0% found this document useful (0 votes)
10 views28 pagesReliabilitymu Notes22pdf
The document discusses reliability engineering, focusing on the systematic analysis of failures in engineering systems, their causes, and methods for improving reliability through both engineering and quantitative analysis. It emphasizes the importance of defining failures in quantitative terms and highlights various factors contributing to system failures, such as poor design, complexity, and inadequate maintenance. Additionally, it covers reliability characteristics, including Mean Time To Failure (MTTF) and Mean Time Between Failures (MTBF), and introduces the concept of the bathtub curve to illustrate the reliability lifecycle of products.
Uploaded by
rahularunmane.im22Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
10 views28 pagesReliabilitymu Notes22pdf
The document discusses reliability engineering, focusing on the systematic analysis of failures in engineering systems, their causes, and methods for improving reliability through both engineering and quantitative analysis. It emphasizes the importance of defining failures in quantitative terms and highlights various factors contributing to system failures, such as poor design, complexity, and inadequate maintenance. Additionally, it covers reliability characteristics, including Mean Time To Failure (MTTF) and Mean Time Between Failures (MTBF), and introduces the concept of the bathtub curve to illustrate the reliability lifecycle of products.
Uploaded by
rahularunmane.im22Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 28
if Engineer and ie Testing
Y ;
‘Systematic analysis of failures and ries o aes the sues, suchas why do
Systems fal: how frequenly failures occur; how to accurately predict the
‘ceurtence of failures: how to minimize or eliminate them: What are the
fconomis, risk loss and safety aspects of system failures, spd the related
teas, I also considers how 10 design the system for beter maintenance
fineness, or bene maintainability.
The concep of reliability engineering ae highly useful for designing,
Ianufgctunng operating, nd maintaining prods and stems. The spectrins
of actives of reliably engineering can be broadly classified into lwo
aweas—engineerng analysis of fares and quantitative analysis of fares,
Engincering analysis of failures tiesto find out the root cause of files of
8 component or sem tased onthe known scenic and engineering thecrce
and concepts. Thereafter. it atempts to find at methods to eliminate tse
auses and subsequently improve the operational reliably, Another related
acovity is mainnabiliy. This is applicable for a repairable system ‘Ths
sums at designing sytem for easier and faster maintenance. In quantitative
failure analysis, varios statistical and mathematical principles ae extensively
Used. Tis is also equally imponant compared to the engineesng fale
‘analysis. Quanfication of files is very essential to find out the presen
condition, establish fture goals, conduct pap analysis, compare products of
sifferent makes, judge design improvements, and do benchmarking. Tis also
ssemal 1 find out the reliability growth during the sages of product
evelopment and design improvement In a nutsel, enginecrng analysis
and saistical analysis are complementary to each other and ae essential for
implementing efecve and successful reliably programme
Quastitatively, reliability of a system of product is defined as the
probably tht the system or product performs its intended farction adequately
for the sued time duration, usually the mission life, under the sated
environmental conditions. Four important terms ae pen this definition,
‘az probablty, intended function ine, and envionment. Reliability is measured
8a probability because the occurence af failure has random or stochastic
nature. Since reliability is defined a probability, twill be wrong to specify
the reibility of single unit or specific product. tis always specified for
the population of products or systems. Imended function of «system refers
tothe design function ofthe system. For example, 2 uid pump 1s designed
4 pump ceran Mud atthe rate of, say, 10 mfminte and 12 m hea. These
te the intended funtion ofthe pump. Sometimes the intended function is
also termed the specified quility of he product, Time refer othe design fe
‘of mission life ofthe system. The term ‘environment refers the eer
condition in which the system normally operates, Temperature, humidity,
salinity. vibration, shocks, acceleration, ee are the examples of he opeating
{environments Sometimes also common to consider the inp prams
Of the syste i he eof environmen
sabe 08
oper Raley Eagan «3
The foes ofthis books on quantatve reality analysis of components
| and systems. Varios quantitative reliability modes dats sources Ie ieting
tnd estimation methods are discussed in detail to help peoctsing engines
8 wel as academicians to quaniativly evaluate reliably craters
of components and systems
1.2. FAILURES OF ENGINEERING SYSTEMS
‘The failure of a system oF a component is the inability of the system or
component to deliver its intended function siisfactoriy. Fires may be
citer paial or complete. For example, a? HP motor has completely filed
if it does not produce any power. ft produces some power (3), 1 HP) hen
the motor has failed partly, Sine the term ‘part fature’ i subjective in
tut, iis required that we clearly define a failure in quantitative temas,
which can be understood without any futher explnation, Some systems
have operating anges. For example, a voltage stabilizer has opeatioal ange
0f 230 ¢ 1 V. Therefore, i i prodices output voltage within this range
then it performs well therwise sha ful. Tet are ote cases of ales.
in which the physical failure ofa component is considered a socess. For
example, consider the ease of a protective fuse wire. If the fase wire bums
{hysiclly destroyed) when there is overond, we say that i a succes. Th
other words, we say that a fase has failed when it does not burs when
overloading occurs. There are oer deviees having mli-model flues, For
example, a mechanical valve may experince flue to clos. fale to open,
flue to top leaking. There may aso be complete or partial flues
‘Again, in many oer cases. a failure is defined based onthe partial
pplication, and not only on the intended design fonction For example
conser a voltage stabilize designed for 230 +18 V output. I we use ths
for an application where we reque only 250 +25 vols, nd iit spplies
this, hen it may be considered a success. On the other hand, the same
stabilize is failure ft doesnot suport an aplication where 230 = 1% V
1s required.
