) time to failure
f (t ) Point estimate of the probability density function of the (operating) time
to failure at time t
n Number of (non-repairable) items in the population that are operational at
the instant of time t = 0
n S (t) Number of (non-repairable) items that are still operational at the instant
of time t (n S (0) = n)
n S (t) n S (t + t) Number of items that fail in the time interval (t, t + t)
Constant failure rate, i.e. the reciprocal of the mean time to failure
(MTTF) when the times to failure are exponentially distributed
Point estimate of the constant failure rate
(t) Instantaneous failure rate
(t ) Point estimate of the instantaneous failure rate at time t
(t1, t2 ) Mean failure rate for the time interval (t 1 , t 2 )
252
253
256
A Asymptotic availability
A(t) Instantaneous availability (availability function), i.e. the probability of the
item being in an up state at the instant of time t
A(t1, t2 ) Mean availability for the time interval (t 1 , t 2 )
A(t1, t2 ) Point estimate of the mean availability for the time interval (t 1 , t 2 )
COI Continuously operating item
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
F U (t) Up time distribution function
G(t) Distribution function of the repair times
G ACM (t) Distribution function of the active corrective maintenance time
G R (t) Distribution function of the times to restoration
IOI Intermittently operating item
M(t) Maintainability function, i.e. the probability of completing a given
maintenance action within time t: M(t) = M(t 1 , t 2 ) for t 1 = 0 and t 2 = t
M (t ) Point estimate of the maintainability function at time t
M(t 1 , t 2 ) Maintainability for the time interval (t 1 , t 2 )
N(t) Number of failures occurring in the time interval (0, t)
N R (t) Number of restorations occurring in the time interval (0, t)
R(t) Reliability function, i.e. the probability of survival until time t
R(t) = R(t 1 , t 2 ) for t 1 = 0 and t 2 = t
R(t 1 , t 2 ) Reliability for the time interval (t 1 , t 2 )
RTi Observed repair time of item i
MACMT Mean active corrective maintenance time, i.e. the expectation of the
active corrective maintenance time
MACMT Point estimate of the mean active corrective maintenance time
MAD Mean administrative delay
MAD Point estimate of the mean administrative delay
MADT Mean accumulated down time
MADT Point estimate of the mean accumulated down time
MAUT Mean accumulated up time
MAUD Point estimate of the mean accumulated up time
MDT Mean down time
MDT Point estimate of the mean down time
MFDT Mean fault detection time, i.e. the expectation of the fault detection time
MLD Mean logistic delay
MLD Point estimate of the mean logistic delay
MMAT Mean maintenance action time, i.e. the expectation of a given
maintenance action time
MRT Mean repair time
259 5 Assumptions
261 In order to derive correct mathematical expressions for the measures defined in
262 IEC 60050-191, the distinction needs to be made between repairable items [191-41-11] and
263 non-repairable items [191-41-12]. The following classes of items are considered separately
264 in this standard:
268 In order to keep the mathematical formulae as simple as possible, the following basic
269 mathematical models are used to quantify dependability measures:
274 The simplest mathematical model for the reliability of a non-repairable item is the random
275 variable, time to failure of the item [191-45-01]. One of the widely used reliability measures of
276 non-repairable items is the instantaneous failure rate, (t) [191-45-06], also referred to as
277 the hazard rate function. It is derived from the distribution function of the time to failure.
278 The expression (t)t is, for small values of t, approximately equal to the conditional
279 probability of failure of an item during the time interval (t, t + t) given that the item has not
280 failed during the interval (0, t].
281 For repairable items, the basic model is a simple renewal process, when the time to
282 restoration of the item may be neglected, or a simple alternating renewal process in which the
283 time to restoration of the item is non-zero. In the latter case, the item alternates between an
284 up state and a down state, and a widely used measure of reliability of the item is the failure
285 intensity, which is equal to the renewal density.
286 The failure intensity [191-45-08] is a measure derived from the expected value of the
287 cumulative number of failures E[N(t)] of a repairable item occurring during the time interval
288 (0, t]. The expression z(t)t is, for small values of t, approximately equal to the probability of
289 failure of the item during the time interval (t, t + t).
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
290 To avoid improper use of these mathematical expressions, which could yield erroneous
291 results, the specific assumptions detailed in 5.2 and 5.3 should be observed.
293 a) At any instant of time, the non-repairable item will be either in an up state or in a down
294 state (see figure 1).
295 b) Unless otherwise stated, when the item is in an up state, it is considered to be operating
296 continuously.
297 NOTE The mathematical expressions given in this sub-clause may not always be valid for IOIs.
298 c) At time t = 0, the item is in an operating state, and is as good as new. Latent faults are not
299 considered, which, if present, may invalidate some mathematical expressions.
300 d) Preventive maintenance, or other planned actions that render the item incapable of
301 performing a required function are not considered.
302 e) The time to failure is a positive and continuous random variable with a probability density
303 function and finite expectation.
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
State
Up state
Down state
0 Time
304
305 Key
306 Time to failure
309 a) At any instant of time, the repairable item will be either in an up state or in a down state
310 (see figures 2 and 3).
311 b) At time t = 0, the item is in an up state, and therefore, R(0) = 1, and is as good as new.
312 Latent faults are not considered.
313 c) Unless otherwise stated, when the item is in the up state, it is considered to be operating
314 continuously.
315 d) Consecutive up times of the item are statistically independent, identically distributed,
316 positive, continuous random variables with a common probability density function and
317 finite expectation.
318 e) In the case of non-zero down-time duration, the consecutive down times of the item are
319 statistically independent, identically distributed, positive, continuous random variables with
320 a common probability density function and finite expectation
321 f) The up times are statistically independent of the down times.
322 g) Preventive maintenance or other planned actions that render the item incapable of
323 performing a required function are not considered.
324 h) Unless otherwise stated, each other random variable (e.g. time to failure, repair time,
325 logistic delay, and so on) considered in the standard, is a positive and continuous random
326 variable with a probability density function and finite expectation.
327 In summary:
333
334 For continuously operating items (COI) the up state coexists with the operating state, and the
335 up time is concurrent with the operating time.
336 The expressions for reliability measures of continuously operating, repairable items may not
337 be true for IOIs (see figure 4).
338 NOTE 1 Models assuming zero time to restoration are used when either the up time of the item is the only time of
339 interest in assessing the performance of the item, or the time to restoration is so short that it is negligible.
340 NOTE 2 All mathematical expressions for the reliability measures relating to the time to failure of a non-repairable
341 item may also be applied to each time to failure of a continuously operating repairable item.
342
N(t)
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
m+1
m
m1
2
1
0
S F,0 = 0 S F,1 S F,2 S F, m 1 S F, m Time
State
Down state
348 Figure 2 Sample realization of a repairable item with zero time to restoration
349
N R (t)
m
m1
m2
2
1
0
S F,0 =S F,0 =0 S F,1 S R,1 S F,2 S R,2 S R, m 1 S F, m S R, m Time
N(t)
m
m1
2
1
0
S F,0 =S F,0 =0 S F,1 S R,1 S F,2 S R,2 S R, m 1 S F, m S R, m Time
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
State
R,1 R,2 R, m 1 R, m
Down state
350
351 Key
352 N(t) Number of failures during the time interval (0, t]
353 N R (t) Number of restorations during the time interval (0, t]
354 S F,1 , S F,2 , S F,3 ... Consecutive instants of failure
355 S R,1 , S R,2 , S R,3 ... Consecutive instants of restoration
356 U,1 U, 2 , U, 3 ... Consecutive up times
357 R,1 , R, 2 , R, 3 ... Consecutive times to restoration
358 Figure 3 Sample realization of a repairable item with non-zero time to restoration
Enabled time
Restoration Failure
Key
359
360 Figure 4 Comparison of an enabled time for a COI and an IOI
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
371 For non-repairable items, R(t 1 , t 2 ) for a given time interval (t 1 , t 2 ), 0 t 1 < t 2 , is equivalent to
372 the reliability R(0, t 2 ) for the time interval (0, t 2 ).
