14

RELIABILITY OPTIMIZATION

14.1 INTRODUCTION
studied that the reliability of a system can be improved considerably by
Uptil so for, we have :

) Introducing redundancy either in the sub systems (i.e. units) or in the elements.
()The element or component redundancy is superiorto subsystem or unit redundancy.
Therefore we see that, reliability of a system can be increased to any level by adding a large
number ofredundant components. But this adding of redundant components has certain constraints
as bellows.
the increase in cost. As we go on increasing redundant components,
(a) The first constraint is
the cost goes on increasing.
6) The second constraint is the space. Enough space may not be available for adding large
number of redundant components.
Besides above two constraints, there is another objection to increasing element redundancy
indiscriminately. To understand this we take following example
Example
n a defence system, a large number of stations are st up to send a signal for launching
missiles to attack energy war planes. This planning is based on the detection, of the approach of the
planes. In this
way the reliability of the launching of missiles increases with each increase ofa
Station. But the probability of getting a false message sent by accident also increases. Therefore we
by increasing the redundant components to large extent, the probability of getting false
a
nat
massage also
increases.
account
Cnce when we analyze the system reliability, we have to take into
it should.
The probability ofthe system functioning satisfactorily at a time when
probability of the system functioning at a time when it should not.
The
Now we assume
that
when it should=p
aThe probability that every component functions every time when there is no signal=r
probability that component does not function
n he
a
Cr Words, a component will function every time; when should.
it
276 Reliablility and Maintenance Engineerina
ng

0p<1
ponent will not function every time when it should
nor

0<r<1
as shown below:
*Ople system of filler comnponents connected in parallel

D
Fig.14.1
Order to function the system satisfactorily it is necessarythat when there iS a signal at,
Component functions and when there is no signal, none of them funetions. The probability
that at
e One of them functions, when signal is given equal to the probability that none of them
fails,whenthe signal
is given = 1-(1- Py"J
Theprobability that none of themfunctions when no signal is given
Therefore system reliability

R 1-(0-p*J .(1)
If there are m components connected in parallel, then the system reliability

R "-(1-p"] .2)
For a system, which does not give a false alarm r= 1

System reliability R =
L1-(0-p)"]
.(3)
Therefore we see that if m increases, the reliability of the
system decreases.
14.2 DEPENDENcY OF R ON M
Consider a system with m camponents and another system
with (m + 1) components.
R 1-(1-p" 4)
and R L1-(1-p*I ..(5)
Now ratio ofthe reliabilitys m

1-(1- pm* r--py+1

1-(1-p" 1-(1-p" (6)
Put 1-p q
 Optimization
iability
Kelat

277
Then r-40++gt..)
(1-q1+q+q +..."-
r++qt..+)
(1+q+q +..q")
(7)
Let m= 1,2, 3 ... then
(1 +q)
- tg+g2 J- ..(8)

..9)

1+
1++q J .(10)
As g isless then one, hence as m increases, 1, decreases.
But we are interested in reliability optimization
i.e. increasing
1.<l, this will depend on p & r In above equations put = 1, so
Ror find a first stage when
1, we get
r(l+ g) =1 ..(11)

.(12)

1+q+q .(13)

and
lakefirst equation with ,-1.4,is the ratio ofthe reliability oftwo systems one with 2 elements
other with
one element.
netore for a given set of valuesforp and r, we use 2 elements if
Cement. In the equation (4) above, for all pairs (p, r), when =I we , ,>
1,otherwise we use only
should use only one
element.

When (p, r) lies between I and ,= ,p = I

we should use two elements. For exampleifp = .85
Si87, use only one component.
miler calculations If = .85 and r between .87 and .98, use two components
can be done for other values of p &r.
Example 1 chemical complex to ring an
:It is decided to install fire sensor in a section of a the
The
n case
Droha of fire. The reliability of the primitive sensor to activate alarm in case offire is 0.7.
nsors
probabil y
the sensor will not activate the alarm in case of no fire is 0.95. How many
*nSors that Wliat is the
to imaximize the reliability that the massage is sent at the right time?
bevoinstalled
Peliability
liability
value? d
 Maintenance Engineering
Reliablility and
278
Solution : The ratios of reliabilites
r(1 + q)

1+q+g)
Pgiven=,7, henceq= -p= 0.3 1

rgiven = .95
we are interested to find, when.