‘Therefore, itis essential for any reliability activity to stat with clear
and objective definition of failures, preferably in quantitative terms, 0 avoid
any confusion at later stages, Tis farther reinforces the ned to clesly
specify the intended function and state the defnion of fale ofa sytem
before tempting to evaluate is reliability
1.3 CAUSES OF FAILURES
“There are many cougex of fuuges of componenig.and 4 + Resabilty Engineering and Lae Testing
combination. leading fo ther failures. Some of these are known and others
unknown due to many reasons. Whatever be the mechanisms and cause of
failures iis sure that al these result in ether decreasing the strength ofthe
‘omponet o increasing the applied load on itor doing both simultaneously
‘When this occurs, the component fils ata point when the load om it excceas
iss strength, That means, in general, all failures ate eto deterorecee of
strength or lack of control on applied lod. ‘There are many mechonrng
acting on the components causing deterioration of steagih; For trample
‘er fatigue fcson,coroson, chemical reaction, oxidation ad chang a
‘hysical or chemical properties are some ofthe eauses leading todetenenaion
fsrength, These ae natural phenomena causing failures, Apart fron thane
there ae other causes of failures. Some ofthe important cauee are diremee
der
Poor design isthe primary and most important cause of unreliability of
8 product or system, Traditional design considers that lsd and stent are
(cteministic and therefore assume a factor of safety to improve design
However, it has heen found that both strength and load have soshantc anor
and therefore probabilistic design techniques must be used to ensure required
‘elabilcy targets. Lack of adequate investment on improved design, incorcet,
‘manufactrng and testing procedutes, and improper material selesgn wht
considering operating environment aso test i fullurcs
CCompiexity is another reason for system failures Complex systems
severally have more number of components, thereby decreasing overall sytem
‘baby. Complex systems ae difficult o understand. operat, and mavsarn
‘compared to simple systems
‘Lack of proper and effective maintenance is an important cause of
‘ystem failure. This s very important especialy for mechanical systems and
for those systems with increasing failure rates during useful lie period. Such
‘ystems require rezula cleaning, lubrication, balancing. alignment and prevents
‘maintenance for achieving beter reliability. The falure rate increases due to
+f Roe
uae
(16
1.9.2 Time-dependent Hazard Models
In his section, var
Ou types of hazard models are discussed ia
collected from the field or that, ‘et, Data
discussed earlier. Hazard rate a en sin
ened ier cae pate sean net pa
A ypical hazard-rate cur .
Hpfenbepeticlin ‘unequal. The discrete nature ofthis cont 3
se ata toe
eee tical function can be found out to sine
cites rw
aoe ‘components operating under ae
ndtons. Some ofthe gency we hatarte modes ae en
‘hazard, non-linear hazard, and gamma hazards St
19.21 Constant hazard model
cma aw mod
inom do be afc »
Th constant bz odecan bo expesedt-e fog
PETAON Of the random
" cs)
pressed in tems of ite eeiabli
ty fen and hazard
i
Cnaptes 1: Raby Engnedting » V9
FIGURE 1.3 Typics hazardate cre.
wer aun
1. Many products. particularly
the simplest hazed
reliability analysis
fy functions ean
where 2 is a constant, independent of time
MeEfoaic components, exhibit tis characterise. This
‘ee model and therefore hasbeen used for many years
Tha prediction. As discussed in Section 1.9, various re
be derived for this model, These ae:
Probability density Function. #0
Reliability, R= ©"
Unrelibiiy, Foo
“The mean time to failure of the item is
pa
aay)
For this model, it can be seen that the probability demity fonction of
ae (eat raurer flows an exponential distibution and the MITE 1
irae the rexel of the hazard ate, Te wsefal Me period (OHO)
ee es wl follow this tp of behavior a shown by Horizont pomien
fof the bathtub cure .
‘An item with constant haz
and enoited functions Faw 19)
rom Figure 14, it may be ned that clibity
constant hazard. model onetulally with is age. Reliability
sant hazard model reduces exponentelly Ral
cemesponding to the age equal 1o,MTTR Ws equal to y/e = O.S07%. 8
arc ate will have the following rllbility
of any item following FIGURE.L4 Reatiity of component with constant fale rae
unreliability is equal to (1-1e) = 0.6321 This indicates that 63.21% of such
‘ems are expected to fil tthe end of MTTF.
1.9.2.2 Linear hazard model
Ua linear hazard mol, the hazard rate varies with cme and bas a linear
relationship. Many components under mechanical stresses fi due to Weat=
out or deterioration. Th hazard rate of such components inreaces with time
“The simplest form of the linear hazard model can be expressed as
eh, 130 19)
where Bis constant and its value i equ o the slope ofthe’ hazand rate
function. The various reliability fesetions can be derived from this as discussed
on
surseal-[f wa]
sap 402)
1) ey 82 2
‘may be noted tha fle) i 2 Rayleigh- density funtion, Inthe bathtub curve,
far many componente porton beyond th seta period might flo
‘model, a .
The mean time to failure is given by
. ' :
hes ein opening + 2
1.9.23 Non-linear hazard model
‘The hazard rate for many components is not always a linely increasing
fnction of time. It may also hive non-linear relationship. tis therefore
‘sel to have a more general form ofthe hazard model as expressed show
sae “12a
where a and B are constants
We get the folowing reliability functions for the no-tines hana
rode!
0
pla Mb + 1) aay
KO = a? expl-at Mb + 1), 4)
‘This probability density Function generates a wide range of curves for
Various sets of « and bis known asthe Weibull model. Iecan be scent
this mode represents constant hazard model for b =O, and inary insula
hazard model for 6 = 1. The parameter @affets the ampiude and b Oh
shape of the hazard rate fonction. Therefore, thee are known a cule a
‘shape parameters, eespectvey.
(MTTF for this model can be derived using the basic Equation (19.
—tles)
bei
ony
Ad)
ar
c1asy
1
we
‘The nontincarhaard model is ed to dy thee
during the weur-out period of the baud curve. Unlke 1
ponents, any of the mechanical components stu woaig athe
{i tine of ter se and omit wet thereat, Ageing peoomenon
‘much ery for mechanical systems, Therefore, this modelcan bg
ively sed for studying neh ayes, 2 22+ Retain tng od ie Tein
1.9.24 Gamma hazard model
“The Gamma hazard move! is another two-parameter model applied in rei
analysis. The follow
fancton
ii
relationship is used to represent the Gamma hazard
so cn
oa we
‘where. and Ze constant having positive values, Varios reibihty functions
associated wih this model are
gaat
" azn
me?
mee 129)
MrTF= = 129)
“This model takes the role ofa constant hazard rate model for a = 1, and
increasing hazard rate model for @ > 1, This model can be used to study the
reliability of @ component which is replaced (a ~ 1) times by identical
‘components, Inherent itis assumed thatthe replacement canbe carted out
without any significant amount of downtime. This is the ease of standby
redundancy in which a tual of ‘a’ components ae used. Each of the “a!