373 More commonly used expressions are the reliability function R(t) = R(0, t), with R(0) = 1, and
374 the conditional reliability R(t, t+ x | t), when no failure has occurred in time [0, t].
t
375 a) R(t ) = exp
0 ( x ) dx =
t f ( x ) dx
376 where
377 (x) is the instantaneous failure rate of the item;
378 f(x) is the probability density function of the time to failure of the item, i.e., for small values x,
379 f(x)x is approximately equal to the probability that the failure of the item will occur
380 during (x, x + x).
381 NOTE If observed failure data are available for n non-repairable items, from a homogenous population, the
382 estimated value of R(t) is given by
n (t )
383 R (t ) = S
n
384 where
385 n S (t) is the number of items that are still operational at the instant of time t (n S (0) = n).
386 The probability that the item will fail during the time interval (t 1 , t 2 ), 0 t 1 < t 2 , is given by
t2
387 R(t1) R(t 2 ) = t1 f (t) dt
388 The conditional reliability, R(t, t + x | t), is defined as the conditional probability that an
389 item can perform a required function for a given time interval (t, t + x) provided that the
390 item is in an operating state at the beginning of the time interval. (See [9] 1) , page 40.)
t+x R( t + x )
391 R(t, t + x | t ) = exp
t (t ) dt =
R( t )
392 b) When (t) = = constant, i.e. when the (operating) time to failure is exponentially
393 distributed
394 R(t) = exp(t)
395 R(t, t + x | t) = exp( x)
396 c) For an item with a constant failure rate of = 1 year 1 and a required time of operation of
397 six months, the reliability is given by
6
398 R(6 months) = exp(1 ) = 0,606 5
12
1 R ( t ) R ( t + t ) f (t )
402 (t ) = lim =
t 0 t R( t ) R( t )
403 For small values of t, (t)t is approximately equal to the conditional probability that
404 failure of the item will occur during (t, t + t), given that the item has survived to time t.
405 NOTE If observed failure data are available for n non-repairable items, from a homogenous population, the
406 estimated value of (t) at time t is given by
nS (t ) nS (t + t )
407 (t ) =
nS (t )t
408 where
409 n S (t) is the number of items that are still operational at the instant of time t (n S (0) = n);
410 n S (t) n S (t + t) is the number of items that fail in the time interval (t, t + t).
411 It should be noted that the estimated value of the failure density function f(t), at time t, is given by
n (t ) n S (t + t )
412 f (t ) = S
nt
________
1) Figures in square brackets refer to the bibliography.
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
413 The probability that the item will fail during the time interval (t 1 , t 2 ) is given by
t1 t2
414 R(t1) R(t 2 ) = exp
0 (t ) dt exp
0 (t) dt
415 b) When the time to failure is exponentially distributed, i.e. (t) = for all values of t,
416 f(t) = exp( t)
417 and
418 R(t) = exp( t)
419
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
NOTE If observed failure data are available for n non-repairable items, from a homogenous population, with
420 constant failure rate, then the estimated value of is given by
n
421 =
n
TTFi
i =1
422 where
423 TTF i is the time to failure of item i.
424 c) For 10 non-repairable items, from a homogenous population, with a constant failure rate,
10
425 the observed total operating time to failures of all the items is i =1TTFi = 2 years. Hence
10 -1
426 = = 5 year
2
427 If the time to failure of a non-repairable item has a two-parameter Weibull distribution with
428 scale parameter > 0 and shape parameter > 0, then
1 t2 1 R( t )
440 a) (t1, t 2 ) =
t 2 t1 t1 (t) dt = t2 t1 ln R(t21 )
441 b) When the time to failure is exponentially distributed
442 (t1, t2 ) =
444 c) Let t 1 = 6 months, R(t 1 ) = 0,8 and t 2 = 12 months, R(t 2 ) = 0,5, then
1 0,8 0,47
445 (6, 12) = ln = ln(1,6)/6 = = 0,078 3 month 1
12 6 0,5 6
1 1 0,2231
447 (0, 6) = ln = ln(1,25)/6 = = 0,037 2 month 1
6 0 0,8 6
450 a) MTTF = 0 tf (t) dt = 0 R(t) dt
451 NOTE If observed failure data are available for n non-repairable items, from a homogenous population, then
452 an estimate of MTTF is given by
n
TTFi
453 MTTF = i =1
n
454 where
455 TTF i is the time to failure of item i.
456 b) When the time to failure is exponentially distributed, i.e. (t) = for all values of t,
1
457 MTTF =
458 c) For a non-repairable item with a constant failure rate of = 0,5 year 1 ,
460 If the time to failure of a non-repairable item has a two-parameter Weibull distribution with
461 a scale parameter > 0 and shape parameter > 0, then
472 hence
476 All expressions in 6.2 are applicable to COIs. Where they are applicable to IOIs, this is stated.
484 a) The reliability of an item for the time interval (t 1 , t 2 ) may be written as (see [9], page 461
485 and [13], page 105):
t1
486 R(t1, t 2 ) = R(t 2 ) + 0 R(t2 t) z(t) dt
487 where
488 the first term, R(t 2 ), represents the probability of survival to time t 2 , and the second term
489 represents the probability of failing at time t (t < t 1 ) and, after immediate restoration,
490 surviving to time t 2 ;
491 z(t) is the instantaneous failure intensity (renewal density) of the item, i.e., for small values of
492 t, z(t)t is approximately equal to the (unconditional) probability that a failure of the item
493 occurs during (t, t + t), and
494 R(t) = R(0, t) is the reliability function of the item
495 R( t ) = t f ( s) ds
496 where f(t) is the probability density function (also referred to as the failure density
497 function) of the times to failure of the item, i.e., for small values of t, f(t)t is
498 approximately equal to the probability that the item fails during the time interval (t, t + t).
499 More precisely, it is approximately the probability that a given time to failure terminates in
500 the time interval (t, t + t), assuming that the time to failure started at time t = 0.
501 NOTE 1 R(t 1 , t 2 ) is also known as the interval reliability.
502 NOTE 2 If observed failure data are available for n repairable items, from a homogenous population, then an
503 estimate of R(t 1 , t 2 ) is given by
n (t , t )
504 R (t1, t 2 ) = S 1 2
n
505 where
506 n S (t 1 , t 2 ) is the number of items that were operational at the instant of time t 1 and did not fail during the time
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
507 interval (t 1 , t 2 ).
508 By setting t 1 = t and t 2 = t + x, one can obtain the asymptotic interval reliability (see [13],
509 page 106):
1
510 lim R(t, t + x ) =
t MTTF x R( s ) d s
511 which, for large values of t, can be used as an approximation of the R(t, t + x), where
514 b) When (t) = and is constant, i.e. when the times to failure are exponentially
515 distributed,
519 c) For a repairable item with a constant failure rate = 1 year 1 , its reliability over six months
520 is given by
6
521 R(t, t + 6) = exp (1 ) = 0,606 5
12
526 a) By definition [191-45-08], z(t) is the derivative of the expected number of failures,
527 Z(t) = E[N(t)], in the time interval (0, t) where N(t) is the number of failures during the time
528 interval (0, t) and E denotes the expectation.
Z (t + t ) Z (t ) dZ (t )
529 z(t ) = lim =
t 0 + t dt
530 From the renewal theory (see [9], page 457), it follows that z(t) may be written as
531 z(t ) = hCTTF
(n)
(t )
n =1
(n)
532 where hCTTF (t ) is the probability density function of calendar time until the nth failure of the
533 item. This may be calculated by the following recursive relationship:
(1)
534 hCTTF (t ) = f U (t )
(n) t ( n 1)
535 hCTTF (t ) = 0 f U ( x) hCTTF (t x) dx , for n > 1
536 where f U (t) is the probability density function of the up times of the item, i.e., for small
537 values of t, f U (t)t is approximately equal to the probability that a given up time of the
538 item terminates during (t, t + t), assuming that the up time started at time t = 0.