Now
T.
95 (1+.3) = 1.235

95 1+ 1.0165
1+3

.227
4, 95 1 .969
1+.3+.09
Hence at m= 3 ,<1
Therefore 3 sensors should be installed to maximise the reliability.
Reliability of system,
R, 1-a-p*)]
=.834
(95L1-(33]
Hence reliability ofthe systemn is 0.834
Ans
Example 2: In a system p= 9 and r be adjusted to a required
dces not allow use of more than 2 components. It is value. The spacè in the system
necessary that the system reliability be at least
0.9. How large should r be ?
Solution : Given p = .9
Therefore q=1 -.9=.1
No of components réstricted =2
system reliability> 9.9, now we findd out
r(1+.1)
i.e
!, t,
 ability
Optimization
279
or

systemreliabiliy R= [1-¢J=r0.1)3]=r?1-.01
Or
9=
99
90 10
or
99 99 11

or

or r> ,95
r should be grater then .95
14.3 OPTIMAL REDUNDANCY
Let
cost per element
C

p =reliability perelement
m = no. of elements in pårallel
The cost and reliability of a parallel system can be expressed as
c Mc, m, method of redundancy)
R Ap, m, method of redundancy)
Thus the problem of optimal redundancy is reduced to the determination of those values and
thosemethods of redundancy for which R is maximum for a given value of C or C minimum for a
given value
of R. A simple method is used to solve the redundancy optimization problem, with cost
and volume as constraints. We make the following assumptions.

The number of redundant elements in each stage is a continuous variable and the final
SOlution is obtained by rounding off the solution to integer values.

The cost or volume at any stage increases linearly with the number of elements
in that
2
stage.
3. The redundance is active
4. The failure of element are independent
5, The probability ofelements
14.4 PROBLEM FORMULATION
operative forthe successful
parallel system at least one element in each stage must be
op .Series the reliability of
with a reliability q, are used at stage then
i,
thOn f the system. If m, elements
Une entire
system having n stages is
 Engineerina
Reliablility and Maintenance ng
280

R i=2Rm) [1-9m)] =

-qm) i)
i=l|
and the unreliability
is

Q-1-20-4)
i=
wnere q"are small and their products may be neglected. "Then

Rl-41-a i=l i=l

2
and Q 2"
i=
...(3)
he total reliability ina multi constraint problem is limited
bythe followingconstants

2iSe; =j = 1,2 ... k. .(4)

Where e, is the ith resource used at the ith stage.
The constraints might represent total cost c
and total volume C, etc. Equation 2 and 3 are
objective functions and equation 4 is a
functions. constraint
Now we mant to find the value of m such that

i) To maximize R =21-g
i) Subject to m, 21 i,-1,2,...n
eiy20 J,-1,2,..k
ey m, Se
Now taking log ofequation (1)

logR 2log(1s)
i=l
Now differentiating

aR-e"088
R dmi 81 1-g"
 geliability
Optimization

281
dR log(g,) A
or dm 1-" i=l

=8 log(4)1-)
Negating products of gm
dR
ddm-8 logg, = Oi =1...n

i.e. og 9 42m2 log42e...

Assuming that m is known.
Other values ma, ma.
m, can be known

log ExH
m y

Where E
g log 9i
H log 91
14.5 RELIABILITY VERSUS
CosST
In
order to realize the effect
of cost on system reliability the cost constraint was increased
steps of 25 and the redundant equipment in
per stage and system reliability were obtained for
step. The
plot of reliability ofcost is shown each
below:
.100

95

90

75

70L
25 50 75 100 125 150 175

Cost
Fig.14.2
 Reliablility and
Maintenance Enginceri
282
step corresponaing ne a significant
to condition ofan
Relia y growth is actually a stap case lunction, each
in cost Will not resunt
in
any increase
in a stage. It shows that beyond a point
nt
fmerease in system reliability.
produce a highly reliabie component. If we want.a
t is 1act that it is very expensive to
ystem

Two
ot reliability .99, this can be constructed in the following
equipment redundant system with equipment
.
wayd
reliability
reliability of at least.8
equipment redundant system with equipment
hree reliability of at last .07
rour
Ther
equipment redundant system with equipment
cquipment reliability is less, we require more and more equipments. Since
wien the increase in the number is balanced by the decrease
reliability equipment costs less, tlhe cost of
at which the
e indivisual equipment cost. However there is a lower limit of equipment reliability
oge cost starts increasing. There is no linear relationship between cost and reliability.
Let c, be cost of equipment at stage i

e
c a,
1-P)
where reliability of ith equipment and a; and b, are constants.
P equipment
The number of equipments required to activate the assigned stage reliability of Ra is
log(l- Ra)
log(- Pi)
The cost of the ith stage is

bi
1-
C a, 1og (-Kai log(l pl - pi)
for optimum value of pi with respect to
cost, differentiate and equating to zero
de
p 0
(l-P) =0
+

b, log (1-P)
The system cost isS

= 2ma, e
i= 1-P)
where p, can be calculated irom above equation and

m
log(-R
log(l-P)
m, are rounded off to the next integers.