‘components has constant hazard rate A. The combined hazard rate is inereasing
‘wih time as per Equation (1.26). a = I indicates that there is only one
‘component
1.9.25 Other general hazard models
“There are some components that experience no failures fr certain initia ime
period. Thereafter, they sat failing with increasing hazard rate The following
model can be used to stidy such components
a) = alt=, 1% (130)
Here, fy is known a5 the location parameter representing the falur-ree
period ofthe component. This mode! «known a the hree-parameer Weibull
‘model and is extensively sed to study many mechanieal system
Similarly, some other components have constant hazard rates for some
Ina petod and increasing hazaré thereafter. The following model
sititin ands anh tahoonmamnmeld:
rege 1 tata tgenrng + 23%
0) = 0e" aap
Where a and are constants.
“The reliability foretions are:
; Ruy = ep [4alone"~ 0) am
fozae o-(Efe" . oj ay
“This model is generally known as the exponential-haard model
“Many other models can be derived and used for studying special even
cof component failures, For studying the exact behaviour of flues. i It
Suggested that we stat from the basic equation of the harard rate fet
tnd then generalize the model for such components,
1.10. RELIABILITY LOGIC DIAGRAMS
etiaity loge dgrams (RLD) te agra drawn with boxes nd
rerun eto lato of componei fan semby
sretjetct i conpeent and ten nse th tna ea
Cemcs"componcats Ths i aso tnowe tthe relay bok ay
caerea component appears as bck estan guar) nthe MLD,
ID testy alee om engin wna gu a
Ship show he pyscalelanoip of ddl components nm
foe cumple, conidet sal ketal tte stem hat esas
Fiche cnty equipment on the cacent xcedh 10 spre,
Bear lps en be aed fori fea wo new ath
4 apes capone) in pl hi ae wl aso pros t
comely Mee pal wept se wes in ps: ba he
‘Tastes congution. On the oer had we po wo ne
Yo amperes aposy each in seis he RLD wil ea pre ong uw
Taeee ing af any ene or bh five wie wil rt te QR
fen overlong Ths champ antes te imprance of RLD
ae 2
“ne knowlege of RLD ofa syst eset if ve want waa
tne reiaiay prams fesse fm taf he onset bent
‘rte rman of RLD toe pf crc
TDs provided we do nt condr thease a ac DOE.
Tee ed nt condr eer yi vlan of net
wimps orf atesoment. The spate condor a Sk
{oo knowaige aout ts Componens), od eb sea
fare data ox expernenal data ofthe compe syne, TE 24+ Rehab Engineering and ie Tesing
‘only for those systems which are already in use (and considerable fel data
‘Savailable) or a number of fete on prototypes have generated sufficient
dda However, reliability atthe preliminary and detailed design slags cannot
be evaluated ty black-box approach. Reliability evaluation and assurance a
Cesign stages are eesebal fr design improvement, and subsequent approval
hd therefore the only method for doing this is the RLD approach. Series,
Paral, series parallel, no-sris-paralel,outof-m, and networks ae some
bf the generally used rcisbiitylopie diagrams. Well formulated methods are
Svalable to study these configurations for reliability evaluation of systems.
‘These are discussed in etal inthe next section
1.10.1 System Reliability Models
‘Asters defined san arrangement of 2 numberof is consent components
‘a some specific manner s that it converts certain easily availble inputs into
the required eurpat, Some systems are simple, consisting of only a few
Components; some others fe comples, consisting of a large number of
Components with miny interconnections. The main concem of a systems
Engineer is to estimate various reliability parameters of the systems he is
teaing with The general approach fr analyzing such systems isto decompose
them into subsystems of convenient size, and then into component. Tis can
bbe done n'a aumber of stages. Components are atthe lowest hierarchical
Ieee and the sytem i tthe highest. Reablity parameters of the components
te evaluated as discussed i Ue previous section. Reliabltes of subsystems
fe thea estimated using the well-known probability laws. Based on the
oneionl relationship (RLD) between the components, appropriate probability
Thus must Be used for this. This procedure is then repeated fo evaluate the
reliability parameters ofthe next higher level ofthe system hierarchy. In this
(ay in a numberof steps we can evaluate the system reliability. complete
Knowledge of the physical structure of the system and the nature of its
fanctionalcelationships is required to eamty out such modelling.
"This section discusses various system reliability models fr certain well
known RLDS. such as series, paral, stand-by, A-out-ofm,non-seres parallel,
nd networks. All models assume thatthe components fail independent of
ach other, 1 the flue of one component does not affet the fiture or
Success of other component
1.10..1 Series systems
Let us consider a sytem having a toll of °t” components. After a careful
snalyis ofthe physical diagram and the fenetiona relationships, we have 10
dow the RED. If the KLD To0ks like dram as shown in Figure 1.8, we
‘Conolude that It in a series system, The important property of the series
CHE
HIGURE 1.5. Series system,
should bein operation. In other words, the system wil fi i any
fof its components fi. The series systems ae also known she
fm, since each link ofthe chain i esental fr is success. Many
Ugyiems in ure can be reduced to such a simple structure
WLR; be the even withthe component fin working condition
(vent representing system success is then evaluated using the
lon law of probability ofall the events EE. Ey. The relay
isthe probability tat all these events occur simultancouly
lly, can be expressed a:
R= PLE, OBOE)
= PAE) PAESED) PAEVEED
{ BAER/E)) iste probaly of occurence of the event with the
ln tha the event has occured. H we assume that he component
3h we get
= PLE) PAE)» PAE 8)
neans the reliability ofa eres system is equal to the produ of
‘tality components, Numerically, the system rhbility wll
he reliability ofthe leas reliable component. tn other words
Ir of a chain Is leser than the strength of the weakest link oft
consider system having thee componcns in series, Th’
‘hose components ace 0.7, 06, and 0.9. Te sytem reliblty
Re = 07 x06 x04
rn the reliability of the lowest component (0.6), TA
wom eiability dereaten if We use more numbet
sytem Therefore, the technique of compovent-miiminaon
Jimsane of achiving higher fllblity of syne, Sym
might also lead to ths a
Ineomponena of a serie stm are dental with probed
9p oc uraiaily a L=p) the
oats 26 + selads
‘The system reliability can also Be evaluated sing fire events F, a8
{allows
Ry = 1 = (Probabibty of the system file)
Since the system fas sf any ofthe components fil we mus use the
‘unjon law of probably 19 mode cis We gt the following relationship:
Reet -RE CEU. UE) aan
‘The time-dependent reiality function can als he evaluated ina simile
Taw
10 ps0
where pi) the probebiiy thatthe component iis oper
a special case if mes to flr of components are exponentially distributed
(eonsant Rand rate. then we have
pin = exp Cap 39)
lexan sof-r£] 140)
The meantime to fale ofthe stem
rode et aay
an
whe 7, she mean fe ofthe componest
om th. ne get he Sse head rata
a ba aay
Mir,
‘Therefore, the system failure ate is also a constant equal tothe sum of,
‘he falure rates of individual components .