(n)
539 NOTE 1 Let U,1 , U, 2 ,..., U, n , be consecutive up times ( U ) of the item. Then hCTTF (t ) is the probability density
540 function of the sum
541 U,1 + U, 2 +... + U,n1 + U,n
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
542 NOTE 2 The instantaneous failure intensity, z(t), satisfies the following integral equation (see [9], page 458
543 and [11], page 54):
t
544 z (t ) = f U (t ) + 0 f U (t s) z(s) ds
545 which may be solved by numerical methods.
546 NOTE 3 If observed failure data are available for n repairable items, from a homogenous population, then an
547 estimate of z(t) is given by
n (t, t + t )
548 z(t ) = F
nt
549 where
550 n F (t, t + t) is the number of failures observed during the time interval (t, t + t).
552 For small values of t, z(t)t is approximately equal to the (unconditional) probability that
553 a failure of the item occurs during (t, t + t).
554 b) When the up times are exponentially distributed,
555 z(t) = U
560 a) By definition [191-45-10], z() is the limit, if it exists, of the instantaneous failure intensity
561 z(t), when time t tends to infinity:
1
564 z() = lim z(t ) =
t MUT
565 Under appropriate assumptions on f U (t), the above equation follows from the renewal
566 density theorem (see [5], page 248, [9], page 462, [10], page 196 and [12], page 367).
567 See Note 2.
568 NOTE 1 Using the elementary renewal theorem (see [9], page 461):
Z (t ) 1
569 lim =
t t MUT
570 but
t
571 Z (t ) = 0 z(s) ds
572 hence
Z (t )
573 z() = lim
t t
574 provided that z() exists (see [7], page 191).
575 NOTE 2 The easiest to check conditions for the existence of z() are the following: MUT < ; f U (t) is a
576 bounded function on [0, +) and tends to 0 as t +.
577 NOTE 3 If observed failure data are available for n repairable items, from a homogenous population, and the
578 time t is large enough, then an estimate of z() is given by
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
n (t, t + t )
579 z() = z(t ) = F
nt
580 where
581 n F (t, t + t) is the number of failures observed during the time interval (t, t + t).
582 For small values of t and large values of t, z()t is approximately equal to the
583 (unconditional) probability that a failure of the item occurs during (t, t + t).
584 b) When the up times are exponentially distributed,
585 z() = U
1 t2
590 a) z (t1, t 2 ) =
t 2 t1 t 1
z(t ) dt
t2
591 The integral t1
z(t ) dt is equal to the expected number of failures of the item in the time
592 interval (t 1 , t 2 ), hence z (t1, t2 ) may be interpreted as the expected number of failures per
593 time-unit in (t 1 , t 2 ).
594 NOTE If observed failure data are available for n repairable items, from a homogenous population, then an
595 estimate of z (t1, t 2 ) is given by
n (t , t )
596 z (t1, t 2 ) = F 1 2
(t 2 t1 ) n
597 where
598 n F (t 1 , t 2 ) is the number of failures observed in the time interval (t 1 , t 2 ).
599 By setting t 1 = t and t 2 = t + x, the asymptotic mean failure intensity can be obtained:
1
600 lim z (t, t + x ) =
MUT
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
601 which, for large values of t, can be used as an approximation of z (t, t + x ) . This equality is
602 a direct consequence of Blackwells theorem (see [9], page 462).
604 z (t1, t 2 ) = U
608 a) MTTF = 0 tf (t) dt = 0 R(t) dt
609 NOTE When all observed operating times to failure of n items, from a homogenous population, are available,
610 then an estimate of MTTF is given by
n
(operating time) i
total operating time
611 MTTF = = i =1
kF kF
612 where
613 total operating time is the aggregate operating time of all n items during a given time period;
614 k F is the total number of failures observed during the given time period;
615 (operating time) i is the aggregate operating time of the ith item during the given time period.
618 c) For a repairable item with a constant failure rate of 0,5 year 1
622 a) MOTBF = MTTF = 0 tf (t ) dt = 0 R(t ) dt
1
624 MOTBF =
628 a) MUT = 0 tfU (t) dt = 0 (1 FU (t)) dt
629 where f U (t) is the probability density function of the up times of the item (including its
630 operating, idle, standby and external disabled times).
1
634 MUT =
638 All expressions in 6.3 are applicable to COIs. Where they are applicable to IOIs, this is stated.
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
646 a) The reliability of a repairable item with non-zero time to restoration for the time interval
647 (t 1 , t 2 ) may be written as (see [13], page 113 and [9], page 177):
t1
0 R(t2 t) v(t) dt
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
649 where
650 the first term, R(t 2 ), represents the probability of survival to time t 2 , and the second term
651 represents the probability of restoration (after a failure) at time t (t < t 1 ), and surviving to
652 time t 2 ;
653 v(t) is the instantaneous restoration intensity of the item, i.e., for small values of t, v(t)t
654 is approximately equal to the probability that a restoration of the item occurs during (t, t +
655 t) (see definition 3.1);
656 R(t) = R(0, t) is the reliability function of the item
657 R( t ) = t f ( s) ds
658 where f(t) is the probability density function of the times to failure of the item, i.e., for small
659 values of t, f(t)t is approximately equal to the probability that the item fails during the
660 time interval (t, t + t). More precisely, it is approximately the probability that a given time
661 to failure terminates in the time interval (t, t + t), assuming that the time to failure started
662 at time t = 0.
663 NOTE 1 R(t 1 , t 2 ) is the (unconditional) probability of failure-free continuous operation of the item in the time
664 interval (t 1 , t 2 ). The expression may not be true for IOIs.
665 NOTE 2 An alternative name of the reliability given by R(t 1 , t 2 ) is the interval reliability.
666 NOTE 3 If observed failure data are available for n repairable items, from a homogenous population, then an
667 estimate of R(t 1 , t 2 ) is given by
n (t , t )
668 R (t1, t 2 ) = S 1 2
n
669 where
670 n S (t 1 , t 2 ) is the number of items which were operating at the instant of time t 1 and operated without failure in the
671 time interval (t 1 , t 2 ).
672 By setting t 1 = t and t 2 = t + x, one can obtain the asymptotic interval reliability (See [13],
673 page 113 and [9], page 183):
1
674 lim R(t, t + x ) =
t MTTF + MTTR x
R( s) ds
675 which, for large values of t, can be used as an approximation of the R(t, t + x), where