‘The stem MITF for ie T) and components MITE (or hie T) are
related by
1
seed aay
r )
Cupar | :alliuiay txgrennng > 27
However, we must be careful in sing these relationships (Equations
[1.42) and (1.43)). These hold good only for those systems having constant
rates forall it components
“The rebabily expression for any general hazard model canbe expressed
noeenl-f, Ene] sy
‘This selationship holds good for any system having different hazard
tions for ils components
(Bos
ailre rates of three components ae 0.065 * 10°, 0.18 * 10°, and
196 * 10" per hour. Evaluate the failure rate, MITF of a system and the
inblty at 500 hours if these components are connected in seis.
olution:
y= 0065 * 10° per nour
Je = 0.18 * 10" per hour
y= 096 * 10° per howe
from this, we get the reliability of each component as:
1, = expt 00065 * 10° * 500) = 0.9680
‘exp 0.48 * 10? * 500) = 09139)
expt- 0.96 * 10° * 500) = 0.6188
Ag= Ay + Aa + Ay = 0.065 * 10 + 0.18 * 107 + 096 + 10° = 1.205,
10° per hour.
L 1
1-1 _, -s30 ous
Ay” 1205 +10
MITE, =
System reliability, Ry = RVRSR
er system’ with mt components. Ifthe success of only one &f the
ponents is suiient to meet the iatended fonction ofthe system. hen it
wn as 8 m-anit parallel system, In reliability we represent such 2
by the parallel RLD as shown in Figure 1.6. These arc m parallel
‘between the IN and QUT terminals, andthe sucess of any one path
sess to suecenily rami information from IN end '9 OUT ers, Elise sea tl can be used for evaluation of the sytem reliably
ir dele in terms of sscesy erent and fae
ERE ipsinen, These arc:
Wye hie UE,
UE)
B= seeaT= RE, NE, 0. 0B) 4s)
pe Pi enn eye
Bol ~ RE) RE) ..rE,) (146)
eo is expressed as
22 Rel Hanes fi - pai asn
= PAD PAE) and ght) = PLE) (1.48)
mut be a ine
es sme reliability, then :
SLaM~ pen” and gem tant 145)
ed find out he minimum numberof Weil
A glo target system relay, Assuming that al
Is Nave the sume reliably value, wo have
. - ‘
__ Set hi orig + 28
logarithm and simplifying, we get
= lesttOe)
Tox)
For example, let us assume thatthe reliability ofeach wnit used in
system is 0.75. We want wo find out the minimum numberof paralet
red to mect the mission reibility of 0.9950. Using Equation (15),
m= 382 = 4 unis, That means we require four unite to saity the
ity target.
Rolibi
ash
‘of a two-unit identical parallel system can be expressed as
Ra app 1.52)
"pis the reliabitty of each unit.
This gives R= 0:96 for p = 08 and R = 0:9 for p = 09. The system
ly of 0.99 ean also be achieved by operating three units having
08 in parallel instead of two units having p = 0.9. This shows that any
of reliability can be achieved by adequate arrangement of units in
|. However, the technical feasibly of such arrangements must be
before final acceptance ofthe design. This s feasible in most of the
le systems, In mechanical systems it may not bea good choice. For
ipl, itis technically not possible to connect two shafs in parallel instead
Fi Alngle shat in a machine
Let us now see how many parallel units are beneficial. Consider 3
having m units in parallel. This means tat there ae (me ~ 1) rundant
none inital unit. Adding one more redundant unit to this system will
In higher syitem reliability, which can be evaluated by the folowing
jon:
Reeilt) ~ 1 = poy (53)
hs we can also evalate the incremental reisbiltystibued to the
Funltedded by We following equation
‘ AR w(t) = Rrnsi(t) ~ Rall)
= (= poy" — 1 - po
= a (1 ~ aor”
Teun be shown tht sing the frst parallel component to an exiting
ihe mos beneficial decision. The benefit fading subsequent paral!
‘progrevely reduces asthe number of pall components
Weight, space, and cost increase with each additional component.
‘tad asaya eguied before taking any decison.
Pun-now oenelder anv wnit-papalielredundamt-eyategs ia. whl all ca
ase . . .
301+ Relay toponeng ae Tsing
Ro eter css)
“The mean ime 1 fire ofthe system is
MITF. «36,
eae
os
MITE of a parte! syem when he nit tbls ae weg an
be elutes y te fone pode,
verre [i fla ogean]
1 1
had daha
1
oe ase)
Fr example, fora two-nit parallel system
139)
all units are similar,
eae es
TW
(1.60)
For an ‘m order parallel redundant system with ‘n° ld (ndependen
und identically.distributed) units baving same hazard rate, we get the follewines
ce ele togreeing + 31
aon
Therefore,
[Now let us consider the evaluation of hazard rate of two-unit pendent
and identical system. The system reibiity 18
Ryn) = 20% eH
amy
Since we know tat
fo, Rw
2° RO)
BO neh ene
i)
Equation (1.63) shows thatthe flue rate of the system is time depondant
even when the falure rates of the units are constant.
es
The fail rates of tee components ae 0.065 + 107, 018 + 10%,
056 + 10° per hour, Evaluate the fare rte, MTF of» stem andthe
reliability at 500 hours if these components are connected in paral,
Solution:
Ay = 0065 + 10° per hour
Jy = 0.18 * 10? per hour
4y = 0:96 * 10° per hour
From this, we get the relly ofeach component ws
ee eee 32_+ Relabiytogineenng and Le Testing
y= exp 018 * 10° + 500) = 09139
y= expt 096 * 10° * 500) = 0.6188
System relia,
Res 1-1 RL RA Ry = 0.9980,
1.10.13 koutof-m systems
‘out-of system i essentially a parallel system where more than one ofits
parallel components i equited to meet the system success, For instance, te
ofthe four engines i a commercial aircraft may be necessary to succesfully
ompete the fight The other two are added to inreas the system eis
Many such systems exist in modem industries, The models developed for g
simple parallel system do not apply to these systems. In fact the pale
system with" components isa special case ofthe kato syste then
= 1 2 kourofm system, m-k unite are redundant units and are dea
for te purpose of improving the system reliability. Sometimes sch tyetene
se known as partial-redundant systems to denote that k> 1. The Fats are
salle bari unas whose survival is a must forthe sucess of the system
‘The binomial distrbution can be used to evaluate the celability for
‘ystems with denial and sastically independent components. fp the
‘Probability of sucess ofeach component. then the probability that exactly
‘out of m components are successful is given by
Pon, oe(T}ea - pm as
Equation (1.64) i valid when we consider only oe case of redundancy,
vi. exactly «sucessful unis, However, itcan be sean that the system meee,
an aso be possible with more than i successes In each case ne mart cha
imo account all combination forthe evaluation of sytem reabity, Equation
(1.65) cam be effetvely used for thn purpose
te ST ra ors (16s)
+ Now consider the cave of oui nit having tenestea pauatshesort
as + Relay npnecing + 33
rn eat eneing +33
Rw a(t] .
(466)
For example, fora 2outof-3 system, we have
ooo
PG - 2)
For example, unit elibi
on.
R
P= OS, we get system elabilty
8 will have a ebay
‘2 0.992, Tis shows that the system elibility decreases wih The formule
for kout-ofm can also be used for reliability evaluation af» sere seme
1 this case, we subsiute k = m. For example, for the above syn If
= 3, it becomes series system with 3 components, Therefor:
Ry Aer
s=[) |r =r i”
08
is the reliability ofa seriessystem with three components It may be nod
thatthe components may be physically in parallel, but in the eeablty vnme
(RLD) they ate in series
The system MTTF can be evaluated as follows:
512
a= fy Roa aan
For the constant failure rate case, we get
a
may te noted that for k= 1, the system becomes « pale
And for k= m, the system becomes a seie-syatem,
‘The unreliability of a Koutof-m system can be expressed ung Ud
| binomial tilee probabties as 7 . ‘
34+ teiabiy Engng and te Tene
In some problems, it may be beter and easier to first evaluate the
system unrelatlit wsing Equation (1.69) and then evalat the system reliability
from tis,
hz @& 8
A system consists of 4 units. If any ofthe the units are operational, the
system i considered operational If he fie rae of each unit's O88 10?
Pe hour. evaluate the system reliability at 500 hours
Solution:
Failure eae of each ui
0.88 © 10° per hour
Reliability of each unit, R= exp(- 088 * 10° * 500)
6440
System reliably. Re = E 4C (0.6440) (1 - O.6440)"" = 05524
wert, = tS 663 hours
ass 107
1.10.14 Stand-by system
Stand-by redundancy i another important ype of redundancy. Tis is geerilly
followed atthe system level. For example, 0 meet the power requirements
uring peak load penod, vo generators ae installed in an industry. Each
{generator is capable enough fo meet the demand. Only one unt is operated
St any given instant of time by Keeping the other one in the stand-by mode
Many manufacturing industries follow this type of redundancy. From system
reliability point of view, tis ype of redundancy is much superior tthe
parallel redundancy. Sometimes, a changeover switch (automate or manval
or relay is used between the tand.by unite for emooth operation of he sytem.
‘The methods discussed inthe previous sections do not old good fr stand-
by systems snc the reability logic digram ofthe system changes depending
(on the position ofthe switch. This type of systems can be best modelled for
reliability sie by the famous Markov models, The basic concepts of Markov
‘models are discussed in the subsequent sections of this chapter.