678 This expression follows from the key renewal theorem (see [9], page 462).
682 and
MTTF
683 lim R(t, t + x ) = exp( x )
t MTTF + MTTR
687 b) When the times to failure and times to restoration are exponentially distributed, then,
688 using either Markov techniques or the Laplace transform, the following is obtained:
R
689 R(t1, t 2 ) = + exp [ ( + R )t1 ] exp [ (t 2 t1)]
+ R + R
690 and
R
691 lim R(t, t + x ) = exp( x )
t + R
1 10 2 1
694 R 0, = + exp ( 12 0) exp 2 0 = 0,606 531
4 12 12 4
1 1 1 3 3
695 R , = 0,510 475; R , = 0,505 693; R , 1 = 0,505 455
4 2 2 4 4
1 10 1
696 lim R t, t + = exp 2 = 0,505 442
t 4 12 4
700 a) By definition [191-45-08], z(t) is the derivative of the expected number of failures, Z(t) = E[N(t)],
701 in the time interval (0, t), including up and down times, where N(t) is the number of failures
702 in the time interval (0, t), and E denotes the expectation, thus
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
Z (t + t ) Z (t ) dZ (t )
703 z(t ) = lim =
t 0 + t dt
704 Using the alternating renewal process theory (see [9], pages 180 and 466), it follows that
705 z(t) may be written as
706 z(t ) = hCTTF
(n)
(t )
n =1
(n)
707 where hCTTF (t ) is the probability density function of calendar time to the nth failure and
708 may be calculated by the following recursive relations:
(1)
709 hCTTF (t ) = f U (t )
(n) t ( n 1)
710 hCTTF (t ) = 0 hCTTF ( x) f U+R (t x) dx , for n > 1
711 where
712 f U (t) is the probability density function of the up times of the item (including its operating,
713 idle, standby and external disabled times). For small values of t, f U (t)t is approximately
714 equal to the probability that a given up time of the item terminates during (t, t + t),
715 assuming that the up time started at time t = 0;
716 f U+R (t) is the probability density function of the sum of the up time ( U ) and the
717 corresponding time to restoration ( R ), and is given by
t
718 f U +R ( t ) = 0 gR (t s) f U (s) ds
719 where gR (t ) is the probability density function of the times to restoration of the item, i.e.,
720 for small values of t, gR (t ) t is approximately equal to the probability that the item is
721 restored from a fault to an up state in the time interval (t, t + t), assuming that a failure
722 resulting in a fault occurred at time t = 0.
723 For small values of t, z(t)t is approximately equal to the (unconditional) probability that
724 a failure of the item occurs during the time interval (t, t + t).
725 NOTE 1 Let U, 1 , R,1 , U,2 , R, 2 ,..., U,n , R,n ... be consecutive up times ( U ) and times to restoration ( R ) of the
(n)
726 item. Then hCTTF (t ) is the probability density function of the sum
728 while f U+R (t) is the probability density function of the sum R,m 1 + U,m for any m > 1.
729 NOTE 2 The instantaneous failure intensity, z(t), and the instantaneous restoration intensity, v(t), fulfil the
730 following simultaneous system of linear Volterra integral equations (see [14], page 193):
t
731 z (t ) = f U (t ) + 0 f U (t s) v(s) ds
t
732 v (t ) = 0 g R (t s) z(s) ds
733 which may be solved by numerical methods.
734 NOTE 3 If observed failure data are available for n repairable items, from a homogenous population, then an
735 estimate of z(t) is given by
n (t, t + t )
736 z(t ) = F
nt
737 where n F (t, t + t) is the number of failures observed during the time interval (t, t + t), where the time scale
738 includes both up and down times.
739 When the up times are exponentially distributed, then (see [4], page 317)
743 b) When the up times and times to restoration are exponentially distributed, Markov
744 techniques or the Laplace transform can be used to yield (see [14], page 200 and [18],
745 page 66):
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
U R U2
746 z (t ) = + exp [ (U + R )t ] = A(t )U
U + R U + R
748 c) For a COI with a failure rate of = 2 year 1 and a restoration rate of R = 10 year 1 , and
749 for arbitrarily chosen values of t = 0, ,, and 1
1 20 4 1
750 z(0) = 2 year 1; z = + exp 12 = 1,683 262 year 1
4 12 12 4
1 3
751 z = 1,667 493 year 1; z = 1,666 708 year 1; z(1) = 1,666 669 year 1
2 4
752
756 a) By definition [191-45-10], z() is the limit, if it exists, of the instantaneous failure intensity
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
757 z(t), when time t tends to infinity:
1
760 z() = lim z(t ) =
t MUT + MTTR
761 which, under appropriate assumptions on f U (t) and g R (t), follows from the renewal density
762 theorem (see [5], page 248, [9], pages 184 and 461, [10], page 196 and [12], page 367).
763 See Note 2.
764 NOTE 1 Using the elementary renewal theorem (see [9], page 461):
Z (t ) 1
765 lim =
t t MUT + MTTR
766 but
t
767 Z (t ) = 0 z(s) ds
768 hence, if z() exists, then
Z (t )
769 z() = lim
t t
770 (see [7], page 191).
771 NOTE 2 The easiest to check conditions for the existence of z() are the following: MUT < , MTTR < , at
772 least one of f U (t) or g R (t) is a bounded function on [0, +) tending to 0 as t +.
773 NOTE 3 If observed failure data are available for n repairable items, from a homogenous population, and the
774 time t is large enough, then an estimate of z() is given by
n (t, t + t )
775 z() = z(t ) = F
nt
776 where
777 n F (t, t + t) is the number of failures observed during the time interval (t, t + t).
778 For small values of t and large values of t, z()t is approximately equal to the
779 (unconditional) probability that a failure of the item occurs during (t, t + t).
780 When the up times are exponentially distributed, then (see [4], page 317)
781 z() = A U
784 b) When the up times and times to restoration are exponentially distributed, then (see
785 6.3.3b),
U R 1
786 z() = lim z(t ) = =
t U + R 1 1
+
U R
788 c) For a COI with a failure rate of = 2 year 1 and a restoration rate of R = 10 year 1 ,
20
789 z( ) = = 1,666 667 year 1
12
793 a) By definition
1 t2
794 z (t1, t 2 ) =
t 2 t1 t1 z(t) dt
t2
795 The integral t1 z(t) dt is equal to the expected number of failures of the item in the time
796 interval (t 1 , t 2 ). Hence z (t1, t 2 ) may be interpreted as the expected number of failures per
797 time-unit in (t 1 , t 2 ).
798 NOTE If observed failure data are available for n repairable items, from a homogenous population, then an
799 estimate of z (t1, t2 ) is given by
n (t , t )
800 z (t1, t 2 ) = F 1 2
(t 2 t1 ) n
801 where n F (t 1 , t 2 ) is the number of failures observed during the time interval (t 1 , t 2 ), where the time scale includes
802 both up and down times.
803 By setting t 1 = t and t 2 = t + x, the asymptotic mean failure intensity may be obtained:
1
804 lim z (t, t + x ) =
t MUT + MTTR
805 which, for large values of t, can be used as an approximation of z (t, t + x ) . This equality is
806 a direct result of Blackwells theorem (see [9], page 462).
807 When the up times are exponentially distributed (see [4], page 317),
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
810 b) When the up times and times to restoration are exponentially distributed, then (see [14],
811 page 200 and [18], page 66)
814 c) For a COI with a failure rate of = 2 year 1 and a restoration rate of R = 10 year 1
1
exp( 12 0) exp 12
1 20 4 4 = 1,772 246 year 1
815 z 0, = +
4 12 144 1
0
4
1 1 1 3 3
816 z , = 1,671923 year 1; z , = 1,666 928 year 1; z , 1 = 1,666 680 year 1
4 2 2 4 4
1 1
817 lim z t, t + = = 1,666 667 year 1
t 4 0,5 + 0,1
820 a) MTTF = 0 tf (t) dt = 0 R(t) dt
821 where R(t) is the reliability function of the item
822 R( t ) = t f ( s) ds
823 NOTE When observed (operating) times to failure of n items, from a homogenous population, are available,
824 then an estimate of MTTF is given by
n
826 where
827 total operating time is the aggregate operating time of all n items during a given time period;
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
828 k O is the total number of failures of the items while operating during the given time period;
829 (operating time) i is the total operating time of the ith item during the given time period.
1
831 MTTF =
1
833 MTTF = = 0,5 years = 4 380 h
2
1
840 MOTBF =
1
842 MOTBF = = 0,5 years = 4 380 h
2
846 a) The instantaneous availability of a repairable item with non-zero time to restoration at an
847 instant of time t may be written as (see [13], page 111, [9], page 175 and [5], page 101):
t
848 A(t ) = 1 FU (t ) + 0 (1 FU (t x)) ( x) dx
849 where
850 F U (t) is the up time distribution function of the item:
t
851 FU (t ) = 0 fU (s) ds
852 which is equal to the probability that a given up time will less than or equal to t;
853 v(t) is the instantaneous restoration intensity of the item.