«In this section, a much simplified prcedie is followed to study the
Feisty of stand-by models by assuming thatthe sith reliably is Unity
(perfect switeh. 1 can be shown that a Gamma model is adequate to sly
‘sch systems. Let there be total m units, out of which one unit 8 kept a8
Mand-y. Let the failure rate ofeach unit be 2 ad its
‘That means,
ability ate
eA gnEP NS
Chee Relay Engine + 35
‘The system reliability can be obtained by using the Gamma model
follows
an ym
2 gn erin
‘The system reiabilty can also be expressed in terms of the unit eb
by substituting =~ Inip) Thus, we get
=p (inp)
R
Fora two-unit system, we get
«can be shown thatthe system reliability of stand-by systam
fs better than that of a parallel sytem provided that the switch reliably I
‘The flare rae of each unt of wo-nit stand-by system is 78 * 10 par
hour, Evaluate the relability ofthis system ater 00 hous of pero
Solution:
0.75 * 10° per hour
Reliability ofeach unit after $00 hours,
xp(- 075 * 10° * 500
System reliability,
Rs = 06973 "01
1.10.15. Series-parallel systems
“Tn series poralel systems, a momber of units are’ arranged in series and auch
rnd ph ree seta ea
santa acto santa ter sa ca
rm nd cl ee
“heal oa
steps. Inthe ft sep the relablity of ech of tw parle abyss I i
36 + selabiny treme nL estos
(He HeH ee}
. t
fe He He }HeHe}
FIGURE 17. Serisparaiel sytem
step the reliability thus evaluated fo each subsystem is sed to evalate the
system eabiltytsing the equation forthe parle confgurstion, The following,
crample ilustrtes the procedure
Ee
Consider Figure 1.7. The ibis ofthe four units ofthe ist subsystem ate
(065, 072,077, and 0.82. The reliabilities of the thee uit in the second
subsystem ae 0.75, 067, and 0.95 Sia. forthe thd subsystem, the unit
rebabiliies are 090, 0.77. 0.68, DBI, and 082. For the Last subsystem, the
uni reablies are 0 95 and 087. We requ evaluating the system liability
Solution:
Let P, represent the reliability ofthe sth parallel subsystem. Therefore,
P, = (065¥0:7290.77,0 82) = 02955
2; = (075)0.67)095) = 04774
P, = (0.90¥0-7940.68)0.81)0.82) = 03817
P= (095)087) = 08265
System reliability
y= 1 (1~ Pd — Pot ~ PX ~ Py = 09605
110.16 Parallel-series system
“The prallel-series system i another system configuration in which a number
cof unit are arranged ina parallel subsystem and such subsystems are arranged
in series The procedure for evaluation of system reliability i similar to that
ofthe series paralel system as discussed before. The only difference is that
in th fist slp we evaluate the reliability ofeach ofthe subsystem by sing
the equitin for pall configuration. Inthe second sep, we evaluate the
eaten sattetettall th niin: fa: postin sentlitensia Ton tollediidlt suaninta
‘A arallel-series system is shown in Figue 1.8 It consist of four subsea
inns, cach hain oe compote
in prs Tee
components abo gen nFigue Weneed wea we san at
Solution:
Let, represent the reliability ofthe ith series subsystem. Therefore,
y= 1 (1-081 ~ 083) ~ 07841 ~ 0.75) = 09985
Pr=1 ~~ 092\1 ~ 081y1 ~ 082) = 09973
P\= 1-1 ~098y1 ~ 091) = 09982
aq eh
FAGURE 18 Pol sees tem
System relay,
y= Py PsP, = 0.9939
may be noted here that he serie pale and
3 Srisparalel and pre-e ome
‘ae similar in nate. Both these types of sytem configurations can ssa He
Sted by aermnve procedures Enon a te pl eso Tt
‘eho is discussed in deal nthe newt sexton oe
1.10.17. Non-series-parallel systems
‘There se many systems which 60 not have loge agra, sich
Serica or parallel or out-of. Such complex systems can be adel
8 non-senies-paralle structure for elibility stoies. The bilge network,
‘shown in Figure 19, isthe simplest nn-serier parallel suctre. This je
consists of five components and they are represented in a deed graph ak
town in Figure 9. Such systems ate popularly known as nework. The
complexity of the network sytem increas a8 the nnber of components
‘anid laterennradiitha Wrermae, tuathine se je
|. our
z GH
FIGURE 1.9 A bridge network
The RLD approach helps us for reliability modelling in many ways.
(One of the most important and frequently fllowed techniques isthe path set
snproach, A path sti set of components frm IN to OUT terminals of the
1nerwork, The success of all ofthese component of path sts esxemil for
'chieving the system sucess. In complex systems, thee could be many such
path seis. This means that each of there path ses forms parallel paths fom
1 to OUT terminals. For the successful operation ofthe system, there shold
‘beat least one continuous path berween TN and OUT terminals. Therefor,
the probability of success of cach path can be evaluated by using the
‘multpication law of probability and tht of the pale! paths bythe union
law of probability
Although te failure ofeach component assumed tobe independent
the file ofthe paths isnot independent since some components ind place
‘in mre than one path That means the pth sets reno statistically independent
‘Therefore, care must be taken to see tht the iderapotance law isnot violated
hile modelling this system. This is explained as shown below:
Let Ey isthe event of occurrence of (1.2)
Es the event of occurence of (23)
Let Ey and are independent events
‘Therefore,
PED = pops
PED = oars
‘Then, probabiliy of joint occurrence of Ey and Eis
PIE, 9 Ey) = POE) PLE) = pips
‘Thats, py = ps for any value of &
Assuming tat the reliability ofthe component i= p, ad the probability
of success of the path j= Rj then from the RLD, shown in Figure 1,10, we
an derive the probability of success of each path as follows:
R= prs x
Kap,
nO
OG
FIGURE 1.10 Lopic diagram forthe ster,
‘Then the clibibty of the entire seueture ie
Re=1~ (= ROU RYd- RY
By Ry + By RAR ~ Roky RyRy + RRA,
(On simplification and using the idempotence law, we get
Re= Pum + Dims + Paps PPPs ~ Pipapws~ PirPv + PIPsPwUN
Pisma~=ps=p
y= 2p spat ap my
the component 5 is bidirectional element, it tani signal i
both directions. The logic diagram forthe system is shown in
re MN
ae f—+ our
FIGURE.1321- Logic diagram of bridge network with bidketorm man,
‘The lability of the bridge network with the above bidretonu element
‘an be derived from the RLD given In Figure 1.11 as shown below.
Rem (1 =~ mpd) (= po (1 ~ ppd A= pwd)
sui. tencelncanmae i iiiilea iate a: a uses 40+ seth Enpneng n e teing
For the case pj = p forall we get
Re= 2p? 4 2p) Spo 26? an
—||—r”rté“‘CSOSO_
davaa we 016 03, 09 O97 08 O68 wey he ee
Pa hc tte ty poke oa
FIGURE 12 Non-ens-pralel system,
Solution:
In this cise, i canbe sen that thee are four nodes which indiate thatthe 3
highes order of minimal pth se is &~ = 3. The following are the path sets,
Path sts of order 1, T; = (6)
Path sete of order 2.73 = 4.31 4
Path sets of order 3, Ty = (14)
snd T= (234)
‘The probability ofeach path set can be expressed as follows
Dee
PIT) » papas
Now, the expression or system reliability can be obtained by combining
the probabilities of these path sets as follows
Ree PUTUBUTD
PUT, UT) = Pia) = PUT) + PCT) = PT, TS) ® Do + Pay = Da
BOT. Tw MTL 0 TO oT. MT. 7. De:
g Coe taba gnawing © AT
‘Therefore
Ry= Tau Ty) =
= P+ Pas ~ PuDe * Pa * PPP
+ PT) — PT Ta)
POD 6 + pos ~
Paps) Ours * Pap. ~ PIP)
= e+ Pas POD + PIP + PPPs ~ PPD. ~ PPD ~
PPDO4+ POPP PPDDS- PPDMs* POIPD:* POEM
+ PEDID.~ PIDOIL
Sebstratng the numerical valves ofthe x
equation, we get the system reliability as Ry
ables ofeach unit his
9866,
1.11 RELIABILITY EVALUATION USING MARKOV MODEL
Markov models are functions of two random variables viz the sate of the
system (x) and the ume of observation (0). Both and rean be either dace
‘or continuous random variables. Therefore, the total numberof combinations
of variables is four. Based on the nature ofthese variables, the Markov
‘models are named as ether Markov chain or Markov process. See Table 17
fo deals
TABLELT”Mactow Models
‘Shalem ware
Time of observation) ‘Type of aed
Discrete Discree Markov chain
Discrete CContinsous Markow pracent
Continuous Discrete Non-Markowian
Continuous ‘Continuous Non-Markoian
In the eli ste, we use only Markov chains and Maroy poses.