854 NOTE 1 The instantaneous availability, A(t), is equal to the probability that the item is in an up state at the
855 instant of time t and is given by the following integral equation (see [9], page 176):
t
856 A(t ) = 1 FU (t ) + 0 f U+R (s) A(t s) ds
857 which may be solved by numerical methods,
858 where
t
859 f U+R (t ) = 0 g R (t s) f U (s) ds
860 is the probability density function of the sum of the up time and the corresponding time to restoration.
861 NOTE 2 If observed up-state data are available for n repairable items, from a homogenous population, then
862 an estimate of A(t) is given by
n {t }
863 A (t ) = U
n
864 where n U {t} is the number of items which are in an up state at the instant of time t.
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
865 If the item operates continuously, then R U (t) = R(t) and f U (t) = f(t).
866 b) When the up times and times to restoration are exponentially distributed, then using either
867 Markov techniques or the Laplace transform, the following is obtained (see [9], page 181,
868 [14], page 200 and [18], page 66):
R U
869 A(t ) = + exp[ (U + R )t ]
U + R U + R
1 10 2 1
873 A(0) = 1; A = + exp 12 = 0,841 631
4 12 12 4
1 3
874 A = 0,833 746; A = 0,833 54; A (1) = 0,833 334
2 4
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
10
875 A = A() = = 0,833 333
12
879 a) The instantaneous unavailability of a repairable item with non-zero time to restoration at an
880 instant of time t may be written as (see [5], page 107]]):
t
881 U(t) = 1 A(t) = 0 (1 GR (t x)) z( x) dx
882 where
883 z(t) is the instantaneous failure intensity of the item;
884 G R (t) is the distribution function of the times to restoration of the item
t
885 GR (t ) = 0 gR (s) ds
886 which is equal to the probability that a restoration of the item is completed by time t. In this
887 formula, g R (t) is the probability density function of the times to restoration of the item.
888 NOTE 1 The instantaneous unavailability, U(t), is equal to the probability that the item is in a down state at
889 the instant of time t.
890 NOTE 2 If observed down-state data are available for n repairable items, from a homogenous population,
891 then an estimate of U(t) is given by
n {t }
892 U (t ) = D
n
893 where n D {t} is the number of items which are in a down state at the instant of time t.
894 b) When the up times and times to restoration are exponentially distributed, then (see [14],
895 page 200 and [18], page 66),
U
896 U (t ) = (1 exp[ (U + R )t ])
U + R
1 2 1
900 U (0) = 0; U = 1 exp 12 = 0,158 369
4 12 4
1 3
901 U = 0,166 254; U = 0,166 646; U (1) = 0,166 666
2
4
2
902 U = U() = = 0,166 667
12
1 t2
906 a) A(t1, t 2 ) =
t 2 t1 t1 A(t ) dt
t2
907 The integral t1
A(t )dt is equal to the expected up time accumulated in the time interval
908 (t 1 , t 2 ), hence A (t1, t2 ) gives the expected fraction of the time interval (t 1 , t 2 ) that the item is
909 in an up state.
910 It then follows that the mean availability, A (t1, t 2 ), and the mean accumulated up time,
911 MAUT, in the time interval (t 1 , t 2 ) are related as
t2
912 MAUT = t1 A(t ) dt = A(t1, t 2 ) (t 2 t1)
913
914 NOTE If observed up times in the interval (t 1 , t 2 ) are available for n repairable items, from a homogenous
915 population, then an estimate of A(t1, t 2 ) is given by
total up time
(up time) i
916 A ( t 1, t 2 ) = = i =1
(t 2 t1 ) n (t 2 t1 ) n
917 where
918 total up time is the aggregate up time of all n items during the time interval (t 1 , t 2 );
919 (up time) i is the total up time of the ith item during the time interval (t 1 , t 2 ).
920 An estimate of the mean accumulated up time, MAUT, in the time interval (t 1 , t 2 ) is given by
n
total up time
(up time) i
921 MAUT = = i =1
n n
922 where
923 total up time is the aggregate up time of all n items during the time interval (t 1 , t 2 );
924 (up time) i is the total up time of the ith item during the time interval (t 1 , t 2 ).
925 Following from the assumptions given in 5.3, the asymptotic mean availability lim A(t1, t 2 )
t2
926 is equal to the asymptotic availability A (see [7], page 191):
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
MUT
927 lim A(t1, t2 ) = A =
t2 MUT + MTTR
928 b) When the up times and times to restoration are exponentially distributed, then integrating
929 A(t) over the time interval (t 1 , t 2 ) gives:
MAUT
A(t1, t 2 ) =
t 2 t1
R U exp [ (U + R )t1] exp [ (U + R )t 2 ]
933 = +
U + R (U + R ) 2 t 2 t1
z (t1, t 2 )
=
U
934 If the item operates continuously, U = .
935 c) For a continuously operating item with a failure rate of = 2 year 1 and a restoration rate
936 of R = 10 year 1 , then
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1
exp( 12 0) exp 12
1 10 2 4
937 A 0, = + = 0,886 123
4 12 144 1
0
4
1 1 1 3 3
938 A , = 0,835 962; A , = 0,833 464; A , 1 = 0,833 340
4 2 2 4 4
939 The mean accumulated up time, MAUT, during the first year may be calculated as follows:
1 1 1 1 3 3
940 MAUT = A(0, 1) 1 = A 0, + A , + A , + A , 1 / 4
4 4 2 2 4 4
1 t2
945 a) U (t1, t 2 ) =
t 2 t1 t1 U (t) dt = 1 A (t1, t 2 )
t2
946 The integral t1
U (t ) dt is equal to the expected down time accumulated in the time interval
947 (t 1 , t 2 ). Hence, U (t1, t2 ) gives the expected fraction of the time interval (t 1 , t 2 ) spent in the
948 down state.
949 It then follows that the mean unavailability, U (t1, t 2 ), and the mean accumulated down
950 time, MADT, in the time interval (t 1 , t 2 ) are related as
t
951 MADT = t12 U (t) dt = U (t1, t2 ) (t2 t1)
952 NOTE If observed down times in the interval (t 1 , t 2 ) are available for n repairable items, from a homogenous
953 population, then an estimate of U (t1, t 2 ) is given by
955 An estimate of the mean accumulated down time, MADT, in the time interval (t 1 , t 2 ) is given by
n
960 Following from the assumptions given in 5.3, the asymptotic mean unavailability lim U (t1, t2 )
t2
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
MTTR
962 lim U (t1, t2 ) = U =
t2 MUT + MTTR
963 b) When the up times and times to restoration are exponentially distributed, then integrating
964 U(t) over the time interval (t 1 , t 2 ) gives:
MADT
U (t1, t 2 ) =
t 2 t1
U U exp [ (U + R )t1] exp [ (U + R )t 2 ]
967 =
U + R (U + R ) 2 t 2 t1
z (t1, t 2 )
= 1
U
969 c) For a COI with a failure rate of = 2 year 1 and a restoration rate of R = 10 year 1
1
exp ( 12 0) exp 12
1 2 2 4 = 0,113 877
970 U 0, =
4 12 144 1
0
4
1 1 1 3 3
971 U , = 0,164 038; U , = 0,166 536; U , 1 = 0,166 660
4 2 2 4 4
972 The mean accumulated down time, MADT, during the first year may be calculated as
973 follows:
1 1 1 1 3 3
974 MADT = U (0, 1) 1 = U 0, + U , + U , + U , 1 / 4
4 4 2 2 4 4
MUT
979 a) A = lim A(t ) =
t MUT + MTTR
MTTF
981 A =
MTTF + MTTR
982 b) When the up times and times to restoration are exponentially distributed, then
R
983 A=
U + R
985 c) For a COI with a failure rate of = 2 year 1 and a restoration rate of R = 10 year 1
10
986 A= = 0,833 333
12
MTTR
990 a) U = lim U (t ) = =1A
t MUT + MTTR
MTTR
992 U=
MTTF + MTTR
993 b) When the up times and times to restoration are exponentially distributed, then
U
994 U=
U + R
996 c) For a COI with a failure rate of = 2 year 1 and a restoration rate of R = 10 year 1 , then:
2
997 U= = 0,166 667
12
1001 a) MUT = 0 tf U (t) d t = 0 (1 RU (t)) d t
1002 where
1003 f U (t) is the probability density function of the up times of the item;
1004 F U (t) is the up time distribution function of the item.
1005 NOTE 1 When the item operates continuously, then, according to the assumption 5.3g) (i.e. no preventive
1006 maintenance):
1007 MUT = MTTF
1008 However, when function-preventing preventive maintenance is permitted, the relation between MUT and MTTF
1009 is more complex, and usually, MUT < MTTF.
1010 NOTE 2 If observed up times are available for n repairable items, from a homogenous population, then an
1011 estimate of MUT is given by
n
total up time
(up time) i
1012 MUT = = i =1
kU kU
1013 where
1014 total up time is the aggregate up time of all n items during given period of observation;
1015 k U is the total number of up times of the items during given period of observation;
1016 (up time) i is the total up time of the ith item during the given period of observation.
1017 EXAMPLE
1018 Consider an item belonging to IOIs, which operates as described below. The item starts
1019 operating at time t = 0 and is required to operate (be in operating state) for fixed time
1020 interval [0, X), X > 0 during which the item can fail with a constant failure rate >0. If the
1021 item did not fail in this interval, it becomes not required for subsequent fixed time interval
1022 [X, X+Y), Y > 0, during which the item cannot fail (i.e. the item is in a failure free idle state
1023 during this interval). At time t = X+Y the next required time starts and the functioning
1024 process repeats as from the initial time t = 0, independently of previous history of the item
1025 functioning process. When the item fail during required time [0, X), it is repaired to the
1026 state as goods as new and then the functioning process repeats again as at the initial time
1027 t = 0, independently of the previous history of the item functioning process. Repair times
1028 are mutually independent and also independent of the previous history of the item
1029 functioning process. See figure 5 below.
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
State
X X X X
Operating
state
Up state
Y Y Y
Idle
state
Down state
Failure
Restoration, the item
continues functioning
as at t = 0
1030
1032 It is clear from the above description that consecutive up times are statistically
1033 independent and identically distributed positive, continuous random variables. Therefore,
1034 to calculate MUT, the up time to first failure of the item can be considered. The up-time
1035 hazard rate function U (t ) is depicted in figure 6.
U (t) X X X X
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
Y Y Y
0
Time
t=0 X X+Y 2X+Y 2X+2Y 3X+2Y 3X+3Y 4X+3Y
1036
for n ( X + Y ) t < n ( X + Y ) + X
1039 U (t ) = , n = 0, 1, 2, ...
0 for n ( X + Y ) + X t < (n + 1) ( X + Y )
1040 Since
t
1041 1 FU (t ) = exp
0 U ( x) dx ,
1042 then by integrating U (t ) , the above formula can be written as:
1 exp( X ) exp( X )
1045 MUT = +Y = MOTBF + Y
1 exp( X ) 1 exp( X )
1046 The second term of the above formula is equal to the expected accumulated idle time to
1047 failure of the item. For = 0,01 h 1 and X = 10 h, we obtain the following values of MUT
1048 for some values of Y:
1049 MUT = 195 h for Y = 10 h, MUT = 290 h for Y = 20 h,
1051 whereas:
1052 MOTBF = 100 h.
1
1054 MUT =
U
1
1056 MUT = = 0,5 years = 4 380 h
2
1060 a) MDT = 0 t gD (t) dt
1061 where gD (t ) is the probability density function of the down times of the item (which are
1062 defined to include the items restoration times, after failure and/or function-preventing
1063 preventive maintenance times), i.e., for small values of t, gD (t ) t is approximately equal
1064 to the probability that the item returns to its up state from its down state in the time
1065 interval (t, t + t), assuming that the down time started at time t = 0.
1066 NOTE If observed down times are available for n repairable items, from a homogenous population, then an
1067 estimate of MDT is given by
n
1069 where
1070 total down time is the aggregate down time of all n items during a given time period;
1071 k D is the total number of the down times of the items in the given time period;
1072 (down times) i is the total down time of the ith item during the given time period.
1073 b) If the down times are exponentially distributed with a parameter D , i.e.
1074 gD (t ) = D exp( D t )
1075 then
1
1076 MDT =
D
1077 NOTE According to the assumptions in 5.3 (any fault is the result of a failure, and no preventive
1078 maintenance), any down time is equal to the time to restoration, i.e.
1079 MDT = MTTR
1080 and, for exponentially distributed down times
1081 D = R
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1
1083 MDT = = 0,01 years = 87,6 h
100
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1085 (Symbol M(t 1 , t 2 ), 0 t 1 < t 2 )
1087 a) The probability that a given maintenance action on an item can be completed in the
1088 time interval (t 1 , t 2 ), assuming that the maintenance action started at time t = 0, is given by
t2
1089 M (t1, t 2 ) = t1 g MA (t ) dt
1090 where gMA (t ) is the probability density function of the time to complete a given
1091 maintenance action for the item, i.e., for small values of t, gMA (t ) t is approximately
1092 equal to the probability of completing a given maintenance action during the time interval
1093 (t, t + t), assuming that the maintenance action started at time t = 0.
1094 In practical applications, the maintainability function, M(t), defined as
t
1095 M(t) = M(0, t) = 0 gMA ( x) dx
1096 is used, with M(0) = 0. This is equal to the probability that a given maintenance action will
1097 be completed by time t, assuming that the action started at time t = 0, i.e. M(t) is the
1098 distribution function of that time. The maintainability, M(t 1 , t 2 ), and the maintainability
1099 function, M(t), are related as follows:
1117 where mMAT (t ) is the number of maintenance action times with duration greater that t, i.e. not finished up to
1118 time t, and mMAT (0) = m.
1119 b) If a given maintenance action times are exponentially distributed with a parameter MA , i.e.
1
1125 MMAT =
MA
1126 c) For a repairable item with MA = 1 000 year 1 , and t 2 t 1 = 16 h, then [simplify by using
1127 hours, rather than years, as the unit]
1 000
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1128 M(16) = M(0, 16) = 1 exp(16 ) = 0,839 021[too many sig. figures]
8 760
1129 also:
1 000 1 000
1130 M(2, 18) = exp(2 ) exp(18 ) = 0,667 758
8 760 8 760
1131 and
1 000 1 000
1132 M(4, 20) = exp(4 ) exp(20 ) = 0,531 453
8 760 8 760
1136 a) By definition,
1 G ( t + t ) G ( t ) g (t )
1137 (t ) = lim =
t 0 t 1 G (t ) 1 G (t )
1138 where
1139 g(t) is the probability density function of the repair time of an item (excluding technical,
1140 logistic and administrative delays), i.e., for small values of t, g(t)t is approximately
1141 equal to the probability that the repair started at time t = 0 will be completed during (t, t +
1142 t);
1143 G(t) is the distribution function of the repair time of the item, i.e. G(t) is the probability that
1144 the repair, started at time t = 0, with G(0) = 0, will be completed by time t:
t t
1145 G(t ) = 1 exp
0 ( x) dx = 0 g( x) dx
1146 For small values of t, (t)t is approximately equal to the conditional probability that the
1147 repair will be completed in the time interval (t, t + t), given that the repair started at time t
1148 = 0 and has not been completed by the instant of time t.