‘The numberof sates in a Markoy model can be found outa follows. Lat
there be m units ina system, each wit having m stats. The tol numberof
slates ofthe system in the model isn", For example, the mechanical vale
used ina hydraulic system will have tree states, viz uly operational ft
‘in the open mode (anable to cose), and failed in te shor mode (unable
open), If two suc valves are used in a system, then the Markov model fot
‘he system of valves will hive nine states. One ofthese 1” sates is kOOR
5 the absorbing sate. The absorbing slate is that state of the system 18
‘which ifthe system enters, it will never come out Tis slate may be
alter finite number of steady states. In oer words, the probability of
continuing a system inthe absorbing sate is unity
“This technique is applicable only for thse components that have WO
oes, vie success and fol. Thy component flue rate myst also be constant with respect time. In other words, the components failure process
must be memory less. tis assumed tat the component hi only two sates:
Sate 0: The component is operational
State 1: The component is fed,
As discussed eae, the Fst step in 3 Markov process isto draw the
state transition diagram of a given system, Let us is model a nonrepeiable
‘component. Such components, after fire, will not be repaired. Instead, it
Wil be epace ty anew component Examples for sch ates ae incandescent
Tams, stelites fuse wires, mises, ete Such stems ae studied using the
vwellAnown Markov models, Let the component e inthe operational sate
(0) a time 1
1.11.1 Reliability Evaluation
‘This seein discusses the procedure for reliability evaluation ofa non repairabe
system or component the Markov model. nally the component is assumed
10 be inthe operational state 0. Let the file rate ofthis component be A
‘The probability that the component will ail and reach state I inthe time
Jnterval Aris 24r This is sraphiclly llrtrated in Figure 113.
° eo) 7
FIGURE 113 Markov model fra single unit system,
Le
al) = the probity thatthe component remains operational at time ¢
P40 Af) = the probability that the component remains operational at
ime re ar
.( = the probity thatthe component i in the filed state at ie ¢
ult + st) = the probability thatthe component is in the failed sate at
time 1+ a
‘The state probabilities can be expressed by the fllowing formulation,
P+ 81) = Probability that he component was in state 0 at and does
ot change to sate I after time A
= Fae = Aa)
e 1 Pat A Pol) = Agar ca
: Chanter Rehabley tagemenng #43
Since the component is non-repaiable, the probability ofits remaining in
state 1 is unity because the component is no epated The probaly of the
‘component being in state I at ime #+ ais the sum ofthe probability that
inal (time the component was i tate I tthe probbiy that was
in state O and then changes to state 1 in time Ae Mathematically, sci he
expressed as
PU + an) = APD + Pun) am
‘The above two equations ae known as the slate transition probabil
‘These can be rewrite 3
Raran - Ro
= ARO
7 ”
Aare) BO ar 7)
x
Ie can be shown that as A» 0, we have
0. _ anny us,
ae
‘This diferemtal equation can be solved for deriving the expres fo
‘Pa, which is the probability that the item isin the upstate ak tne 1. Hy
efnition, Pi) isthe reliably at time
Intro
4.10
Where cy and cy are constants of integration. The values of then onsale
‘can be obtained by substituting the inital conditions.
Intaly, when = 0, the component isin state O and, than)
PaO) = 1 =e 4
‘The reliability ofthe component a times the probability thal the component
Is in state 0 at vie 1. Therefore,
Doom on $4 + fetabin eogeenn and ile Tsing
Similar methodology ean also be developed for modelling repairable
“stems and components. In such cases, an addtional transition Fate must De
used to take into consideration the repa activities. Repair activities bring the
‘tem from the failure state 1 to the sucess sat , with a rensiion rte sal
te the fepair rate Instantaneous and steady sate availabilities can also be
‘Stulusted By this method. This is beyond the scope of this book and is
(Gerefore. not discussed ere
The sate trans
ion diagrams for afew common systems are presented
me TaNe 18
TABLE LB. State Transtion Diagrams
Seen ‘State tansiion diagrams
“— SD
on ae Oe Rea
—— 1 ifeabity geting +245,
2 Evaluation of MTTF
ren cn ao wd or vanion of MTT of aye Let |
sate transition matrix fora sytem having total n™ mater of ac
eee 7 wl ba") man et repent enya ony
the sume order of T: Let the absorbing sate be ‘a Now the MUTE
sm can be evaluated by the equation
am
‘Dy i the determinant obtained fom the matrix [-P] after deleting he
ow and column, besides the ath tow and column. D, is the determnent
eleting the ath cow and column from the matin {1-P]
“The method is ilustrated for a wo-nit redundant system. The sate
lon matrix for the system ie given by
Died, Asta 9
BR at
{allure rate ofthe unit be A = 0.88 * 10% per hou, Thesefore,
MITF Is equal
MITe=
3
ITTF, » (= 25900 hover
MITE = Saar "| 46 + etait ages ade Tog :
‘with repair, Lethe repr rate Be = 0.08 per Nous The system MTTE is
gual 19
w+34 008 + 3*058+ 1
MTR,
Base
1.2110 hours
‘This also shows that a rwo-uit redundant system with cepa performs much
‘ener compared to that without repair The readers are encouraged to do
similar exesses on ther typeof systems and derive the equation for eiblity
and MTTF using Markov models,
1.12 RELIABILITY IMPROVEMENT TECHNIQUES
‘One of the main objectives of caring out rliability sues to assess the
present status of our sytem or the system under design and development
That ist check whether out systems are able o meet the design and mission
Tequirements or nt. A gap analysis is then cared out and trateges ae
‘lane to improve rebbliy. Reliab canbe improved by various ecniques,
Some ofthe most important and widely used wechniques are briefly discuseed
here. The following methods are used o improve the reliability of engineering
components and systems
(i) Usage of ener components
(in) System simplification (removing complexities)
(i) Deeaing
(iv) Redundancy
(0) Coatoling the environment
(i) Maintenance
(i) Removing easy furs by burnin
(it) Worst case design (to minimize sensitivity to variations in parameters
and operating condition)
(ix) Process control and quay assurance
(x) Human relailiy
1.12.1 Usage of Better Components
‘Components are the bulding blocks of sytem. Component reliability can
‘be improved by the fllowing methods
+ High quality evel (design and manufacturing)
‘+ Minimum variations in specifications
+ Lower fire rales
++ Factor of safety (PS) margin,
Cube retbiy egmeig + 47
‘These atibutes cane achieved by the following stioes
+ Good design (engineering)
+ Selection of proper materials
+ Adequate technology of manufacturing
Suit quality cont
+ Continuous updation and improvement
+ Proper vendor evaluation and seletion
‘Asan initial step, the ertical components in the sytem canbe idenied
and improved. Reliability improvement of other components can be ie Ih
the next stage. This isa costly affaic but quite efecive to a eran exten.