1149 NOTE If observed repair data are available for n repairable items, from a homogenous population, the
1150 estimated value of (t) at time t is given by
nR (t ) nR (t + t )
1151 (t ) =
nR (t )t
1152 where
1153 n R (t) is the number of items that are still under repair at the instant of time t (n R (0) = n);
1154 n R (t) n R (t + t) is the number of items with repair completed in the time interval (t, t + t).
1155 It should be noted that the estimated value of the repair density function g(t), at time t, is given by
n (t ) nR (t + t )
1156 g (t ) = R
nt
1
1165 MRT =
1172 c) For 10 repairable items, from a homogenous population, with a constant repair rate, the
10
1173 observed total repair time of all the items is i =1RTi = 5 hours. Hence
10
1174 = = 2 h 1
5
1175 If the repair time of a repairable item has a lognormal distribution with scale parameter m
1176 and shape parameter > 0, then (see Table B.2)
1 (ln t m) 2
1177 g (t ) = exp
t 2 2 2
1178 and
t
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1179 G (t ) = 0 g( x) dx
1180 hence
g (t )
1181 (t ) =
1 G (t )
1182 EXAMPLE
1183 Assuming that an item has a mean repair time (MRT) of 1,5 hour and repair time variance
2
1184 (VRT) of 0,16 hour , in order to compute the repair rate, the parameters m and of the
1185 lognormal distribution of repair times have first to be determined. According to the results
1186 given in Table B.2,
2
1187 MRT = exp m +
2
1198 a) MRT = 0 t g(t) dt = 0 (1 G(t)) dt
1199 where g(t) is the probability density function of the repair time of an item, and G(t) is the
1200 distribution function of the repair time of the item.
1201 NOTE 1 From the definition of the repair time
1202 MRT = MACMT MTD
1203 where
1204 MTD is the mean technical delay;
1205 MACMT is the mean active corrective maintenance time.
1206 NOTE 2 If observed repair times are available for n repairable items, from a homogenous population, then an
1207 estimate of MRT is given by
n
1213 b) If the repair times are exponentially distributed with a parameter , i.e.
1214 g (t ) = exp( t )
1215 then
1
1216 MRT =
1
1218 MRT = = 0,001 years = 8,76 h
1 000
--``,,``,,,,,,,`,,,`,,,`,,```-`
1222 a) MACMT = 0 (1 G ACM (t )) dt = 0 t g ACM (t) dt
1223 where
1224 g ACM (t ) is the probability density function of the active corrective maintenance times of an
1225 item (including technical delay and repair time, but excluding logistic and administrative
1226 delays), i.e., for small values of t, g ACM (t ) t is approximately equal to the probability
1227 that the active corrective maintenance of the item is completed in the time interval (t, t +
1228 t), assuming that the active corrective maintenance started at time t = 0;
1229 G ACM (t) is the distribution function of the active corrective maintenance time of the item,
1230 i.e. G ACM (t) is the probability that the active corrective maintenance, started at time t = 0,
1231 will be completed by time t:
t
1232 G ACM (t ) = 0 g ACM ( x) dx
1233 NOTE 1 By definition of the active corrective maintenance time:
1234 MACMT = MRT + MTD
1235 where MTD is the mean technical delay.
1236 NOTE 2 If observed active corrective maintenance times are available for n repairable items, from a
1237 homogenous population, then an estimate of MACMT is given by
n
1239 where
1240 total active corrective maintenance time is the aggregate active corrective maintenance time of all n items
1241 during a given time period;
1242 k ACM is the total number of active corrective maintenance actions on the items during the given time period;
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1243 (active corrective maintenance time) i is the total active corrective maintenance time of the ith item during the
1244 given time period.
1245 b) If the active corrective maintenance times are exponentially distributed with parameter
1246 ACM , i.e.
1248 then
1
1249 MACMT =
ACM
1250 c) For a repairable item with the mean technical delay MTD = 5 h and the mean repair time
1251 MRT = 9 h:
1252 MACMT = 5 + 9 = 14 h
1256 a) MTTR = 0 t gR (t) dt
1257 where g R (t ) is the probability density function of the times to restoration of the item, i.e.,
1258 for small values of t, gR (t ) t is approximately equal to the probability that the item is
1259 restored from a fault to an up state in the time interval (t, t + t), assuming that a failure
1260 resulting in a fault occurred at time t = 0.
1261 NOTE 1 The mean time to restoration (of a faulty item), MTTR, may be written as the sum of the expected
1262 values of its constituent times:
1263 MTTR = MFDT + MAD + MLD + MACMT
1264 = MFDT + MAD + MLD + MTD + MRT
1265 where
1266 MFDT is the mean fault detection time;
1267 MAD is the mean administrative delay;
1268 MLD is the mean logistic delay;
1269 MACMT is the mean active corrective maintenance time given by
1270 MACMT = MTD + MRT
1271 where
1272 MTD is the mean technical delay;
1273 MRT is the mean repair time.
1274 NOTE 2 If observed times to restoration are available for n repairable items, from a homogeneous population,
1275 then an estimate of MTTR is given by
n
1277 where
1278 total time to restoration is the aggregate time to restoration of all n items during a given time period;
1279
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
k R is the total number of times to restoration of the items during the given time period;
1280 (time to restoration) i is the total time to restoration of the ith item during the given time period.
1
1284 MTTR =
R
1
1286 MTTR = = 0,01 years = 87,6 h
100
1290 a) MAD = 0 t g AD (t) dt
1291 where g AD (t ) is the probability density function of the administrative delay during a time to
1292 restoration of a faulty item, i.e., for small values of t, g AD (t ) t is approximately equal to
1293 the probability that the delay ends in the time interval (t, t + t), assuming that the delay
1294 started at time t = 0.
1295 NOTE If observed administrative delays are available for n repairable items, from a homogenous population,
1296 then an estimate of MAD is given by
n
1298 where
1299 total administrative delay is the aggregate administrative delay of all n items during the given time period;
1300 k AD is the total number of administrative delays during the given time period;
1301 (administrative delay) i is the total administrative delay of the ith item during the given time period.
1302 b) If the administrative delays are exponentially distributed with parameter AD , i.e.
1303 g AD (t ) = AD exp( AD t )
1304 then
1
1305 MAD =
AD
1
1307 MAD = = 0,001 years = 8,76 h
1 000
1311 a) MLD = 0 t gLD (t) dt
1312 where gLD (t ) is the probability density function of the logistic delay during a maintenance
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1313 time of a faulty item, i.e., for small values of t, gLD (t ) t is approximately equal to the
1314 probability that the delay ends in the time interval (t, t + t), assuming that the delay
1315 started at time t = 0.
1316 NOTE If observed logistic delays are available for n repairable items, from a homogenous population, then an
1317 estimate of MLD is given by
n
1319 where
1320 total logistic delay is the aggregate logistic delay of all n items during a given time period;
1321 k LD is the total number of logistic delays during the given time period;
1322 (logistic delay) i is the total logistic delay of the ith item during the given time period.
1323 b) If the logistic delays are exponentially distributed with parameter LD , i.e.
1324 g LD (t ) = LD exp( LD t )
1325 then
1
1326 MLD =
LD
1
1328 MLD = = 0,001 years = 8,76 h
1 000
1329
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1330 Annex A
1331 (informative)
1332
1333 Performance aspects and descriptors
1334
Performance Random Probabilistic Modifiers
aspects variables descriptors
Predicted
Time between Probability
failures density
function Estimated
Reliability Number of failures
in interval (t 1 , t 2 )
Reliability Extrapolated
function
Time to Instantaneous
restoration
Failure
Maintenance Preventive rate Asymptotic
support maintenance time
Renewal
function
Up time Renewal
density
Maintainability Down time
Expectation
Variance
Standard
deviation
p-fractile
1335
1336 NOTE A mathematical operation on a random variable results in a basic measure. The addition of a modifier to a
1337 basic measure results in a specific measure.