‘While atemping to improvement reliability of components we must hep in
‘mind two important factors. These are the effects of reliability improverseat
fon cost, and the technical or manufacturing feasibility of each improvement
rean be shown that the cost of components increases when its iil) 6
improved. The cost fuction hasan exponential relationship to the relabiliy
‘That means, fer atsining ceri evel of ibility, if furtherimprovene
are ted the cost wl nceas exponentially as shown in Figure Lt. Teron
an adequate wade-ff policy must be planned before attempting w improve
Component reliability. The second factor is the present techaical and
‘manufacturing capability available inthe shop. Fuber in many cases
found that even with further investment it may not be posible tw peels
perfect components. This is due to various controllable and uncontllae
random factors affecting the production proces during various sige of
production.
FIGURE 1.26 Reality and cos 48+ eatin Engneeng ad de
112.11 Cost of reliability improvement
Reliability and cost can he related using the following equation
Cw oct cs
where
C= design and development cost of component
= lability ofthe equipment
‘and b are postive constants
‘The level of operational reliability is to be decided by cost constrains,
snd operational requirements,
Le
1206; r= 085
C= ach = gi?
aT
We get = (66) omes
Les
be
0s. eve,
This means thatthe cost of relibiity improvement fom 0.610 0.85 is
approximately § umes. At higher levels of reliably, the cost of improvement
lwzreates exponentially and prohibitively high. There are also practical
sitficutes in manufacturing for vey high reliabilities im many situations. In
son. there ae economic as well as design and manufacturing constants
‘hile thinking about very high reliability. Since the production of a perfect
‘component is almost impossible, andthe cost of part improvement is very
Tuzh, his approach may not be profitable for larger complex system. In other
words itis impossible to achieve any level of igh reibilty by corpponent
level improvements alone. We require to apply other methods for reliability
improvement apart from trying to improve component reliability. Te system
reliability improvement methods are discussed inthe following sections.
1.12.2 System Simplification
System simplication is done during design stage. Most systems can be
signed in many diferent ways. The intended funtion isthe same. 80,10
simpliy a system, creative thinking and design is necessary. The numberof
components and alto the variety of components in the design shouldbe as
few as posible. Usage of standard, jterchangeable pats in very Important
‘The desion chaoui ha naatle concen cnlarchangesble paris la very important.
’ Scraper ti tngnemng + 49
All components used in the design should be ofthe requted rebrand
these must be easily avilable :
1.12.3. Derating
'Deratng isthe reduction in electrical and thermal seses applied to 3 pat
inorder to decrease the part flare rae, which enhance the equipment lath.
Derating canbe defined as a design srategy to improve the relabiity uf
at by operating the par at steses vale less than te maximum rated vale
The rated parameters can be vollage, current, power, force, torque, spec,
temperature, humiy, water hea, et. Ihs ome smilies othe trail
‘esi strategy of factor of safety (FS)
imum lowabe
Fstor of sf = Masia aio 11490)
‘As aie of thumb, most of the electronic components are wed thie
50% rated screngths or FS = 2. In mechanical systems, the factor of silety
is less (about 15105) or more (above 8) depending on the aplicain, The
benefit of derating shouldbe studied before taking the decision of how mich
to derate. The benefit varies from component to component and also Im
pplication o application. Derating will result i higher eos of the system.
Therefore, «trade-off policy should be planed
1.12.3.1 Importance of derating
1, Deatng the most tet tol or the designe to dese he
fale rates of part This cn elo compenste fr my of the
‘arabe lett i any design
2. Alleleewoni pars prodiced nan sey ine ae ot ide
Sue ierences and variations exit rom ep oe. Prope
i deaing wl lp to compere fr these prio pa vrai
Sd simi hie pat othe supmen els
3. Blecroni parts with ntl manufactrer' par rs yb
pchsel fom differen sop While thes er are clecly
Inercanged there maybe sgn ference nds, ei
and manfotng process. Dering wil pt compet
these difernces. oe
4 The designer wil yo aici the varios els at enone
eal ettenes owiehe equipment ay be bjt 1
{ue to anticipate propery the input of al of thr vale,
eng an provide an adonl margin of se
Tei alto apparent prs a the epoca cen parame
see not over ter nr fe, Proper dean vl
‘not completely stable over their entire life, Proper derating will AelabityEngneing and Ue Test
1423.2 Effect of derating on part stress reliability
prediction
During the useful life ofan electronic pr, its elit sa function ofboth
{he electrical and the thermal suesss lo which the parti subjected. Increase
te thermal sees directly increases the Junction temperature, which will
Increase fur rte according tothe mathematics! model of fare rte