1339
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1340 Annex B
1341 (informative)
1342
1343 Summary of measures related to time to failure
1344
1345 Table B.1 Relations among functional measures of time to failure
1346 of continuously operating items
dF (t ) 1 dF (t )
F(t) 1 F(t)
dt 1 F (t ) dt
f (t )
t
f(t) 0 f ( x ) dx t f ( x ) dx
t f ( x ) dx
dR(t ) 1 dR(t )
R(t) 1 R(t)
dt R(t ) dt
t t t
(t) 1 exp
0 ( x) dx (t ) exp
0 ( x) dx exp
0 ( x) dx
NOTE Similar relationships hold among functional measures of any random variable, for example time to first
failure, up time, down time, time to restoration, corrective maintenance time, repair time.
1347
1348
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1349
1350
1351
1352 Table B.2 Summary of measures for some continuous probability distributions of time to failure of
1353 continuously operating items
Reliability
Probability density Failure rate Expected
Distribution Range (survival) Variance
function f(t) (t) value MTTF
function R(t)
>0 1 1
Exponential exp ( t ) exp ( t )
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
t0 2
1
Weibull
> 0, > 0
t0
(
( t ) 1 exp ( t ) ) (
exp ( t ) ) ( t ) 1
(1 +
) 1 2 1
(1 + ) 2 (1 + )
2
> 0, > 0 ( t ) 1 f (t )
Gamma
t0 ( )
exp ( t ) t f (u) du
R( t ) 2
>0 k 1
( t )k 1 ( t ) i f (t ) k k
Erlang k {1, 2, ...}
(k 1) !
exp ( t ) exp ( t ) i! R( t ) 2
t0 i =0
k {1, 2, ...} kt 2 kt 2 2
Rayleigh kt exp exp
kt 1
t0 2 2 2k k 4
x 1 t
(x) is the complete gamma function defined as ( x ) = 0 t e dt , x > 0.
1354
1355 Annex C
1356 (informative)
1357
1358 Comparison of some dependability measures for continuously operating items
1359
1360 Table C.1 Comparison of some dependability measures of continuously operating items with
1361 constant failure rate and restoration rate R
R
Reliability R(t 1 , t 2 ) exp( t 2 ) exp( ( t2 t1)) + + + exp[ ( + R ) t1] exp[ (t2 t1)]
R R
1 1 1
Mean time to failure MTTF
1 1
Mean operating time between failures MOTBF
R 2
Instantaneous failure intensity z(t) exp( t ) + exp[ ( + R )t ]
+ R + R
R
Asymptotic failure intensity z() 0
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
+ R
R
Instantaneous availability A(t) exp( t ) 1 + exp[ ( + R )t ]
+ R ( + R )
exp( t1) exp( t 2 ) R exp[ ( + R )t1] exp[ ( + R )t2 ]
Mean availability A(t1, t 2 ) 1 +
(t 2 t1) + R ( + R )2 t2 t1
R
Asymptotic availability A 0 1
+ R
Instantaneous unavailability U(t) 1 exp( t ) 0 (1 exp[ ( + R ) t ] )
+ R
exp( t1) exp( t 2 ) exp ( + R )t1] exp[ ( + R )t2 ]
Mean unavailability U (t1, t 2 ) 1 0
(t 2 t1) + R ( + R )2 t2 t1
1362
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
Copyright Deutsches Institut fr Normung e. V.
Provided by IHS under license with DIN Licensee=Bogazici University/5964815002
No reproduction or networking permitted without license from IHS Not for Resale, 01/07/2014 09:07:19 MST
E DIN EN 61703:2013-11 (QWZXUI
61703/Ed2/CD IEC (E)
55
1363 Annex D
1364 (informative)
1365
1366 Software dependability aspects
1367
1368 Items containing software are prone to design failure due to the activation of a latent design
1369 fault resulting from human error during software development.
1370 In test or operation, these latent faults may be activated in response to certain trigger
1371 circumstances.
1372 Such failures tend to be transient (i.e. the item becomes operable again if the trigger is
1373 removed) and systematic (i.e. the item will fail in a similar manner if the trigger is encountered
1374 again or if deliberately reproduced).
1375 Corrective maintenance of software requires a modification to remove the fault. This can
1376 usually be performed away from the installation after normal operation has been resumed.
1377 An important attribute of items containing software is the ability to recover to an upstate
1378 following the failure of the item due to a software fault. This is measured by the time to
1379 recover, which comprises three aspects:
1385 Availability with regard to software is a combination of its reliability and the ability to recover.
1386 Some modes of failures will have a zero time to recovery, for example giving wrong results for
1387 a calculation, and do not affect availability.
1388 Maintainability of the software is measured by the time to effect a modification. This time is a
1389 combination of time to diagnose (subdivided into time to localize and time to identify the fault),
1390 active maintenance time (to devise the modification) and re-testing time (time to re-run earlier
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1392 Maintenance of software requires the movement of information rather than of people and
1393 parts, and logistic delays are due to queues within the support system.
1394 The overall failure intensity (rate of occurrence of failures) for items containing software is the
1395 sum of the failure intensity of the hardware plus the failure intensities resulting from
1396 processes of activation of all individual latent design faults.
1397
1398 Bibliography
1399
[2] Ascher, H.R., Feingold, H., Repairable System Reliability: Modeling, Inference,
Misconceptions and Their Causes, New York, Marcel Dekker, 1984.
[3] Asmussen, S., Applied Probability and Queues, Second Edition, New York, Springer-
Verlag, 2003.
[4] Aven, T., Reliability and Risk Analysis, London, Elsevier Applied Science, 1992.
[5] Aven, T., Jensen, U., Stochastic models in reliability, New York, Springer-Verlag,
1999.
[6] Barlow, R.E., Proschan, F., Mathematical Theory of Reliability, New York, Wiley,
1965. Reprinted: Philadelphia, SIAM, 1996.
[7] Barlow, R.E. Proschan, F., Statistical Theory of Reliability and Life Testing:
Probabilistic Models, New York, Holt, Rinehart and Winston, 1975. Reprinted with
corrections: Silver Spring, MD, To Begin With, 1981.
[8] Beichelt, F., Franken, P., Zuverlssigkeit und Instandhaltung: mathematische
Methoden, Berlin, VEB Verlag Technik, 1983.
[9] Birolini, A., Reliability Engineering: Theory and Practice, Sixth Edition, Berlin,
Springer-Verlag, 2010.
[12] Feller, W., An Introduction to Probability Theory and Its Applications, Volume II,
Second Edition, New York, Wiley, 1971.
[13] Gnedenko, B.V., Belyayev, Y.K., Solovyew, A.D., Mathematical Methods of Reliability
Theory, New York, Academic Press, 1969.
[14] Henley, E.J., Kumamoto, H., Reliability Engineering and Risk Assessment,
Englewood Cliffs, Prentice Hall, 1981.
[15] Heyman, D.P., Sobel, M.J., Stochastic Models in Operations Research, Volume I,
Stochastic Processes and Operating Characteristics, New York, McGraw-Hill, 1982.
[16] Resnick, S., Adventures in Stochastic Processes, Boston, Birkhuser, 1992 (4th
printing, 2005).
[17] Ross, S.M., Stochastic Processes, Second Edition, New York, Wiley, 1996.
[18] Villemeur, A., Reliability, Availability, Maintainability and Safety Assessment, Volume
1, Chichester, Wiley, 1982
1400 Further information on the treatment of software aspects of dependability is available from the
1401 following references:
1402 BS 5760: Part 8:1998, Reliability of systems, equipment and components. Part 8. Guide to
1403 assessment of reliability of systems containing software.
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---
1404 IEC 61713:2000-06 Software dependability through the software life-cycle processes
1405 Application guide.
1407 ____________
1408
--``,,``,,,,,,,`,,,`,,,`,,```-`-`,,`,,`,`,,`---